Walk around Glasgow as I did for a week this summer and the buses advertise their corporate identities. First Glasgow, McGill’s and Stagecoach have distinct liveries, and other operators serve parts of the surrounding region. It can seem an odd way to organize public transport. London has numerous private bus companies too, but passengers largely encounter one Transport for London network, with common fares, information and service standards. Glasgow’s different buses are the visible consequence of a much more consequential institutional distinction. In one city, the public authority decides what passenger network it wants and hires companies to operate it. In the other, commercial operators have historically made many of those network decisions themselves.

The explanation begins with the Transport Act 1985, which deregulated most local bus services outside London from October 1986. Until then, much of Glasgow’s bus system had been operated through the public Strathclyde transport organization. Deregulation separated the operating business from the public authority and gave companies broad freedom to register commercially viable routes and timetables. The former public operating business was subsequently privatized, while operators serving surrounding districts competed, expanded and consolidated. The companies visible today are the descendants of that restructuring, not the product of a contemporary decision to divide Glasgow into separate passenger networks.

The early results were striking. Research into the initial effects of deregulation in Strathclyde recorded 83 operators across the region, twenty more than before the reforms, and more than 2,000 notified changes to unsubsidized services in the period examined after the initial regulatory standstill. Those figures describe the wider region nearly forty years ago, not the number of operators in Glasgow today. They nevertheless capture the scale of the experiment. Bus companies could respond to commercial opportunities by adding, withdrawing or modifying services without waiting for a public authority to redesign the network as a whole. The expectation was that competition and commercial initiative would produce a better and more efficient transport service.

London pursued competition differently. Instead of allowing companies to determine much of the passenger network, its transport authority retained control over routes, frequencies, fares and service standards. Private companies compete for contracts to operate services that the authority specifies. The operators still have to manage drivers, vehicles, maintenance, depots and day-to-day reliability, and they have strong incentives to control costs. The difference is that they are competing to deliver portions of an integrated public service rather than independently choosing much of the service being sold to passengers. Both models use private-sector expertise and competition. They simply place competition at different levels of the system.

Why does that matter to the person standing at a bus stop? Consider someone travelling across Glasgow to an evening shift, perhaps using two services operated by different companies. Each operator may be running its part of the journey competently. The first bus may be frequent and reliable, but that is worth considerably less when the second runs hourly or finishes before the shift ends. An individual route can be economically rational for the company operating it while the complete journey remains inconvenient, expensive or impossible. For passengers without access to a car, the consequences can extend well beyond a longer wait. The ability to reach employment, education and healthcare depends on how the services fit together, not merely on how efficiently each bus is operated.

Glasgow has not ignored those interfaces. ZoneCard provides travel across participating buses, ScotRail and Subway within selected zones, offering a common ticket for people whose journeys cross operator or transport-mode boundaries. Strathclyde Partnership for Transport, or SPT, also supports services that the commercial market would not otherwise provide. These are useful and longstanding forms of integration. They do not mean that SPT specifies the complete commercially operated regional bus network as Transport for London does. Nor does a common ticket determine which destinations should be connected, what evening frequency is adequate or whether an important transfer should be redesigned. Fare coordination solves part of the passenger’s problem. Network planning is a larger responsibility.

Competition itself is not the villain in this story. Studies of Britain’s first decade of deregulation found substantial reductions in the real cost of operating buses. That matters because labour, maintenance, purchasing and vehicle utilization are major costs, and money spent inefficiently cannot be spent on additional service. London also achieved substantial operating-cost reductions through competitive tendering while retaining public network control. The distinction challenges a familiar claim about the British experiment: improvements in operating efficiency do not establish that private companies must also be responsible for deciding the metropolitan route network. Competition can discipline the production of bus services without requiring the passenger network itself to be the product of separate commercial decisions.

Edinburgh and Dublin provide further evidence that the choices are not simply public versus private. Edinburgh developed a comparatively coherent service around Lothian Buses, a dominant municipally owned operator working within Scotland’s deregulated framework. Dublin’s National Transport Authority specifies a common passenger network while using both the publicly owned Dublin Bus and privately operated Go-Ahead Ireland to deliver services. Neither arrangement is identical to London’s, and each has its own institutional history. What they demonstrate is that operator ownership, commercial competition and responsibility for the passenger network can be combined in several ways. A city does not have to choose between a public monopoly and a collection of independently planned private services.

For Glasgow, the practical question is what belongs under common metropolitan responsibility. Someone needs to be accountable for whether routes connect useful destinations, frequencies and operating hours support real journeys, fares work across the system and service changes improve the network rather than simply individual routes. Operators still need freedom to perform the tasks they understand best, and public authorities need enough competence and funding to exercise the responsibilities they assume. Different company names on buses need not matter very much to passengers when the services behind them form a coherent whole. They matter considerably more when corporate boundaries determine the choices available to someone trying to cross the city.

Glasgow’s regional transport authority is now examining a different allocation of those responsibilities. SPT approved the development of bus franchising in March 2024, adopted a regional bus strategy in September 2025 and commissioned a formal franchising assessment in 2026. The work concerns the wider twelve-authority Strathclyde region, not just Glasgow City. Under a potential franchise, SPT could specify the passenger network while operating companies competed to deliver services. That would not necessarily require public ownership of First Glasgow, McGill’s or the other companies. As of October 2026, however, the statutory assessment is still under way, with independent audit and statutory scrutiny ahead. The proposed reform is a live policy choice, not an improvement already delivered.

Franchising would not guarantee better service. A public authority can design poor routes, set inadequate frequencies, make expensive contracting decisions or fail to respond to passengers just as commercial operators can make mistakes. The potential benefit lies in giving one accountable institution the responsibility and means to improve the complete network, with operators competing or otherwise being engaged to deliver it. Glasgow’s debate therefore concerns much more than bus-company ownership, paint schemes or the mechanics of procurement. It is about whether decisions affecting the same passenger journey should be made independently by several companies or coordinated across the region they collectively serve.

The buses on Glasgow’s streets provide a surprisingly useful window into Britain’s forty-year transport experiment. The enduring lesson is not that public buses are necessarily better than private buses, or that competition has failed. It is that competition in operating buses and responsibility for designing a passenger network are fundamentally different things. The complete journey is what passengers need. The operating company is one of the organizations required to deliver it.

The Bus Is Not the System — Metropolitan Transit Governance examines that distinction across Britain and Ireland, including the economic evidence, passenger outcomes and the growing significance of electrification infrastructure. The full 84-page Strategy Report and supporting Evidence Book provide the detail for transport authorities, operators, policymakers and investors. For access to the Evidence Book or related advisory work, contact Michael Barnard.

This is the first of five articles examining the report’s findings. Next: how a bus network can appear more productive while providing less service.

Subscribe now

Transit agencies have now had enough practical experience with zero-emission buses that procurement tells us more than another technology comparison. Agencies have tested vehicles, rebuilt depots, installed charging and refuelling equipment, trained technicians, negotiated energy supplies and learned which assumptions survive daily service. Manufacturers have had years to improve their products, while governments have funded demonstrations and early fleets. Battery-electric and hydrogen fuel-cell buses can both carry passengers in regular service. The more useful measure of maturity is what transit agencies are prepared to buy repeatedly and at fleet scale.

That purchasing evidence no longer supports the familiar framing of battery-electric and hydrogen as parallel zero-emission technologies whose eventual roles remain broadly unresolved. The global 2026 procurement register I assembled contains 388 sourced records across 81 countries, spanning firm orders and awards, open tenders, completed procedures whose outcomes remain uncertain, framework quantities and other identifiable stages of procurement. The pattern is overwhelmingly battery-electric.

The distinction between procurement stages is important because bus purchasing produces several public records around the same vehicles. A tender for 100 buses, the subsequent award and an announcement that the first vehicles have arrived can all describe the same purchase. Options may never be exercised. Funding approvals do not necessarily become bus orders. The register treats firm orders and awards separately from tenders and other forward-looking procurement signals instead of adding everything into one headline total.

On that conservative basis, the firm procurement in the register contains 13,165 battery-electric buses and 65 hydrogen fuel-cell buses. That is a little over 200 battery-electric buses for every hydrogen bus. Among the firm acquisitions captured for those two zero-emission technologies, about 99.5% are battery-electric.

The forward procurement evidence points in the same direction. Open procedures in the register contain thousands more battery-electric buses, while the identifiable hydrogen pipeline remains small. That matters because deliveries and registrations can lag procurement decisions by years. A bus entering service today may reflect a technology choice made when subsidy programs, vehicle prices and operating experience looked quite different. Current tenders and awards provide a more immediate view of what agencies are choosing after absorbing the lessons of the past several years.

The register is not a census of every bus purchased everywhere. Procurement transparency varies widely, terminology is inconsistent across countries and Chinese awards do not map neatly onto European tender procedures or North American multi-year contracts. Some purchases will inevitably be missing. Closing a 13,165-to-65 gap, however, would require a very large hydrogen procurement market that has escaped the available public record while battery-electric orders remained observable across dozens of countries. I have found no evidence of one.

Nor does convergence mean that every transit agency has selected battery-electric or that no more hydrogen buses will be bought. It means that the evidence is no longer balanced enough to justify treating the two as similarly established fleet pathways. Battery-electric buses now have a much larger commercial reference class: many agencies, many suppliers, repeated orders and deployments under very different operating conditions. That accumulated experience should matter when the next agency makes the same decision.

The complete workbook contains the individual procurement records, source register, classifications, country and regional fields, fleet denominators and calculations behind the analysis. Transit agencies, policymakers, investors and researchers who want access to the workbook or want to discuss the findings can email me.

The headline hydrogen number also turns out to be much less broad than it first appears. Most of the 65 firm hydrogen buses in the global register come from a single Chinese city where the municipal government is deliberately expanding hydrogen applications, financing the buses and developing the surrounding hydrogen supply chain.

Read more

The reaction to my recent Briefing on Ireland’s National Heat Study and Nigel Banks’ Fabric Fifth work included a familiar objection from someone working in social-housing retrofit in the UK. Deep retrofit, they argued, is not merely about saving energy. Warmer homes improve comfort and health, especially for people on low incomes and those already vulnerable to illness.

That is true, and the previous Briefing on SEAI’s National Heat Study and Fabric Fifth said so explicitly. SEAI accounted for people taking some efficiency improvement as warmer rooms rather than lower energy consumption and treated that additional comfort as a legitimate health and wellbeing benefit. Cold, damp and mould are genuine health problems, and social housing contains plenty of households struggling with them. The latest English Housing Survey found about 23% of social-rented households could not keep comfortably warm in winter.

But the exchange exposed something that was largely missing from my article as well. We were discussing healthy homes almost entirely as a winter problem. That leaves out an increasingly important part of the health risk. A harmonized Nature Medicine analysis estimated 67,873 heat-related deaths across 32 European countries in summer 2022, 50,798 in 2023 and 62,775 in 2024. These are epidemiological estimates rather than counts of death certificates saying heatstroke, but the scale is no longer ambiguous. Extreme summer heat is killing tens of thousands of Europeans in bad years, particularly older people and people with existing health problems. Many of those are precisely the populations invoked when making the winter-health case for retrofit.

The usual health ledger is full of cold homes, winter cardiovascular and respiratory stress, damp, mould, fuel poverty and excess winter mortality. Overheating, inability to cool a bedroom overnight, heat stress, disrupted sleep and heat mortality appear much less consistently. A home that is healthy in January and dangerous in July has not solved the thermal-health problem. Once that is recognized, the objective starts to look less like minimizing heating demand and more like maintaining safe indoor conditions throughout the year.

That matters to the Fabric First Trap because the argument has never been that insulation is bad. Cheap loft insulation, draught repair, ventilation fixes and other worthwhile building measures should be done. Some buildings genuinely require substantial remediation, and deep retrofit can be justified on comfort, maintenance, peak-demand, health or economic grounds. The trap comes when evidence that better fabric is beneficial gets turned into a sequencing rule: therefore extensive fabric work should precede electrification. The benefits of an intervention do not by themselves establish the order in which every intervention must occur.

Once summer heat and indoor air are put on the same health ledger as winter warmth, the sequencing question looks rather different. So does the role of the heat pump.

Read more

I was listening to Tom Raftery’s Climate Confident podcast conversation with carbon sequestration firm Graphyte’s founder when I started making a list of questions. The concept is appealingly physical compared with many engineered carbon-removal proposals: take waste biomass that would otherwise decompose or burn, dry it, make it resistant to water and microbes, and put it underground where the carbon should remain for a very long time. There is biomass going in, blocks going underground and a relatively direct measurement problem between the two. But drying takes energy, the blocks are wrapped in something waterproof, fresh forest residues are very different from dry agricultural wastes, and the boundaries around a carbon-removal process can make a great deal disappear. By the end of the conversation I wanted to know what those details did to the claimed removal.

If you are considering an investment in Graphyte and want to go beyond the public claims, email me. I have a full workbook behind this analysis covering the carbon, process, feedstock, logistics and competing-pathway assumptions, and can walk through the diligence in the context of an actual investment decision.

There was another question in the background before I opened a spreadsheet. For years I have increasingly organized energy and climate pathways around sequencing: use a scarce resource for the things it does particularly well before consuming it in a lower-order or terminal use. When contributing to TenneT’s 2050 target-grid scenario work in the Netherlands, that discipline showed up repeatedly. Electricity used directly for transport, heat and industry generally deserves consideration before turning the same electricity into hydrogen or synthetic fuels and accepting the conversion losses. In technoeconomic assessments across industrial technologies, fuels, infrastructure and carbon-management proposals, the same pattern keeps appearing in different clothes: define the useful service, identify the real alternatives, widen the system boundary and look for the point where a valuable resource gets consumed. My work on carbon capture and sequestration had already pushed me toward the same view from the other direction. Concentrated biogenic or industrial CO₂ streams produced after useful work has been done are inherently more interesting to me than systems that consume useful energy or material simply to manufacture something to sequester.

So I did not come to Graphyte looking for the idea of cascading biomass before burial. I came to it wondering whether Carbon Casting might be an exception.

Graphyte’s process takes residual biomass, dries it, forms it into dense blocks, wraps and isolates those blocks from water and oxygen, and places them into engineered storage. Its Loblolly project in Arkansas has been operating since 2024 using rice hulls, while Ponderosa in northern Arizona is intended to use wildfire-risk forest thinnings. Recent Isometric removal records from Loblolly provide something unusually useful for diligence: operating data on electricity, feedstock moisture, polymer use and transport that can be connected to physical tonnes rather than inferred from a generic process diagram.

The immediate diligence therefore started much more narrowly than the strategic question. How much does drying reduce the removal? How much fossil carbon is embedded in the polymer system being buried along with the biomass? What happens when logistics and the rest of the operating system are included rather than looking only at the finished block? Those were ordinary technoeconomic questions about energy, mass and system boundaries, and they looked potentially damaging when I started, particularly for freshly harvested forest material.

Instead, Graphyte survived them reasonably well. Dry rice husk requires almost no incremental drying heat. Plastic wrap emissions are measurable but modest compared with the carbon being stored. Wet forest thinnings carry a much larger drying penalty, but they also contain substantially more carbon per dry tonne. Adding the visible process burdens did not make Carbon Casting fall apart.

That result brought the pre-existing strategic question to the foreground. The interesting boundary was no longer around the dryer or wrapper. It sat before Graphyte received the feedstock at all. Carbon Casting was taking a useful biological resource and moving it immediately to a terminal use where its other value propositions are taken permanently out of circulation. The analysis had to test the question I had brought to it from the start: what happens if the biomass does useful work before we bury what remains?

Below the paywall: I follow the same dry tonne of rice husk and forest thinnings through Carbon Casting and several productive alternatives, track where their carbon, nutrients and minerals go, and test whether direct burial really deserves priority.

Read more

Verdox has done something worth taking seriously. In a two-month test at Hydro’s Sunndal smelter, the company reported capturing CO₂ from exhaust containing only about 1% CO₂, including exposure to contaminants found in smelter gas. Hydro’s investment and continued work with Verdox give the trial a serious industrial counterparty. The published account establishes a relevant field test, but it does not provide the gas mass balance, complete electricity boundary, pressure drop, product specification, electrode replacement record or installed cost needed to price a commercial facility.

I have spent more than a decade assessing dozens of CCS technologies and projects, from direct air capture to industrial flue-gas systems. That experience makes the Sunndal trial interesting, and it makes me follow the gas, energy and money all the way to permanent storage before putting an investable cost on it.

The first test is the gas volume: at 1% CO₂, a facility must move an enormous amount of mostly non-CO₂ gas for each tonne captured. Verdox advertises roughly 1.5 GJ, or 417 kWh, per tonne for its electrochemical process. Investors need to know which fans, pretreatment, switching, purge, drying and compression loads are inside that number, and how long a finished electrode lasts in the actual exhaust. A result measured at the cell cannot pay the bills for an entire capture, delivery and storage chain.

Aluminum adds a second question. Its carbon anodes produce most of a smelter’s direct CO₂ emissions, and both Hydro and Rio Tinto are developing ways to change the smelting process while they investigate capture. A capture plant attached to an existing potline may have a useful role; its investment life must be tested against changes in the process that produces the CO₂. The answer depends on the particular smelter, power supply, storage access and timing of alternative technology.

To make that diligence inspectable, I built a Verdox carbon-capture workbook that separates company claims, peer-reviewed cell results, engineering assumptions and complete-system cost scenarios. It follows the gas from additional flue-gas cleaning to a compressed CO₂ flange, then tests illustrative routes to geological storage. It uses Flyvberg reference class forecasting against the class to determine the likely costs. The assumptions can be changed; the central case is a screening estimate rather than a Verdox quote or a project forecast. Paid readers get the findings, key assumptions and questions for their own diligence below. Access to the working workbook can be arranged separately for a live investment or project decision.

Behind the paywall: the modeled cost through storage and the assumptions that can overturn it; the plant needed to process a one-percent stream; and whether a cleaner biogenic CO₂ stream or a changed aluminum process is the better use of capital. The evidence that would change the investment case is more revealing than any single energy claim.

Read more

The argument for bringing liquefied natural gas to Oʻahu is that it would replace oil for a while, lower bills, and give the island time to reach a renewable electricity system. The last step matters. Hawaiʻi law requires electric utilities to reach 100% renewable electricity generation by 2045. LNG cannot be the fuel at the end of that bridge.

The Hawaiʻi State Energy Office’s fuel study presents LNG as a possible near-term bridge. It points to imported green ammonia and hydrogen among the technologies that might eventually replace fossil gas, alongside local biodiesel production. The law requires renewable electricity; it does not require ammonia. Yet if a new gas plant and import system depend on a later switch to green ammonia, the price of that switch belongs in the decision made today. JERA’s proposed Oʻahu LNG plant remains subject to regulatory approval, and no green ammonia supply contract or conversion project has been approved.

That far shore looks costly. The source and calculation transparent Hawaiʻi electricity cost spreadsheet I developed estimates about 21 cents per kilowatt-hour leaving a new LNG plant, assuming high utilization. Its green ammonia case costs about 95 cents on the same power-plant basis. Renewable electricity must first be used to make hydrogen abroad; the hydrogen is turned into ammonia for shipping, then delivered to Oʻahu, turned back into hydrogen, and used to generate electricity. Each conversion consumes energy and adds equipment and operating costs. The figures are modeled scenarios, not bids or future retail rates, but the gap is too large to dismiss as a small detail of a fuel transition.

The idea that green ammonia would replace LNG as a shippable fuel depended on green hydrogen being cheap. That was never going to happen, the projects that reached final investment decision in the past three years have made that clear, and now realistic projections for 2040 and 2050 are still above $5 per kg, nowhere near the $1 per kg that so many hydrogen spreadsheets were using a few years ago.

Households can see what that risk means. For an Oʻahu detached home using 700 kilowatt-hours a month, the workup estimates that a 6-kilowatt rooftop solar system, a 10-kilowatt-hour battery, and daytime water-heating control would cover about three quarters of annual use. The modeled installation costs about $29,000, or $24,000 if the household qualifies for the assumed $5,000 state tax credit. At the September 2026 rates used in the model, its utility bill falls from about $322 to $63 a month. Including the annualized equipment cost, upkeep, and remaining grid purchases, its total electricity cost is about $241 a month. The simple cash payback is 8.4 years. Roof conditions, financing, actual usage, and tax-credit eligibility will change the result.

Now suppose the modeled ammonia generation replaced the same amount of oil generation and its entire additional cost flowed through Oʻahu electricity sales. The example home’s grid-only bill would rise to about $627 a month. Under that scenario, the solar and battery system’s simple payback falls to about 4.4 years. It makes the exposure clear: a family that can produce and shift much of its own electricity has substantial protection against an expensive imported-fuel pivot. A family that rents or cannot afford the installation deserves access to shared solar, storage, and bill savings rather than being left to bear that risk alone.

And note the “high utilization” caveat related to LNG and green ammonia electricity prices. The more families and businesses that put solar and batteries on their properties and the more that timeshift water heating and EV charging to the daytime, the lower the utilization of the generation plants. Their high capital costs get spread over a lot fewer kWh and so the cost per kWh has to rise. This might be fine if an alternative didn’t exist. But it does.

In the TFIE Strategy Briefing whitepaper The Clean Energy Future Hawaiʻi Can Actually Build, I set out a pathway built around rooftop and parking-canopy solar, batteries, managed vehicle charging, daytime water heating, efficient cooling, suitable wind, grid upgrades, and a limited locally sourced biomethane reserve for difficult periods. Commercial roofs and parking areas can produce electricity near businesses; flexible cooling and charging can use more of it when the sun shines. Industry can electrify suitable equipment and avoid paying to turn renewable electricity into imported fuel and back again. Those applications require their own site-level economics and grid planning, but they deserve to be compared with the full LNG-to-ammonia sequence, not merely with today’s oil bill.

The State Energy Office acknowledges that local renewables remain important, but its alternative-fuels study did not examine their buildout in comparable detail. Before committing Hawaiʻi to an LNG bridge, the state should cost the proposed exit from it and the local route alongside it. For households able to install solar and batteries now, the model already shows a payback at current prices. An ammonia-powered future would make waiting much more expensive.

Email me to engage me for consultation about a more realistic Hawai’ian energy future. Subscribe to TFIE Strategy Briefing for costed alternatives to Hawaiʻi’s imported-fuel future.

Subscribe now

A recent New York Times newsletter says fatal crashes involving e-bikes and electric scooters in New York City have reached a record high this year. “Record” is a compelling headline word, especially when the vehicles are relatively new. But it directs attention to what the victim was riding and away from what else was involved. The most recent complete-year NYC DOT crash report records 17 people killed while riding e-bikes in 2025. Fifteen died in crashes involving a motor vehicle. Of five people killed while riding stand-up scooters, three died in motor vehicle crashes. Those numbers do not assign fault to a driver in every case. They do mean that “e-bike deaths” and “scooter deaths” are often also deaths in collisions with much heavier vehicles.

For pedestrians, the imbalance is starker. The city recorded 112 pedestrian deaths in 2025, 105 in crashes involving motor vehicles. A comptroller analysis of 2020–23 found motor vehicles involved in 434 of 449 pedestrian deaths, while e-bikes and stand-up scooters were involved in six. Scooters on sidewalks and reckless e-bike riding are real problems, including injuries that a fatality count misses. But a run of stories that makes electric two-wheelers the face of pedestrian danger gives readers a distorted view of what kills people on New York’s streets.

Small annual totals make the rhetoric of records particularly volatile. One more stand-up scooter rider death on top of the five in 2025 would raise that count by 20 percent. At the same time, electric Citi Bike trips exceeded 29 million in 2024, before counting any privately owned e-bikes or scooters. More riding can produce a higher death count without proving that a given ride has become more dangerous. The city should investigate every death and the fast machines involved in many of them; readers should also be told that millions of ordinary journeys sit behind these small fatality totals.

Some of those machines are plainly a safety concern. In an August 2026 enforcement announcement, the city said that 45 e-bike rider deaths from 2017 through 2025, 54 percent of the total, involved devices capable of exceeding 25 mph. Fourteen stand-up scooter rider deaths, 52 percent, involved devices capable of exceeding 20 mph. Capability is not proof of the speed at impact or the cause of each crash, but the prevalence of prohibited fast devices makes enforcement against their sellers and dangerous operation a sensible response. It does not remove the need to protect riders of legal devices from cars and trucks.

In New York’s East Bronx scooter pilot, riders named cars, car-share and ride-hail among the trips they would otherwise have taken. A 48-city travel-diary study estimated a median reduction of about 0.15 car miles per shared scooter trip and 0.25 per bike-share trip, including e-bikes. Even a fraction of a car mile matters when trips number in the millions. Some replace walking or transit, so it would be wrong to credit every electric ride with removing a car. There is nonetheless good evidence that these services displace some driving.

A car trip replaced by an e-bike or scooter no longer burns gasoline or produces tailpipe emissions along that route; the electric trip uses far less energy. New York’s Department of Environmental Protection estimates that motor vehicles account for about 28 percent of local nitrogen oxide emissions and 11 percent of local fine particulate emissions. Traffic pollution is a health issue as well as a carbon issue: the city health department estimates that traffic-related fine particles contribute to roughly 320 premature deaths annually. Electric bikes will not erase that burden, and shared scooters have manufacturing and fleet-service impacts, but replacing combustion-vehicle miles means lower emissions and cleaner air for the people breathing beside those streets.

The same substitution changes what a street sounds like. An electric bike or stand-up scooter has no idling engine or exhaust note, and a replaced car or gasoline moped is one less source of engine noise. New York’s noise rules address the health and quality-of-life effects of loud engines and mufflers. Tires, horns and the rest of traffic do not vanish because someone chooses an e-bike; the benefit comes when enough motor vehicle trips are actually avoided. It is still a benefit rarely mentioned in an account of what these devices cost the city.

They also buy time and support work. A study following Citi Bike’s launch found lower bus ridership on routes exposed to bike share than on comparison routes. It predates the electric fleet and does not measure today’s busiest-hour crowding, but it shows that a bike can substitute for a bus when that is the faster or more convenient choice. For delivery workers, the device is a means of earning a living. An NYU study of New York couriers found certified e-bikes competitive on five-year ownership costs and documented the importance of charging and battery access to workers’ choices. In freight delivery, the city’s microhub pilot replaced more than 3,000 truck trips in its first year by transferring goods to handcarts, cargo bikes and small electric vehicles. Cargo bikes did only part of that work, but the result shows why smaller vehicles have economic value in a dense city: they can deliver goods without sending a truck down every block.

“It bleeds, it leads” is a poor accounting system for transport. Fatal crashes deserve investigation, and dangerous riders, illegal fast devices and sidewalk conflicts deserve a response. The next record-high headline should also name the motor vehicles involved, distinguish riders from pedestrians, and put the fatalities beside the scale and value of electric travel. New York can reduce the harms from e-bikes and scooters while making streets safer from the vehicles implicated in most of its rider and pedestrian deaths. Cleaner air, quieter streets, less driving and useful work belong in that story too.

Subscribe to TFIE Strategy Briefing for more transport analysis.

On August 12, 2026, Heart Aerospace put a very large battery-powered airplane into the air from Plattsburgh International Airport in upstate New York. The X1 weighed more than 25,000 pounds, had a 106-foot wingspan and flew for 27 minutes. It climbed to 1,100 feet above ground level and put more than a megawatt through its electrical propulsion system while taxiing, taking off, climbing, maneuvering and landing. There were no hydrogen tanks, fuel cells or gas turbines aboard as propulsion insurance. For 27 minutes, batteries flew an aircraft roughly the size of the regional machines Heart ultimately wants airlines to operate.

That matters because aviation electrification has spent much of the past decade surrounded by distractions. Billions went into electric vertical takeoff and landing aircraft whose economics, certification burdens, infrastructure requirements and constrained payload-range combinations made them poor substitutes for existing transportation. My expectation has long been that aviation will electrify first through comparatively boring fixed-wing aircraft operating short, repetitive routes. My 2021 interview with Heart founder Anders Forslund reinforced that view. Heart’s original ES-19 was not an attempt to reinvent the airplane. It was a conventional-looking 19-seat regional aircraft with electric motors, lithium-ion batteries and a claimed 400-kilometre range.

The important complication is that Heart’s commercial programme has not actually been purely electric since 2022. The company replaced the ES-19 with the 30-seat ES-30 and added two turbogenerators, describing the configuration as reserve hybrid. The propellers remained electrically driven, but fuel-powered generators could keep the electrical system supplied after the battery reached its operational limit. Heart claimed 200 kilometres of battery-electric range, 400 kilometres at full passenger load using the hybrid system and as much as 800 kilometres with 25 passengers.

That first ES-30 was a genuine series hybrid. Fuel went into a turbine, the turbine drove a generator, electricity moved through the electrical system and an electric motor finally turned the propeller. There is an unavoidable efficiency penalty in doing that rather than connecting a turbine through a gearbox directly to the propeller. Generator, power-electronics and motor efficiencies can all be high, but they multiply rather than disappear.

Heart then changed direction. In 2024 it moved to what it called Independent Hybrid propulsion, with two battery-electric propellers and two separate conventional turboprops. At the aircraft level it was effectively an independent parallel hybrid. The turbines no longer generated electricity. They directly drove their own propellers, while a separate pair of electric motors supplied battery-electric propulsion. That solved one problem while making another highly visible: the aircraft had four nacelles, four propellers and two complete forms of propulsion. An electric flight carried unused turbines, gearboxes, propellers and fuel-system machinery, while a long hybrid flight carried an electrical propulsion system powerful enough to fly the aircraft independently. Heart had avoided the conversion losses and some of the integration problems of series hybridization by creating an aircraft with a lot of duplicated equipment.

The newly revealed ES-36 reverses that decision. Heart has returned to series hybridization, but now with only two nacelles and two propellers, both permanently driven by electric motors. Each nacelle contains an electric propulsion motor and its own turbogenerator system rather than the aircraft carrying another pair of mechanically driven thermal propellers. Heart says the redesign adds six passenger seats and 1,415 pounds of payload while retaining the preceding aircraft’s maximum takeoff weight and battery capacity, and reduces wingspan by about 11 feet.

The architecture has zigzagged because Heart has repeatedly changed where combustion sits in the propulsion chain. The 2022 ES-30 separated combustion from propeller drive by using generators, the 2024 Independent Hybrid put turbines mechanically back onto two propellers, and the ES-36 separates them again. In the current architecture, gas turbines no longer drive propellers mechanically. Their function is to make electricity when the mission requires more energy than Heart considers practical to carry in batteries.

That leaves the interesting contradiction at the centre of the aircraft. X1 has demonstrated that batteries can provide enough power to fly a regional-aircraft-scale machine, yet Heart still considers turbines necessary for a commercially flexible airplane beyond about 200 kilometres. The explanation is not simply that batteries are “too heavy.” It lies in a distinction that gets blurred constantly in electric-aviation discussions: the amount of power an aircraft needs at one moment is not the same thing as the amount of energy it needs over the whole flight. How Heart exploits that distinction determines what the ES-36 actually is at 200, 600 or 1,200 kilometres—and how much better batteries could change the answer.

The interesting question, then, isn’t whether the ES-36 is “really electric.” It’s how quickly its fuel-burning portion shrinks as batteries improve, which missions actually benefit most, and whether the hybrid machinery still earns its place once the comparison is made against an efficient turboprop rather than an easy-to-beat regional jet.

Read more

A floating data centre that makes its own electricity, cools itself with the ocean, produces drinking water for a coastal community and can sail away when a hurricane approaches is almost perfectly designed for the current AI infrastructure hype cycle. Optimal Transit says its Kraaken platform can deliver 100 MW of continuous power, reserve 60 MW for AI computing, send as much as 40 MW ashore and produce about 30 million litres of freshwater every day, all from a vessel only about 91 metres long. None of the constituent ideas is imaginary. Ocean thermal energy conversion (OTEC) is real, ammonia Rankine cycles are real, liquid-cooled computers reject useful quantities of low-temperature heat, and offshore vessels, desalination plants, subsea pipelines and disconnectable mooring systems all exist. That is precisely why Kraaken deserves more scrutiny than a proposal built around an obviously impossible technology.

A large part of my work is taking energy and infrastructure propositions apart into their denominators: mass flow, energy flow, civil works, operating constraints and the equipment that has to connect one to the next. Kraaken turned out to be unusually interesting because almost every attractive feature depends on the interfaces among several real technologies rather than on any one of them. The first question is where the electricity comes from. Optimal Transit’s answer is a proprietary development of OTEC it calls Digital Ocean Thermal, or DOT. Conventional OTEC exploits the small temperature difference between tropical surface water and deep ocean water. As I laid out in my recent assessment of OTEC itself, the underlying thermodynamics work; the difficulty is the denominator. With perhaps 25°C water at the surface and roughly 5°C water at depth, enormous quantities of water and heat have to move through the plant to produce comparatively modest quantities of electricity.

Optimal Transit says DOT changes that equation dramatically. Its own published process diagram says the technology reduces the cold-water pipe by 70%, reduces the cold-water pumps and condenser by 70%, and allows a turbine about 70% smaller. It also routes data-centre waste heat through something it calls the AHEB, likely for ammonia heat exchange booster, to “supercharge” the ammonia vapour and says that higher vapour pressure is part of the reason the turbine can shrink. Those claims cannot really be considered independently. If less cold water flows through the plant, less heat can be carried away unless the water temperature rise changes; if the same flow passes through a much smaller pipe, velocity and pressure loss increase; if the condenser really has 70% less heat-transfer area while rejecting the same heat, something else in the heat-transfer equation has to increase substantially; and if the turbine becomes much smaller because its inlet conditions have changed, the source of the additional pressure and enthalpy has to appear somewhere in the cycle balance.

That raises some deceptively simple questions. What does “70% smaller cold-water pipe” mean: diameter, cross-sectional area, an individual pipe in a multi-riser bundle, flow capacity, mass or cost? If the answer is multiple smaller risers, that can make individual pipes easier to manufacture, but it does not make condenser heat duty or total seawater flow disappear. The same ambiguity exists around the condenser and pumps. Are they 70% smaller in physical volume, heat-transfer area, installed power, equipment count or cost? Those are very different propositions. A 70% reduction in physical condenser area at unchanged heat duty, for example, requires more than three times the product of heat-transfer coefficient and usable temperature difference.

Then there is the AHEB. A liquid pump can establish a higher pressure in a Rankine cycle, while heat exchangers supply the enthalpy needed to heat and vaporize the working fluid. But the temperature of the heat source limits the pressure at which that vaporization can occur. Twenty-five-degree ocean water cannot boil ammonia at a saturation temperature approaching 45°C, and server heat at 45°C is a finite low-grade heat stream, not a new primary energy source. I also gave Kraaken a more generous test: suppose none of the data centre’s electricity comes from the OTEC plant. Supply all 100 MW externally and give DOT the entire resulting 100 MW of 45°C server waste heat. That removes the feedback-loop objection entirely and leaves a cleaner question: how much useful work can that heat possibly contain against a 5°C sink, and how much heat still has to be rejected afterward?

The physical scale raises another set of questions. Optimal Transit describes a vessel roughly 91 metres long and “50,000 long tons,” although the public material does not identify whether that tonnage means displacement, deadweight or something else. Published OTEC reference platforms of 50–100 MW are roughly 198–285 metres long, while Kraaken additionally has to contain a large AI installation, electrical systems, desalination, propulsion and whatever machinery enables storm departure. And storm departure itself is not the same as unplugging a laptop. If one enormous cold-water pipe becomes a dozen or more kilometre-class risers, what remains suspended when the ship leaves? What happens to the moorings, manifold, shore power connection, freshwater line and fibre? Currents can vary in both speed and direction with depth, so the vessel and its riser field do not necessarily see the same environmental forces, and a small net stationkeeping force can hide much larger opposing loads distributed along the submerged structure.

The proposition may contain genuinely clever engineering. But by the time the public claims are followed through the water flows, heat rejection, ammonia cycle, platform dimensions and offshore interfaces, the central diligence question is no longer whether the technologies are individually real. It is whether the complete system closes.

I rebuilt the public Kraaken proposition from the bottom up: seawater flow, pipe hydraulics, condenser duty, ammonia state points, waste-heat exergy, riser arrays, stationkeeping loads, platform dimensions and shore interfaces. Several claims remain possible in principle. Others require improvements much larger than the public explanation accounts for. Below are the calculations, the published OTEC reference designs, Optimal Transit’s own claims and the specific engineering numbers that would have to exist for Kraaken’s 100 MW proposition to close.

Read more

Replacing a boiler and comprehensively renovating a house are different projects, with different budgets and timetables. Ireland’s National Heat Study, published by the Sustainable Energy Authority of Ireland in February 2022, examined household decisions about both. Its conclusion was explicit: support focused on replacing fossil heating could deliver more immediate emissions reductions than a fabric-first approach. Fabric improvements remained important, but the study proposed concentrating on those needed to support low-carbon heating rather than making extensive renovation a prerequisite. Heat pumps featured prominently in every decarbonisation scenario.

That distinction is the centre of my Fabric First Trap argument. Building fabric means the walls, roof, floor, windows and other elements separating indoors from outdoors. Insulation and draught-proofing can reduce heat loss, but reducing heat demand is not the same intervention as changing the source of heat. The sequencing problem arises when extensive renovation consumes the money or time available for replacing fossil heating. It does not arise merely because someone insulates a roof or repairs a draughty door.

SEAI approached the issue through simulation, not a trial comparing installation sequences in otherwise identical homes. Its modelling represented more than 680 heat-demand archetypes across residential, services and industrial sectors, combining physical characteristics with costs and consumer behaviour. Even with modelled grants covering 60% to 80% of capital costs, expensive fabric measures attracted limited uptake. That did not mean households rejected insulation: in the Balanced scenario, 83% of the residential building stock received some form of heat-demand reduction measure. Projected uptake concentrated on inexpensive, less disruptive work rather than extensive renovation everywhere.

The spread in modelled paybacks explains much of that result. For detached, oil-heated dwellings under its 2030 assumptions, Figure 7 of the National Heat Study gives average consumer paybacks of seven years for draught-proofing, nine for cavity-wall insulation and fourteen for roof insulation. Solid-wall insulation reaches 28 years, high-efficiency glazing 40 and floor insulation 229. SEAI’s calculations also assign a notional value to additional comfort. The useful finding is the enormous variation between measures under the model’s assumptions, not a claim that every floor-insulation project takes two centuries to repay its cost.

SEAI also adjusts energy savings for comfort-taking. People who previously rationed heating may use some of an improvement to obtain warmer rooms rather than taking the entire benefit as reduced fuel consumption. The report recognises that as a health and wellbeing benefit while keeping it separate from energy and emissions savings. A warmer home can be a successful outcome without delivering all the carbon savings calculated for an unchanged indoor temperature.

The measured retrofit studies in my review resist a single explanation too. The Michigan Weatherization Assistance Program evaluation by Meredith Fowlie, Michael Greenstone and Catherine Wolfram found savings below engineering projections without finding significantly higher indoor temperatures in weatherised homes. The evaluation of Wales’s Arbed programme, meanwhile, reported warmer homes alongside a 37% reduction in average daily gas use in its monitored sample. The implications differ: model overprediction cannot always be blamed on occupants seeking more warmth, and warmer homes can coexist with substantial energy savings. Neither study directly compares heat-pump-first with insulation-first installation.

There is another economic effect in SEAI’s analysis. Where a well-installed heat pump supplies useful heat more cheaply than the oil boiler it replaces, avoiding another unit of heat demand saves less money. Additional fabric measures tend to have longer paybacks, all else equal. That is conditional on energy prices and actual system performance, not a promise that every heat pump lowers bills. Nor does it settle sequencing by itself: the comparison must also account for installation costs, any reduction in required equipment capacity, planned building work and the consequences of delaying replacement.

Nigel Banks of Octopus Energy approaches the same question through practical delivery. In his (Deep) Fabric Fifth update, he explains that heat pumps, smart operation and inexpensive measures ideally form a coordinated package. His point is not that insulation must literally be the fifth task. It is that urgent boiler replacement need not wait for wall insulation, while repairs, ventilation and suitable basic measures remain part of the plan. Solar and storage also feature in his framework, although his UK tariff and cost examples require separate assessment before being applied in Ireland.

Our published conversation about heat pumps and fabric-first sequencing brought those practitioner concerns together with my review. What is striking here is the consilience of evidence: measured retrofit outcomes, practitioner experience and national energy-system modelling use different methods and data, yet converge on the same distinction. Fabric should enable low-carbon heat, not automatically stand in front of it.

The distinction is not exclusive to critics of fabric first. Alan Clarke and Kate de Selincourt’s Passivhaus Trust paper on the timing of heat pumps in retrofit examines when to install a heat pump within a staged retrofit while retaining the importance of fabric and ventilation for health and comfort. Lower heat loss can also reduce required heating capacity and winter electricity demand. SEAI’s own study leaves extreme-weather system performance for further investigation, so its household payback figures cannot settle the full system value of fabric improvements during a prolonged cold spell.

There is now an important development beyond the 2022 report. SEAI’s July 2026 heat-pump guidance removes the fixed heat-loss threshold for heat-pump grants under the One Stop Shop and Community Energy programmes, with flexibility under the Warmer Homes Scheme where its team advises it. Instead, those routes use a designer- and installer-led approach. The heat-loss assessment remains; what has changed is whether one numerical threshold determines eligibility.

The distinction matters because Better Energy Homes retains the 2.3 W/(m²·K) heat-loss-indicator requirement before grant payment. SEAI explains that the programmes receiving additional flexibility place integrated whole-house design responsibility with the provider, while individual upgrades do not necessarily have that same coordination.

This is not a general removal of retrofit conditions. The affected programmes still consider heating, fabric and ventilation together, and the One Stop Shop programme retains its post-works BER requirement. The 2026 change is better described as a move away from one rigid heat-loss eligibility threshold in selected programmes than as SEAI abandoning fabric requirements.

The technical reasoning is interesting too. SEAI’s 2026 heat-pump bulletin draws on UK field measurements and interim Irish pilot findings to emphasise sizing, system design, installation, commissioning, controls and operation alongside building heat loss. Better fabric reduces the quantity of heat required. It does not, by itself, determine how efficiently a heat pump supplies that heat. A building assessment and a heating-system design answer different questions.

That gets much closer to the useful principle hiding underneath arguments about Fabric First and Fabric Fifth. A house should be repaired where it has defects. Cheap and effective fabric measures should be done where they make sense. Heat loss should be measured and the heating system properly designed. Deep retrofit can be entirely justified for comfort, health, peak demand, building renewal or economics in a particular property.

What should not be assumed is that every home must complete the expensive parts of that programme before it stops burning fossil fuel. SEAI’s own national modelling was making that distinction in 2022. Its policy guidance is now beginning, selectively, to reflect it.

If you value independent, evidence-led analysis of building heat, electrification and the economics of the energy transition, consider a paid subscription to TFIE Strategy Briefing.

Subscribe now

HyFlux has a technically interesting idea wrapped inside an investment proposition I think has a very low probability of reaching its stated aviation ambitions. The small UK company wants to exploit the extreme cold of liquid hydrogen already carried by a hydrogen aircraft to enable fully superconducting electric motors with much higher specific power than conventional machines. That is not pseudoscience: NASA Glenn Research Center’s High-Efficiency Megawatt Motor programme is developing a 1.4 MW partially superconducting machine with a 16 kW/kg target and 99% efficiency, while NASA’s newer cryogenic-machine work treats substantially higher specific power as a serious research objective rather than fantasy. HyFlux’s proposition therefore deserves technical diligence rather than dismissal.

The problem is that HyFlux is not asking investors merely to back an interesting motor; it describes its propulsion approach as the “only feasible solution” for zero-emission aircraft propulsion while its own recent investor communications say the company is moving toward multi-megawatt scale and beginning conversations with suitable investors. My conclusion is much more adverse: HyFlux may eventually build an excellent superconducting motor and still fail to create an important aviation business.

I first assessed hydrogen passenger aviation as a complete system in 2023 and rebuilt that assessment in September 2026 after another three years of airframe design, certification activity and airport-infrastructure research. Some of my earlier objections did not survive better evidence, which is useful because the remaining case is stronger without them. Regulators are developing certification pathways, cryogenic tanks have more viable geometries than simplistic spherical-tank arguments implied, and there is no serious basis for saying that hydrogen passenger aircraft are inherently incapable of flying or being certified. The real question is commercial: can a safety-equipped aircraft preserve a useful passenger, baggage, range and reserve mission after the hydrogen installation is included, move through certification and repeat production, and meet a sufficiently developed hydrogen network at enough airports to sustain scheduled operations? When I put those gates together in a judgmental model and deliberately give several of them generous assumptions, the resulting pathway remains extremely weak, because the aircraft, fuel infrastructure and fleet all have to mature on compatible timelines rather than merely proving that each can exist independently.

HyFlux’s value depends heavily on that larger pathway succeeding, which makes the company unusual from an investment perspective. A startup developing an excellent conventional inverter, controller or electric motor could potentially sell into several expanding electrification markets even if one airframe programme disappeared; HyFlux’s aviation proposition becomes most valuable in a world where commercial liquid-hydrogen aircraft already exist in meaningful numbers. The company is therefore making a second-order bet, not simply that superconductivity works but that an entire fuel and aircraft architecture with weak current commercial prospects becomes important enough to need its particular version of superconducting propulsion. The public story makes that proposition sound deceptively compact—use a very cold fuel to keep a motor superconducting, obtain exceptional power density, then scale toward multi-megawatt flight—but an investor should immediately ask what exactly is being scaled, what support systems belong inside the kilograms used for the headline power-density claim, how an airline handles a propulsion system that prefers to remain tens of degrees above absolute zero, and what has actually been demonstrated between the earlier 100 kW research stage and the multi-megawatt architecture HyFlux now discusses.

Those questions matter more because the alternatives are not waiting for hydrogen to become ready. My aviation pathway through 2100 does not assume that today’s aircraft and fuel system remain frozen while hydrogen improves: battery-electric aircraft move upward from smaller payloads and ranges, hybrids extend that reach, conventional airframes can consume increasingly expensive sustainable liquid fuels without replacing their fundamental architecture, rail removes some journeys where it works well, and higher fares alter discretionary demand. Hydrogen aviation therefore has to become preferable to an evolving portfolio of alternatives, not merely preferable to fossil kerosene under a sufficiently high carbon price or more electricity-efficient than synthetic e-kerosene. That broader context is why I find HyFlux unlikely to achieve its aviation ambitions even while taking its motor research seriously; too much of the commercial case depends on several conditions outside the company’s control becoming favourable together, while the public record still describes a company working through component and subsystem development rather than one scaling a mature propulsion product.

Below the paywall I start with the motor rather than the company history: what “fully superconducting” actually means, why four coupled technical integrations sit inside a fifth aviation constraint, what HyFlux’s >20 kW/kg claim tells an investor and what it leaves out, how a superconductor can quench and why thermal state becomes an operating and maintenance issue. I then put even a technically successful motor back inside the liquid-hydrogen aircraft and airport system, compare that market with the alternatives developing around it, assess whether HyFlux has the organizational and financial scale to execute the programme, and only then trace the more complicated 100 kW, corporate and IP lineage. The provenance is interesting; the technology and market risks are much bigger.

Read more

Ocean thermal energy conversion, usually shortened to OTEC, is one of those energy technologies that never entirely goes away. The proposition is appealing. Tropical oceans have warm water at the surface, typically around 25°C or more, while water roughly a kilometer down can be around 4°C to 5°C. Put a heat engine between the two reservoirs, vaporize a working fluid such as ammonia with the warm water, run the vapour through a turbine, condense it with the cold water and repeat. The fuel is free, the temperature difference is available day and night, and the underlying thermodynamics are entirely legitimate. The basic process generally needs a temperature difference of roughly 20°C or more to become useful.

The interesting question is not whether OTEC works. It does. The question is why a technology first demonstrated as a net electricity source almost half a century ago still has no utility-scale commercial fleet. Makai Ocean Engineering has spent decades developing OTEC and deep-water systems, and its Hawaiian research installation has operated with a 105 kW turbine-generator. Yet Makai still describes a multi-megawatt offshore demonstration operating for several years as an important step before very large commercial systems become readily financeable. That is a revealing deployment record for a technology whose fundamental physics have been understood for generations.

The answer starts with the quality of the energy resource. A heat engine operating between 25°C water and 5°C water has a theoretical Carnot ceiling of only about 6.7% between those seawater temperatures. Real equipment has to maintain temperature differences across heat exchangers, operate turbines and generators, and run pumps, so practical cycle efficiencies are only a few percent. A detailed 100 MW net OTEC design study for Indonesia places conventional Rankine-cycle efficiency in the 3% to 5% range and assumes substantial internal electricity consumption. Low efficiency by itself is not the indictment. The heat differential in the ocean is free. The problem is what low efficiency means physically: a modest amount of useful electricity requires processing several gigawatts of heat and moving enormous quantities of seawater.

That Indonesian design makes the denominator tangible. To deliver 100 MW net, it requires 235 cubic metres of cold deep water every second and another 470 cubic metres of warm surface water every second. Together that is 705 m³/s, or 705,000 liters every second. An Olympic-size competition pool holds about 2.5 million liters, so the plant circulates the equivalent of one Olympic pool every 3.5 seconds, about 17 every minute and more than 24,000 every day. The pool comparison uses the 2.5-million-liter competition pool at Montreal’s Olympic complex as the reference.

The plant is not consuming all of that water. It returns most of it to the ocean after passing through the system. But “not consumed” is very different from “does not require infrastructure.” Every liter has to enter through an intake, pass through pipes and heat exchangers with low enough pressure losses that pumping does not consume the plant’s output, and then be discharged again. On an electricity basis, the reference design moves roughly 25 cubic metres, or about 25 tonnes, of seawater for every net kilowatt-hour it produces. That is a much more useful denominator than the total amount of thermal energy stored in tropical oceans.

The cold-water intake is where the abstraction becomes civil infrastructure. An IRENA engineering review put a 100 MW-class cold-water pipe at roughly 10 m in diameter and about a kilometer long, while Makai gives a similar scale for a future 100 MW plant. A 10 m diameter is an enormous pipe, extending downward for roughly a kilometer and remaining connected to a plant that must continuously move hundreds of tonnes of cold seawater every second.

The heat exchangers are similarly dominated by scale because several gigawatts of thermal energy must cross them to produce only 100 MW of net electricity. Makai describes the heat-exchanger installation for a 100 MW plant as roughly 6 m high and occupying about 930 m². Better working fluids, heat-exchanger designs and materials can reduce costs and parasitic loads, but they do not alter the basic requirement to move enormous quantities of heat across a temperature difference of only about 20°C.

The kilometer depth itself is not equivalent to pumping water vertically one kilometer from a mine. Hydrostatic pressure inside and outside the intake largely balances. Pumping loads come mainly from friction, screens, bends, heat exchangers and other pressure losses. But when hundreds of cubic metres per second are moving continuously, even small increases in pressure loss translate into large electrical loads. Large pipes and relatively low velocities are necessary precisely because parasitic pumping can otherwise consume a material fraction of gross generation.

Putting OTEC offshore trades one form of civil infrastructure for another rather than eliminating it. A vertical intake can be shorter than running a pipe from shore down a sloping seabed, but then a kilometer-scale riser hangs beneath a floating structure and has to tolerate currents, vessel motion, corrosion, fatigue, storms and stationkeeping loads. In that Indonesian 100 MW design, accommodating the machinery and seawater systems drove the platform toward Suezmax tanker scale— roughly 275 m long and 48 m wide, nearly three soccer pitches from bow to stern. Offshore engineering can certainly build systems at that scale, but the ability to engineer complex subsea infrastructure does not make it cheap.

The deployment history is the other hard denominator. Mini-OTEC in Hawaii produced net electricity in 1979. Later demonstrations in Hawaii, Japan and elsewhere proved additional components and integrated systems, and Makai’s 105 kW plant became the first closed-cycle OTEC system to send electricity to a US utility grid. Those are legitimate engineering achievements. What has not followed is the pattern seen with technologies that found strong economic niches: increasingly large commercial plants, repeated orders, standardized equipment, competitive suppliers and a growing base of operating data. Solar, wind and batteries have moved into deployment measured in hundreds of gigawatts per year. Utility-scale commercial OTEC deployment remains absent.

There is another denominator that matters just as much as seawater flow. The relevant resource is not the total area of tropical ocean with a 20°C temperature gradient. It is the much smaller subset of coastlines where kilometer-deep water lies close enough to substantial electricity, cooling or water demand to justify the intake, export and shore infrastructure. A vast theoretical resource can shrink quickly once distance to depth, grid connection, landfall, maintenance access and actual customers are included.

OTEC still has a plausible niche. The strongest case is a steep-sided tropical island where deep cold water lies close to shore, land is constrained, electricity is expensive because imported fuels dominate, and several valuable services might share deep-water infrastructure. The thermal gradient is available day and night, which is a genuine advantage over variable generation considered in isolation. If commercial OTEC is going to establish itself anywhere, those unusually favourable locations are the obvious places to do it.

Even in Hawaiʻi, which is close to a textbook best-case geography for OTEC, the competition is formidable. My 2026 assessment, The Clean Energy Future Hawaiʻi Can Actually Build, found that a fully electrified civilian Oʻahu economy would require roughly 6,000 GWh of electricity a year, while the island’s screened solar resource is large enough to exceed that demand substantially, with batteries, flexible demand, modest wind and small amounts of firm reserve addressing the timing and reliability problem. And Hawaiʻi Island already has firm geothermal generation at Puna, further narrowing OTEC’s prospective niche there. OTEC therefore is not competing with imported diesel in isolation. In one of the places where its physical resource and demand match is most attractive, it has to beat a portfolio of cheap modular solar, batteries, demand flexibility, selective wind, direct seawater cooling and, on the Big Island, geothermal—technologies that are already commercial and can be deployed incrementally rather than as a single enormous piece of marine infrastructure.

But the coproduct argument needs the same denominator discipline as the electricity claim. Cold deep water is genuinely valuable for cooling. If cooling is the service required, using cold water directly avoids first converting a few percent of a small thermal gradient into electricity. In 2014, when I assessed the broader family of ocean-energy technologies, I concluded that deep-water cooling was probably the most useful ocean-energy application.

Freshwater can also be a useful coproduct in some OTEC configurations, especially on water-stressed islands, but desalination does not inherently require OTEC. Reverse osmosis is already a mature technology that produces freshwater using electricity without first requiring a kilometer-deep intake and a thermal power cycle. Cooling, electricity and freshwater therefore need to be tested against their simplest mature alternatives before integration is assumed to create value. Shared infrastructure is useful only when the savings from sharing it exceed the additional complexity.

The strongest OTEC proposition I can imagine therefore remains a steep tropical island with very deep water immediately offshore, unusually expensive electricity, significant cooling or desalination demand, constrained land and favourable financing. The empirical test is straightforward. Build a multi-megawatt plant there, meter the net electricity after every pump and auxiliary load, operate it for several years, publish the complete installed cost, maintenance record and availability, then build another one commercially without heroic bespoke engineering or subsidies. That would materially change the evidence.

Until then, OTEC belongs in the category of technologies that are physically real and potentially useful in narrow circumstances without being serious candidates for broad energy-system deployment. The oceans contain an extraordinary amount of thermal energy, but that is the wrong denominator. The relevant denominator is how much seawater, machinery, heat-exchanger area and deep-ocean infrastructure are required for every useful MWh delivered.

The ocean has plenty of thermal energy. The denominator is why OTEC remains a niche.

OTEC is real engineering, but that does not make it a major energy solution. Subscribe to TFIE Strategy Briefing for more denominator-first assessments of which technologies scale, which stay niche, and why.

Subscribe now

The Port of Tilbury put a hydrogen fuel-cell reach stacker into service this month. Hyster and Tilbury describe the machine as a pre-production trial, with 32 kilograms of hydrogen onboard and enough stated endurance for a 12-hour shift. I assumed it would join a small but recognizable population of hydrogen reach-stacker trials around the world, perhaps high single digits and conceivably low teens once Chinese prototypes and poorly publicized demonstrations were included. That expectation turned out to be generous. After following actual machines rather than announcements, I could substantiate only the earlier Hyster prototype at Valencia and the new Tilbury machine as hydrogen reach stackers that have clearly performed real terminal work. Valencia’s pilot concluded in 2025; Tilbury’s has just begun. Hyster explicitly describes Tilbury as building on the completed Valencia pilot.

The more consequential finding was on the battery side. My assembled estimate from manufacturer disclosures, production data, named fleets and orders puts the global battery-electric reach-stacker population at roughly 1,500 machines, with something like 1,200 to 1,800 a reasonable uncertainty range. This is not a census and I would not pretend otherwise, but several independent pieces fit together well. Work led by Sahar Rashidbeigi at APM Terminals and carried into the Zero Emission Port Alliance put the worldwide reach-stacker population at about 9,300 machines. A commercial market estimate puts 2025 electric reach-stacker production at about 460 units, while the same research publisher estimates total 2025 port reach-stacker production at about 2,460 units. Those two estimates imply electric machines were already approaching one-fifth of current production. Against a stock of roughly 9,300, an electric installed base somewhere around 1,500, or roughly 15% to 16%, no longer looks aggressive. It looks like what might be expected when fleet turnover meets improving economics.

Reach stackers are a good machine for exposing bad intuition about electrification. A large one can weigh 70 tonnes or more before it picks anything up, then grab a loaded ISO container weighing around 40 to 45 tonnes with a telescoping boom, move it around a yard and stack it several containers high and more than one row deep. The enormous ship-to-shore cranes get the photographs, but reach stackers do much of the flexible work once containers are on land: repositioning boxes, dealing with exceptions, feeding trucks and trains and working in terminal areas where fixed gantries are inappropriate. In my earlier work on port-equipment electrification and later in From Quay To Sea, the recurring point was that ports have to be understood as logistics and energy systems, not collections of isolated engines. Reach stackers make the argument almost embarrassingly concrete. Looking at a 70-tonne machine lifting a 40-tonne box encourages the thought that it must require a high-energy-density fuel. Looking at what the machine actually does leads somewhere else entirely.

It travels slowly on flat, paved surfaces over short distances inside a bounded industrial site. It needs very high peak power to accelerate, lift and operate hydraulics, but much less average energy than its size suggests. It brakes frequently, lowers heavy loads from which some potential energy can be recovered, waits while containers and vehicles are positioned, and repeatedly returns to predictable parts of the terminal. The machine does not need enough stored energy to cross a continent. It needs enough energy to get through the next natural charging opportunity. That distinction between power and energy is crucial, and real operating data show just how large it is. New Zealand freight operator Reliance Transport reports that its 74-tonne SANY electric reach stacker consumes only about 33 kWh per operating hour while routinely getting through full eight-hour shifts and handling around 100 container moves a day during busy periods. A machine capable of enormous instantaneous loads is averaging energy consumption equivalent to only about 33 kW over the working hour.

The battery-electric fleet looked surprisingly large until I checked the economics. Then the surprise mostly disappeared. What operators are buying is not merely a lower-carbon reach stacker, but increasingly a lower-cost one.

Read more

When Norwegian transit authority Ruter sent a Chinese Yutong electric bus into an isolated mine for cybersecurity testing, the investigation was responding to two very different fears. One was explicitly geopolitical: could a Chinese-built bus passing through sensitive locations such as Lutvann, headquarters of Norway’s intelligence service, function as a rolling visual-surveillance platform? The other was technological: could a manufacturer with remote diagnostic and software-update access reach systems important enough to disable the bus? The Lion Cage material framed both “Lutvann” and “Kill Switch” as scenarios to investigate. The visual-surveillance scenario was not supported by what the investigators found. The remote-access scenario was.

That distinction largely vanished as the story travelled. The Lutvann concern had attracted political attention because Yutong buses had entered a military facility associated with Norway’s intelligence services, raising obvious questions about the cameras and sensors increasingly built into modern vehicles. But Lion Cage found that the Yutong’s exterior camera system was isolated from the manufacturer’s online systems and concluded that the Lutvann scenario, as it related to images and video, was not possible with the bus. The interior surveillance system had its own operator-side connectivity rather than a covert pathway to Yutong. The most explicitly espionage-focused part of the investigation therefore produced a reassuring result: the bus was not sending the exterior imagery that had made the Lutvann story alarming.

The second finding was real and important. The Yutong was a modern connected vehicle with over-the-air software updating and remote diagnostics, and Ruter found an internet-connected path through the manufacturer’s systems into operationally important parts of the bus, including the battery and power-supply control environment. Sufficiently privileged access could therefore potentially interfere with operation and render the vehicle unusable. My reaction when I first encountered that result, however, was much less dramatic than the headlines: of course it could. A manufacturer that can diagnose a vehicle remotely and send software updates to it necessarily has a communications path from outside the vehicle to systems inside it. Cars, trucks and buses have been becoming networked computers for years. The meaningful security questions are what that path can reach, what it can change, how it is authenticated and logged, and who can sever it. None of those questions is inherently Chinese.

Read more

A forecast that survives the failure of its predicted mechanism without losing much confidence is no longer functioning very well as a forecast. That is increasingly how I see a recognizable intellectual lineage running from the peak-oil movement of the early 2000s through Richard Heinberg and The Oil Drum to Nate Hagens’ The Great Simplification, with Simon Michaux providing an especially revealing modern example. The original claim was not merely that petroleum was finite, mature oilfields decline or conventional crude production would eventually peak. Those were and remain reasonable observations. The consequential claim was that an imminent petroleum-supply constraint would leave industrial societies without adequate substitutes, force economic contraction and begin a substantial simplification of modern civilization. The constraint did not arrive as forecast. Instead of that experience substantially reducing confidence in the expected outcome, a portion of the movement found new reasons why essentially the same outcome remained imminent.

Richard Heinberg’s The Party’s Over said in 2003 that global oil production would peak within the next few years, after which industrial societies would have progressively less net energy available, with managed contraction presented as the sensible response. By 2010, Heinberg was saying that conventional economic growth was effectively over and that aggregate economic activity would probably never again exceed its 2007 level. The global economy subsequently grew far beyond that level. Heinberg eventually acknowledged the uncomfortable evidence. In a 2018 retrospective on The End of Growth, he noted that both US and global GDP had risen rather than continued to fall, but interpreted the result largely as postponement rather than falsification: unconventional energy, financial interventions and debt had bought time, while the ultimate encounter with limits remained certain. That may be a hypothesis worth investigating, but it does not validate the forecast that growth had already ended.

The distinction matters because “growth cannot continue infinitely on a finite planet” and “this particular constraint will terminate economic growth within the period I am forecasting” are entirely different propositions. The first is almost definitionally true if the timeframe is long enough and growth is defined in sufficiently material terms. The second is a prediction with timing, mechanisms and opportunity costs attached to it. Utilities, governments, investors and households make decisions over decades, not geological eternity. Being correct that something cannot continue forever does not make someone correct about what prevents it from continuing in 2008, 2018, 2028 or 2038. A forecasting tradition that repeatedly retreats from a specified mechanism and timeframe toward the proposition that limits must eventually assert themselves is steadily moving from empirical prediction toward an unfalsifiable worldview.

Nate Hagens did not independently rediscover this framework years later. He was inside the movement. He became a major contributor and eventually managing editor of The Oil Drum, the most technically serious online centre of peak-oil analysis. In 2010 he explained that the Institute for the Study of Energy and Our Future, ISEOF, had been incorporated in 2007 as The Oil Drum’s nonprofit corporate parent. That same organization now produces The Great Simplification. This is not merely intellectual resemblance. There is direct institutional continuity between the peak-oil project and Hagens’ current media platform.

That history makes the important question less whether Hagens sincerely believes the current version of the story than whether the framework has learned enough from the failure of its earlier version. The more revealing evidence is what happened when the oil constraint failed to produce the expected economic result—and why Simon Michaux’s minerals argument fit the successor story so perfectly.

Read more

A couple of months ago I wrote about battery-electric garbage trucks as a heavy-duty electric vehicle story hiding in plain sight. Refuse collection is unusually well suited to batteries: relatively short and predictable routes, constant stopping and starting, lots of regenerative braking, fixed depots and long overnight parking periods. The people buying these vehicles aren’t trying to make transportation exciting. They’re trying to get Tuesday’s route completed. Hydrogen garbage trucks received only a couple of paragraphs because their record was already dominated by trials while battery-electric fleets were scaling. Then Leipzig’s experience with 16 hydrogen refuse trucks crossed my screen, and it was worth going back to see whether I had been too dismissive. I hadn’t been.

Leipzig bought the FAUN BluePower trucks in 2024, enough to make up roughly a third of its collection fleet. Two years later, their reported deployment rate is 45.6%, compared with about 80% for conventional refuse trucks. Technical faults, maintenance and repairs keep them off the road, and conventional replacement vehicles supplied by the manufacturer cover the missing work. Leipzig already has dedicated hydrogen refueling infrastructure, so the usual explanation that the pilot failed because nobody built a station doesn’t apply. This is a substantial fleet with purpose-built fueling, public subsidies and two years of operating experience, and the trucks still aren’t available often enough for normal municipal service.

The relevant financial comparison isn’t with diesel. A city decarbonizing refuse collection is choosing between zero-emission alternatives, and battery electric is the obvious benchmark. The stronger cost literature does not show hydrogen and batteries naturally converging on similar total costs. A 2026 Nature Communications study of real-world European heavy-truck utilization found hydrogen beating batteries only when favorable assumptions for fuel-cell development and hydrogen prices were paired with unfavorable assumptions for battery trucks. A separate Swiss assessment of demanding bus duties, including infrastructure and the additional vehicles required to deliver the service, again found battery electric the lowest-cost zero-emission option. These aren’t refuse-truck invoices from Leipzig, so they can’t tell us the city’s exact cost premium. They do tell us that hydrogen needs an unusually friendly set of assumptions to win a lifecycle-cost comparison.

The refuse-truck evidence makes the economic problem less abstract. Researchers in Helsinki measured total battery-electric refuse-truck consumption at about 2.3 kilowatt-hours per kilometre, including collection work, and recorded a February shift of more than 120 kilometres and 500 container lifts without intermediate charging. Hydrogen operators have had to pay for a more complicated support system. Arnhem needed conventional backup because its hydrogen truck could not reliably replace a normal vehicle, while Herten experienced months-long waits for parts and heavy dependence on the specialist supplier. Leipzig adds the most damaging cost variable of all: expensive capital that spends much of its life unavailable while another truck does the work. A total-cost model that assumes a vehicle delivers its scheduled workload is not describing a fleet that manages 45.6% deployment.

Europe has spent years trying to turn hydrogen refuse trucks into an ordinary product. The Hydrogen Waste Collection Vehicles in North West Europe project, HECTOR, ran from 2019 through 2023 with seven trucks at seven locations. Its objective was to create operating experience and provide a basis for wider deployment. The Refuse Vehicle Innovation and Validation in Europe project, REVIVE, ran from 2018 through 2024 and ultimately deployed 11 trucks across seven European cities. The vehicles collected garbage and the projects generated years of technical reports, demonstrations and follow-up announcements. What they did not generate was the wave of larger repeat orders that would indicate a technology graduating into routine fleet procurement.

That is a textbook example of what I described in Hydrogen Has Narrative Density. Electrification Has Deployment Density. Firsts, pilots, funding awards, partnerships, demonstrations and small orders each generate a news event. Mature deployment does not generate a separate story for every additional machine. In the bounded six-sector dataset behind that analysis, battery technologies produced only 1.8 times as many positive milestone announcements as hydrogen, while the physical deployment differences were routinely one, two or three orders of magnitude larger. For trucks, the deployment gap was about 29 to one. Narrative volume compressed the physical difference between the technologies.

North America supplied an almost perfect refuse-truck example. New Way Trucks and Hyzon developed a hydrogen refuse truck, put it through successful trials with waste companies and attracted positive operator comments. GreenWaste then signed a conditional agreement in October 2024 for 12 vehicles promoted as North America’s first commercial hydrogen refuse-truck order. The sequence looked compelling: prototype, first trial, more trials, customer endorsement, commercial order. Then Hyzon collapsed before supplying the fleet. At WasteExpo in 2025, New Way said it still wanted to continue the programme but was looking for another fuel-cell supplier or a buyer of Hyzon’s assets. The truck had proven it could collect garbage. The product ecosystem failed before commercialization.

The European projects show the same problem in less dramatic forms. Arnhem suffered low vehicle reliability and station outages. Herten had component failures, long waits for spares and insufficient reliability to retire the diesel truck it was supposed to replace. Bielefeld bought seven hydrogen refuse trucks and then lost its practical refueling option when regional stations closed. Moving the vehicles to the city’s hydrogen bus station required resolving technical and funding restrictions, with a solution finally implemented in April 2026. A functioning fuel-cell truck without fuel is still a truck that cannot collect garbage. Hydrogen adds another chain of dependencies—fuel production or delivery, compression, storage, dispensing, specialist maintenance and a thin parts network—and every link has to work on collection day.

Freiburg is the strongest counterexample, and it is a real one. The city has 22 fuel-cell refuse trucks across its municipal fleet, a second hydrogen station for redundancy and plans for local hydrogen production. Freiburg proves that a city determined to operate a hydrogen refuse fleet can build enough infrastructure around it to make the system function. It also illustrates the price of doing so: specialized trucks, redundant refueling and a local fuel-production project for a fleet of 22 vehicles. That is not evidence of an intrinsically competitive technology. It is evidence that sufficient institutional commitment and infrastructure can make one work.

Battery-electric refuse trucks have had failures as well, including a particularly poor early vehicle in Mobile, Alabama, charging and reliability problems in Nottingham and a troublesome conversion in Somerset. The difference is no longer subtle: successful battery trials turned into fleets, while successful hydrogen trials usually turned into more trials. Westminster committed £20 million to 45 electric refuse trucks and a dedicated charging depot. Copenhagen was running 86 electric refuse trucks by 2023. Republic Services had more than 250 battery-electric collection trucks operating by September 2026 and was heading toward 300. Its management says suitably selected routes are achieving one-for-one replacement of conventional trucks and completing full working days without midday charging. Battery refuse collection has moved from proving that the machines can work to deciding which routes and depots to electrify next.

Of course, compared to China those Western battery-electric fleet numbers are already small. China registered at least 7,900 new pure-battery garbage trucks in 2025 across seven refuse-specific categories, including 2,269 battery-electric compactor trucks, the closest direct analogue to the rear-loading refuse vehicles being discussed here. Shanghai alone registered 302 battery-electric compactors in that single year, with Chengdu adding 174 and Changsha 151. China also has hydrogen refuse trucks, so this isn’t a case where one technology was tried and the other ignored. The cleanest national hydrogen figure I found recorded 59 fuel-cell garbage trucks in 2023, within 159 fuel-cell sanitation vehicles of all kinds. Individual hydrogen refuse deployments since then have generally been measured in single digits or low tens. China has tested both pathways inside the same manufacturing ecosystem, and the market has chosen battery electric at vastly greater scale.

That Chinese evidence makes the narrative-density problem particularly obvious. Seven HECTOR trucks can sustain a four-year European demonstration programme. Eleven REVIVE trucks can generate six years of project activity. A single New Way/Hyzon demonstrator can produce a series of firsts, trials and partnership stories before a conditional 12-truck order becomes another headline. Meanwhile one country can add thousands of battery-electric garbage trucks in a year without anything like thousands of international stories. The difference between narrative density and deployment density is not an abstract media critique. It changes how mature the technologies appear.

Pilots are supposed to graduate. Hydrogen refuse trucks have repeatedly demonstrated that fuel cells can propel a truck down a street, operate the compactor and finish a collection route. That question was settled years ago. Commercialization requires something harder: reliable daily dispatch, affordable fuel, available technicians, durable suppliers, spare parts, sensible infrastructure costs and customers who come back to buy much larger fleets. On those measures the hydrogen refuse-truck story remains weak. Battery electric has moved well beyond it.

Leipzig matters because most of the usual excuses have already been removed. Sixteen trucks are not an experimental singleton. They have dedicated refueling infrastructure. Public subsidy reduced the purchase burden. The fleet has had two years to settle into service. Yet conventional vehicles still have to cover for it while reported deployment remains at 45.6%. The original hydrogen story was about 16 zero-emission garbage trucks entering service. The useful story arrived two years later, when someone finally published the denominator.

Tuesday morning still comes. The garbage still has to be collected. In Leipzig, the hydrogen trucks are failing that test more often than they are passing it.

For more evidence-first analysis of what is actually scaling in the energy transition, subscribe to TFIE Strategy Briefing.

Subscribe now

In 2021 I argued that globalization was essential to sustainability. International trade spreads technology, creates manufacturing scale, raises productivity and gives countries access to resources and capabilities they could not efficiently reproduce within their own borders. I also suggested that deeper economic integration would bring a “lovely byproduct of global peace.” That part was too optimistic. Russia invaded Ukraine less than six months later. The pandemic had already demonstrated the fragility of concentrated supply chains, while the years since have brought US-China trade restrictions, export controls, industrial subsidies and a much sharper concern with economic security. Globalization did not abolish geopolitics, and economically efficient concentration does not necessarily produce resilience.

The sustainability argument nevertheless looks stronger to me now, although with a different qualifier. Depending heavily on one unreliable supplier for a continuously consumed commodity is dangerous. An international system with several suppliers, several routes, domestic alternatives and enough redundancy to survive a disruption has a different risk profile. What I have come to call strategic interdependence accepts that countries will depend on one another while paying much more attention to the structure of those dependencies.

Energy makes the distinction unusually clear because electrification changes what crosses borders. Fossil-fuel economies repeatedly purchase things that disappear when used. An LNG cargo is burned and another cargo has to arrive. Oil imported this month contributes nothing to next month’s mobility. Clean-energy trade increasingly moves solar panels, batteries, electric vehicles, transformers, power electronics and other durable equipment that continues producing useful energy services for years after the trade transaction is finished.

The geopolitical reaction has included real attempts to localize supply chains, some of them sensible. China’s dominance of several stages of solar, battery and clean-technology production is an obvious vulnerability. Yet the International Energy Agency’s 2026 technology review finds that international trade in major clean-energy technologies reached an all-time high in the second quarter of 2025 despite higher tariffs and defensive industrial policies. In its stated-policies scenario, the value of net trade in the six clean technologies it models more than doubles from about $290 billion in 2025 to $620 billion in 2035. Emerging economies have also gone from receiving less than 5% of Chinese EV exports in 2020 to nearly 40%.

There are good reasons to make battery cells in Europe, solar modules in India, transformers in North America and wind components in additional regions. Diversity of supply, inventories, alternative shipping routes and some domestic capacity in critical technologies reduce disruption risk. But requiring every country to reproduce every stage of every clean-technology supply chain would throw away much of the manufacturing scale and specialization that made solar panels, batteries and EVs inexpensive enough to spread globally. The relevant question is how concentrated a dependency is, how quickly alternatives can substitute for it and what happens when a supplier or route disappears.

Electricity interconnection shows how far that reasoning has moved into energy-security policy. I explored this recently in As Geopolitics Fragments, Electricity Interconnection Gains Strategic Value. Four years after Russia’s invasion exposed the consequences of European dependence on Russian gas, European countries have not responded by trying to turn themselves into electricity islands. The European Commission’s current grid programme explicitly treats resilience and security of cross-border energy infrastructure as objectives. Its Energy Highways initiative is intended to remove interconnection bottlenecks, improve security of supply, integrate more renewables and reduce fossil-fuel dependence.

Ukraine provides a considerably harder test than normal electricity-market modelling. After Russia invaded in February 2022, Ukraine and Moldova accelerated their synchronization with the Continental European electricity system. ENTSO-E completed the emergency synchronization in March 2022 while Ukrainian grid operators were already working under wartime conditions. Cross-border electrical infrastructure in that case was not merely a mechanism for finding a lower wholesale price. It provided another source of system support while domestic infrastructure was exposed to extraordinary disruption.

Interconnectors themselves can be damaged or attacked. Neighbouring countries can experience similar weather, exports can become politically contentious during shortages and one enormous cable can become another single point of failure. The security value comes from the portfolio around the connection: several routes and counterparties, substantial domestic generation, storage, flexible demand and an internal grid capable of moving electricity where it is required. Strategic energy interdependence does not replace domestic resilience. It broadens the resources available when part of the domestic or international system fails.

The transition from fuel imports to productive-asset imports is already visible at national scale. Pakistan provides an unusually clean example. The State Bank of Pakistan’s 2026 assessment estimates that decentralized solar capacity may have reached roughly 50 GW by the end of 2025. It estimates cumulative foreign-exchange savings from reduced energy imports at about $7.1 billion over six years and calculates that the installed solar base could reduce the country’s 2026 energy-import bill by another $5.2 billion to $7.8 billion. Most strikingly, its staff calculate that the cumulative dollar cost of imported solar panels has now been broadly offset by the foreign exchange saved through lower fuel imports.

Pakistan imported large quantities of manufactured equipment, much of it from China, and that equipment is reducing its requirement to buy fuels abroad. The panels are imported, but the sunlight is Pakistani. Once installed, those foreign-manufactured assets turn a domestic resource into useful electricity without requiring another shipload of fuel.

Electric vehicles produce the same structural change in transport. The IEA estimates that the global EV fleet displaced about 1.7 million barrels of oil per day in 2025. China alone accounted for roughly one million barrels per day of that displacement and is also the world’s largest oil importer. Ethiopia represents a much earlier-stage version of the same logic: its government banned imports of gasoline and diesel vehicles in 2024 explicitly as part of a strategy to promote electric mobility and reduce fuel imports. The IEA reports that Ethiopia’s oil-import bill had exceeded $4 billion in 2022 and that around 100,000 EVs had subsequently been deployed, although charging and maintenance infrastructure remain constraints.

For an oil-importing economy, replacing an internal-combustion vehicle with an imported EV does not eliminate international trade. It changes the trade relationship. Instead of importing a machine and then importing thousands of litres of fuel over its operating life, the country imports more of the capital up front and powers it from an electricity system that can contain increasing shares of domestic wind, solar, hydro or other generation. Batteries and vehicles still have international supply chains and replacement requirements, but they do not create the same continuous dependence on a commodity whose value disappears during combustion.

Africa makes this transition particularly interesting because imported clean technology can help create the market for later industrial development. In Africa’s Solar Boom Is Hiding In The Import Data, I argued that inexpensive solar, falling battery costs, diesel displacement, electricity access and transport electrification were starting to reinforce one another rather than developing as isolated technology markets.

The hardware flows are already material. Sustainable Energy for All estimates that Africa imported more than $12 billion of solar photovoltaic modules, lithium-ion batteries and assembled EVs between 2022 and 2024. Solar-module imports rose from more than $1.2 billion in 2022 to nearly $2 billion in 2023 before easing to $1.6 billion in 2024, with more than 90% coming from China. Battery imports reached an estimated $1.8 billion in 2024, while imports of assembled passenger EVs exceeded $500 million.

That degree of concentration is a vulnerability, but it is not the whole story. Imported equipment creates installation businesses, maintenance capability, financing experience, logistics networks and customers. Solar and batteries can displace diesel generation and improve electricity reliability for mines, telecom sites, farms, warehouses and factories. Electric motorcycles, buses and other vehicles add useful electricity demand and create businesses around charging, financing, maintenance and battery services. As those markets become larger and more predictable, regional assembly and manufacturing become more plausible.

The World Bank’s 2026 work on African electric mobility makes the feedback loop explicit. It argues that electrifying transport can cut petroleum imports and foreign-exchange exposure while vehicle charging can provide anchor demand for solar mini-grids. Higher utilization improves the economics of the electricity asset and can reduce unit electricity costs for surrounding users. At the same time, the mini-grid gives electric motorcycles and other vehicles access to locally generated energy that replaces daily purchases of petroleum.

This is the African flywheel I have been watching. Cheap imported technologies enable deployment. Deployment builds useful electricity demand and local service industries. Better economics support more generation and storage. Electrified transport increases electricity utilization while reducing exposure to fuel imports. Larger markets justify better logistics and eventually more regional assembly and manufacturing. International trade can therefore be part of the process by which countries gain domestic productive capability rather than evidence that they lack it.

None of this implies that African economies should remain overwhelmingly dependent on one external manufacturing country. Supplier diversification, African assembly and manufacturing where scale supports them, stronger regional electricity networks and larger continental markets all improve resilience. But requiring local self-sufficiency before deployment would reverse the sequence that made most successful industrial ecosystems possible. Markets, skills, suppliers and institutions develop partly through doing the work.

That is where my 2021 argument needs its largest correction. Trade can raise the cost of disruption and create substantial benefits from cooperation, but it does not make geopolitical conflict disappear. Dependencies can be weaponized. Companies can optimize supply chains until they become brittle. Governments have legitimate reasons to care about critical suppliers, chokepoints and domestic industrial capability.

The sustainability argument for globalization survives because the physical transition remains international. Excellent renewable resources, hydro reservoirs, critical minerals, manufacturing centres, engineering expertise, capital and demand do not line up conveniently with national boundaries. Building resilience means diversifying those relationships and strengthening domestic capabilities where they matter, not deliberately constricting the networks through which technologies, electricity and knowledge move.

The old global energy system moved enormous quantities of coal, oil and gas from concentrated geological deposits to consumers that had to keep buying them. The emerging system increasingly moves solar modules, batteries, EVs, transformers, power electronics, minerals and electricity. Some of those supply chains are too concentrated today and should become more diverse. But much of what crosses the border now leaves behind productive capacity that can reduce the next import bill.

Globalization is still essential to sustainability. Five years of geopolitical disruption have clarified the condition attached to that claim. A resilient transition requires countries to produce more useful energy from their own resources while maintaining several ways to obtain the technologies and energy they cannot efficiently supply themselves. It means stronger electrical links to reliable neighbours, more diverse industrial supply chains and progressively less dependence on imported fuels that must be purchased again every time they are burned.

Subscribe to TFIE Strategy Briefing for decision-grade analysis of energy, infrastructure, industrial strategy and the transition.

Subscribe now

Ports are being asked to make thirty- and fifty-year capital decisions while several of the assumptions that justified their existing infrastructure are changing at once. Ships are getting larger, requiring deeper channels, larger cranes and more capable tug fleets. Climate change is worsening some of the weather and water conditions that interrupt port operations and, just as importantly, the shipping and inland transport networks around them. Ports themselves have to decarbonize, both because of regulation and because customers increasingly care about the carbon intensity of supply chains. Meanwhile, some of the largest cargo streams moving across their quays are entering structural decline. None of those issues is news to a competent port manager. The strategic difficulty is that they interact, with several increasing capital requirements while others reduce the cargo volumes and revenues available to pay for them.

My maritime freight projections have had this tension embedded in them for years. Roughly 40% of maritime freight tonnage today is coal, oil, petroleum products, LNG and LPG, while raw iron ore adds roughly another 15%. Fossil-fuel cargo declines as the energy system electrifies, while raw iron ore comes under pressure from slower growth in steel-intensive infrastructure, increasing scrap availability, electric arc furnaces and more iron reduction occurring closer to mines and inexpensive renewable electricity.

In my updated shipping fuel projection, the marine liquid-fuel requirement falls from roughly 425 million tonnes of fuel-equivalent in 2030 to about 180 million tonnes in 2050 and around 70 million tonnes by 2100. Those figures are about maritime energy rather than directly about port revenue, but the cargo-side logic matters enormously because coal, crude oil, petroleum products, LNG and raw iron ore have been paying for channels, berths, rail connections, harbour services, land and shared infrastructure. Those costs do not disappear proportionally when the tonnes do.

At the same time, the ships carrying the durable parts of maritime trade are becoming more demanding. Container ships in particular have grown dramatically, increasing berth, channel and crane requirements as well as windage during manoeuvring. In a recent sponsored Lloyd’s List discussion, Svitzer chief executive Kasper Friis Nilaus connected larger vessels, greater windage, constrained channels and difficult weather with rising towage requirements. My subsequent assessment, Tugboats May Be Where Port Electrification Comes Together First, found that the interesting part of Svitzer’s TRAnsverse design was not merely bigger bollard-pull numbers, but better directional control and dynamic force at the speeds where tugs actually work around large vessels. It reinforced something that has been sitting in my maritime projections for years: tugs are unusually good candidates for electrification because they combine short distances, repeated duty cycles, predictable charging opportunities and enormous power requirements for relatively short periods. The port of the future can therefore handle fewer tonnes of fossil bulk while requiring more capable equipment to handle the ships and cargoes that remain.

Climate disruption adds another claim on the same capital base. A 2023 Nature Climate Change study of 1,320 ports found that for roughly two-thirds of them, modeled delays arriving through disrupted trading partners exceeded their own direct weather-related downtime risk. That finding matters because a port can spend heavily hardening its quays, drainage, electrical systems and access roads and still lose traffic because another port in the service rotation is closed. China’s 2026 typhoon season offered a practical demonstration as Shanghai, Ningbo and Yantian suffered successive interruptions, carriers omitted calls and changed rotations, and some Chinese ports that had not themselves closed subsequently experienced large increases in cargo dwell. The port fence is not a meaningful boundary for climate resilience when ships, containers and schedules are networked globally.

The same problem runs inland. Vancouver’s rail and highway connections were severed during the 2021 atmospheric-river floods even though the port itself survived. Durban restored much of its port operation after its 2022 flooding while damaged rail and road connections continued constraining freight. Low Rhine and Danube water during 2026 reduced barge carrying capacity while ports remained open, leaving vessels to carry less cargo or requiring more sailings to move the same amount. Panama demonstrates the ocean-routing equivalent, with drought-related canal restrictions reducing transit and loading capacity and pushing some freight toward longer or more expensive alternatives. The climate fingerprint differs between those cases—human-caused warming has a much clearer role in aggravating European drought conditions than could be established for Panama’s 2023 rainfall deficit—but strategically they expose the same problem. Freight capacity can deteriorate substantially without anything inside the port being destroyed.

Western Pacific typhoons make that network exposure particularly consequential. Total typhoon counts do not have to increase for risk to ports to rise. A 2016 Nature Geoscience analysis of landfall-prone Northwest Pacific typhoons found intensification of 12–15% over the preceding 37 years, with the proportion reaching Category 4 or 5 doubling or tripling in the groups affecting East and Southeast Asia. Subsequent research has found northward shifts in parts of western Pacific cyclone exposure and a roughly threefold increase in rapid-intensification events close to coastlines globally between 1980 and 2020. The attribution and regional patterns require care, and the evidence does not say that every Asian coastline will see more typhoons. It does say that basin-wide storm counts are a poor proxy for port risk when intensity, tracks, rapid intensification, rainfall and higher coastal water levels are changing around a region that contains a disproportionate share of the world’s major container ports.

Port authorities are not oblivious to this. Rotterdam’s strategic work explicitly anticipates large reductions in fossil cargo, greater climate-resilience requirements, industrial transformation and continuing competition for containers. Newcastle, still heavily dependent on coal exports, is pursuing containers, clean-energy industries, automotive traffic and other non-coal business. Port Waratah Coal Services in Newcastle is already contemplating a future in which coal demand can be served through one terminal instead of two and aligning that possibility with lease expiry, asset life and rehabilitation provisions. The interesting weakness in port strategy is therefore not failure to notice that the world is changing. It is the collective arithmetic of what ports expect to replace the disappearing business.

Rotterdam expects containers to become more important and expects to maintain or strengthen its competitive position. Newcastle wants a major container terminal. Other ports serving overlapping hinterlands have their own expansion plans. There is historical evidence that these ambitions do not always add up: the European Court of Auditors found neighbouring ports investing in similar capacity without adequately testing whether their shared hinterlands contained enough traffic. The problem does not require port planners to be unaware of their competitors. Every port can know exactly who its competitors are and still build a business case in which it wins market share from them. If several competing ports make the same assumption, their individual strategies can each appear coherent while the combined strategy is impossible. Containers can replace declining bulk for particular winners, but they cannot replace declining bulk for the port sector simply because they appear in everyone’s diversification strategy.

The same arithmetic problem appears in green bulk. Ports routinely envisage hydrogen, ammonia, methanol, biofuels, captured carbon, biomass, circular raw materials and other transition commodities replacing some of the coal, crude oil, LNG and petroleum products moving through their terminals today. There will certainly be new commodity flows, but my projections find nothing resembling a tonne-for-tonne substitution of green molecules for the fossil energy system they replace. Electrification removes enormous quantities of fuel from the energy and transportation systems instead of replacing every tonne of fossil fuel with a tonne of another molecule, while the long-term shift toward scrap and electric-arc steelmaking puts additional pressure on raw iron-ore movements. A port strategy in which declining coal and oil are replaced by both rapidly growing container traffic and enormous new quantities of green bulk can be plausible for an individual winner. It becomes considerably harder to believe when every competing port expects to be that winner.

That makes electrification unusually attractive because its value does not depend strongly on which of those forecasts proves correct. I developed the port-side pathway systematically in From Quay To Sea: A Port Decarbonization Roadmap, where the sequence starts with the straightforward electrical substitutions in ground equipment and vehicles, moves through harbour craft and shore power, and extends outward into coastal and blue-water shipping. The ordering is important. Ports do not have to begin decarbonization by betting on which future maritime molecule wins. They can begin by removing diesel from equipment with known operating patterns, building electrical capacity and operational competence while moving into progressively harder applications.

The inland side is now producing the same result. In my upcoming European port inland bulk fleet transformation and electrification roadmap, the starting point is not today’s fleet with every diesel engine replaced by an electric motor. Cargo transition changes the denominator first. Declining fossil and other exposed bulk flows reduce the fleet that ultimately has to be replaced, while persistent dry- and wet-bulk routes are worked through vessel cohorts, replacement and retrofit windows, route energy, direct charging, battery exchange and terminal constraints. The emerging pathway is overwhelmingly electrical, with the port increasingly acting as an electrical-logistics platform connecting vessels, terminals, grid capacity, battery infrastructure and scheduling rather than as a filling station for a collection of new molecules. That work extends the From Quay To Sea roadmap beyond the waterfront and makes the same point from the other direction: electrical infrastructure can serve the port estate, harbour craft and the inland vessels connecting the port to its hinterland.

This is also why the competitive case I made earlier—that electrified ports will have an advantage as maritime trade changes—looks stronger rather than weaker as the other pressures accumulate. Electric cranes, yard tractors, material-handling equipment, drayage trucks, harbour craft, inland vessels and tugs reduce energy consumption, maintenance requirements, local pollution and exposure to carbon pricing. Shore power improves the emissions profile of vessel calls while helping justify electrical infrastructure that can also serve terminal equipment, charging, batteries and industrial customers. Ports with abundant, reliable and reasonably priced electricity also become more attractive locations for industrial processing as heat, vehicles and machinery electrify. The port and inland-shipping roadmaps increasingly look like parts of the same system rather than independent decarbonization exercises.

Crucially, those benefits survive forecasting errors elsewhere. A port can be wrong about whether ammonia, methanol or hydrogen becomes the largest traded green molecule and still benefit from efficient electric cargo handling. It can win less container traffic than expected and still lower the cost of handling every container it does receive. A battery-electric tug remains useful whether the ship alongside is carrying containers, food, steel products or machinery. A substantial grid connection initially serving yard equipment can subsequently support trucks, harbour craft, inland vessels, shore power and battery storage. Grid connections, substations, energy-management systems and charging infrastructure therefore support multiple plausible futures instead of depending on one commodity thesis being right.

That does not mean electrifying every legacy asset. A coal terminal with a short remaining commercial life should not automatically receive electrical infrastructure designed for another forty years, just as it should not automatically receive expensive climate adaptation intended to protect it into the 2070s. The investment case is strongest in the durable shared parts of the port: electrical distribution, general-cargo and container handling, road and rail interfaces, harbour craft, inland-vessel interfaces, tugs and other services whose usefulness survives changes in cargo composition. Ports have to decide which assets deserve protection and renewal, which require additional capability for bigger ships, which declining cargoes can continue paying their way, which replacement businesses are realistically theirs to win and which facilities should be converted or retired.

That is a much harder strategic problem than either a conventional growth plan or a conventional climate-adaptation plan. Ports are clearly aware of most of its components. The danger is that their answers do not add up across the sector, with too many strategies relying simultaneously on winning the same container growth and handling the same projected flood of green bulk while adaptation and ship-size requirements keep raising the capital needed to remain competitive. Electrification does not remove that uncertainty, but it reduces the penalty for getting the commodity forecasts wrong. It lowers the cost and carbon intensity of whatever durable freight business a port actually ends up handling, builds infrastructure useful across multiple modes and cargoes, and avoids making every capital decision dependent on correctly predicting the future molecule. In a period when so many other port investments require getting the future right, electrification is one of the clearer no-regrets bets.

Subscribe to TFIE Strategy Briefing for evidence-based analysis of how electrification, climate risk and shifting freight flows are reshaping ports, shipping and industrial strategy.

Subscribe now

Landon Frigault, a Sustainable Energy Engineering student at Simon Fraser University, dropped an OceanX story into my LinkedIn messages over the weekend. A Chinese floating wind turbine had reportedly gone through a severe typhoon and kept spinning, which was enough to make me open the link. I expected another instance of China building a wind turbine at a scale that would have sounded implausible a decade ago, but the machine itself quickly became more interesting than the weather story. OceanX carries two turbines on one floating structure, and both rotors sit downwind of their supports.

Both choices were familiar to me from work I had done years ago. I had worked through the multi-rotor question in an old Quora answer, looking at why vertically stacking smaller rotors on a mast generally sacrifices swept area and puts more of the remaining rotor area into weaker wind closer to the ground, while putting one rotor behind another introduces wake losses, turbulent inflow and ugly cyclic loading. Modern three-bladed horizontal-axis turbines had converged on their familiar form for good engineering reasons, and most arrangements involving extra rotors looked worse after swept area, wind shear, wakes, structure and maintenance were included.

Downwind turbines had their own history. They offer some attractive characteristics, including natural alignment tendencies and more freedom for blades to deflect away from the support, but they historically paid for them every time a blade crossed the tower wake. The abrupt velocity deficit and turbulence behind the tower periodically unload the blade, contributing cyclic structural loading and, on some machines, the characteristic low-frequency thump that helped make downwind designs unpopular.

OceanX combines those two design choices on a 16.6 MW floating turbine, but it does so in configurations that differ markedly from the ones I had been considering years ago. The two 182 metre rotors sit side by side rather than one above or behind the other. Their supports are extraordinarily slender compared with ordinary wind-turbine towers and are held within a network of substantial pretensioned stays. The complete floating structure turns relative to its mooring system, keeping the rotors, towers and stays in approximately the same relationship to the wind as direction changes. Mingyang’s technical paper on OceanX makes clear that these are not independent curiosities bolted onto a conventional floater. They form a coupled structural and aerodynamic design.

The typhoon claim itself turned out to be both less extraordinary and more useful than the headlines suggested. OceanX reached its Yangjiang site in August 2024 and Super Typhoon Yagi arrived in September. The turbine did not achieve grid connection until December 11, 2024, so descriptions implying that it continued feeding electricity into the grid throughout Yagi are chronologically impossible. Whether its rotors continued turning unloaded at particular points is a different question. The instrumented storm response is much more valuable. Mingyang reports nacelle wind above 41.5 metres per second, significant wave height of 6.5 metres, a maximum wave of 9.8 metres and nacelle inclination varying by only about zero to three degrees. Afterward, its engineers compared measured floating-body responses with simulations prepared before the storm and reported close agreement, no obvious abnormal resonance, closely matching upstream and downstream float drafts, and no obvious change in platform attitude or draft.

OceanX’s published normal cut-out wind speed is 25 metres per second and its survival specification is 57 metres per second, so Yagi did not expose the machine to a wind speed beyond what a typhoon-rated offshore turbine is supposed to survive. The reported wind was nevertheless far beyond normal generating conditions. Dynamic pressure scales with the square of velocity, making the ambient pressure associated with 41.5 metres per second roughly 2.8 times that at 25 metres per second before blade pitching and other load-reduction measures are considered. At the same time, the platform was responding to a 6.5 metre significant sea and individual waves approaching ten metres. A floating wind turbine has surge, sway, heave, roll, pitch and yaw available to it before the flexibility of towers, blades and mooring lines is included, and OceanX adds two widely separated rotors, a large stayed V structure and a rotating mooring connection. In that context, the roughly zero-to-three-degree reported nacelle inclination response is considerably more informative than the category attached to the storm.

The storm data made the rest of the machine worth taking seriously without resolving the two design choices that had initially made me skeptical. Two turbines still mean two nacelles, two drivetrains, two hubs and six blades. Downwind rotors still pass repeatedly through disturbed air created by the structures ahead of them. The thin towers and stays may reduce the tower-shadow problem enough to alter lifetime economics, or they may shift structural cost and fatigue into components that are less obvious in photographs. Whole-platform yaw eliminates conventional nacelle yaw systems but places more importance on a rotating connection below the waterline. My assessment therefore started to diverge: the severe-weather response looked better the more closely I examined it, the downwind arrangement became more technically interesting, and I remained unconvinced that two rotors were necessary.

OceanX does not overturn most of the objections I had made years ago about multi-rotor or downwind machines. Its designers have instead changed many of the conditions that created those objections. The rotors are beside one another instead of stacked or tandem. The stays may allow supports slender enough to greatly reduce the wake that historically penalized downwind machines. The complete structure weather-vanes rather than asking two nacelles to yaw independently. Almost all of those potential advantages, however, could also exist with one larger rotor. Following the load path from blades through towers and stays into the floating platform and its moorings made me substantially more interested in downwind floating wind, while leaving the case for two rotors unresolved.

Read more

Green Lightning really does make nitrogen fertilizer from air and electricity. But nitrogen atoms do not multiply. At the concentration reported by Washington State University Extension, a 100-gallon-per-day Green Lightning machine puts about 161 pounds of newly fixed nitrogen into its water over a year. That is the physical quantity that matters: nitrogen that was in the atmosphere and has actually been converted into a form plants can use.

That is what made the claim particularly striking when it surfaced after I published Farm Electrification Should Follow The Work. On-farm nitrogen production would fit the electrification thesis beautifully if the numbers worked: air and water are available everywhere, electricity increasingly is as well, and avoiding industrial ammonia production and fertilizer transport would have obvious attractions. The problem is not whether electricity can fix nitrogen. The problem is whether Green Lightning makes remotely enough of it to support what farmers are being told it can replace.

The marketing claim needs to be kept separate from the measured nitrogen. Green Lightning and Canadian distributor Nytro have described each gallon as the equivalent of about three pounds of conventional nitrogen, and Nytro president Chris Nykolaishen was still using three pounds of nitrogen equivalence per gallon in February 2026. At 100 gallons per day, that becomes 109,500 pounds of claimed conventional-nitrogen replacement per year.

Those two numbers sound as though they are different ways of measuring the same thing, but they are not. About 161 pounds is a mass of actual nitrogen atoms. About 109,500 pounds is an assertion about agronomic performance. At the WSU-reported concentration, the claimed fertilizer equivalence is roughly 680 times the quantity of nitrogen actually produced. Calling that larger number “equivalent nitrogen” does not create the missing nitrogen any more than making a truck more fuel-efficient creates diesel.

This distinction matters because crops physically incorporate nitrogen into biomass. Nitrogen is a constituent of amino acids, proteins, nucleic acids and chlorophyll, and better fertilizer timing, placement or uptake can reduce losses but cannot make more nitrogen atoms enter a crop than are available from all of its nitrogen sources. If a few thousandths of a pound of newly fixed nitrogen accompanies a crop response normally associated with several pounds of fertilizer nitrogen, the additional nitrogen has to have come from somewhere else.

There is a straightforward alternative explanation for apparently successful reduced-fertilizer trials. Agricultural soils can supply substantial nitrogen through mineralization of soil organic matter, while residual fertilizer, manure, previous crop management and other sources can contribute more. A field can therefore maintain yield after a large fertilizer reduction without the replacement treatment having supplied the missing nitrogen. That is why fertilizer-substitution trials need controls, soil nitrogen accounting and repeated seasons rather than simply observing that a treated crop grew well.

The plasma chemistry itself does not rescue the equivalence claim. Green Lightning’s patent describes a gliding-arc plasma reactor that oxidizes atmospheric nitrogen, using a pathway descended from the Birkeland-Eyde process industrialized more than a century ago. Plasma nitrogen fixation is real chemistry, but historically its disadvantage has been energy consumption, not whether it can make nitrate at all.

North Dakota State University’s assessment puts modern Haber-Bosch production at roughly 800–1,100 kJ/mol and the old Birkeland-Eyde process at roughly 2,500–3,000 kJ/mol, while saying current non-thermal plasma approaches remain roughly five to ten times less energy-efficient than Haber-Bosch. Researchers are actively trying to improve those numbers, but there is no overlooked physical pathway that turns a small amount of electrical energy into enormous quantities of fixed nitrogen.

Green Lightning’s earlier marketing illustrates how easily actual nitrogen and claimed equivalence can become conflated. NDSU preserved an older Green Lightning graphic advertising “110K lbs of N/yr” alongside an 1,100-watt electrical requirement. The current six-head machine is rated at 100 gallons per day, and 100 gallons multiplied by three pounds of claimed nitrogen equivalence and 365 days gives 109,500 pounds per year, almost exactly the old 110,000-pound figure.

That strongly suggests the older number was an annualized fertilizer-equivalence figure, despite being labelled simply as nitrogen. NDSU showed why reading it as actual fixed nitrogen makes no physical sense: producing roughly 100,000 pounds of nitrogen annually from a continuous 1.1 kW input would require about 19.47 kJ/mol, compared with NDSU’s stated theoretical minimum of roughly 200 kJ/mol for making nitrogen oxides. The literal interpretation would therefore beat the theoretical energy minimum by roughly an order of magnitude.

The useful number remains the amount of nitrogen actually in the water. WSU reported approximately 530 ppm nitrate-N, or 0.00442 pounds of nitrogen per gallon. A machine making 100 gallons per day at that concentration would produce about 0.442 pounds of nitrogen per day, approximately 161 pounds per year. Those 36,500 annual gallons are simultaneously being credited, at three pounds of conventional-N equivalence per gallon, with replacing 109,500 pounds of conventional nitrogen.

Nytro says newer systems make more nitrate. Its 2026 Ag in Motion submission says recent optimizations have doubled, tripled and even quadrupled nitrate concentrations, while its own operating guidance says nitrate concentration varies and the product has no guaranteed analysis. The public material does not establish that WSU’s 530 ppm value is the exact baseline for those claimed improvements, so a direct current comparison would be inappropriate.

A generous sensitivity test still shows the scale of the problem. Quadrupling the WSU-reported concentration to 2,120 ppm would increase actual nitrogen production to roughly 645 pounds per year from a 100-gallon-per-day machine. That is an appreciable improvement over 161 pounds, but it remains about 170 times smaller than the 109,500 pounds of claimed annual conventional-N replacement. Better equipment narrows the gap without coming close to eliminating it.

The crucial empirical question is therefore simple: do controlled field trials show that this very small amount of newly fixed nitrogen reliably substitutes for the much larger quantity of conventional nitrogen claimed? The independent results available so far do not demonstrate that multiplier.

In a 2025 Carrington Research Extension Center corn trial summarized by NDSU, Green Lightning treatments under conventional tillage fell into an intermediate statistical group and were not distinguishable from either the conventional urea-ammonium nitrate (UAN) treatment or the untreated check. That portion of the trial therefore demonstrated neither equivalence nor failure. It showed exactly why a crop growing after treatment is not enough evidence to determine where its nitrogen came from.

The strip-till portion produced a clearer result. Conventional UAN yielded about 211 bushels per acre and occupied a separate statistical group, while Green Lightning treatments yielded roughly 195 to 200 bushels per acre and were statistically grouped with the approximately 189-bushel untreated check. One site-year cannot settle the technology, but this trial did not show Green Lightning delivering the agronomic equivalent of the conventional nitrogen treatment.

Precision Planting tested complete fertilizer replacement even more directly in 2024. Its 100% Green Lightning treatment applied 120 gallons per acre across six applications and produced 235.9 bushels per acre, compared with 280.7 bushels under the conventional UAN program. The 44.8-bushel loss exceeded the 30.7-bushel loss the researchers calculated the cheap Green Lightning treatment could tolerate while still breaking even financially.

Precision Planting noted that hard water and difficulty producing a consistent Green Lightning product may have contributed, and proposed reverse-osmosis-treated water for later testing. That is a legitimate qualification on one trial rather than evidence that every future system must perform the same way. But reliable fertilizer production is also part of the proposition farmers are buying, and Nytro’s current instructions now specify reverse-osmosis-quality water for optimized operation.

A 2025 pasture experiment associated with the University of Missouri and Ohio State offers another useful check. The North Central Soil Fertility Conference proceedings report that nitrogen was below detectable levels in the Green Lightning material used during the first application and note technical problems with the machine that may have limited nitrate production. In April, Green Lightning plots produced 20.6% less forage than the untreated control, a marginal statistical result, and 32% to 39.4% less than the conventional nitrogen treatments. The equipment problems limit what can be concluded about future machines, but the experiment did not demonstrate the claimed fertilizer-replacement effect.

Nytro cites more favourable results, and those should be considered rather than ignored. Its 2026 Ag in Motion submission says more than a dozen small-plot trials conducted during 2025 found that replacing 50% to 75% of synthetic nitrogen could save $35 to $60 per acre without reducing yield. The public summary, however, does not provide the individual trial protocols, soil nitrogen measurements, nitrate concentration of the applied Green Lightning product, replication or statistical results required to distinguish nitrogen supplied by the machine from nitrogen already available in the field.

Formal testing is still underway. Olds College describes its Nytro evaluation as a three-year program beginning in 2025 and continuing through 2028, examining wheat, barley and canola yields, biomass, nutrient uptake and soil effects. RealAgriculture likewise reported in February 2026 that work involving Olds College and Lakeland College was still being used to establish application rates.

Commercial sales have moved faster than that validation process. In March 2025, Saskatchewan Hansard recorded an MLA quoting Nykolaishen saying Nytro was already selling machines and collecting data during what he called a “soft release.” By February 2026, Nykolaishen told RealAgriculture that approximately 150 six-head units were operating across about 65 farms, while larger containerized systems were starting to ship.

These are significant capital purchases rather than inexpensive experimental devices. Nytro currently lists its six-head machine at C$70,000 and its 30-head containerized system at C$349,000, including delivery, installation and onboarding. Farmers are therefore buying commercial equipment while the multi-year trials intended to establish how much conventional fertilizer it can reliably replace are still underway.

The correction is simpler than the technology. Green Lightning makes nitrogen, and that nitrogen can be measured. Nitrogen atoms do not multiply. At the WSU-reported concentration, a 100-gallon-per-day system produces about 161 pounds of newly fixed nitrogen annually; even a hypothetical fourfold improvement raises that to only about 645 pounds. If a field maintains yield after tens of thousands of pounds of conventional nitrogen are removed from the calculation, the rest of the crop’s nitrogen must come from soil or another source unless controlled experiments show otherwise.

So far, the independent field evidence has not demonstrated the enormous agronomic multiplier on which Green Lightning’s fertilizer-equivalence claim depends. The plasma chemistry is real, but the commercial proposition requires something much more consequential: convincing evidence that a comparatively small mass of newly fixed nitrogen can reliably substitute for vastly more conventional fertilizer nitrogen. Not all farm electrification is created equal.

Subscribe to TFIE Strategy Briefing for evidence-based assessments of technologies being sold as climate solutions.

Subscribe now