What Drove The Hydrogen Madness? The Detour Was Avoidable

The problems were known long before the latest hype cycle.

In May 2023, a senior executive responsible for decarbonization at a multibillion-dollar transportation company asked me, with considerable exasperation, what drove the “madness on hydrogen.” They had an engineering degree, an MBA, serious strategy experience and direct responsibility for making a large transportation business less carbon intensive. Hydrogen proposals kept arriving for applications where the energy balance and economics looked awful. That conversation became What Drives This Madness On Hydrogen?, in which I described a many-headed hydra of fossil-fuel interests, technology companies protecting investments, governments protecting industries and revenues, financiers following advisers and enthusiasts reinforcing the whole structure through confirmation bias, familiarity and loss aversion.

Three years later, I would keep the broad diagnosis but change its weighting. I put too much explanatory emphasis on individual psychology and not enough on organized economic interests, long-range corporate strategy, competence preservation, lobbying and institutional processes capable of producing adverse evidence and then preventing it from terminating a favoured pathway. The cancellations, missing customers, cost overruns and operating failures that have accumulated since 2023 are useful because they validate earlier analysis, not because they revealed problems that could not have been understood beforehand. By the time I wrote the Madness article, the case against a general-purpose hydrogen economy had already been made repeatedly from first principles by credible engineers, scientists and energy analysts, and I had spent several years independently reaching much the same conclusions through work on synthetic fuels, transportation, hydrogen demand and proposed North African hydrogen exports.

I certainly was not first. Ulf Bossel, Baldur Eliasson and Gordon Taylor had laid out the thermodynamic problem in The Future of the Hydrogen Economy: Bright or Bleak? in 2003, tracing the electricity required to produce, compress, transport, store and use hydrogen and showing why a hydrogen infrastructure was intrinsically much more energy intensive than using high-quality electricity more directly. The US National Academies examined the complete hydrogen supply chain in 2004, including production costs, efficiency, distribution, storage and end use, and warned that the barriers were formidable and that even an aggressive transition would take decades to have substantial effects. Joseph Romm, who had served in senior US Department of Energy roles, brought the argument to a much wider policy audience with The Hype About Hydrogen, warning during the previous hydrogen boom that promises of future hydrogen vehicles risked deferring technologies capable of reducing emissions much sooner.

The warnings continued as the current hype cycle emerged. Chemical engineer Paul Martin was publicly dissecting fuel-cell vehicle energetics by 2017 and by March 2019 was working through the much larger implications of making hydrogen from renewable electricity. Cambridge engineering professor David Cebon brought the same first-principles approach to heavy transport, heating and storage, eventually framing the strategic choice as a hydrogen economy versus an electron economy. By 2022 Jan Rosenow could publish a review of 32 independent studies of hydrogen heating and find that none supported widespread use for space and water heating. This was not a tiny contrarian literature that appeared after deployments began failing. Serious analysts had been repeatedly pointing to the same structural problems.

I happened to be fortunate in one respect. Instead of receiving a future green-hydrogen price from a trusted analytical source and treating it as an input, several assignments forced me to build the chain myself. In 2019, my assessment of Carbon Engineering’s air-to-fuel proposition required putting direct-air carbon capture, electrolytic hydrogen, synthesis losses and electricity consumption into the same calculation. The completed cost comparison with direct use of electricity found the synthetic-fuel pathway dramatically more expensive than putting the electricity directly into a battery vehicle even when I made assumptions favourable to the process. Hydrogen was not a detail in the result; making it and then processing it again was one of the principal reasons the energy and cost chain deteriorated so badly.

By 2020 I was explicitly arguing that the latest hydrogen-economy cycle was mostly hype while retaining hydrogen’s legitimate industrial uses. In 2021 my hydrogen-demand projection was already taking most road transportation, building heat and routine energy storage out of hydrogen’s future growth story. The 2022 Northern Africa work then forced the supply economics into the foreground. Europe was imagining Morocco, Algeria, Egypt and other countries making enormous quantities of renewable hydrogen cheaply enough to export north as a new energy commodity. My conclusion in Hydrogen Can Be Green, But Won’t Be Cheap was that those countries would generally obtain more value by using renewable electricity domestically and exporting electricity where practical. Green hydrogen could certainly be produced. There was no convincing basis for expecting it to become remotely as cheap as the export, transport and energy-carrier narratives required.

Michael Liebreich provides a useful example of how even a very strong analyst could reach the right conclusion about hydrogen demand while inheriting the wrong assumption about hydrogen supply. His 2020 work was already sharply skeptical of most proposed uses, and his subsequent Hydrogen Ladder became one of the most influential ways of explaining why direct electrification should come first. On supply, however, BloombergNEF’s analysis assumed steep electrolyzer learning, extraordinarily cheap renewable electricity and green hydrogen eventually approaching roughly a dollar per kilogram in favourable markets. BloombergNEF was a source Liebreich had every reason to trust, but that meant much of the complete electrolysis-plant economics sat inside an institutional black box rather than being rebuilt from the components upward.

Liebreich opened that box much faster than most institutions. Within a few years he was interrogating electricity prices, utilization, hydrogen transport, ammonia conversion and full project economics much more aggressively. His later Clean Hydrogen’s Missing Trillions challenged the very low electricity prices and capital costs required to make major hydrogen strategies work, and his subsequent analysis has moved to the conclusion that green hydrogen has no plausible pathway to the affordability once assumed because electrolyzer stacks are only one component while electricity and heavy electrical, chemical and civil engineering do not have five- or tenfold cost reductions available. The episode is instructive precisely because Liebreich is such a capable analyst. If someone able to tear apart the assumptions can temporarily inherit the wrong answer from a source he reasonably trusts, ministers, executives and journalists consuming the headline numbers are even more vulnerable.

The core supply problem was straightforward. Electricity dominates the variable cost of electrolytic hydrogen, so operators want it to be extremely cheap. The cheapest wind and solar electricity is intermittent, however, and restricting production to the lowest-price hours leaves a capital-intensive industrial facility idle for much of the year. Raising utilization requires overbuilding generation, combining resources, adding transmission or storage, or purchasing electricity during increasingly expensive periods. Cheap electricity and high utilization both make hydrogen cheaper, but they cannot simply be assumed independently at their most favourable values. My 2022 analysis called this an economic seesaw because pushing one side down tends to push the other up.

The electrolyzer stack was also repeatedly confused with the electrolysis plant. A real facility needs transformers, rectifiers, water treatment, pumps, cooling, controls, purification, compression, piping, buildings, safety equipment and substantial grid infrastructure. Much of that comes from mature industrial supply chains that have already experienced decades of cost optimization. Solar modules and battery cells are very different products: standardized, produced in enormous quantities, amenable to extensive factory automation and accumulating huge numbers of manufacturing doublings. A site-engineered chemical and electrical plant does not acquire the cost curve of a solar module merely because both technologies are associated with decarbonization.

By 2025, actual project costs were confirming what the reference class had suggested. My comparison of major institutional electrolyzer-system forecasts with observed project costs found earlier projections undershooting real-world evidence by roughly 60% to 300%, with major organizations progressively revising their assumptions upward. The embarrassing aspect was not that forecasts can be wrong. Electrolysis was not a mysterious new industrial process, balance-of-plant engineering was familiar, learning-curve analysis was mature and reference-class forecasting existed precisely to prevent analysts from assuming that a favoured technology would enjoy exceptional cost decline merely because advocates expected huge scale.

Storage and distribution made the cheap-green-hydrogen proposition harder again. Hydrogen leaving an electrolyzer has to be compressed and stored, moved through pipelines or trailers, liquefied at considerable energetic expense, or converted into another molecule and processed again. Pipelines become cheaper per kilogram when enormous volumes flow consistently, which generated one of the recurring circular arguments of hydrogen policy: cheap hydrogen would produce demand, demand would fill pipelines, pipelines would lower delivery costs and those lower costs would make hydrogen cheap. Similar reasoning surrounded “surplus” renewable electricity, which was treated as though enormous quantities of nearly free clean power would sit around waiting for electrolyzers rather than being competed for by batteries, heat pumps, electric vehicles, transmission, industry and other flexible loads capable of extracting considerably more useful work from each megawatt-hour.

The distinction between green hydrogen and cheap hydrogen was therefore never peripheral. Expensive green hydrogen can decarbonize applications that genuinely require hydrogen molecules. Ammonia, methanol and some prospective iron-production pathways can justify paying a premium because hydrogen has chemical value. A general-purpose hydrogen energy economy required something much more ambitious: hydrogen had to consume multiples of the electricity used by direct-electric competitors, pay for additional conversion equipment and carry the cost of a new storage and distribution system while somehow reaching customers at a competitive price. The recurring $1 and $2 per kilogram forecasts were not harmless optimism about a component cost. Vast portions of the proposed hydrogen economy disappeared if those prices did not materialize.

Up to this point, the story is about why cheap green hydrogen was never a credible foundation for a new energy economy. The more uncomfortable part is why fossil-fuel companies and automakers had long-range reasons to keep that future politically alive, and why some public institutions ended up telling executives and policymakers almost the opposite of what their own technical evidence showed.

Read more

Hydrogen occupies a remarkably large share of the energy-transition conversation for a technology with remarkably little deployment behind it. Fuel-cell trucks, hydrogen buses, hydrogen trains, hydrogen ships, hydrogen storage projects and one-off demonstrations recur through corporate press releases, government announcements, trade publications and mainstream coverage as significant branches of the transition. The cumulative impression is of a technology progressing alongside battery-electric and direct-electric alternatives across multiple sectors, with each new first, pilot, project, partnership or procurement adding another piece of apparent commercial momentum.

JCB’s recent Hydromax speed record provides an unusually clear example of how that happens. One purpose-built hydrogen-combustion record car completed two timed runs at Bonneville and averaged 406.320 mph. The achievement was real and independently verified by the FIA, but its commercial evidence was narrow: one vehicle, no fleet, no customer utilization, no operating economics and no repeat procurement. Ground News nevertheless grouped 127 news sources around the completed record, after an earlier pre-attempt publicity cycle had already generated an 88-article cluster. Those counts include wire syndication, partner amplification and independent reporting rather than 127 independent validations, but that is precisely the point: the underlying engineering event remained one commercially irrelevant car while the number of opportunities to encounter it as evidence of hydrogen progress multiplied dramatically.

The deployment data show the same distortion at market scale. Across the bounded 2022–2025 announcement register behind this analysis, battery-electric technologies actually produced somewhat more positive milestone announcements than hydrogen: 206 versus 114 across six common sectors. Among announcements representing scaled deployment or repeat procurement, the difference widens to 168 battery milestones versus 55 for hydrogen. Yet even that three-to-one maturity-weighted announcement advantage bears little resemblance to the physical markets underneath it, where battery and direct-electric deployment exceeds hydrogen by factors ranging from roughly 20 to 10,800 in the directly comparable cases.

Freight trucks show the same compression. The announcement register contains 26 hydrogen milestones and 39 battery milestones, a ratio of 1.5 to one. China had about 18,000 fuel-cell heavy trucks at the end of 2025, compared with about 366,000 electric heavy trucks, a stock difference of roughly twenty to one. Using the broader medium-plus-heavy category, the best reconstruction is approximately 24,000–25,000 fuel-cell trucks against an electric fleet above one million, placing the deployment difference above forty to one.

Transit buses look similarly close in announcements and similarly distant in deployment. There are 47 battery milestones and 23 hydrogen milestones in the register, almost exactly two to one. Hydrogen buses are not merely prototypes in this sector; substantial fleets have been purchased and put into service, and transit is the one common sector in the normalized dataset where hydrogen announcements reach commercial stages comparable with battery announcements. Even so, the physical fleet comparison is about 15,000 fuel-cell buses globally against more than 680,000 electric buses in China alone, a deployment gap greater than 45 to one.

Maritime pushes the discrepancy much further. Thirteen hydrogen ferry and harbour-craft milestones sit beside 27 battery milestones, again approximately two to one. DNV’s operating-fleet evidence shows seven hydrogen-fuelled vessels against more than 1,300 battery-equipped vessels, a difference of roughly 186 to one. The battery figure includes hybrid installations, so it is not an exact propulsion-for-propulsion comparison, but the architectural qualification is far too small to reconcile a two-to-one announcement ratio with an operating-fleet difference approaching two orders of magnitude.

Passenger cars are more extreme still. The separate passenger-car extension contains five hydrogen milestone roots and six battery roots, which would make the technologies appear almost evenly represented if the announcements were all a reader saw. Against those announcements are approximately 16,000 fuel-cell passenger vehicles in the 2025 sales denominator and about 13 million battery-electric cars. The physical-flow difference exceeds 800 to one.

Grid storage provides the clearest demonstration of the effect. A strict review of hydrogen’s stationary-power announcements leaves only four cases from 2022 through 2025 in which hydrogen genuinely functions as an electricity-storage vector or integrated storage medium: H2Évora, HYFLEXPOWER, Denham and EMEC. There are 36 battery-storage announcement roots, so batteries already lead hydrogen nine to one in the information stream. The corresponding physical scale is approximately 10 MW for the hydrogen demonstration denominator against 108,000 MW of battery-storage additions in 2025, a difference of about 10,800 to one. Those figures compare power rather than energy capacity, duration or equivalent storage service, so they are a measure of physical market scale rather than storage performance.

The graph is the central finding of the research. Hydrogen and battery announcement counts occupy roughly the same part of the logarithmic scale, while their deployment numbers progressively separate by one, two, three and finally four orders of magnitude. Corporate communications, government PR, industry amplification and journalism are taking markets with radically different physical scale and turning them into streams of discrete announcements that are much closer in frequency. The result is an information environment in which hydrogen appears far more prevalent than its installed base, sales, operating fleets and deployed capacity justify.

This does not require journalists to fabricate hydrogen successes or communications departments to publish false statements. Most of the events are real. A hydrogen bus enters service, a truck completes a demonstration, a train sets a distance record, a storage project begins operating or a government awards funding. Each becomes a legitimate news event, but a news event has no natural mechanism for carrying the scale of the market behind it. One hydrogen demonstrator and another hundred thousand battery vehicles can each produce a press release, a trade article, a government announcement and social-media posts. Once converted into stories, radically unequal quantities of physical activity become much more nearly equal units of information.

The maturity of the underlying events compounds the effect. Across the normalized four-year sample of 140 eligible events in six common sectors, battery events averaged 5.97 on the seven-stage commercial-development ladder compared with 4.40 for hydrogen. About 82% of battery events were at scaled deployment or repeat procurement, compared with about 40% of hydrogen events. Hydrogen announcements also carried more future/readiness framing, averaging 1.35 on the zero-to-two measure compared with 0.88 for batteries. The media and PR system is therefore giving substantial visibility to a hydrogen event population that is both much smaller in physical scale and materially earlier in commercial development.

The research did not support every mechanism I initially expected to find. Novelty language occurs on both sides, with battery companies and customers readily describing first plants, largest orders, records and technical milestones. A normalized one-root-per-event comparison also found only a small difference in explicit utilization evidence, and source-lineage work did not show that hydrogen publicity is uniquely dependent on copied press releases. Those findings matter because the distortion does not depend on a special conspiracy of hydrogen-friendly journalism or a uniquely aggressive hydrogen PR machine. It emerges from the ordinary operation of an information system that treats discrete events as news while largely discarding the denominator behind them.

The stronger conclusion is therefore about the proportions produced by the system. Battery technologies generated more announcements in the period studied and vastly more commercial deployment, but the announcement advantage is tiny compared with the deployment advantage. Hydrogen receives announcement-scale visibility while batteries and direct electrification receive deployment-scale capital, hardware and useful work. That difference is large enough to make technologies separated by orders of magnitude in the physical economy appear to remain serious peers in the public information stream.

Below the paywall, the analysis turns from the scale mismatch to the mechanism behind it: how firsts, pilots, partnerships, repetition and institutional amplification can make a small market feel much larger and more mature than it is. I trace that effect through the announcement corpus, commercial-stage data and source genealogy, then show the three checks that separate genuine market formation from a dense stream of activity: restore the physical denominator, collapse correlated publicity back to the underlying event, and follow what actually happened next.

Read more

In May 2025, I wrote that Poland’s coal generation had fallen from more than 90% to 63% since 1990, “all due to renewables.” That sentence aged badly in two ways. The original CleanTechnica article was directionally right about the decline of coal, but final 2024 data subsequently put coal at just 56.2% of gross electricity generation, with renewables at 29.4%. Then 2025 pushed coal down again, to 52.7%, while renewables rose to 31.4%.

The second correction matters just as much. It was too simple to say renewables were responsible for all of coal’s decline. Solar and wind are doing much of the structural displacement, but gas has been gaining share as well. In both 2024 and 2025, natural gas recorded the largest increase in generation share, rising by more than two percentage points each year. In 2025, hard coal still generated 57.6 TWh and lignite 33.5 TWh, but gas was up to 24.4 TWh, onshore wind supplied 23.8 TWh and solar 20.3 TWh. Poland is replacing coal with a mixture of renewables and gas, not executing a clean one-for-one coal-to-renewables swap.

What has changed most since that earlier article is the slope. Coal supplied more than 80% of Polish electricity in 2015. By 2021 it was still at 72.5%, a decline of 10.4 percentage points over six years. In the next four years it lost another 19.8 percentage points, reaching 52.7% in 2025. Coal supplied less than half of Polish generation in five separate months during the year, and in June renewables produced more electricity than coal over a full month for the first time. This is no longer a slow diversification story measured over several decades. Most of the recent change has occurred in a handful of years.

The monthly numbers also show why declaring victory would be premature. In May 2026, hard coal still supplied 30% of electricity entering Poland’s power system and hard coal plus lignite supplied nearly half. Solar, meanwhile, supplied a remarkable 21% for the month and wind another 11%. Battery storage capacity reported to Poland’s Energy Market Agency had passed 1 GW. This is what an electricity system in the middle of a transition looks like: renewables can dominate particular hours and become major sources over whole months while coal remains materially important when weather, demand and system conditions change.

That leads to the more interesting Polish coal story now. Coal generation is falling faster than Poland can stop relying on coal capacity. Annual electricity generation measures energy, in MWh or TWh. Power-system adequacy asks a different question: whether enough dependable MW are available during the difficult hours when demand is high, wind is weak, solar is absent, imports are constrained or several things go wrong at once. A coal plant can therefore run fewer hours every year, lose market share and become increasingly uneconomic as an energy producer while still being considered useful insurance against those stress periods.

Poland’s capacity market makes that distinction unusually visible. In September 2025, the supplementary auction for the 2026 delivery year contracted 7.58 GW of capacity obligations. The important detail is not that all 7.58 GW was coal, because it was not reported that way. It is that plants emitting more than 550 kg of CO₂ per MWh were allowed to participate under a special derogation. Similar supplementary auctions are planned for 2027 and 2028, and the European Commission’s exemption allowing this high-emitting capacity to receive support runs only through December 31, 2028.

That can look contradictory if the only metric being watched is annual coal generation. Poland is adding renewables quickly enough to push coal down the merit order, then paying to keep some high-emitting capacity available because the rest of the system is not yet capable of replacing all of its reliability function. Those are not opposite policies so much as different stages of the same transition. The energy transition has run ahead of the capacity transition.

Poland is already seeing what has to come next. Renewable generation grew enough in 2025 that 1.4 TWh was curtailed, twice the 2024 amount. Almost all of that curtailment was attributed to balancing rather than physical grid congestion. During many sunny periods, Poland now has more low-marginal-cost generation than its relatively inflexible power system can conveniently absorb. At other times, it still wants coal plants standing by. Building another solar farm solves the first problem only if the electricity can be shifted, exported or turned into useful demand, and it does not by itself solve the second.

This is why transmission, interconnection, batteries, flexible demand, district heating, smart EV charging and other forms of load shifting increasingly matter as much as another tranche of generation capacity. Poland’s old coal fleet was built to be the electricity system. The replacement is a mesh of generation, grids, storage and responsive demand. That replacement can use much less fuel, but it requires more coordination across the system.

The physical transition is becoming harder to miss. On July 10, 2026, Poland’s first offshore wind farm, Baltic Power, delivered its first electricity to the grid. At that point 54 of its 76 turbines had been erected. When fully commissioned, the project will provide about 1.2 GW, producing roughly 4 TWh a year, equivalent to around 3% of current Polish electricity demand. Offshore wind is moving in the same direction as the already much larger solar buildout: technologies that were peripheral to the Polish system a few years ago are becoming material pieces of it.

The near-term outlook no longer requires heroic assumptions. The International Energy Agency now expects Polish renewable generation to grow about 13% annually through 2030, overtaking coal on an annual basis in 2028 and reaching about 53% of generation by 2030. Over the same period it expects coal generation to fall around 11% a year. Gas is forecast to grow, not disappear, which reinforces the point that Poland’s electricity decarbonization is progressing rapidly without yet being complete or entirely clean.

There is a useful denominator for watching Poland from here. Do not track only the percentage of electricity generated from coal. Track that alongside the amount of coal capacity the system still believes it needs during its hardest hours. The first number is already falling very quickly. The second will tell us whether grids, storage, flexible loads, interconnection and replacement generation are arriving fast enough to let the plants themselves disappear.

Poland has therefore moved into a more difficult and more interesting stage than the one I described in May 2025. It no longer needs to demonstrate that wind and solar can take large quantities of generation away from coal. They already are. The test now is whether Poland can build enough flexibility around them that coal stops being valuable even as insurance.

Getting coal out of the energy mix comes first. Getting coal out of the power system is the harder second act.

For more system-level analysis of grids, generation and the energy transition, subscribe to TFIE Strategy Briefing.

Subscribe now

Global hydrogen demand exceeded 100 million tonnes in 2025, and almost all of it was still consumed in the same places hydrogen has been used for decades: refining and industry. Even the IEA’s latest hydrogen-demand accounting, despite projecting a considerably larger future role for hydrogen than I do, describes the existing commodity principally in millions of tonnes and shows refining and industrial uses continuing to dominate demand. Low-emissions hydrogen remains tiny by comparison. The reason the mass unit fits so naturally is straightforward: an ammonia plant, refinery or methanol plant does not buy an abstract quantity of energy called hydrogen. It buys a chemical feedstock by mass and uses the molecule in a specific industrial process.

A 2023 article of mine argued that expressing hydrogen in MWh and TWh was distorting energy-transition discussions. There is a technical qualification worth making. A megawatt-hour is dimensionally a unit of energy, so hydrogen’s heating value can certainly be converted into MWh without violating physics or mathematics. The analytical problem starts when dimensional validity is treated as though it made the choice of unit neutral. Electricity is overwhelmingly generated, traded, stored and consumed in kWh, MWh and TWh, while hydrogen is overwhelmingly manufactured, transported and consumed as kilograms and tonnes of an industrial molecule. Putting both into the language of MWh encourages a comparison between energy commodities before hydrogen has demonstrated that it has won the energy markets being imagined for it.

Natural gas is an imperfect analogy because its economic purpose is much more closely tied to its chemical energy. Buyers overwhelmingly want gas for heat or for conversion into electricity and mechanical work, which makes calorific units closely related to the service being purchased even though volume and mass remain important. Hydrogen’s present market has a different structure. Ammonia producers need hydrogen atoms, refineries use hydrogen for desulfurization, hydrocracking and hydrogenation, and methanol and other chemical processes consume it as feedstock. Those industrial balances naturally resolve into kilograms and tonnes because that is what plant capacity, contracts, process ratios and physical deliveries are built around. Converting the same flows into TWh can be mathematically clean while making the economic role of the molecule less clear.

This sits beside a denominator problem I addressed recently in Stop Steering With The Wrong Energy Metrics. That assessment separated primary energy, final energy and useful energy because they describe different boundaries in the energy system: raw supply entering the system, energy crossing the customer boundary, and the heat, motion, light or work the economy ultimately receives. The hydrogen problem begins one step earlier. Before choosing among energy denominators, it is worth asking whether the commodity being discussed is principally an energy commodity at all. Hydrogen overwhelmingly is not today, and kilograms and tonnes describe its industrial market more directly than MWh or TWh.

That distinction has become more important as the hydrogen-for-everything narrative has collided with actual deployment. My current Hydrogen Demand Will Shrink, Not Become The New Oil model begins with existing commodity demand and then asks what survives the transition rather than allocating hydrogen into imagined future sectors first. It has global demand at about 116 Mt H₂/year around 2020, declining to roughly 61 Mt around mid-century and about 36 Mt by 2100 as refinery-linked demand falls and proposed transport, heating and generic-storage uses fail to grow enough to replace it. The remaining pool is concentrated in fertilizer, methanol, selected industrial chemistry and residual refining rather than a broad new fuel economy.

The deeper problem with high-hydrogen scenarios is not simply the unit printed on the axis. A hydrogen application is not automatically a hydrogen market. Space heating belongs to a heat market, electricity generation belongs to electricity and capacity markets, and grid storage belongs to flexibility and reliability markets. Hydrogen can compete as an input to any of them, but additional hydrogen demand exists only after customers in those markets choose it over the alternatives. Starting with a modeled number of TWh reverses that causality by assuming the application first and calculating the molecule afterward, which can make an allocation exercise look much more like demonstrated market formation than it really is.

Below the paywall, the analysis turns from the unit question into the decisions different audiences actually have to make. Investors get a way to distinguish real hydrogen demand from modeled or policy-created demand by following tonnes, offtake quality, utilization and the economics of the market hydrogen is trying to enter. Policymakers get a clearer basis for setting targets around physical demand and useful services instead of headline TWh allocations that can smuggle assumptions into the denominator. Analysts and strategists get the EEX and EIA cases as concrete examples of how unit choices and accounting boundaries can reshape the story without changing the underlying physical system, along with a practical rule for choosing metrics that match the market or service being evaluated. The value is not in proving that hydrogen can be converted into MWh; it is in knowing when that conversion clarifies a decision and when it quietly manufactures a market that does not yet exist.

Read more

Methane thermolysis, commonly called methane pyrolysis, has an attractive proposition for lower-carbon hydrogen. Split methane without oxidizing its carbon and the outputs are hydrogen and solid carbon instead of hydrogen and CO₂. Hazer Group is now providing a useful test of what that chemistry means at industrial scale. The Australian developer and KBR recently signed a paid study with a large Japanese power utility examining plants producing 30,000 and 300,000 tons of hydrogen a year. The larger configuration is only a pre-feasibility study, not a project decision, but at that scale one characteristic of methane thermolysis becomes impossible to treat as a footnote: roughly every ton of hydrogen brings three tons of solid carbon with it.

That ratio is not peculiar to Hazer. It follows from the chemistry of methane thermolysis itself. Methane is CH₄. Splitting it into its constituent elements produces hydrogen and carbon, and the mass balance is about three tons of carbon for every ton of hydrogen. Hazer’s particular process produces a graphitic carbon that the company is developing as a commercial product, which is a meaningful distinction from generic carbon black or amorphous carbon. But at the chemistry level the important point is simpler. A 300,000-ton-per-year methane-thermolysis hydrogen plant creates roughly 900,000 tons of solid carbon every year, about 2,500 tons every day.

That is also why methane thermolysis deserves more serious consideration than many hydrogen propositions. It does not create a concentrated process CO₂ stream that then requires capture, compression, transport and geological storage. Hazer has moved beyond laboratory work into an operating demonstration plant and has completed a commercial-scale process design package with KBR. The question is not whether the chemistry works. It is what happens when the chemistry reaches industrial scale.

The usual answer is that carbon and graphite are valuable products. They certainly can be. Hazer is testing its graphite in steel, batteries, asphalt, concrete and other applications and is developing purification, pelletization and customer qualification pathways. But the relevant denominator is not whether markets for carbon products exist. It is whether enough customers want this particular carbon, at the required grade, price and location, at the same rate that hydrogen customers determine methane-thermolysis production.

There is also a hype-cycle element to the renewed attention. Methane thermolysis is neither new chemistry nor an exotic industrial concept. Splitting methane into hydrogen and solid carbon has been understood for a long time, while carbon production, high-temperature gas processing and solids handling are all established industrial disciplines. Hazer’s catalytic implementation may prove commercially better than earlier approaches, but it is arriving during overlapping hydrogen and carbon-management investment cycles in which avoiding a CO₂ stream sounds particularly attractive. That makes it worth separating technical merit from attention.

An ongoing TFIE analysis of hydrogen press coverage versus actual deployment of competing technologies is finding the same pattern repeatedly: hydrogen pathways receive disproportionate attention for pilots, demonstrations and proposed plants while incumbent or electrified alternatives often accumulate vastly more real-world deployment with less fanfare. Methane thermolysis deserves assessment on its industrial economics and mass balance, not on the amount of attention currently available to anything combining hydrogen with a carbon-management story.

That makes the most useful next question where methane thermolysis has a genuinely strong industrial fit rather than simply an attractive story. Steel is probably the strongest possible pairing because it can actually use both products, and Hazer is sensibly pursuing exactly that proposition. It is working with POSCO and is part of the M Resources proposal for Whyalla, where hydrogen could support direct reduction of iron while Hazer graphite could be used as a recarburizer in electric arc furnace steelmaking. That is a much better fit than an ammonia plant, refinery or power station, all of which can consume large quantities of hydrogen while having little intrinsic use for the resulting solid carbon.

A modern hydrogen-DRI steelworks provides a useful scale test. Stegra’s Boden project in Sweden is designed around 2.1 million tons of DRI feeding an initial 2.5 million tons of finished steel annually. At roughly 54 to 58 kg of hydrogen per ton of DRI, that requires about 113,000 to 122,000 tons of hydrogen a year. Supplying that hydrogen through the Hazer process would create roughly 340,000 to 365,000 tons of graphite.

Hydrogen-DRI does not eliminate carbon from the electric arc furnace. Carbon is still useful for final steel chemistry, FeO reduction, slag foaming and process energy. Recent modelling of carbon-free H₂-DRI cases uses roughly 18 to 25 kg of injected carbon per ton of steel, although actual requirements vary with metallization, iron feed, steel grade and furnace operation. Applied to a 2.5-million-ton steelworks, that is around 45,000 to 63,000 tons of carbon annually. Even this unusually favorable pairing therefore consumes only a fraction of the carbon produced alongside its hydrogen. Roughly 280,000 to 320,000 tons a year, around 80% to 90% of the Hazer graphite, would still need other customers.

Those other markets are not empty spaces waiting for carbon. Steelmakers already choose among petroleum coke, anthracite, natural graphite, synthetic graphite and other carbon products, with biochar increasingly being developed as a lower-carbon alternative. Battery manufacturers need highly engineered graphite and can choose among natural, synthetic and increasingly recycled sources, with qualification depending on purity, morphology, particle size, surface characteristics and electrochemical performance. Concrete and asphalt are harder still to count as enormous graphite markets because carbon is an optional additive rather than a fundamental ingredient. Hazer may prove competitive in some of these uses, but each is a distinct materials market with incumbent suppliers, specifications, qualification requirements and logistics.

That exposes the structural difference between methane thermolysis and conventional carbon supply. Existing carbon supply chains are largely decoupled from the customers they serve. A steelmaker can contract for the types and quantities of carbon it needs, switch suppliers, alter grades or increase biochar content without forcing an unrelated plant to manufacture something else. A battery-material producer can blend natural and synthetic graphite and increase recycled content as economics and qualification permit. Carbon production responds to carbon demand.

Methane thermolysis couples the two businesses chemically instead. Hydrogen demand determines how much solid carbon is produced. Hydrogen’s difficult transportation characteristics also encourage production close to the hydrogen customer, meaning the hydrogen market largely determines where the carbon appears. The much larger solids stream must then be stored, handled, graded or pelletized as required, qualified into multiple applications and transported to enough customers to clear production continuously. At commercial scale, a methane-thermolysis hydrogen facility is therefore also a substantial carbon-materials and logistics business.

None of that means Hazer’s process is a bad idea. A site with biogas or suitable natural gas, a durable industrial hydrogen requirement and one or more nearby carbon consumers could be attractive, particularly if Hazer graphite displaces higher-emissions synthetic graphite, coke or other fossil carbon. Hazer may also prove unusually good at processing its graphite into useful grades and developing markets for it. That would distinguish Hazer commercially.

It would not change the structural constraint shared by methane-thermolysis processes. Hydrogen output and carbon output are chemically coupled while hydrogen and carbon demand remain economically independent. Steel, probably the best dual-product industrial customer available, substantially improves the match and still leaves most of the carbon requiring other buyers. When the co-product is three times the mass of the headline product, its market is part of the process.

Subscribe to TFIE Strategy Briefing for more denominator-first transition analysis.

Subscribe now

In September 2022, six months after Russia invaded Ukraine and while European gas and electricity prices were dominating headlines, I argued that energy independence was the wrong lesson to take from the crisis. European countries had not made a mistake by depending on each other. They had made a mistake by allowing too much of a critical energy supply to depend on Russia, an increasingly unreliable counterparty with enormous leverage over a continuously consumed commodity. The better objective, I argued, was strategic energy interdependence: broad connections among reliable trading partners, multiple routes, diverse forms of generation and enough redundancy that the failure of one supplier or corridor would be inconvenient instead of economically destabilizing.

That argument was closely tied to electrification. Earlier in 2022, while assessing European plans to manufacture hydrogen in North Africa for export, I had reached a simpler conclusion. Morocco, Algeria and Egypt had good reasons to use green hydrogen domestically for fertilizer and other existing hydrogen demand. Europe had much weaker reasons to recreate the fossil-fuel trading system by converting renewable electricity into hydrogen, transporting the molecules and accepting the conversion losses and infrastructure cost. North African wind and solar could increasingly be shared directly with Europe through high-voltage direct-current transmission. My shorthand at the time was “moving electrons, not moving molecules”.

Four years later, the July 2026 HVDC-VSC project inventory from RTE International provides an unusually useful test. Filtering RTE’s global future-project list for schemes that cross national borders, whether subsea, underground, overhead or mixed, produces about 60 prospective international VSC projects. Europe and its immediate neighbourhood dominate the list. If worsening geopolitics were pushing countries toward electrical autarky, this is one place where the retreat should be visible. Instead, cross-border transmission remains a substantial development category in precisely the regions where energy security has become more politically salient.

Fossil-fuel import dependence and electrical interdependence expose countries to different risks. A gas pipeline or LNG terminal remains useful only while fuel keeps arriving, leaving the buyer continuously exposed to commodity prices, supplier behaviour and global fuel markets. An electrical interconnector joins two systems that already contain generation, storage, demand and other connections. Power can generally move in either direction, and a country with several interconnections can draw at different times on hydro, wind, solar, nuclear, storage and dispatchable generation spread across different jurisdictions and weather systems. Domestic resources remain part of the portfolio. The aim is not independence from trade but diversification of the ways useful energy can reach consumers.

I kept returning to that idea after the immediate 2022 crisis. In early 2023 I argued that HVDC connections among renewables-heavy countries could mitigate geopolitical energy risks, and later treated strategic interdependence as a grid-investment criterion. Countries insisting on greater electrical self-sufficiency would generally need more generation and storage than countries able to share resources, while the connections themselves still had to be assessed for counterparties, route redundancy and disruption risk. Strategic interdependence never meant replacing one concentrated dependency with another.

The most striking result from the 2026 review came from looking for projects being stopped by geopolitical conflict. Lithuania, Latvia and Germany are still advancing the Baltic-German PowerLink toward EU Project of Common Interest status after the Baltic states separated their electricity systems from the Russian sphere and while regional infrastructure security has become a major concern. The war is obviously part of the strategic context, but it is not suppressing that project.

Policy language increasingly makes the broader rationale explicit. ACER links stronger European interconnection with greater cross-border use of renewables, lower fossil-fuel dependence, flexibility and security. Britain’s Clean Flexibility Roadmap treats interconnection as part of the country’s security-of-supply portfolio alongside storage and demand flexibility. These institutions are not describing international transmission solely as a mechanism for electricity-market arbitrage.

Most major interconnectors take many years to develop, so the causality should not be exaggerated. Renewable balancing, wholesale-price differences, offshore wind, resource adequacy, congestion and market integration can justify HVDC links without a war anywhere nearby. Russia’s invasion did not cause every project expected in the 2030s. The evidence supports a narrower and more useful conclusion: geopolitical shocks have added security and resilience value to infrastructure that often already made sense for economic and decarbonization reasons. Four years after I argued for strategic energy interdependence, the project pipeline and policy language provide substantial confirmation of the direction without requiring every cable to be reclassified as a geopolitical project.

Below the paywall is the professional layer: the exact cross-border HVDC denominator and maturity screen, regional distribution, conflict-exposure cases, the actual causes of project attrition, the VSC-versus-LCC technology boundary, resilience failure modes, implications for regulators and investors, and the evidence that would strengthen or weaken the strategic energy interdependence thesis.

Read more

The Financial Times supplied an interesting prompt this week. Short sellers have made an estimated $2.1B betting against Oklo, NuScale and Nano Nuclear, while $30.3B of their combined market value has disappeared since the three small modular reactor developers peaked in October 2025. The FT describes the rise and collapse as a hype cycle. I have been writing for years that the underlying SMR economics did not support the story investors were being sold, so the obvious question was how much money investors might have saved by treating that analysis as investment diligence rather than merely an argument about energy technology.

I went back through my published work on hydrogen, electric air taxis and SMRs, matched the dates of substantive warnings against subsequent valuations, and tried to make the calculation hostile to my own thesis. The result is about $95B of shareholder value subsequently erased from roughly $119B that remained exposed in the core set, close to 80%. Include X-energy, which listed years after my SMR economic critique was public and has since lost another $5.8B from its post-IPO surge, and the broader tally reaches roughly $100B.

That needs an immediate caveat. $100B of lost market capitalization is not $100B of cash physically fed into companies and burned, nor could every shareholder have sold simultaneously at the reference valuations. Markets do not work that way. The useful counterfactual is at the individual investor level: someone considering or holding these investments could read the analysis, decide that the system economics did not justify the valuation, and keep the money somewhere else. Avoiding a 70%, 90% or 100% loss preserves capital just as surely as picking another stock that rises.

The calculation also refuses some tempting numbers. Nikola briefly approached a $30B valuation in 2020, but I am not claiming that entire collapse. My broad warning that the “hydrogen economy” was mostly hype was published December 23, 2020, after Nikola’s great speculative peak. Nikola ended 2020 worth about $5.9B, so that is the useful starting point for this exercise. It later filed for bankruptcy with its equity worth only tens of millions. Counting the earlier peak would produce a much better marketing number and a much worse analysis.

Hydrogen is still the largest part of the bill. After that December 2020 assessment was public, the hydrogen enthusiasm cycle carried Plug Power to around a $40B valuation, Ballard Power above $10B and FuelCell Energy to roughly $7B. Their combined market value today is only around $6B. Add the roughly $5.9B of Nikola value that still existed at the end of 2020 and the hydrogen portion of the exercise accounts for approximately $59B of subsequent shareholder-value destruction. The original argument did not depend on anticipating interest rates or investor sentiment. It was about delivered hydrogen cost, conversion losses, infrastructure, utilization and competition from direct electrification. Those problems subsequently became commercial problems.

The financial mechanism became even clearer with time. In March 2026 I described the sector’s recapitalization cycle: investor enthusiasm sends the limited set of listed hydrogen stocks upward, the companies issue equity or convertible securities into the enthusiasm, the anticipated profitable markets fail to arrive at the expected scale, cash disappears, valuations fall, and management waits for the next policy or technology narrative. Plug Power’s accumulated deficit has passed $8B, Ballard’s exceeds $2B and FuelCell Energy’s is around $1.8B. This is not a young industry encountering the ordinary losses of commercialization. Some of these companies have been repeating versions of the cycle for decades.

The eVTOL case provides an unusually clean dated receipt because I wrote the investment conclusion explicitly. On November 24, 2021, I calculated that the publicly traded urban-air-mobility pure plays had already fallen from $27.92B at their peaks to $11.82B. I called the sector vastly overvalued and told investors in the companies to “cut your losses”. I do not count the $16.1B that had already disappeared when I wrote that. Four months later I checked again and found that cumulative destruction had risen to about $21B. Another $5B had disappeared after the warning, roughly 42% of the remaining $11.82B in only a few months.

SMRs close the loop. In 2021 I argued that small modular reactors were mostly bad policy because making reactors smaller surrendered economies of vertical scale while assuming manufacturing economies that did not yet exist. In 2023, after NuScale’s UAMPS project collapsed, I wrote explicitly about the investment and SPAC dynamics behind SMR enthusiasm. More recently I tested the proposition that serial production would rescue costs and found that even generous Wright’s Law assumptions do not rescue the economics. Then came the 2025 AI-and-nuclear stock mania. Oklo, NuScale and Nano Nuclear subsequently lost $30.3B from their October peaks. X-energy, which listed in April 2026, has lost another $5.8B since its initial surge.

There is a common analytical error underneath all three categories. Investors tend to value the interesting component and assume away the expensive system around it. A fuel cell works, but a hydrogen transportation business also needs hydrogen production, compression or liquefaction, distribution, storage, refueling infrastructure, high utilization and a delivered cost that beats batteries. An eVTOL can fly, but a profitable air-taxi system also needs certification, low maintenance, high aircraft utilization, vertiports, acceptable noise and downwash, airspace integration and enough wealthy passengers making trips that existing helicopters, cars and transit do not already serve better. An SMR can sustain fission, but the investable product also includes licensing, fuel, factories, construction, security, financing, waste management, repeat orders and electricity cheap enough to compete with the alternatives.

That distinction between a component and a system has been the useful investment signal. None of the original warnings required knowing whether Plug Power would peak in January 2021, whether eVTOL SPACs would crack before certification, or whether AI enthusiasm would send Oklo toward a spectacular valuation before investors rediscovered reactor timelines and capital requirements. They required asking whether the entire product could become competitive, repeatable and financeable. When the answer depended on several difficult things becoming cheap, abundant and easy at the same time, the valuation deserved a much higher discount than the market was applying.

Market capitalization is still only one measure, so I ran a second test. Across the companies where the filings allowed a reasonably comparable reconstruction, at least $11.5B of operating cash was consumed after the relevant warnings, before counting several billion more spent on factories, equipment, facilities and infrastructure. Adding only readily identifiable physical-capital expenditures puts the conservative floor around $14.5B of cash consumed or deployed. Not all of that physical capital is worthless, and some assets will retain salvage or alternative-use value. But the gap between the roughly $100B market repricing and the smaller cash figure is useful: investors first paid extraordinary prices for expectations, then companies converted part of those expectations into actual cash and spent it pursuing the business cases.

Fusion is deliberately absent from that ~$100B calculation. I wrote in 2021 that fusion power had once again receded into the future, focusing not on whether fusion reactions could be produced but on whether whole facilities could ever deliver competitive net electricity. In 2024 I included fusion among the questionable capital allocations in my review of Breakthrough Energy Ventures. Commercial fusion companies had raised more than $11.5B in private capital by mid-2026, before Commonwealth Fusion Systems added another $1B in July. Helion has raised about $1.5B and was valued at $15.5B in its latest financing.

I am counting none of that as capital destroyed. Those investors may eventually make excellent returns, lose most of their money, or spend another decade discovering which. Most of the major fusion companies are private, their latest financing rounds generally imply rising rather than collapsing valuations, and there is no liquid-market repricing that permits the same calculation available for Plug Power, Joby or Oklo. More than $12B of private fusion investment belongs in a different column today: capital at risk, outcome unresolved. If it eventually produces competitive electricity, the investment thesis will have been vindicated. If it does not, there will be another dated set of receipts to examine.

That exclusion matters because this is not an exercise in collecting technologies I dislike and calling their funding losses. Fusion is a technology I have been publicly skeptical of for years, billions remain invested at high private valuations, and I am refusing to count them because the evidence does not yet support the claim. NuScale provides another version of the same lesson. After its UAMPS project failed and I wrote in November 2023 that its SMR proposition remained economically weak, its shares subsequently went through another enormous speculative rise. Anyone using my analysis as an instruction to maintain an unhedged short position would have discovered that weak long-run economics and short-run stock prices are different things.

The more useful investment interpretation is simpler. An investor does not have to predict the top of a bubble to benefit from recognizing that the story underneath it is weak. There is no requirement to short the company, trade options or identify the exact month reality will catch up. Declining to put capital into a business whose economics require unrealistic system assumptions is an investment decision. Keeping $100 that would otherwise become $20 leaves $80 available for the next decision.

That is why this exercise is more interesting to me than an “I told you so” scorecard. Hydrogen, eVTOL, SMR and fusion companies are very different technically. The repeated analytical question sits upstream of the technologies themselves: does the entire system survive contact with physics, infrastructure, realistic competitors, financing and repeatable commercial demand? Investors repeatedly paid very large sums before answering that question. Roughly $100B in subsequent public-market destruction, another $14.5B or more of real capital consumed or deployed, and more than $12B currently at risk in private fusion companies suggest that the question is financially consequential.

Subscribe to TFIE Strategy Briefing for the deeper professional layer: denominator checks, capital filters, comparator cases and decision context for technology and infrastructure bets before narrative momentum becomes an expensive position.

Subscribe now

A couple of years ago I looked at the very different economics of decarbonizing domestic freight in China, Europe, India and the United States. The premise was straightforward. Freight does not move the same way everywhere. Some economies have extensive and heavily used rail systems, some have enormous highway networks, some make much greater use of rivers and coastal shipping, and some combine all three at scale. Electrification was going to arrive in those inherited systems rather than replacing them with a globally standardized freight model.

Returning to that analysis with substantially more research behind it produced better numbers. Some of the earlier estimates mixed scopes or leaned too heavily on incomplete data, especially for India and for the treatment of domestic coastal shipping. The dataset below is the one I would use today. More importantly, the improved data reinforce the underlying point: the economics of freight electrification depend heavily on the transport system a geography already has.

The first problem in making the comparison is deciding what counts as domestic freight. Road and rail are relatively straightforward. Water is not. China has large inland-waterway traffic, large coastal domestic shipping volumes and still larger ocean trade. For a comparison of domestic freight, the first two belong in the denominator and international ocean freight does not. Similarly, if the EU-27 is being treated as a single economic geography, freight shipped from Spain to the Netherlands or Finland to Germany is internal water freight in much the same sense that cargo moving between Chinese provinces along the coast is internal to China.

With those boundaries made explicit, China’s 2025 freight statistics work out to about 44% road, 20% rail and 36% domestic water across those three modes. For the EU-27, recombining European Environment Agency 2023 data gives roughly 54% road, 12% rail and 34% internal water freight, although the underlying mode statistics are not perfectly harmonized and should not be presented as though the EEA had published that exact modal split.

India is much more road-heavy. The NITI Aayog modelling baseline, which is a model rather than consolidated measured freight statistics, gives about 69% road, 23% rail and 8% water when its road, rail and water categories are normalized against one another. The United States has a less cleanly harmonized source stack, but the resulting reconstruction is much stronger than the estimate I used previously. It puts the U.S. at roughly 53% road, 36% rail and 10% water across the same three-mode comparison, with somewhat lower confidence than the other rows. The important structural feature is the unusually large role freight rail retains compared with Europe and India.

Those differences are not trivia. They determine where electricity has to enter the freight system. China can electrify enormous truck fleets while still moving very large volumes by rail and water. India has a road-dominant freight economy but is simultaneously building capacity into an almost completely electrified railway. Europe already possesses an extensive railway and waterway system, yet trucks still do most inland freight work. The United States combines a continent-scale road network with a freight railway system that remains commercially important and overwhelmingly diesel powered.

The direction of travel is different as well. China reported about 140,000 new-energy heavy-truck sales in the first half of 2026, up 78.6% year on year. “New energy” is a broader Chinese regulatory category than battery electric, so it would be wrong to label all of those trucks BEVs, but the scale is already industrial rather than experimental. China is also explicitly targeting new-energy heavy trucks at around 40% of annual heavy-truck sales by 2030, while building charging and swapping infrastructure along major freight corridors.

India is pushing in another direction at the same time. Almost all of its broad-gauge railway is now electrified, and around 2,800 kilometers of Dedicated Freight Corridors have been completed. By early 2026 those corridors were handling roughly 480 freight trains a day, according to India’s Ministry of Railways. India has plenty of trucks to electrify, but it also has a credible route for increasing freight movement on infrastructure that already draws electricity directly from the grid.

Europe presents almost the inverse policy problem. It already has more than 200,000 kilometers of rail and a substantial electrified share, yet the long-term inland modal direction has not been toward rail. On Eurostat’s territorial freight series, road gained about 3.3 percentage points of share from 2014 to 2024 while rail and inland waterways slipped slightly. Meanwhile, electrically chargeable trucks above 3.5 tons reached 4.2% of EU registrations in 2025, almost twice their 2024 share but still a small part of the market. Europe cannot wait for modal shift to make road freight disappear. It has to electrify the trucks too.

The United States is the longer-duration case. Its freight railways are good at moving enormous volumes long distances with relatively little energy, while trucks provide flexibility, direct service and speed that rail frequently cannot match. As battery costs and charging infrastructure improve, that competitive boundary moves. National Renewable Energy Laboratory modelling finds zero-emission trucks capable of total-cost-of-driving parity or better across market segments by 2035 under continued technology improvements. That is a 2030s economic signal, not a prediction that rail traffic should already be falling today.

This is consistent with my earlier CleanTechnica analysis of electric trucks versus North American freight rail, which found that electric trucks were already lower-carbon per ton-mile than diesel rail in eight U.S. states and argued that falling battery, energy and maintenance costs would progressively shift the road-versus-rail economics. The point was not that rail would suddenly lose its role, but that a competitive assumption which had been stable for decades was starting to move.

The common direction is electrification, but there is no common modal endpoint. Countries are electrifying the freight systems they actually have, while expanding, shrinking or competing between modes where local economics permit it.

Below the paywall I go from the public comparison to the professional layer: how inherited road, rail and water systems change the economics of electrification, why battery swapping is becoming freight infrastructure for Chinese trucks and commercial vessels while rail is only beginning to test the same modular-energy idea, and where that leaves infrastructure and capital decisions through the 2030s.

Read more

The argument in my recent Briefing article, Scrapping A Working Gas Car Can Be The Greener Choice, generated a useful counterpoint. One reader has a hand-me-down 1995 Dodge Ram from his father and intends to keep repairing it until it rusts away, someone makes him stop driving it or gasoline finally becomes unaffordable. He accepts that the climate crisis is real, but, he wrote, “we won’t be one-at-a-timing a solution to this.”

His decision makes economic sense. The Ram is paid for, he knows how to keep it running and gasoline remains affordable to him. Spending hundreds or occasionally thousands of dollars repairing a familiar truck can be much easier to justify than spending $50,000 on a new electric pickup. If the climate solution requires him to make a financially foolish decision for his household, the problem is not that he has failed to understand climate change. The available replacement and the policy around it have failed him.

The climate arithmetic points strongly the other way. Depending on engine and drivetrain, a 1995 Ram 1500 gets roughly 12 to 16 miles per gallon. Driven something like the US norm, that means around 7 to 9 tonnes of CO2 every year from gasoline combustion alone. An efficient electric pickup charged on an average US grid would be around 2 tonnes. Replacing the Ram could therefore avoid roughly 5 to 7 tonnes of CO2 every year, before counting the emissions from extracting, refining and transporting its gasoline.

The lifecycle research I examined in the previous Briefing matters because it tested the obvious objection. Manufacturing a replacement EV also causes emissions, so perhaps the environmentally responsible choice is to keep an existing combustion vehicle until it dies. Campbell and Geyer explicitly modeled that manufacturing penalty against the years of combustion emissions avoided by retiring functioning vehicles earlier, and across 92% of their modeled scenarios, early retirement reduced emissions. An old pickup getting somewhere around 14 mpg and driven regularly is close to the sort of vehicle policy should be trying hardest to retire.

But the reader’s line about not solving the problem “one at a time” is the most interesting part, because that is exactly how vehicle decarbonization works. National fleets do not turn over as fleets. Every year millions of people separately decide to repair, sell, scrap or replace individual vehicles. A mechanic decides whether another repair is worthwhile, a family decides whether its aging SUV has another few years in it, and a contractor decides what replaces a work truck. Those individual decisions aggregate into the national fleet. There is no separate fleet-level mechanism waiting somewhere above them. The policy question is whether millions of the highest-emitting vehicles remain in service for years longer because replacing them is a bad household investment.

For this Ram owner, that household calculation could look quite different in another three or four years. Today’s $50,000-plus electric pickups are not compelling replacements for someone happily maintaining a simple old truck. Slate is now offering a deliberately basic electric pickup starting at about $25,000, with a projected 205 miles of range and a strong emphasis on owner customization and repairability. Ford, meanwhile, is developing a four-door midsize electric pickup on its Universal EV Platform with a targeted starting price of about $30,000 and launch planned for 2027.

Neither should be counted as a proven mass-market product yet. Slate is still preproduction, and Ford has not released final specifications for its truck, but they signal something important. The relevant comparison is moving away from a paid-off old pickup versus a $50,000 to $80,000 electric one. In a few years it could be a paid-off old pickup approaching its fourth decade versus a new electric truck costing $25,000 to $30,000, which is a much more interesting household decision.

At those prices, operating costs start to matter. Using a roughly 14 mpg Ram, normal American annual driving, conservative gasoline and electricity prices and a reasonable efficiency assumption for a smaller electric pickup, the EV comes out about $220 a month cheaper in energy and maintenance, or around $2,600 a year. Most of the difference is simply replacing a lot of gasoline with much cheaper electricity, with lower maintenance adding to the advantage. That does not mean this owner should rush out and borrow $30,000, because a paid-off vehicle has a substantial economic advantage precisely because the capital cost is already sunk. The missing piece is bringing the upfront replacement cost down far enough that the monthly operating savings matter to the purchase decision.

That was the policy problem I explored in the earlier scrappage article. The United States currently has no federal new- or used-EV purchase credit; those incentives ended in September 2025. But Democrats have previously proposed point-of-sale cash incentives tied to surrendering an operating gasoline vehicle, with extra help for lower-income households and provisions for used clean vehicles. A future Democratic administration and Congress could reasonably return to some combination of EV incentives and vehicle scrappage, although no one should count today on a particular program or dollar amount.

A better program would target the emissions opportunity much more precisely than the old generic EV tax credit did. Someone replacing a relatively efficient five-year-old car should not receive the same public support as someone retiring a heavily driven 14 mpg pickup. Scrappage value should rise with fuel consumption, recent mileage and expected remaining life, while lower-income households should receive more help and used EVs should qualify fully. For a particularly valuable retirement case such as this Ram, a combined point-of-sale EV incentive and targeted scrappage payment in the range of $10,000 to $15,000 would be defensible. Applied to a $25,000 Slate or roughly $30,000 Ford, that could leave something like $12,000 to $20,000 to finance or pay, followed by about $220 a month in lower operating costs.

That is a very different proposition from telling someone to throw away a useful paid-off truck and spend $50,000 for the climate. Keeping this Ram running today may be the financially sensible choice. If it lasts another three or four years, the emergence of inexpensive electric pickups, combined with a sensible future scrappage program, could make replacing it the financially sensible choice instead.

And that gets back to “one-at-a-timing” the problem. The United States will not wake up one morning and replace its combustion fleet. It will do it when an old Ram in one driveway, an aging SUV in another and millions of other vehicles reach individual decision points, the electric replacement makes more sense and the aging internal combustion vehicle is scrapped instead of being sold to a lower tier of buyers. Vehicle turnover has always happened one vehicle at a time, and good climate policy makes the cleaner choice win more of those decisions.

For practical analysis of transport, energy and climate policy, subscribe to TFIE Strategy Briefing.

Subscribe now

Earlier this year for Redefining Energy, I sat down with Cornelis “Case” Plet, CTO of Grid Systems Integration at GE Vernova, for a wide-ranging discussion about HVDC and what happens as the grid becomes increasingly dominated by power electronics. We talked about the emergence of 2 GW HVDC as a standard building block, multi-terminal and multi-vendor systems, grid-forming inverters, the very different approaches being taken in China, Europe and North America, and why the shortage of experienced power engineers may be a bigger constraint than equipment manufacturing. What follows is a lightly edited transcript of that conversation, cleaned for readability while staying as close as possible to what Case and I actually said.

Listen to the podcast on Spotify

Listen to the podcast on Apple

Michael Barnard [MB]: Welcome back to Redefining Energy. I’m your host, Michael Barnard. Today I’m speaking with Case Plet, the CTO of Grid Systems Integration within GE Vernova’s Electrification business. Case, welcome back to the discussion.

Cornelis Plet [CP]: Thank you. It’s great to be back again, Michael.

[MB]: It’s been a couple of years since we last talked, so people don’t necessarily know who you are and what you’re up to today. You’re an expert on HVDC and electrification, so why don’t you start with the journey that got you to being the CTO at GE Vernova?

[CP]: In some ways, it’s a dream come true for me, for sure. From a fairly early age, I developed an interest in electrical engineering. It’s probably one of the least tangible subjects for me, very abstract. It fascinated me that you can calculate things that you can’t see or touch or smell and do useful stuff with it. That led me to focus on STEM-related topics, physics mostly, and ultimately to choose to study electrical engineering. I studied at Imperial College in London, where I did my undergraduate degree, and, probably as many engineers will have experienced, after three years of studying you think you know everything. So I decided to quit my studies and join Shell, where I was lucky enough to join the first Dutch offshore wind farm project as an intern.

That’s where I got exposed to power engineering. Power engineering is something that really stuck with me because of its scale. I like big things and the logistics involved with them. It often becomes quite political because those projects have a social role in society as well, supplying power to people like you and me. I also realized when I joined that project that I knew nothing. I remember the very first day, when the lead engineer drew out the single-line diagram of the wind farm, and I could barely recognize any of the symbols he was using. That made me think, well, maybe I do need to continue my studies.

After I worked on that project for a year and learned a lot about high voltage, logistics, and all the equipment and technology involved in realizing such an offshore wind farm, I decided to go back to university with renewed focus. I finished my master’s degree and got the opportunity to do a PhD. I stayed on for another three years, focusing on converter control and protection. I had the luxury of choosing my own topic and my own supervisor. Maybe at the time it wasn’t as critical a topic as it is today, but we were focusing on how converters behave in the case of a grid fault.

[MB]: Everybody became an expert on that overnight last year after the Iberian Peninsula thing. It was amazing how many people were all of a sudden PhD-level power engineers.

[CP]: Yeah, AI-powered, which wasn’t around when I was doing my PhD. In fact, we really struggled at the time to get some of the findings published because a lot of reviewers weren’t seeing this as a real problem yet. We didn’t have inverter-dominated grids at the time. But it did set me on a path toward being interested in inverter-based transmission, so HVDC. A gentleman I work closely with today, Colin Davidson from GE, came to Imperial College and gave a lecture on HVDC LCC, line-commutated converter technology. I remember thinking, now that’s cool. That is really the biggest scale of power electronics, probably the highest complexity in terms of control. That is something I’d quite like to do.

So I applied for a job with GE Alstom at the time, but was unlucky and didn’t get it. I decided to move back to the Netherlands, live with my partner there, and get a job at KEMA. At that time in the Netherlands, if you wanted to do something serious with high voltage, the KEMA short-circuit laboratories were the place to go. These are the most powerful short-circuit laboratories in the world, and this is really where cutting-edge technology would be tested and validated. Of course, sometimes it didn’t work out and you would hear a mighty bang, which is always entertaining. But that company got bought by a Norwegian company, DNV, so I went with it.

I moved to the advisory arm, where I did a lot of failure investigations, especially into cables, but also other kinds of what we call primary equipment. That’s where I really got an appreciation for high-voltage engineering and for the quality control necessary to make something that can be hundreds of kilometers long. You can’t have the tiniest thing wrong with it, because the smaller the defect is, the more dangerous it becomes inside the insulation, for example. That makes it a very fascinating component to learn more about. After doing that for a few years, I felt I’d learned enough about cables and wanted to go back to systems.

I got the opportunity to lead a European Union-funded research project called PROMOTioN, which is a bit of a heavy-handed acronym for Progress on Meshed Offshore HVDC Transmission Networks. This was a very large, multi-year research, development, and demonstration project to show that the technologies needed to build multi-terminal meshed HVDC grids, such as HVDC circuit breakers, the control and protection systems needed to do it, and HVDC gas-insulated switchgear, were at a sufficiently high technology-readiness level to build such grids. On top of that, we started working on what was needed from all the non-technical aspects to start creating such grids. One of the things that makes HVDC very interesting is that, by its nature, it is typically applied over very long distances: hundreds, sometimes thousands, of kilometers. That means you’re always crossing borders of some kind, with different people who own the equipment, who have to invest in it and get a return on it, different regulatory models, and different political regimes with different ideals that can change every four years. So that makes HVDC projects kind of like a toy of these large storms that can happen.

[MB]: This is in the developed world and in the developing world, but there’s one country where it’s less of a concern.

[CP]: Yes. If you’re building it within a country, and I think that also goes for the Western world, it is much easier. Then we don’t have this issue with multiple owners and multiple regulatory frameworks.

[MB]: Outside of the United States.

[CP]: I think within the United States it’s fair to say that, within a state, it can be done more easily than if you have to cross states, or especially if you have to cross different ISOs. There’s also that part to it. We digress a little bit there, but that very notion is a challenge to building multi-terminal, multinational HVDC grids.

That was also one of the work packages in the PROMOTioN project: to chart out the issues, the gaps, and the areas where we could get countries to come closer together by, for example, aligning on what we call a benefit and how we measure it, what we call a cost and how we measure it, and then how we can start thinking about sharing it. I don’t want to say that we fixed or solved that problem, but we at least made it quite visible and took a first step toward some of the thought processes that we see today in planning the offshore grid in Europe. It was a super interesting project. I cut my teeth on that project and learned a lot about HVDC technology.

[MB]: What years did that project run?

[CP]: That was five years ago now, and then the five years before that. It started in 2015 and finished in 2020, roughly.

[MB]: Because now we’re starting to do the work for the HVDC mesh grid in the North Sea, right?

[CP]: Yep.

[MB]: So all the work you led there was preparatory for that. And, of course, as you probably remember, I was asked to assist with editing the second edition of Super Grid Super Solution, Eddie O’Connor’s book with Kevin O’Sullivan, on exactly that: a meshed HVDC overlay grid for Europe.

[CP]: I think we’re still a long way from that holy grail of a meshed, truly multinational overlay grid. But we can see the small building steps and building blocks appearing. The Germans are, in one way, taking a lead in creating the first true multi-terminal hubs with HVDC circuit breakers, but they’re still looking at it from a single-vendor perspective. So they’re tackling the HVDC circuit-breaker problem before tackling the multi-vendor problem. In the UK, there are some developments toward tackling the multi-vendor problem first, but leaving the DC circuit breaker probably for later. And then we see some actual multi-terminal projects happening. The first one was taken into operation, I think, in 2024 or 2025: the Caithness-Shetland-Moray link-up in Scotland.

We see multiple initiatives appearing now that use multi-terminal HVDC technology, but all with slightly different technology implementations. I think what we’ll first see is some of those projects being built and operational experience being gained. Based on that operational experience, we’ll find out what an HVDC grid code could look like, for example. Then you’ll see gradual standardization come in that might eventually lead to such a pan-European overlay supergrid in HVDC.

[MB]: I’m certainly hoping that the proposal for an EU grid architect comes through.

[CP]: Yes.

[MB]: Because right now it is the ENTSO-E club, and I’ve actually had personal experience of how that starts to evolve because of the work I did with TenneT Netherlands last year, assisting them with their 2050 decarbonized Netherlands target scenario. I was making it more pragmatic than it was so they could have a target grid. But that was all within the Netherlands. The way their scenarios had evolved involved dealing with a bunch of very parochial distribution system operators, industrial stakeholders, and other stakeholders who were basically saying, ‘Yeah, we’re just going to stay the way we are.’ Europe needs a grid architect, and there’s a proposal to get one in place. ENTSO-E formally doesn’t like it, but I—

[CP]: Can’t say too much about that.

[MB]: No, I’m not asking for an opinion from your position, but I do—

[CP]: —see that if you don’t have a multilateral grid-planning platform, then it will be very difficult to achieve true multilateral grid planning. That’s basically what we’re seeing in the US. We are a bit better at it in Europe, but there are definitely still steps to be made to make that a truly European grid plan that then also gets executed as such.

[MB]: For all the challenges that emerge from that organization, at least it exists and was doing the work.

[CP]: Exactly.

[MB]: That doesn’t exist in the United States.

[CP]: No, I still don’t think there’s a common understanding in the United States of how the benefits of a grid should be measured and quantified, for example, let alone an inter-regional grid-planning platform.

[MB]: And being in Canada, it’s like we have some movement in that direction, but it’s some movement in that direction. That’s about all I can say about it so far. It’s far from baked, and it’s kind of painful.

[CP]: But from an HVDC grid-planning perspective, I do see change in the United States within one state or one ISO realm. Recently in Ontario, an HVDC line was greenlighted, really drawing on the ability of HVDC to bring quite large amounts of power into city centers with minimal space use. Similarly, Dominion recently approved a couple of HVDC links to essentially meet their data-center demand. And, of course, recently we saw HVDC being included in the grid plans in CAISO. I think that was last year, or perhaps two years ago already. That means HVDC is finally beginning to become an accepted tool in the regular grid planner’s toolbox rather than something proposed by developers who then struggle to get through all the regulatory hoops to get it built.

[MB]: And to be clear, that’s a developed-world perspective. In Pakistan, they have the North-South HVDC that was put in with Chinese help. There’s a fair amount of HVDC in India. A lot of HVDC links are emerging in ASEAN, Singapore, and Malaysia so they can share their renewable resources.

[CP]: The Asian part of the world will see very strong growth in HVDC and, I suspect, will also drive a lot of the technology development in the near future.

[MB]: Last time we spoke, one of the things we talked about was that China now has its own entire technology stack. They leveraged a lot, but the thing about them is that they’re doing the biggest projects and the most projects in the world, so they’re learning all the lessons. Just as you spent your year learning power electronics on a wind farm, their engineers have deep experience.

[CP]: There’s no better school than actually doing something and learning from the mistakes that you make to get better the next time around.

[MB]: Hopefully, the mistakes other people make that you get to be part of fixing. Let’s pivot a bit. Tell me about Grid Systems Integration and GE Vernova overall, because it was only relatively recently that it was just GE. How do you refer to it internally: the divorce, the separation?

[CP]: I don’t really know. We don’t talk much about it, to be honest. Essentially what happened is that GE, which traces its roots back to Thomas Edison and being the original DC advocate, got into financial trouble and was restructured a few years ago into three entirely separate companies. As far as I understand it, pretty much the only things they have in common are that they’re still called GE something and we’re still using the original GE logo. So we have GE HealthCare, GE Aerospace, and then GE Vernova, which is the part I’m in. Vernova stands for ‘ver,’ something with green, and ‘nova,’ new energies.

GE Vernova itself is split into three major parts. We have the Power branch, where we build nuclear power plants, gas power plants, and hydro. We have Wind, where we produce the wind turbines. And then we have the Electrification business, which is the part I’m in now. Within Electrification, we have several different businesses. We produce the high-voltage primary equipment, the transformers, the switchgear, those things. We have a business focusing on grid automation: all the digital systems needed to make the grid work, from protection to system control, and the software needed to analyze what is happening in the grid. We have Power Conversion, where we make medium-voltage, megawatt-sized power-electronic drives, whether for vessels, connecting batteries to the grid, or supplying power to data centers. And then we have Grid Systems Integration, which is the part I’m part of.

Within GE, that is where we do HVDC, FACTS, and AC substations, as well as the services needed to keep those things running. What we really are is almost like an EPC arm. We are an engineered-equipment package vendor. We basically buy equipment from our other colleagues—the transformers, switchgear, and substation automation—and engineer and integrate it all into a solution that we can sell to our customers: developers, utilities, industry, whoever needs a transmission solution.

Within our scope, we don’t only buy equipment from our colleagues. There are some things our colleagues don’t make, like the HVDC power-electronic valves or the FACTS power-electronic valves; the control and protection platforms, consisting of the hardware, the computers on which the algorithms run, and the control and protection algorithms themselves; as well as converter transformers. The reason converter transformers are part of our scope is that we have the actual factory for the special type of transformers needed for HVDC converter stations within Grid Systems Integration. That means we can control the factory slots and capacity, the quality, and the design rules needed to make sure we can optimize the transformer design within the HVDC system design and have full control over it.

That, in a nutshell, explains Grid Systems Integration. We deliver HVDC, FACTS, and AC-substation solutions, and we make the three core products that go into those solutions: the power electronics, the control and protection, and the transformers.

[MB]: And now you’re the CTO of that organization. How big an organization is it? I mean, having worked as a major systems integrator with one of the biggest technology firms in the world, I get the job.

[CP]: Yes.

[MB]: I wasn’t a CTO with them. Talk about the CTO role, because that’s an interesting role. It’s not necessarily a program manager responsible for delivery. So what is the CTO role, and how is it structured inside GE Vernova with you in it?

[CP]: For me, it’s a dream job. I get to play a key role in developing technology that I’m passionate about. We have a couple of different hats on. The first is to drive our technology strategy within the company and make sure that we have a product portfolio—the valves, the control and protection systems, and, increasingly importantly, the control and protection algorithms as a separate product—that enables us to deliver competitive solutions. For example, if our customers indicate a need for a higher power rating for a converter station, I have to make sure that within our product portfolio we have a valve product capable of delivering it. If not, I have to identify that way ahead of time so we can start developing a new valve that can do it.

Identifying what the market really needs, what is possible from a technology perspective, trends in technology development, and how those might improve the performance of our technology and products is part of the role my team plays every day, translating that into proposals for new R&D projects. This includes developing new control and protection features that complete our library of different control functionalities, enabling us to expand the number of applications in which we can implement our HVDC solutions. For example, we are now commissioning an offshore wind farm in the North Sea with a symmetrical monopole. That is an offshore wind farm connecting to shore, which is a very different application from a project connecting two points within the same AC grid with a bipolar solution. You need a whole additional set of control features for that, which need to be added to what we call our software trunk, our library of control features.

Continuously expanding this and making sure we have as complete a set of control and protection features as possible—to address the different potential applications, grid codes, and regions of the world, and to keep up to date with changes in those grid codes, such as the addition of synchronous grid-forming functionality—is all part of the product roadmaps we establish and use to shape our R&D portfolio.

[MB]: Last time you gave me a strong primer on LCC versus VSC. What I’m hearing you say is that GE Vernova is exploring having those technologies in-house, but doesn’t today—

[CP]: We have those technologies in-house, both LCC and VSC. For LCC, we have a very significant installed asset base. But, of course, the technologies themselves keep evolving, especially when it comes to control and protection. On the one hand, we need to keep up with technology developments themselves. For example, we can buy processors with much greater processing power for less money these days. What does that mean for our control platform? Can we get something more reliable, that takes up less space and is easier to maintain? We use those kinds of developments to come up with new versions of products that we already have in our portfolio.

The other thing is that we’re expanding the number of different kinds of applications for existing technology. To do that, we need to add features to our control library that enable this expansion into different applications. For example, if you have a symmetrical monopole system and want to use the same technology to develop a bipolar system, additional control functions need to be added to the software trunk, our library of control features, to make that work. In the same way, if the grid code changes and requires new functionality like synchronous grid-forming, we have to add that to our library of control functions.

[MB]: In the context of this, and without asking you to unveil any secret sauce, you’ve been in the role for a while. Are there any clear trends you’re seeing that you expect everybody is going to have to adapt to?

[CP]: What we’re seeing, for sure, is the trend toward the 2-gigawatt standard design at 525 kV. This is almost a real standard building block now, especially in Europe, but we also see it happening elsewhere in the world. Different grids and geographies may lead to different kinds of standard designs. The other standard design we now see appearing is at the 3-gigawatt level. In the US, of course, there are several projects that use 3 gigawatts at 500 kV, but we now also see this appearing in India for VSC-related technology. So that is definitely a technology trend that everybody in the market is moving toward.

From a control and protection perspective, we’re seeing a move toward having synchronous grid-forming capability included in the requested set of control features or control strategies to deal with a changing AC grid. We see an AC grid where many existing conventional power plants are being replaced with inverter-interfaced generation. That changes the dynamics of the grid and changes the way stable operation is guaranteed. It means that this inverter-interfaced generation is increasingly being asked to play a role in guaranteeing that stability. That is really what synchronous grid-forming capability does: the converters themselves can very quickly adjust their output based on things happening in the grid to help maintain grid stability. This is a completely new control feature that hasn’t been implemented before, so it has to go through a lot of development and testing to make sure it meets the grid-code requirements and really works.

[MB]: Of course, when I spoke to Mark O’Malley a couple of years ago—and I know you know Mark—the example I came up with was that it was basically a herd of rabbits out there. There were inverters everywhere, but they were grid-following, not grid-forming. The capability was there, but the standards weren’t. Mark and his efforts were strongly focused on the emerging grid-control standards. Have the standards emerged to the point where there is now coherence around grid-forming inverters?

[CP]: No, I wouldn’t say there’s coherence. We see that standards have emerged in some parts of the world and are emerging in other parts, but there are still differences between them. That essentially means our implementation of synchronous grid-forming may be a little different in one part of the world compared with another, which adds to the complexity of the control and protection algorithm library we have to maintain.

[MB]: Okay, so we’re starting to see standardization. Certainly when I talked to Mark, the obvious ability of VSC technology to be grid-forming was there. It’s power, and it creates a sine wave.

[CP]: Yeah.

[MB]: Now we’re standardizing on 2-gigawatt HVDC, so we’re getting modularity in one aspect, which is good, but we’re not quite at modularity and standardization in other things. I talked to Bent Flyvbjerg a couple of years ago as well, around the time his book was coming out, and one of the conversations we had was about how you get as modular as possible. Transmission is already one of the major megaproject types most likely to hit schedule and budget, but making it better is part of that increasing modularity.

[CP]: I think you’re definitely hitting an interesting point there. We are standardizing, if you can call it standardization, on power levels, voltage levels, and converter configuration. These are all 525 kV, 2 gigawatts, mostly bipolar converters, and often with a metallic return, which is definitely a very important point. It helps the supply chain build capacity around those characteristics. These are physical characteristics.

What we’re not seeing as much standardization on yet, but what no doubt will have to come if we’re moving toward a truly inverter-dominated grid, is how we define control modes and the performance of those control modes. Today that is described to some extent in grid codes and in our customer specifications, but it will probably be implemented a little differently by all the different vendors. They will all meet the requirements, but the implementations might be slightly different.

This is where the topic of multi-vendor interoperability comes in. InterOPERA, the follow-up to the project that I led, is very soon going to deliver proof that we can actually build functioning grids with HVDC converter stations from different vendors. But it probably also requires a lot of work toward defining how we implement control modes so that we can be confident they’ll work from the start, without having to go through very cumbersome iterative testing where everybody continues to tweak their control approach a little bit, without knowing what the other people are doing, until it eventually works.

[MB]: The burning question in my mind is: is this a mostly solved problem inside China, or are they still iterating through this as well?

[CP]: What the Chinese are showing is that multi-vendor operation is not a technical challenge. It can be done as long as information can be shared about control systems, interfaces, and definitions, and as long as the architecture of the grid allows that within whatever intellectual-property protection looks like in China.

In Europe, we’re choosing a different paradigm. Of course, we don’t like to share the detailed information of our control and protection system with our competitors, and vice versa. We spend a lot of money developing this. There’s a lot of intellectual property in there that gives us our competitive edge. So we have to find a way of making it work without exchanging all of that information, and that means we need a different approach.

In China, the way HVDC systems are designed is that the end user takes a much bigger role in the system design and in defining exactly how it should be implemented. That means you’ll have similar implementations between different vendors, which simplifies integrating the technology of different vendors. What you’ll also sometimes see is that the higher-level control and protection systems are delivered by one vendor and the lower-level systems by another. It’s typically the higher-level control and protection between different converters that determines whether they’re going to work together. So if those are from the same vendor, it’s easier to achieve multi-vendor interoperability among converter stations than if they are from different vendors.

[MB]: There’s a structural thing that really ties back to the PROMOTioN project. They’ve solved more of those problems simply because of their government model. Remind me: they have a single transmission system operator for all of China?

[CP]: They have two: China Southern and China State Grid.

[MB]: That is a much less complex TSO landscape than we’ve got in the West.

[CP]: And the TSOs also have a stake in some of the OEMs developing this technology. So they own part of the technology and can therefore help align and exchange information that would otherwise be protected by IP between those vendors, which of course helps the integration.

[MB]: Not actually replicable in Europe or the United States, I would say.

[CP]: Well, we have a different system, which has its own pros and cons, and we have to take a different route. But it’s certainly not technically impossible. In fact, multi-vendor systems have been around for a long time for LCC systems. We’ve been doing this for decades. There’s less information that needs to be shared, which makes them easier to integrate. But it shows that it can be done if the will is there and if the business case is there.

[MB]: Let’s pivot a little bit. This might not be directly in your wheelhouse because you’re setting strategy and research agendas, but right now the world is going through this massive spate of electrification. The supply chains are stretched. Timelines for getting technologies are stretched. That probably means, I would assume, GE Vernova’s order book is absolutely full and the place is humming. Is there an expectation of relief? Is manufacturing capability coming online to deal with this?

[CP]: I think it’s fair to say that relief is already here somewhat. Last time we spoke, I think, was when we found ourselves really on this ramp-up in terms of projects being announced and large tenders being signed, vendors getting their order books filled very, very quickly, and developers and utilities scrambling to get a production slot to meet their timelines. However, some things have changed a little since then. First of all, we’ve seen a slight slowdown in the uptake of HVDC, notably in the United States, due to all kinds of rules from the current administration. But also in Europe, some of the focus has shifted toward Ukraine and everything that comes with it.

It’s led to a speeding up of some project plans, but also probably a pivot away of some of the funds that were dedicated to solving the energy transition toward supporting Ukraine and bolstering our security at home. So we see that demand for HVDC systems has dropped off a bit. It is still significant and still growing, but not quite as fast as before. At the same time, what all vendors are really seeing is a bull market for HVDC. It is growing faster than we’ve ever seen before. It is increasingly becoming part of the regular transmission planner’s toolbox, and that means we’ll probably see increasing growth for a while to come.

Of course, we’re all investing in capacity, and some of that capacity is beginning to come online. At the very least, we know when it’s going to come online and can take that into account in what we bid for. So some of the pressure is beginning to come off, which we also see in the competitive landscape around us. At the same time, I would not expect really huge amounts of additional capacity to come online any time soon on top of what has already been invested in. We also want to see what is going to happen to the HVDC landscape. Will this growth persist or not? And what will really happen when all the factories being built today open their doors and increase capacity even further?

[MB]: Certainly from the UK, the Black Sea Interconnector from Georgia to Romania is approved in principle, but it’s not going anywhere as long as there’s a live war in Ukraine.

[CP]: There are a few other projects like that as well, where the plan is there and it’s approved in principle, but now we have to go and find financing. Or it’s in the plan, but we realize the offshore wind market is not as strong as it used to be, so we’re postponing some of the offshore wind connections. That doesn’t mean they won’t happen. It just creates a little uncertainty around exactly when. Of course, we don’t want to be building factories that will then be empty, so we adjust what we do based on a risk profile, as do all the other competitors in the market. I would imagine it is a different—

[MB]: —geopolitical landscape than when we last spoke and you were North America-based. And now, I’m just going to say, you’re a lot closer to one problem area and a lot farther away from another problem area.

[CP]: Exactly. Except both have really developed into problem areas. I think the biggest challenge we’re facing is the ability to staff projects: the ability to hire experienced and skilled engineers who can start contributing to product development and project delivery at the scale and growth levels the industry is still demanding. Even though the growth levels have come down a little, they’re still not low enough to match what is available in terms of human support on the other end. I think that remains one of the challenges: getting enough skilled engineers into this field to work on all the projects that need to be done.

As an industry, we’re pivoting in ways to deal with that problem. More standardized solutions are one way; they reduce the engineering hours per project. We’re also looking at what AI can do for us to help speed things up. But in the end, we still need people to do the work. That is, I think, the main challenge that probably most vendors and OEMs in this business are facing today: getting the people, getting them trained, and getting them to start contributing to real projects.

[MB]: So I have a habit of, every once in a while, looking at a different domain and looking at the percentage of engineers who are trained in China, or Chinese nationals trained externally, versus the rest of the world. I’m going to assume that for power engineering, the ratios are still extraordinarily weighted toward Chinese nationals going through many of these programs. Is that a fair assumption, and does that impact GE Vernova’s workforce composition?

[CP]: First of all, I’m not sure how many of the students would be Chinese or from other parts of Asia. But when I go and give lectures, which I often do at many European universities, there is a very large number of foreign students in the classroom, as opposed to UK or Dutch students, for example, in the two countries where I’ve lived and where I lecture. So I would probably agree that we have an issue in terms of the number of people from homegrown schools, let’s say, opting for an engineering degree, and that we have a lot of foreign students coming to pursue those degrees in European universities, who then also, of course, join GE to contribute here.

[MB]: It is fascinating to me. One of the things I observed in the United States over the past year was that a full 50% of foreign students in the top universities were Chinese nationals. Of course, after their education, many of them stayed. Traditionally, that’s happening a lot less now, but they’re also going back. The other thing I observed recently was that Nature released its highest-cited universities. One of the striking things was that American universities had almost fallen off. I aggregated Europe and China and compared them with the United States, and it was China, Europe, United States, in that order, which was quite substantially different. The Chinese institutions have just shot up the ranks. The focus on education there has shot up, the United States has declined, and Europe has stayed very strong.

Obviously, the number of foreign students in US universities has plummeted in the past year, so that’s going to have a significant impact. You mentioned AI. The people I know who follow AI papers say it’s something like 60% to 80% Chinese researchers. They’re just dominating so many fields of STEM. And speaking of students and the new workforce, it’s Engineering Month. So let’s talk about Engineering Month. Is this a GE Vernova thing? When I looked at it, it seemed to be a North American pattern. But you’re doing it based out of Paris, so tell people about Engineering Month in general and then GE Vernova’s interpretation of it.

[CP]: I don’t actually know if this is a thing beyond GE Vernova. Certainly in my previous role, we did not have a celebration of Engineering Month. It’s something I really bumped into when I joined GE Vernova, what is it now, eight months ago? And really, what it is is a celebration of engineering excellence. At the core of everything we do is technology and the ability to engineer technology into useful products that enable competitive solutions. One of the things I probably underestimated a little until I joined here is the amount of detail, the depth, the creativity, and the persistence it takes to create products that can control the amount of power needed to feed entire cities on a microsecond-by-microsecond basis. What this week is about is really highlighting what it takes to do that.

What are all the different disciplines, the different phases of a project, and the different phases of product development? What are all the different things we need to think about? And how do we place that in the greater context of what we’re trying to achieve: creating energy security, creating a clean and sustainable power supply, enabling safety, helping economies and people thrive, ultimately, and the role that engineering plays in that?

[MB]: You know how I feel about engineers. I’m not an engineer. At some point or other, some guidance counselor missed the switch, because I should have been one. But I spend a lot of time dealing with engineering disciplines and engineers. And so I’m privileged, honored, and very grateful that you invited me to speak to GE Vernova’s engineers. I’m looking forward to it.

[CP]: Looking forward to it.

[MB]: And on that note, I guess it’s kind of the end of our time together for today. So the question I would ask you is the question of underestimation. Do you think the industry is underestimating some particular part of the problem set that’s going to cause problems over the next decade?

[CP]: The fundamental change the grid is going to undergo in the next decade or decades is the shift away from being a grid supplied by rotating generators whose characteristics we understand very well, and on which essentially the entire grid design is based, to a grid supplied by distributed inverter-interfaced resources. That is a fundamental shift in the way we operate grids, but also in the way we design grids and guarantee stability in their operation.

I wouldn’t say we’re underestimating what that shift really means. I think people are quite aware of what it means. But are we really ready for that shift, and do we fully understand what the solution is? Things like synchronous grid-forming are part of the solution, but is that enough? There is still a bit of an open end there, which would be great to learn more about with future guests in future episodes.

[MB]: Certainly, talking with Mark O’Malley was an eye-opening experience. It was a tremendous opportunity for me. I’m so privileged that so many people have shared amazing stuff with me. And on that note, Case, thank you so much for spending an hour with me. I’m looking forward to talking to your engineers in a few weeks, and I’m sure we’ll talk again soon.

[CP]: It was my pleasure, as always. Thanks a lot for the opportunity, Michael. I appreciate it.

If you found this conversation useful, listen to the full episode of Redefining Energy, subscribe wherever you get your podcasts, and subscribe to the TFIE Strategy Briefing for more analysis on grids, electrification and the technologies reshaping the energy transition.

Subscribe now

This essay was triggered by Bryan Cheang’s Montreal Economic Institute critique of Mark Carney’s proposed Canada Strong Fund. Cheang’s central objection deserves engagement: Norway accumulated petroleum revenues before investing them, while Singapore accumulated reserves before establishing GIC. Canada proposes to capitalize a C$25 billion investment fund while carrying government debt, so whatever label Ottawa uses, the assets have to earn enough to cover financing costs, operating expenses, investment risk and the opportunity cost of using the money elsewhere.

Singapore makes the argument less tidy. GIC is only one of its two major public investment institutions, and it is not the one most relevant to what Carney appears to be attempting. Temasek was incorporated in 1974 with 35 companies worth S$354 million that the Singapore government had created or acquired during its industrial development. The state retained ownership while professional commercial management was separated from day-to-day policymaking. Temasek does not prove that public investment is inherently wise, or that Canada can reproduce Singapore’s institutions, but it establishes a more important point: public wealth need not begin as oil royalties or a pile of accumulated cash. Governments can help create productive capabilities, retain ownership, impose commercial discipline and compound the resulting wealth.

The most revealing complication comes from Cheang himself. His peer-reviewed paper, What Can Industrial Policy Do? Evidence from Singapore, is considerably more nuanced than the MEI article. He treats Singapore as a serious challenge to the Austrian proposition that successful industrial planning is impossible and accepts that its industrial policies contributed materially to development. His remaining case is about trade-offs: state-led development may have constrained local entrepreneurship and small-business discovery relative to an unknowable market-led alternative. That is a worthwhile argument about institutional capacity, commercial discipline and opportunity cost. It is no longer an argument that governments cannot create productive wealth.

Once that concession is made, the interesting question changes. It is not whether government or markets should run the economy. It is how governments can set strategic direction, build common capabilities and solve coordination problems without suppressing the distributed knowledge, experimentation and competitive pressure that make markets useful.

The Hayek I Had Misunderstood

That question led me back to Friedrich Hayek and to a correction in my own understanding of him. I had repeatedly encountered extreme positions attributed to Hayek, checked what he actually wrote and found more nuance, which left me inclined to blame the harder market fundamentalism associated with his name mostly on later libertarians and free-market institutions. That was too generous to Hayek himself.

His central economic insight remains powerful. Much of the knowledge required to coordinate an economy is local, tacit, temporary and distributed among millions of people. Prices let that knowledge influence decisions without requiring a minister, executive or investment committee to possess the whole picture. Hayek also distinguished centralized direction from what he called planning for competition: government could establish institutions and rules that allowed competitive discovery rather than prescribing every result. This leaves considerably more room for public goods, institutional design and social provision than the caricature in which markets know everything and governments know nothing.

It also supplies one of the best questions that can be asked of industrial policy: how could officials know that the company, technology or production model they favour will outperform alternatives they have not considered?

The later Hayek carried that insight into much harder political territory. His concern about the limits of economic knowledge expanded into hostility toward deliberate attempts to shape distributive outcomes, a rejection of social justice as a coherent objective and a willingness to subordinate democracy to his conception of liberal order. The qualifications around his notorious comments on temporary authoritarian government matter, but so does the fact that he made them. Those positions were not simply invented by the institutions that later claimed his intellectual legacy.

The evolution was not a clean division between a sensible young Hayek and an extreme old one. The Road to Serfdom was already a forceful political argument, while the older Hayek retained more nuance about legitimate government activity than many of his followers. The better description is that a powerful epistemic warning widened and hardened. What is worth retaining is the knowledge problem, competition as discovery and suspicion of institutions claiming more certainty than the evidence permits. What should be rejected is the leap from government cannot know everything to government should therefore avoid collectively choosing strategic, environmental or distributive outcomes. The knowledge problem is a constraint on how government should act, not a universal veto on acting.

Keynes, Schumpeter and the Symmetry of Failure

Keynes supplies the necessary counterweight, although industrial policy was not his primary subject. His contribution here is the recognition that non-intervention is itself a consequential policy choice. Markets can produce collapsing investment, prolonged unemployment and socially destructive outcomes without quickly returning to an acceptable equilibrium. Watching that happen is still a policy decision.

The 2008 financial crisis demonstrated the point through revealed preference. Stephen Harper’s Conservative government had defined itself partly around fiscal restraint and scepticism about activist government, yet its 2009 Economic Action Plan introduced major stimulus when the alternative appeared to be a much deeper recession. Faced with an actual collapse rather than an ideological debate, tough-love fiscal conservatives reached for collective demand management.

Keynes and Hayek were also less alien to one another than later political arguments imply. Keynes told Hayek after reading The Road to Serfdom that he agreed deeply with much of its moral and philosophical argument. Their important disagreement was over the inference Hayek drew: Keynes did not accept that moderate planning, public investment or social provision necessarily led down a road toward totalitarian control.

The useful synthesis is therefore complementary. Keynes explains why governments sometimes have to act because markets can produce destructive outcomes or fail to supply essential capabilities. Hayek explains why the resulting intervention should preserve decentralized information, competition and adaptation rather than assuming that officials possess a complete answer in advance.

Schumpeter supplies a final correction that industrial-policy critiques routinely omit. Markets produce enormous numbers of losers. Failed firms, obsolete factories, stranded capital, bankrupt investors, displaced workers and discarded technologies are part of the process he called creative destruction. It makes little analytical sense to describe failed private investments as healthy experimentation while treating failed public investments as proof that government cannot allocate capital.

The meaningful comparison has to count failure on both sides and ask what remains afterward. Did failures produce useful knowledge, productive capability or competitive pressure? Did successful investments elsewhere compensate for them? Were social costs or strategic vulnerabilities excluded from the private-market calculation? A government programme in which every supported company succeeded would itself be suspicious, suggesting that assistance flowed only to safe incumbents, losses were repeatedly refinanced or politically favoured firms were protected from competition. A credible industrial strategy should tolerate losers without repeatedly rescuing them.

Taken together, Keynes, Hayek and Schumpeter suggest a practical rule. Government should identify public purposes it can justify, build common platforms, correct specific failures and establish durable constraints such as carbon prices, reliability requirements or strategic-security objectives. It should then leave as much technological, corporate and operational discovery as possible to competing firms, customers and investors. Where public money carries meaningful risk, the public should receive an appropriate share of the return. Where the evidence turns against a project, cancellation has to remain a genuine outcome.

Carney’s agenda provides a live test of that synthesis. Transmission and shared infrastructure create arenas in which many firms and technologies can compete. The Canada Strong Fund is formally designed to invest commercially, take minority positions and retain public upside. The proposed west-coast oil pipeline has followed a different path: governments created the public development vehicle while its route, binding shippers, full financing and production case remained unresolved.

Canada no longer needs to decide whether it has industrial policy. It already does. The harder question is whether that policy uses markets to discover viable answers—or progressively rearranges public risk until a predetermined political answer appears commercial.

Read more

GeoPura has been promoting a striking maritime first at the Port of Tilbury on the Thames. Its HPU2 hydrogen power unit was placed on a floating platform and used to provide rapid charging to a commercial electric vessel, delivering 300 kW dockside from a system built around five Ballard fuel cells. The accompanying language stretches outward from that event toward shore-to-ship power, rapid vessel charging, port electrification, grid constraints and energy resilience. At an April hydrogen showcase at Tilbury, attended by more than 100 delegates, it made for an excellent demonstration: a barge, a hydrogen generator, a commercial vessel and a collection of technologies sufficiently unusual to merit photographs and headlines.

What it did not demonstrate was a maritime charging business. There is no stream of vessels using the floating system, published utilization, a charging tariff, repeat customers or a fleet whose operating model depends on it. The demonstration vessel was also not hydrogen-powered. It was battery electric. Strip away the event language and GeoPura demonstrated that a hydrogen generator can make electricity which can be fed through a charger into a battery. That is technically unsurprising. The interesting question is why hydrogen needs to be in that chain at all.

The engineering problem GeoPura says it is addressing is real. Ports can have substantial electricity supplies while lacking enough capacity at a particular berth to deliver a short burst of several hundred kilowatts or a few megawatts. Distribution upgrades can take years. Cables have to cross busy industrial sites, substations need capacity, and a ship may require much more power for an hour than the local connection can continuously provide.

Those characteristics describe an energy-storage problem unusually well suited to batteries. A modest grid connection can charge a stationary battery slowly between vessel calls, while the battery supplies the much higher charging rate when a ship arrives. The grid sees a manageable load; the ship sees a high-power charger. If the berth itself is awkward to electrify, the battery can be containerized or mobile. For suitable cargo operations, the battery can go one step further and move onto and off the vessel, separating the time required to charge batteries from the time the vessel spends alongside. That is why my review of maritime battery studies concluded that the practical battery boundary is moving outward, with port infrastructure increasingly becoming the constraint rather than battery chemistry.

GeoPura inserts a substantially longer chain. Renewable electricity makes hydrogen in an electrolyzer. The hydrogen is compressed, stored and transported or produced nearby, supplied to the HPU2 and converted back into electricity by fuel cells. The HPU itself combines fuel cells, battery storage and power electronics before electricity finally travels through the charger to the vessel battery. GeoPura’s own hydrogen efficiency explanation says that around 3.1 MWh of electricity used for electrolysis results in approximately 1 MWh of usable electrical output later through a fuel cell.

There is also a rather obvious question missing from the floating-charger story: how exactly does the hydrogen get to the barge? GeoPura’s normal supply model is to move compressed hydrogen by road in MEGC tube trailers or manifolded cylinder packs. Tilbury has announced an on-site green hydrogen production facility, but even hydrogen made inside the port still has to get from the electrolyzer to a mobile HPU sitting on the water. Does a tube trailer drive to the quay and refuel it through a high-pressure transfer system? Are cylinder packs moved onto the floating platform? Does the port install a dedicated hydrogen pipeline to the berth, undermining part of the claim that this avoids fixed infrastructure? Or does the barge periodically stop charging vessels and move somewhere else to refuel? A mobile electrical generator does not make its fuel logistics disappear. It merely moves the infrastructure problem upstream, while a battery buffer needs the electrical connection the port ultimately requires anyway.

That makes the Tilbury demonstration an unusually pure example of the hydrogen detour. Electricity is converted into hydrogen, moved through a specialized fuel chain and converted back into electricity, with a battery incorporated in the generator architecture before the electricity reaches another battery in the vessel. A battery buffer skips most of that equipment and most of those energy losses. This is exactly why my maritime fuel economics put electricity first: once useful propulsion energy rather than kilograms of fuel becomes the comparator, direct electricity has a formidable efficiency advantage wherever the operating geometry permits it.

There are places where this tradeoff can still be rational. GeoPura’s strongest application remains displacement of diesel generators where useful grid electricity genuinely is unavailable: temporary construction, events, remote sites and backup power where customers value quiet operation and elimination of local combustion emissions. GeoPura itself primarily describes the HPU family as generators for temporary, off-grid and grid-support applications. I described that as one of the remotely defensible hydrogen-for-energy niches when I assessed Ballard’s proposed acquisition of GeoPura. The company has real equipment, customers and operating experience. The problem was never whether fuel cells could generate electricity. It was whether the economics justified building a very large hydrogen production, distribution and HPU business around them. A port berth where the final requirement is high-power battery charging is a considerably weaker extension of that niche.

While hydrogen projects continue to generate firsts, battery maritime systems increasingly generate operating records. Zero Emission Services began commercial operation of the Alphenaar in the Netherlands in 2021 using standard interchangeable 20-foot energy containers. Its newer LFP ZESpacks have 2.9 MWh of capacity and charge at up to 1 MW. The operating model allows vessels to exchange packs instead of sitting alongside waiting for them to recharge.

China is pushing the same systems logic further. CATL’s Jining 6006 electric cargo vessel carries two containerized battery power sources totaling 3.919 MWh, has a stated 230 km range and can swap batteries in about 15 minutes. More important than the specifications is the architecture: CATL describes an integrated vessel, shore charging and operations system rather than a fuel product looking for places to be used.

These systems are not free of infrastructure requirements. Ports need electrical capacity, battery inventory, charging stations, power electronics, safety systems and operating procedures. None of that disappears because the word battery is involved. My maritime battery analysis found that ports increasingly become the constraint as viable battery range expands. The issue shifts from whether batteries can propel a vessel to how many megawatts can be supplied at the quay, how charging loads are managed and how batteries move through the logistics system. Buffering storage turns sharp vessel loads into steadier grid demand, while swapping allows packs to charge for hours without holding a vessel at berth.

That is what a system looks like. It accumulates common infrastructure, repeat users and operating experience. Each additional electric vessel can use some combination of the same grids, substations, chargers, batteries, power electronics and control systems already serving electric trucks, port equipment, buildings and stationary storage. This is the broader point in my practical pathway for maritime decarbonization: electrify fixed and shorter routes, hybridize the next band, build port electrical infrastructure early, and reserve molecules for the portions of shipping that actually need them.

Tilbury itself illustrates the contrast. In February, two months before the hydrogen showcase, a 5 MW electric HGV charging hub opened at the port, with 16 ultra-rapid charging positions capable of serving 16 electric trucks simultaneously. That is ordinary high-capacity electrical infrastructure intended for repeated commercial use. The fact that the same electrical capacity does not automatically extend to every quay is an infrastructure problem. It does not turn hydrogen into the obvious solution to the missing cable.

The timing gives the Tilbury demonstration another layer. It took place on April 30. Ballard announced its proposed £275 million acquisition of GeoPura on June 23. Ballard was already supplying fuel cells to GeoPura, so nothing about the acquisition enabled the Tilbury demonstration. Ballard itself describes the transaction as building on an existing partnership and combining its fuel cells with GeoPura’s hydrogen production, logistics and stationary-power business. In my analysis of the transaction, that existing relationship was central: buying GeoPura does not create the supplier-customer relationship. It internalizes it and wraps it in a much larger integrated-hydrogen-company narrative.

Ballard’s acquisition announcement makes that narrative explicit. It says the combination will maximize revenue per deployed megawatt through multiple customer touchpoints and give Ballard access to the stationary power market and recurring energy-as-a-service revenues. GeoPura, meanwhile, presents the same basic HPU platform across construction, film and television, events, data centres, EV charging, utilities and maritime.

There is nothing improper about selling the same equipment into multiple sectors. The investor problem comes when every place a generator can physically operate is treated as evidence of another scalable market. Put an HPU beside a film set and it is clean production power. Put one beside an EV charger and it becomes hydrogen-enabled electric mobility. Put one beside a data centre and it becomes resilient digital infrastructure. Put one on a floating platform and it becomes maritime electrification. The machine has not changed very much. The headline has.

That distinction matters for Ballard because the acquisition is being justified partly through addressable-market expansion while GeoPura itself remains a capital-intensive growth business. My earlier assessment found a company with substantial previous capital raises and asset-backed debt facing an ambition to deploy thousands of HPUs requiring billions more in investment. Ballard gets a new growth narrative extending beyond transport markets in which hydrogen has repeatedly failed to achieve broad commercialization. The Tilbury story does not prove that maritime charging belongs in that addressable market. It proves that the equipment can be placed on a barge.

This pattern is familiar from passenger rail. My recent global audit, Hydrogen Trains Have Headlines. Battery Trains Have Higher Utilization, found that hydrogen trains receive international attention for national firsts, world firsts and power records, while much of the evidence for battery trains sits in regional transport reports and routine operator notices. Germany already has substantially more battery-electric passenger trainsets operating than hydrogen ones, a larger battery orderbook and better practical fleet utilization in the evidence I could reconstruct, yet hydrogen launches remain disproportionately visible.

Hydrogen naturally produces publicity because every deployment is unusual. A new production arrangement, tube trailer, storage installation, dispenser or fuel-cell application can be announced as a project milestone. Electrification becomes less newsworthy as it succeeds. The first battery ferry is an event. The hundredth charging session is operations. A transformer feeding a battery behind a quay has little visual drama, especially when the same battery chemistry and power electronics are already appearing in trucks, buses, grid storage and industrial equipment.

That difference in visibility can badly distort perceptions of technological progress. Hydrogen transportation has accumulated enormous numbers of pilots without establishing broad commercial markets. In my review Hydrogen Transport Has Been Contained, Not Commercialized, I assessed 174 hydrogen transportation firms and projects. Of the 106 surviving entries, only three qualified as durable commercial niches, all in material handling and tiny, mostly legacy niches at that. Forty-seven remained demonstrations, pilots or proofs of concept, while another 36 were subsidy-shaped transport niches.

The point is not that demonstrations are useless. They are supposed to resolve uncertainties on the way to repeat deployment. When a sector continues producing firsts without producing enough seconds, thirds and hundredths, the demonstrations themselves become evidence about the weakness of commercialization.

The same standard should apply at Tilbury. How many vessels use the system after the showcase? How many megawatt-hours does it deliver in an ordinary month? What does that electricity cost at the vessel compared with grid power buffered through batteries? What utilization is required to recover the capital tied up in the hydrogen production, transport and fuel-cell chain? Does another port buy the system after seeing those numbers?

Until those answers exist, there is very little maritime significance to the demonstration. Hydrogen successfully generated electricity. An electric vessel successfully accepted it. Both technologies were already known to do those things.

My latest shipping energy pathway finds a much larger electric maritime system emerging as fossil cargo declines and the liquid-fuel requirement shrinks. Deep-sea shipping still has a hard liquid-fuel remainder, but charging an electric vessel at a berth is not part of that hard problem. Ferries, inland vessels, port craft, working vessels and much short-sea activity have exactly the operating patterns that make electricity attractive.

GeoPura has demonstrated a way to charge one of those vessels using hydrogen. The maritime transition will be demonstrated when charging them is so routine that nobody thinks it deserves a press release.

For analysis that separates transition infrastructure from transition publicity, subscribe to TFIE Strategy Briefing.

Subscribe now

On August 12, Heart Aerospace’s X1 demonstrator lifted off from Plattsburgh International Airport in upstate New York, marking one of the most consequential electric-aviation flights to date. The 106-foot-span aircraft weighed more than 25,000 pounds at takeoff and flew for 27 minutes entirely on batteries, with its electric propulsion system delivering more than a megawatt at peak. Heart describes X1 as the largest battery-electric aircraft ever flown. The record matters less than what the flight represents: a full-scale aircraft approaching commercially relevant size took off, maneuvered and landed using batteries and electric motors alone. After years of redesigns, delays and hard engineering, Heart has moved regional electric aviation another step from projection toward physical reality.

In 2021, I interviewed Heart founder and CEO Anders Forslund, when the company was developing the 19-seat, fully electric ES-19 after closing a $35 million Series A. Forslund was already clear that this was nowhere near enough money to develop a commercial aircraft. His benchmarking and bottom-up estimate was around $500 million to reach early production and put the first airplanes into the air, with successive physical milestones required to unlock successive financing rounds. The number belongs to a different aircraft programme now, but the financing logic looks prescient. Aerospace eventually has to become hardware, and hardware has to perform before investors finance the next, considerably more expensive stage.

Heart has hardly followed a straight line. The ES-19 became the 30-seat ES-30, pure battery propulsion became hybrid-electric, the hybrid architecture changed again and a commercial-service target of 2026 became 2031. X1 is not the production propulsion configuration either: its four propellers were electrically driven from batteries, while the intended ES-30 combines electric propulsion with conventional turboprops. Heart is late and still has an enormous distance to travel through certification, manufacturing and airline operations. What has remained stable is the problem it is trying to solve. Regional airlines, airports, routes and passengers already exist, and short regional flights are an obvious place to exploit electricity’s efficiency and low energy cost if batteries can carry enough of the mission.

Aviation did not lack speculative transition capital while Heart was grinding through that process. It spent extraordinary amounts on two propositions whose central problems were visible well before most of the money arrived: urban eVTOL air taxis and hydrogen-powered passenger aircraft.

Read more

In May, I wrote about what happened after Poland heavily subsidized hydrogen buses. Cities discovered that grants covering much or all of the vehicle cost did not produce cheap transportation. They still had to procure hydrogen, secure refuelling infrastructure and absorb the risks of depending on a specialized fuel supply chain. By April 2026, Poland had 153 hydrogen buses registered, 140 in service and another 107 contracted, enough deployment to move the discussion beyond demonstrations and pilot projects. The results documented by CEE Bankwatch included high delivered fuel costs, failed tenders, station losses, fuel-quality problems and municipalities switching planned hydrogen purchases to battery-electric buses.

That assessment mostly looked at the transit-agency side of the market. Poland’s hydrogen strategy had a broader industrial-policy logic. Subsidized buses would create hydrogen demand, demand would support refuelling stations, stations would help justify hydrogen production, and production would support a domestic hydrogen industry. The experience of Polenergia shows what happened on the supply side. Poland’s largest private energy group spent several years developing renewable-hydrogen projects, secured technology partners and public support, and in one case won a 15-year municipal fuel tender. It nevertheless concluded that hydrogen no longer justified substantial additional investment and began withdrawing from the sector.

The most useful case is H2HUB Nowa Sarzyna, in southeastern Poland. Polenergia planned a 5 MW renewable-hydrogen installation capable of producing roughly 500 tonnes annually, together with distribution and refuelling infrastructure. It contracted Norwegian electrolyser manufacturer Hystar to supply and commission the plant, and the International Finance Corporation agreed to support development spending and part of the electrolyser purchase cost. Poland’s environmental fund separately offered support for hydrogen filling stations at Nowa Sarzyna and the Rzeszów bus depot. By October 2024, Polenergia had a building permit and all eight Hystar electrolyser stacks had completed factory acceptance testing. The project had therefore progressed substantially beyond the memoranda, feasibility studies and future production targets that make up much of Europe’s hydrogen pipeline.

It also had something many hydrogen projects lack: an identified long-term customer. MPK Rzeszów was procuring hydrogen for 20 fuel-cell buses, and in October 2024 Polenergia won the tender. The proposed supply agreement was worth about PLN120 million and would have run for 180 months, giving the project a 15-year municipal offtaker. The quantities were meaningful without being sufficient to absorb the plant’s entire production, making Rzeszów the sort of anchor customer around which additional transport or industrial demand could theoretically develop. This was close to the market-development sequence Poland’s hydrogen strategy envisioned: public procurement created durable demand, an established energy company built local renewable supply, and dedicated infrastructure connected the two.

Polenergia nevertheless decided in January 2025 that it could not conclude the Rzeszów agreement. Its disclosure cited legal issues associated with the tender as well as the risk that it would not be able to deliver hydrogen on schedule. The decision came while the company was already reassessing its hydrogen strategy. Its 2025–2030 corporate strategy subsequently called for a gradual withdrawal from hydrogen transportation, a subsidized hydrogen-storage research project at Nowa Sarzyna was discontinued, and the H2HUB investment was impaired. Later financial disclosures cited the development of the green-hydrogen market, the project’s investment-risk profile and limited possibilities for obtaining financing. The important point is not that one tender collapsed, but that an established energy company looking at the complete investment case decided that the project no longer met its requirements despite having progressed unusually far toward operation.

Polenergia’s much larger H2Silesia project makes the wider strategic shift clearer. The proposed 105 MW plant was intended to produce around 13,000 tonnes of renewable hydrogen annually for industrial and transport customers. It qualified under the European Hy2Infra programme for a state-aid ceiling of €142.77 million, and in June 2025 Poland’s Bank Gospodarstwa Krajowego recommended PLN618.3 million in non-repayable support under the National Recovery and Resilience Plan. Even with that prospective subsidy, Polenergia continued to make investment conditional on securing hydrogen sales contracts, acceptable project economics, financing and final corporate approval. After Polenergia sought changes intended to reflect current market conditions and its revised strategy, BGK concluded that the alterations could not be accommodated within the funding arrangement. The grant agreement was not signed, H2Silesia was impaired and subsequent company reporting listed the project as abandoned.

The reasons Polenergia gave shareholders are more useful than another theoretical comparison of electrolyser and battery efficiency. In its responses to shareholder questions, it identified rapid improvements in battery vehicles and growing competition from Chinese manufacturers as reducing the attractiveness of hydrogen vehicles in low-emission transport. It also pointed to slow development of the regulatory mechanisms expected to support renewable hydrogen and to customers delaying purchasing decisions. Polenergia was not retreating from the energy transition generally. Its strategy directs substantial capital toward offshore and onshore wind and other electricity assets where revenues can be secured through mechanisms such as power purchase agreements and contracts for difference. Hydrogen was being asked to compete internally for capital against technologies with clearer demand, simpler infrastructure and more bankable revenue streams, and it lost that competition.

Rzeszów then had to deal with the consequences from the other end of the supply chain. Its 20 hydrogen buses were still arriving, but the prospective local renewable-hydrogen supplier was gone. A subsequent fuel procurement attracted a single offer from PAK-PCE Stacje H2, part of the ZE PAK/Polsat Plus hydrogen business. The contract was worth about PLN9.64 million for an estimated 117 tonnes of hydrogen through the end of 2027, above the city’s initial budget, with the first buses supplied through containerized refuelling infrastructure while permanent facilities are developed. Instead of creating enough predictable demand to bring competitive local production into operation, Rzeszów became a relatively small captive hydrogen customer dependent on an external supplier and dedicated logistics.

That outcome matters because Poland has not lacked either hydrogen policy or public funding. The country subsidized hydrogen buses to establish demand, supported refuelling infrastructure, funded electrolyser projects and offered substantial grants to larger renewable-hydrogen developments. The Nowa Sarzyna project added an established energy company, a contracted electrolyser supplier, international financial involvement, permitting and a 15-year municipal customer. Those conditions addressed many of the explanations usually offered for why green-hydrogen projects fail to reach final investment decisions. They still did not create an investment case attractive enough for Polenergia to continue.

This does not imply that renewable hydrogen has no future in Poland. Poland already consumes large quantities of fossil-derived hydrogen in refining, chemicals and other industrial processes, and replacing that existing hydrogen with progressively lower-carbon supply remains a substantial decarbonization task. Those consumers also provide large, concentrated demand that does not first have to be created through subsidies for new end-use technologies. Urban buses are different because a mature direct-electric alternative already exists. My May assessment argued that creating additional hydrogen demand in public transport before decarbonizing Poland’s existing industrial hydrogen consumption was poor sequencing. Polenergia’s withdrawal strengthens that argument by showing that the hoped-for supply response was not simply waiting for municipalities to buy enough buses.

The original Polish industrial-policy proposition was that subsidized transport demand could help bootstrap a competitive low-carbon hydrogen ecosystem. The bus programme did create customers that require hydrogen for the next decade or more, but Polenergia’s experience shows that creating demand is not the same thing as creating economical supply. Even with substantial public support and a long-duration customer, prospective local production could remain difficult to finance and unattractive relative to competing investments. Poland’s hydrogen bus experience therefore now provides evidence from both sides of the market: municipalities discovered that subsidized vehicles could leave them with expensive and fragile fuel supply, while a prospective supplier discovered that subsidized demand did not necessarily make producing that fuel an attractive business.

If decision-grade assessments of transport technologies, infrastructure and industrial policy are useful to your work, subscribe to TFIE Strategy Briefing. Paid subscribers get the deeper evidence, scorecards and procurement analysis behind the public arguments.

Subscribe now

Clean Energy has won a $27.6 million contract to build a second hydrogen station for the Orange County Transportation Authority as OCTA expands its fuel-cell bus fleet from 10 buses toward 50. The press-release reading is straightforward: a growing zero-emission fleet needs more fuel capacity. But OCTA already opened a substantial hydrogen station at its Santa Ana base in 2020, sized to fuel roughly 40 to 50 buses per day. The total cost for the various refueling systems has risen to $108 million, well over $2 million per bus.

That first station did not simply wear out. OCTA owned much of the installation, but Air Products owned the leased 18,000-gallon liquid-hydrogen tank and vaporizers. After the parties failed to reach a new commercial agreement, Air Products removed that equipment in January 2026 and the station became unusable. OCTA is now planning a roughly $19 million replacement and capacity expansion at Santa Ana while separately paying Clean Energy for the new Garden Grove station.

The operating consequences were already visible before the equipment was physically removed. OCTA’s ten fuel-cell buses traveled 270,462 miles in 2024 and only 14,232 miles in 2025, a decline of almost 95%, amid hydrogen fueling problems and reliance on off-site fueling. The agency sent buses to a commercial Shell station near Long Beach and moved toward temporary mobile fueling while trying to restore depot capability.

OCTA also operates battery-electric buses, which gives the case an unusually useful internal comparator. Its chargers have had reliability problems too, with OCTA reporting equipment availability around 80%, but the agency also reported no lost bus deployments because charging infrastructure was unavailable. The charging infrastructure also cost a fraction of the hydrogen refueling infrastructure, about $6 million. The hydrogen buses faced a different exposure because they depended on a dedicated fuel chain with few substitutes when a storage, supply or contracting link failed.

That is the Achilles heel of hydrogen for transit. A transit agency needs hydrogen that is affordable at the nozzle, genuinely low-carbon across its production and delivery chain, and reliably available every day for buses expected to operate for 12 to 15 years. Hydrogen bus programs keep struggling to secure all three conditions at once.

Low-carbon hydrogen starts with low-carbon electricity, but electrolysis consumes far more electricity than putting that electricity directly into a battery bus. The resulting hydrogen still has to be compressed or liquefied, transported or produced on site, stored, temperature-managed, dispensed and maintained to demanding purity and safety standards. Centralized production can improve electrolyzer utilization but adds logistics. On-site production reduces some logistics while adding electrolyzers, compression, storage and often poor utilization of expensive equipment. Those steps remain in the delivered fuel cost regardless of who owns them or which grant paid for them.

The Garden Grove award does not eliminate that exposure. The $27.6 million figure is not simply the construction price of a station. It bundles design-build work, facility modifications, fuel supply and 18 months of operations, maintenance and training. OCTA’s procurement documents also specify that at least 33.3% of the hydrogen supplied during commissioning and training must come from renewable sources. They do not establish that OCTA has secured 100% renewable hydrogen at an affordable long-term delivered price for the fleet’s operating life.

Orange County is not an exceptional case. SunLine Transit in California has spent a quarter century trying to make hydrogen transit work and has cycled through early electrolysis, reformers, a large PEM electrolyzer and, most recently, a truck-fed liquid-hydrogen station. Major refueling-system expenditures across those generations add up to about $27 million in 2026 dollars for a fleet of roughly 31 to 32 hydrogen buses. Compressor failures, station outages, delivered backup hydrogen and repeated rebuilding persisted despite exceptional institutional experience with the technology.

Poland provides the same signal under very different market and policy conditions. By April 2026 it had 153 hydrogen buses registered, supported by more than €120 million in grants for buses and refueling infrastructure, so this was no longer a collection of tiny pilots. Cities then encountered high delivered hydrogen prices, sparse fueling infrastructure, unprofitable station operations and fuel-contamination events that grounded buses. Several municipalities redirected planned hydrogen procurements toward battery-electric buses as the operating economics became clearer.

Taken together, SunLine and Poland make it difficult to explain OCTA as one bad vendor relationship. Hydrogen transit repeatedly turns a bus purchase into a long-duration dependency on specialized fuel production, storage, transport, dispensing, maintenance and counterparties. Battery-electric buses also need infrastructure, but their depots connect to an electricity system that already serves every city and is expanding for buildings, industry and vehicles. Hydrogen requires transit agencies to establish and sustain a second energy-delivery system for a comparatively small fleet.

Public subsidies can hide that difference at the point of purchase. OCTA’s program, like SunLine’s and many European hydrogen bus programs, has been supported heavily by government funding. Grants can pay for expensive buses and stations, but they cannot remove conversion losses, supply contracts, specialized maintenance or equipment replacement cycles. Once the buses are purchased, another station or another subsidy can become defensible simply because abandoning the existing assets would crystallize the earlier loss.

That dynamic deserves more attention in transit procurement. More buses improve station utilization, while more station capacity protects the bus investment, allowing each successive decision to reinforce the previous one. The relevant comparison is not whether the next hydrogen project can win grant funding. It is whether the entire fuel system can deliver low-carbon passenger-kilometers more cheaply and reliably than battery-electric buses over the fleet’s life.

OCTA’s response is now additional redundancy: rebuild Santa Ana, add Garden Grove and use temporary fueling in between. For an agency already committed to 50 hydrogen buses, that is understandable operational risk management. It also demonstrates how much infrastructure has to sit behind a supposedly simple advantage such as fast refueling.

Transit procurement should treat hydrogen supply as part of the drivetrain rather than as a separate facilities line item. An agency that cannot show how affordable, genuinely green hydrogen will reach its depot reliably for the full service life of the buses has not finished the propulsion-system procurement. OCTA’s experience is larger than one six-year-old station becoming unusable. The specialized fuel chain behind hydrogen buses keeps proving to be the weakest part of the system.

Transit procurement reality lives in operating costs, infrastructure uptime and service delivered, not grant-funded purchase orders. Subscribe to TFIE Strategy Briefing for more evidence-led analysis of what survives contact with operations.

Subscribe now

China produced 45.02 million tonnes of primary aluminum in 2025, 2.4% more than the year before. During the first half of 2026 it produced another 23.19 million tonnes, up 3.8% year over year. Those numbers complicate the argument I made earlier this year that China’s aluminum-sector carbon dioxide emissions had probably peaked around 2024, because one of the expected contributors to that peak—saturation of primary aluminum output near China’s nominal 45-million-tonne capacity constraint—has not yet appeared in the production statistics.

The more interesting possibility is that the emissions thesis may survive the production surprise. Aluminum is sufficiently dominated by electricity that absolute sector emissions do not have to move in lockstep with tonnes of metal. If average emissions per tonne decline faster than production rises, China can continue making more primary aluminum while its total aluminum emissions flatten and eventually fall.

That turns China’s aluminum industry into a cleaner test of industrial decarbonization than it looked like six months ago. A peak caused by declining production would be straightforward. A peak occurring while output is still reaching records would indicate that the structure of production itself had changed enough to break the historic link between industrial growth and emissions growth.

Read more

Power to Hydrogen is celebrating an important engineering milestone at the Port of Antwerp-Bruges: its first industrial-scale AEM electrolyzer, a half-megawatt system intended to produce high-purity hydrogen in a real port environment. The company calls it commercial-scale and emphasizes that the hydrogen will go to real customers. There is nothing wrong with celebrating the hardware. Moving anion-exchange-membrane electrolysis out of smaller stacks and into industrial equipment is technically interesting, particularly if P2H2 can deliver the lower capital cost and renewable-following performance it claims.

But “commercial” is doing more work than it first appears. P2H2’s own project description calls Antwerp’s NextGen site a demonstration district. The Port is funding site preparation, utility partners from the Free Electrons program are providing the capital for the electrolyzer build, and P2H2 received a €900,000 European Regional Development Fund grant to support deployment. The hydrogen is expected to serve industrial operations at the port and local refueling. That makes this a legitimate industrial demonstration with useful output. It does not make it evidence that a self-supporting hydrogen market has arrived. Further, the customers is actually customer, Holthausen.

The more revealing commercial signal is coming from banks. Holthausen is an unusually good case because it removes most of the excuses normally attached to hydrogen financing failures. The Dutch family business dates to 1945 and built its name supplying and transporting industrial gases. Its Holthausen Energy Points business has been producing hydrogen for years, operates hydrogen stations in Groningen and Amsterdam, fills cylinders and trailers, has actual customers, and sits beside an established gas-distribution network. This is not a startup arriving at a bank with a slide deck and a projected total addressable market.

When Holthausen wanted to install larger electrolyzers, it went looking for debt. According to regional development agency NOM, a round of discussions with banks produced nothing, and even the company’s then-house bank would not finance the expansion. Holthausen told Triodos why: its hydrogen production and filling-station operations had been losing money for years, and the losses were getting larger. NOM adds another useful detail. Buying hydrogen from elsewhere let the company keep growing revenue, but there was almost nothing left as profit.

That is close to a natural experiment in hydrogen bankability. The lenders could see actual operating history rather than modeled production, projected utilization and hoped-for customers. They were not being asked to take technology risk on an unknown company. They were being asked to accept management’s thesis that larger electrolyzers would turn a loss-making hydrogen operation into a profitable one. Conventional banks, including the incumbent bank, looked at the numbers and declined.

The eventual capital stack tells the rest of the story. Triodos Bank, whose sustainability mission aligned with the project, agreed to finance it but preferred that another investor come in alongside it. NOM then invested and became a shareholder. Holthausen also received support through the Dutch OWE hydrogen subsidy scheme, which NOM explicitly says helped close the cost gap between expensive green hydrogen and cheaper gray hydrogen. In other words, the expansion became financeable after conventional debt was supplemented by mission-aligned lending, public development equity and government subsidy.

Holthausen is particularly exposed to the part of the hydrogen story that has weakened fastest: transportation. It does have a more defensible merchant-gas niche filling high-purity hydrogen cylinders for laboratories and industrial customers, but its stations and production strategy were built in significant part around vehicles. My review of now 181 hydrogen-mobility companies and projects found 65 had failed, dissolved or abandoned hydrogen, while only three survivors occupied somewhat durable commercial niches, all in material handling. Belgium’s own hydrogen refueling network has similarly struggled with a denominator measured in very few vehicles per station. Hydrogen transportation has been contained, not commercialized.

One Dutch company does not prove what every European banker thinks. Fortunately, we do not have to rely on Holthausen alone. The European Commission’s 2025 Innovation Fund knowledge report says hydrogen projects continue to struggle to secure long-term offtake, that customers show little willingness to pay the green premium, and that the absence of a liquid market undermines price discovery. It says the lack of long-term offtake can prevent non-recourse project finance, that some projects unable to obtain support from parent companies or banks turn instead to grants, and that banks often view counterparties across the hydrogen value chain as unacceptably risky.

The IEA’s 2026 Global Hydrogen Review says much the same thing from another direction. New low-emissions hydrogen offtake agreements were roughly flat in 2025 at about 1.7 million tons, and only around 20% of the newly signed volume was backed by firm contractual commitments. Demand, not electrolyzer manufacturing capacity, remains the missing piece. That is the part a lower-cost AEM stack cannot fix.

There is an almost perfect institutional joke hidden in the name European Hydrogen Bank. It is not a commercial bank. It is an EU financing mechanism that pays fixed premiums per kilogram of qualifying hydrogen, with more than €1 billion awarded in the latest auction round. Its purpose is to bridge the gap between what clean hydrogen costs and what customers will pay. Europe has created something called a bank because market finance alone has not been enough to create the market policymakers expected.

The important caveat is that banks are not refusing all hydrogen. In July, the European Investment Bank agreed to lend OMV €450 million toward a €600 million, 140 MW green-hydrogen plant in Austria. The hydrogen will travel through a dedicated 22-kilometer pipeline to OMV’s Schwechat refinery, where it will replace fossil-derived hydrogen already required by the refining process. That is a large hydrogen project with a known industrial user, captive demand, a defined physical connection and a public-policy lender. It is almost the inverse of the speculative hydrogen economy.

That distinction is the one I have been making for years. Hydrogen measured against alternatives shrinks toward applications where the molecule itself is required. It does not disappear. Refineries, ammonia and some chemical processes have real hydrogen demand that must be decarbonized. What keeps failing is the larger idea that cheap clean hydrogen will create a new economy of road vehicles, heating, power generation, generalized energy storage, merchant fuel distribution and entirely new industrial demand.

P2H2 may have built a very good electrolyzer. AEM may reduce capital costs, improve dynamic operation and avoid some of the material constraints of incumbent technologies. Those are worthwhile engineering goals. But no improvement inside the electrolyzer creates a customer willing to pay enough for the hydrogen that comes out of it.

That is why the most important hydrogen signal in Europe may not be another stack leaving a factory or another port calling itself a hydrogen hub. It may be the credit committee saying no to an established industrial-gas family whose hydrogen business it can inspect in detail. Europe’s conventional banks appear to be pricing hydrogen less as a new economy and more as a set of bounded applications that require strong offtake, policy support or both. As always, be very leery of hydrogen headlines as they usually sound a lot better than the reality.

For more transition analysis grounded in what is actually being financed, built and used, subscribe to TFIE Strategy Briefing.

Subscribe now

China has crossed an electric-vehicle milestone that is less flashy than another record sales number and more important. New urban buses are now essentially 100% electric. In many Chinese cities, the procurement question is no longer diesel versus battery. It is which electric bus replaces the electric bus already in service. The International Energy Agency says almost 70% of China’s more than 680,000 electric buses were deployed before 2020 and sees the recent rebound in sales partly as replacement of those older fleets.

That is what a mature transition looks like. The first big electric cohorts are aging out and the replacement cycle remains electric. Policy has not disappeared; its role has shifted toward renewing the fleet. China’s 2025 national programme explicitly subsidized replacement of new-energy city buses and even complete traction-battery replacement, and the equipment-renewal programme continues in 2026.

The useful denominator is therefore not “Chinese EVs.” It is the share of new sales inside each transport use case. City buses, passenger cars, heavy trucks and intercity coaches are all road vehicles, but operationally they are separate markets. They run different distances, stop in different places, have different charging options, earn or save money in different ways and face different competitors. Put them on one national EV curve and those distinctions vanish. Plot sales share by use case and the transition becomes much easier to read.

Passenger cars are already well into the mainstream. Electric cars accounted for just over half of Chinese new-car sales in 2025. China also had nearly 700 electric car models available by year end, more than the number of conventional models, and the IEA estimates that nearly 70% of battery-electric cars sold there were cheaper than comparable conventional cars even before incentives. The car curve is still rising, but electric is no longer the alternative drivetrain in China. It is one of the normal choices buyers see across the market.

Heavy trucks are farther down the curve and moving faster. Electric heavy-freight trucks reached about 29% of sales in 2025, roughly double their 2024 share. The strongest uptake is in applications with short or predictable routes around ports, mines, steel plants and industrial hubs, where high utilization makes fuel and maintenance savings matter quickly and charging or battery swapping can be organized around the work. Heavy trucks still trail cars in sales share, but their current slope is much steeper.

Intercity coaches are the revealing exception. Battery-electric coaches were only about 10% of Chinese coach sales in 2025. ICCT counted roughly 13,700 of them, up 66.7% from 2024, so this is no dead market. But it would be a mistake to assume that the destination is simply the city-bus curve shifted a decade to the right. Coaches have a harder duty cycle and, more importantly, they compete with a mode China electrified long ago.

China ended 2025 with more than 50,000 km of high-speed rail, and its railways carried more than 4.5 billion passenger trips during the year. High-speed rail reaches 97% of Chinese cities with more than 500,000 urban residents, with one-to-two-hour travel possible across many 500 km city clusters and roughly four-hour trips across 1,000 km corridors. Battery coaches can replace diesel coaches where road service remains useful, but the larger growth path for intercity passenger mobility does not have to be another road vehicle. Much of it is already rail.

For countries that are years behind China, that changes the forecasting problem. A low electric share today does not imply a slow linear climb toward China’s current share. China’s heavy-truck market shows how quickly a commercial segment can move when vehicle capability, infrastructure and operating economics align. Conversely, electrification does not require every vehicle category to reach the same endpoint. If rail takes more of the intercity work, coaches can electrify while the coach market itself remains bounded. The objective is not to maximize the electric share of every vehicle label. It is to provide mobility and freight with less fossil fuel.

The historical bus data deserve one caveat. Chinese sources have used overlapping definitions such as “new energy bus,” and successive IEA editions have changed classifications. The IEA’s 2026 edition explicitly revised China’s historical bus data using CATARC data to better match its current definitions. That is why the chart separates urban buses from coaches and why some older points are reconstructed rather than presented as a pristine official series. The exact shape in a given early year is less certain than the large pattern: city buses surged first and are saturated, cars followed, heavy trucks are now accelerating, and coaches remain much less electrified.

China’s transport transition is therefore a poor advertisement for one master EV adoption curve and a good advertisement for looking at useful work. Fixed-route urban buses went early because batteries fit the duty cycle and policy pushed hard. Cars became mainstream as models multiplied and prices fell. Heavy trucks are taking off first where commercial duty cycles reward electrification. Coaches are beginning to move in a market where electrified rail has already captured much of the attractive longer-distance growth. The lesson for countries behind China is not to draw a straight line from their current EV share to China’s. It is to identify which use cases are ready to climb next.

For more analysis of where electrification is actually scaling, subscribe to TFIE Strategy Briefing.

Subscribe now

Ontario undoubtedly needs more electricity. Demand reached 145.6 TWh in 2025, up 4.4% in a single year, and IESO’s 2026 Annual Planning Outlook has reference-case net demand reaching 250 TWh in 2050. The uncertainty around that number is enormous: its low-demand case is 207 TWh and its high case 297 TWh. That range alone should make planners wary of behaving as though the 2050 generation mix is already known. More importantly, IESO now distinguishes clearly between energy and capacity requirements. After accounting for actions already under way, it sees more than 8 TWh of additional annual energy need emerging in 2032 and more than 12 TWh by 2035, while the remaining incremental summer capacity requirement does not appear until 2035 and is initially about 950 MW.

Those numbers point toward two overlapping but different jobs. Ontario needs generators that produce additional annual TWh, and it needs a portfolio capable of ensuring that enough electricity is available at difficult hours while continuously maintaining frequency, reserves, ramps and other operating requirements. A reactor can provide enormous quantities of low-carbon electricity and dependable capacity over long periods. A wind or solar farm supplies energy with a very different hourly profile. A battery supplies almost no net annual energy but can shift it through time and respond extremely quickly. Hydroelectric reservoirs can provide both energy and flexibility. Transmission can make generation that is plentiful in one part of the system useful somewhere else. Demand response can eliminate a requirement to generate at a particular hour without producing a single electron itself. Treating all of these resources as competitors for one vaguely defined product called “power” is how bad capacity planning starts. IESO itself says an adequate and operable grid must simultaneously provide energy, capacity and essential reliability services.

Ontario’s recent gas trajectory is a useful demonstration of what happens when those products are not separated clearly enough. Transmission-connected gas and oil generation increased from 9.7 TWh in 2020 to 31.4 TWh in 2025, more than tripling in five years, while nuclear output declined from 87.8 TWh to 78.3 TWh as major refurbishment programs and outages worked through the fleet. Gas consequently rose from 7% to 19.3% of transmission-connected generation. That increase cannot reasonably be blamed on nuclear alone: demand increased, weather varied, exports and imports changed and market conditions matter. But IESO specifically says planned shutdowns of multiple nuclear units tightened 2025 supply conditions and that natural gas was increasingly relied upon to maintain reliability. Its March 2026 near-term outlook likewise says Ontario can absorb Pickering going offline for refurbishment partly because gas-plant upgrades and new battery projects are entering service.

Gas is doing more than filling an energy hole left by unavailable reactors. It is also doing jobs that baseload nuclear generally does not. IESO’s operability work says large hydro and gas generators provide both inertia and primary frequency response, whereas baseload nuclear supplies inertia but normally cannot provide primary frequency response because it operates at full output rather than changing output to counter frequency deviations. Hydro and gas currently provide most operating reserve, while hydro, gas and interchange do much of the system’s load following. IESO explicitly anticipates that battery storage and other technologies can take over increasing portions of these services as Ontario reduces its reliance on gas.

This distinction is central to the nuclear discussion because a nuclear-heavy grid is not a grid consisting only of nuclear plants. Very large reactors create very large contingencies when they trip, disappear for long planned outages during refurbishment and maintenance, and are economically happiest running at high output most of the time. None of that makes nuclear bad technology. It means that the hydro, gas, storage, transmission and controllable demand surrounding the reactors are part of the system architecture that makes nuclear useful. Ontario’s current architecture has allowed gas to play too much of that complementary role, and reducing electricity-sector emissions requires replacing much of it with non-emitting flexibility rather than simply adding more low-carbon baseload generation.

At the same time, flexibility cannot manufacture the 8 TWh IESO says Ontario starts needing in 2032. Batteries can move wind, solar, hydro or nuclear electricity from one hour to another. Demand response can move or eliminate load. Efficiency can genuinely reduce the amount of energy required, and IESO says expanded efficiency programs materially reduce its 2050 demand forecast and save $12.2 billion in system costs. But after those savings there is still an annual energy gap, and somebody has to generate those TWh. IESO lists on-time nuclear refurbishments and new nuclear, additional dispatch from the existing gas fleet, further efficiency and cost-effective new renewables among the options for meeting it.

That puts wind and solar much closer to the centre of the Ontario electricity discussion than provincial politics has allowed for most of the past decade. IESO has now returned to buying them after more than ten years without a major new-build renewable procurement. The first LT2 energy window ultimately executed contracts for 1,115 MW of projects with expected annual production of 2.37 TWh, and IESO has more procurement windows planned. When the projects were initially selected, the agency said competitive bidding had produced pricing 21% below the previous large-scale Ontario renewable procurement. The selected projects also required municipal support and substantial Indigenous equity participation, addressing two of the major political defects of the earlier Green Energy Act model.

Ontario’s long renewables hiatus was not the result of wind and solar losing a neutral economic competition. The province’s first renewables boom certainly had real policy problems. Early feed-in tariffs bought solar when the technology was much more expensive, siting policy alienated municipalities, demand stopped growing as expected and Ontario sometimes had surplus baseload electricity that forced wind curtailment. Those mistakes provided ideal raw material for populist politics. Doug Ford’s Progressive Conservatives turned wind turbines and electricity bills into wedges, cancelled 758 renewable contracts after taking office and even terminated projects that had already been materially developed. The cancellations also increased political risk for anyone signing a long-term infrastructure contract with the province.

The cost narrative attached to that political campaign was badly distorted. Ontario electricity prices had been held down by previous political interventions and eventually had to recover toward more economically realistic levels, while large legacy costs from the province’s nuclear build remained embedded in the system. I have written before that roughly half of the consumer price increase being blamed politically on renewables was associated with servicing nuclear debt, with much of the remainder associated with restoring rates after earlier suppression and repairing the grid. Ontario Hydro entered its 1999 restructuring with tens of billions of dollars of debt, much of it associated with nuclear investment and overruns. Wind and solar were visible and politically convenient targets; stranded nuclear liabilities and years of artificially low prices were much less useful on a campaign sign.

That episode deserves remembering, but not dwelling on, because the more important point is what Ontario should build now. The system that produced surplus electricity fifteen years ago is disappearing. Annual demand in 2025 was already 10.5 TWh higher than in 2020, IESO expects substantial further growth, and its 2026 outlook says surplus baseload generation is no longer the planning problem. Ontario is moving from an era in which it sometimes had too much inflexible electricity into one in which it expects to need additional TWh before it needs a large increment of new peak capacity. Wind and solar are particularly well suited to that first problem because they can be procured in repeatable increments and built on timelines measured in a small number of years rather than committing the province today to the exact generating fleet it will require in the 2040s.

Their limitations are real and should be accounted for rather than waved away. Wind production varies and is generally stronger in Ontario’s winter; solar is strongly daytime and seasonal. More variable generation increases regulation and load-following requirements because forecast output never perfectly matches actual output and because solar sunset or wind ramps change net load. IESO explicitly models those operability effects. Existing Ontario wind and solar also do not currently provide primary frequency response, although inverter-based resources can increasingly be designed to provide fast frequency services. None of this means Ontario gets to add arbitrary amounts of wind and solar with no balancing cost. It means that their very inexpensive energy has to be evaluated together with the incremental cost of the flexibility and transmission required to make it useful.

Ontario is now procuring that complementary system at meaningful scale. The latest LT2 capacity window contracted another 640 MW of new battery storage, bringing contracted battery capacity to more than 3.5 GW by 2030. More interesting than the quantity is the procurement curve: IESO says those batteries came in 36% cheaper than its expedited first long-term battery procurement and 16% below its first regular long-term procurement. Repeated procurements are doing exactly what modular technology markets are supposed to do: revealing current prices, creating supplier competition and allowing the province to update the next decision instead of forecasting one technology cost decades ahead.

Batteries do not eliminate the need for hydro, transmission or longer-duration resources. IESO notes specifically that batteries providing operating reserve or load following eventually have to recharge and can run into duration limits during sustained high-demand or low-renewable periods. Ontario therefore also has a Long Lead-Time procurement for resources including long-duration storage and hydro, is maintaining more than 1,000 MW of Northern Ontario hydro through new contracts, and has more than 1,800 kilometres of transmission planned or under development. Interties with neighbouring systems add another source of diversity and frequency support. The useful grid is a portfolio, not a battery monoculture attached to a wind farm.

Demand itself is becoming part of that portfolio. At Ontario’s 2025 annual peak, Capacity Auction resources reduced demand by 316 MW, Peak Perks delivered about 200 MW and the Industrial Conservation Initiative reduced demand by roughly 1,600 MW. IESO now has more than 320,000 homes and small businesses participating in Peak Perks, capable of providing more than 200 MW of demand reduction, while a similar program is being introduced for commercial and institutional buildings. Those numbers are still modest beside a provincial grid with peaks approaching 25 GW, but they precede most of the electrification that creates the really large flexible resource: millions of EV chargers, heat pumps, water heaters and other digitally controlled loads.

The emerging architecture is therefore reasonably clear. Wind and solar can provide substantial amounts of the incremental clean energy. Hydro supplies both energy and valuable flexibility. Batteries increasingly handle fast response, reserves and shorter-duration shifting. Demand response and smart electrification reduce the amount of physical capacity required for infrequent peaks. Transmission and interties increase the value of all of them by broadening the geographical resource pool. Gas currently fills gaps across both energy and operability, but its role should shrink as the non-emitting portfolio becomes deeper. None of these statements requires believing that an Ontario grid can or should become 100% wind, solar and batteries.

Nor does this architecture imply abandoning nuclear. Ontario’s existing reactor fleet is a major low-carbon asset, and recent refurbishment performance deserves recognition rather than being forced into an anti-nuclear narrative. Darlington’s refurbishment was completed ahead of the overall schedule and below budget, while Bruce Unit 3 returned seven months early and under its refurbishment budget. Extending productive reactors on existing sites with existing transmission, workforces and operating organizations is a very different proposition from choosing how much entirely new nuclear generation to build.

The Darlington SMR program belongs somewhere between those categories. Ontario is already physically constructing the first unit, including installation of the first reactor foundation, and plans four units totalling about 1,200 MW. Whatever one thinks about the economics, those projects are now sufficiently advanced that Ontario should learn from what they actually cost, how quickly subsequent units can be delivered and how they operate in the provincial system. Real construction data from Darlington will be more valuable for sizing the next nuclear program than another decade of vendor forecasts.

The much larger future nuclear options are where sequencing becomes critical. Ontario is advancing Wesleyville near Port Hope as a site that could hold as much as 10,000 MW of generation, while Bruce C has entered a roughly $300 million pre-development phase. Preserving sites, conducting environmental work and maintaining a nuclear development option can be sensible given the range in IESO’s long-term demand outlook. It is not the same thing as concluding that every prospective reactor should be constructed. At Wesleyville alone, the potential capacity is roughly ten times the residual 2035 capacity requirement IESO currently identifies after accounting for actions already in flight. Those numbers refer to different years and the system requirement grows beyond 2035, so the comparison is not a claim that Wesleyville is ten times too large. It demonstrates how much uncertainty remains between preserving the option and knowing the required fleet size.

Ontario consequently has a sequencing opportunity that is much more valuable than choosing an ideological winner between nuclear and renewables. The first step is efficiency because avoided electricity requires neither generation nor flexibility. The second is to procure large amounts of the fastest competitive clean generation available, which in present Ontario procurement means wind and solar will have a substantial role. In parallel, batteries, hydro, demand response, transmission and interties should be expanded so that more variable energy displaces gas rather than simply adding variability to it. Existing nuclear should be refurbished when justified, and the Darlington SMRs should be allowed to generate real evidence about the newest nuclear technology Ontario has chosen.

After several procurement cycles, the province will know much more than it does now. It will know whether the 2050 demand trajectory is bending toward 207, 250 or 297 TWh. It will know actual Ontario wind and solar prices rather than arguments about fifteen-year-old feed-in tariffs. It will know how much battery costs fall, how many hours of storage are economically useful, how much EV and building load can be controlled, where transmission remains constrained and how much gas has actually been displaced. It will also know whether the first Darlington SMR and its successors are meeting their own cost and schedule promises. IESO itself says the heightened uncertainty around demand supports an iterative planning and procurement approach.

There will almost certainly still be a requirement for dependable low-carbon generation. The scale is what is uncertain. If Ontario moves toward the high-demand case, variable generation reaches practical economic limits, gas must be driven close to zero and a large persistent energy deficit survives after efficiency, storage, hydro, transmission and demand flexibility are counted, additional large nuclear may be an entirely rational investment. If demand lands nearer the reference or low cases while repeated renewable and storage procurements continue to get cheaper, the economically justified new nuclear fleet could be substantially smaller.

The mistake would be to reverse that order and make the slowest, largest and least reversible investment the starting assumption. Ontario does need generation, and flexibility alone will not produce the missing TWh. It also needs far more flexibility, and new nuclear alone will not provide it. Wind and solar address much of the first problem quickly; storage, hydro, transmission and controllable demand address much of the second; the combination can push gas back toward the margins. Nuclear should then compete for the large, durable clean-energy requirement that remains.

Ontario’s Conservative government deserves criticism for spending years turning renewables into a political wedge and helping destroy the development pipeline it now needs to rebuild. But that is history, not the organizing principle for the next thirty years of electricity investment. The stronger indictment would be to repeat the error in another form by allowing political enthusiasm for nuclear to determine the generation mix before the system requirement has been discovered. Ontario has enough time to build the fast things first, learn from them, preserve its nuclear options and make the very large capital decisions with better information.

That is not hesitation. It is competent infrastructure planning.

For more analysis of electricity-system planning, nuclear economics, renewables, storage and grid flexibility, subscribe to TFIE Strategy Briefing.

Subscribe now

Richard Rumelt’s Good Strategy/Bad Strategy provides a useful lens for Boeing’s decision to move Wisk Aero, Insitu and SkyGrid into Archer Aviation. This is a Rumeltian assessment of both sides of the transaction, starting with his kernel of diagnosis, guiding policy and coherent action. Before deciding whether Boeing is giving away valuable businesses or Archer is assembling a powerful new aerospace company, the useful questions are what critical problem each company faces, what policy follows from that diagnosis and whether the actions reinforce it. On that basis, the two sides of the deal look quite different.

Under the transaction announced August 10, Boeing is transferring autonomous passenger eVTOL developer Wisk, profitable military drone company Insitu and airspace and autonomy software business SkyGrid to Archer. Boeing receives Archer shares equal to 19.75% of Archer’s pre-closing Class A share count, gains board representation and retains access to Wisk’s core autonomous-flight technology. The straightforward interpretation is that Boeing is streamlining while Archer gains technology, revenue and scale. Some of that is true, but Boeing’s history makes it important not to confuse selling businesses with having a strategy.

I’ve written before about Boeing’s decades of financialization and the outsourcing, loss of engineering authority and erosion of institutional capability that followed. Boeing did not end up in its current difficulties because the world stopped wanting airliners. It ended 2025 with a record $682 billion backlog, including more than 6,100 commercial aircraft, while still working through production stability, quality and certification problems of its own making. That makes its reacquisition of Spirit AeroSystems, an operation Boeing spun out two decades earlier, strategically significant. Critical manufacturing integration was once treated as something Boeing could push outside its corporate boundary. Bringing it back inside reflects the recognition that some capabilities are fundamental to solving Boeing’s actual problem.

Wisk is difficult to put in the same category. I described Wisk as a dead-end distraction in 2024, after Boeing had already put hundreds of millions of dollars into the autonomous air-taxi venture. Boeing needs to restore reliable execution on commercial aircraft and major defense programs for markets with demonstrated demand. It does not need to own an autonomous passenger powered-lift program whose eventual market remains highly speculative, so transferring Wisk while retaining useful autonomy access is consistent with the challenge Boeing actually has to solve. Insitu is less obvious. It generates more than $200 million annually and is profitable in a defense market Boeing says is core, while SkyGrid’s autonomy and airspace capabilities could also have future value. Given Boeing’s history of throwing organizational capability over the wall in pursuit of financial neatness, I would not assume that disposing of all three businesses is strategically wise simply because Wisk plainly does not belong.

Read more