I just listened to her conversation on Odd Lots from CityLab and I want to amplify a few things — and push back on one.

First, the amplification: her point about talent hemorrhaging out of public institutions is not talked about enough. I watched it happen in clean energy. The moment a technology gets hot, the government agencies that built the foundational science — DOE national labs, DARPA, ARPA-E — lose their best people to companies paying 2x salaries. We then wonder why procurement is slow and contracts are badly written. You can’t write a good contract for a technology you no longer understand.

At the Loan Programs Office, we tried to solve this by hiring people who actually understood project finance and technology risk — not just lawyers and generalists. It worked. We closed over $107 billion in loans and conditional commitments. Not because we had more budget, but because we had the right capabilities. Mazzucato would call this “dynamic capabilities.” I just call it hiring people who have years of experience that want to contribute to their country.

Second: the consultant critique is a great insight. Consultants have lots of tactical capabilities but the buck has to stop with a government employee. The process have to be owned by the government. We did that with our liftoff reports where we had government employees own the work product but consultants provide financial data and other insights. McKinsey can write you a climate strategy. They cannot tell you how to get people to use it to deliver real outcomes. That judgment only comes from doing hundreds of deals. You can’t outsource your learning curve.

Now the pushback: I think the “mission-oriented” framing, as much as I love it in theory, can become an excuse for governments to avoid the harder work of execution. I’ve seen a lot of beautiful mission documents. We created a great one at LPO - "To be the premier, catalytic public financing partner for high-impact energy and manufacturing investments to advance America's energy and economic future." What’s rare is the agency that can actually underwrite a deal, manage a portfolio, and recover value when things go wrong. The moon mission worked not just because NASA had a mission — it worked because they built an institution capable of executing it.

The good news? The IRA proved that when you design the right financial tools — transferable tax credits, loan guarantees, direct pay — the private sector shows up. We were private sector led, government enabled. The government didn’t tell folks what to do, we recognized folks that were able to get 90% of the way and provided them the last 10%. We need government to be a sophisticated enough counterparty that the private sector finds real value in a partnership.

That’s the real state capacity question for AI, for clean energy, for housing: not just what’s the mission, but do we have the institutional muscle to be a true partner?

My answer is: America has been explaining why we can’t do big things for far too long. We can build it. We’ve done it before. But it requires treating government expertise with respect and learning how to partner in ways that can accomplish big things. SpaceX and Tesla both required close government partnership. The next 100 companies will need the same.

The US energy debate has a favorite pastime: picking sides. Renewables versus gas. Nuclear versus everything. Data centers as climate villains or progress heroes. Politicians and pundits trade talking points while the actual grid quietly ages into inefficiency.

Here is what almost nobody is talking about. The United States operates roughly 600 gigawatts of natural gas capacity. More than 100 GW of that is over 30 years old. The best modern combined-cycle plants achieve heat rates below 7,000 Btu/kWh. Older plants routinely exceed 10,000 — burning 30 to 40 percent more gas per megawatt-hour. Every hour they run, they waste fuel and emit unnecessary pollution not because gas is inherently dirty, but because the hardware is obsolete. In 2025, the normal retirement pace slowed sharply as aging plants were held together as a bridge for AI data centers. That bridge is weaker than it looks.

The scale of what is being built makes the inefficiency consequential. Global AI data center capacity reached approximately 30 gigawatts by the end of 2025 — comparable to peak power demand in New York State. That figure is growing at 3.4 times per year, doubling every seven months. Epoch AI and EPRI project US AI capacity will exceed 50 GW by 2030, approaching five percent of the country’s entire generation capacity. The EIA confirms that increased generation for data centers comes primarily from existing gas plants, not new ones. AI demand is being met by aging baseload plants burning fuel inefficiently, backstopped by peakers that rarely run.

Not all of this compute needs gigawatt-scale concentration. Roughly ten percent — the final frontier training runs that require thousands of processors in extreme synchrony — genuinely cannot be split across sites without capability loss. Amazon’s Project Rainier in New Carlisle, Indiana, the first operational gigawatt-scale AI campus, came online in late 2025 at 1.1 gigawatts and $35 billion in capital cost. OpenAI’s Stargate spans seven US sites totaling more than 9 gigawatts of planned capacity. Epoch AI and EPRI project the largest individual training runs will reach 4 to 16 gigawatts by 2030. These sites are non-negotiable for frontier capability and are already being built. But we don’t really need that many of them.

The remaining sixty-five to seventy-five percent of AI compute — inference, fine-tuning, experiments, and synthetic data generation — can run in facilities well below 100 megawatts, distributed wherever power is cheapest. This is where the aging fleet problem bites hardest. That distributable majority will go to the cheapest available power, which today means aging baseload plants burning fuel at rates the industry should have left behind a decade ago. Fixing the efficiency of the distributed two-thirds of AI compute is the problem grid policy can actually influence. Latency is largely irrelevant for LLM inference — generation takes seconds, not milliseconds — so capital is already voting accordingly: FERC’s 2025 data shows MISO growing at 43 percent annually, ERCOT and the Southeast close behind, and Northern Virginia slowing. Energy cost beats proximity every time.

The hardware transition makes the efficiency gap worse before it gets better. Racks averaged under 8 kilowatts in 2024; current frontier systems require more than 100 kilowatts, with the next generation pushing significantly higher. This is a rip-and-replace cycle, not an upgrade. Most existing colocation facilities cannot deliver more than 25 kilowatts per rack without reconstruction, and air-cooling caps out around 30 kilowatts — everything beyond requires liquid cooling and a specialist labor pool approximately one-hundredth the size of the air-cooling workforce. The Stargate buildout’s rational response — using the grid most of the time and have on-site gas microgrids — solves a timing problem and provides crucial backup to the grid at times of stress.

Two underappreciated tools can change this picture. Tyler Norris of Duke’s Nicholas Institute found that if data centers curtailed compute for just 100 hours per year — less than one percent of annual hours — the existing US grid could absorb up to 100 gigawatts of new demand without any new generation capacity. Curtailment events average two hours each, well within battery storage capability to bridge. AI training is batchable; inference can shift nodes; Google has already agreed to implement demand response in Indiana and Tennessee. Separately, co-located generation is already winning on economics: Amazon paid $87 million for land in Oregon permitted for 1.2 gigawatts of solar and 7.2 gigawatt-hours of storage next to an existing data center; Google acquired Intersect Power for $4.75 billion, a developer whose entire model is co-locating data centers with generation and storage. Battery costs have fallen 45 percent in two years to below $80 per kilowatt-hour, and the solar, wind, and battery industry has safe-harbored over 200 GW of additional generation coming online by 2030. Power generation stocks have returned 289 percent over two years against 25 percent for the S&P 500. Capital understood the bottleneck before policy did.

The opportunity is concrete. Replace the worst 100 gigawatts of inefficient aging gas with modern combined-cycle plants. Unlock stranded grid headroom through modest demand flexibility. Site the distributable sixty-five percent of AI compute on cheap co-located solar generation in the Midwest, Texas, and the Southeast — 250 GW of safe harbored solar and wind already. The power problem is real but solvable: Epoch AI finds power bottlenecks add months, not years, to build-out timelines, because chips cost ten times more than electricity and companies willing to spend on chips will pay a premium for power. The question is not whether the US finds enough electricity for AI. It will. The question is whether it builds that capacity wisely or patches aging plants together for another decade and calls it a strategy.

Data, not dogma.

Subscribe now

The Original Framework

Rosner’s central argument was deceptively simple: when a large load connects directly to a co-located generator at the same site, their combined impact on the transmission system is far lower than if each were planned separately. Physics, not policy, drives the savings. His December 2025 PJM concurrence directed the grid operator to create two new transmission services — Firm and Non-Firm Contract Demand — letting co-located loads like data centers pay based on actual net grid withdrawals rather than gross capacity. In exchange, they accept the risk of curtailment if they exceed what they’ve contracted for. The math works cleanly, and the prose is unusually accessible for a FERC filing. Rosner wrote what amounted to a plain-language explainer because he understood the policy would only stick if people understood it.

In January 2026, he followed with a concurrence on SPP’s voluntarily-filed High Impact Large Load framework, which pushed the same logic even further. SPP’s Load Limited Resource Interconnection Service caps a generator’s grid output to the forecasted hourly demand of its co-located load, hour by hour — meaning the two are matched in real time, minimizing transmission system impacts and cutting interconnection timelines by more than half. Rosner used the occasion to send a direct message to every other grid operator in the country: SPP did this voluntarily, through a Section 205 filing, without waiting to be told. Others should follow.

The Questions the Framework Leaves Open

The elegance of Rosner’s framework rests on an assumption worth scrutinizing: that the co-located generator actually runs consistently. Nothing in either concurrence requires it to operate as baseload or near-baseload. A generator running only 200 to 500 hours per year — essentially in a peaking or backup role — could satisfy the regulatory structure while the data center leans on the grid the vast majority of the time. If that pattern becomes common, the transmission cost savings Rosner promises may materialize on paper while the underlying resource adequacy problem quietly worsens. PJM would be planning less capacity on the assumption the co-located generator is serving the load — but if that generator rarely runs, the reserve margin math falls apart.

Subscribe now

This connects to a broader set of arbitrage questions that neither concurrence fully resolves. Under FERC Order 745, demand response resources receive full locational marginal price compensation for curtailing consumption — the same price paid to generators. A sophisticated data center operator could potentially receive reduced transmission charges under the new Non-Firm Contract Demand service, on the basis that its co-located generator nominally covers most of its load, while simultaneously receiving Order 745 demand response payments when it curtails what is technically grid load. The generator running minimally justifies the reduced transmission charges; the curtailment of notional grid load generates the demand response payment. Neither regulator sees the full picture.

Batteries add another layer. A co-located battery storage system in Illinois could simultaneously receive cost-of-service recovery through an ICC-approved distribution tariff, participate in PJM wholesale markets under Orders 841 and 2222 for capacity, energy, and ancillary services, and qualify for compensation under Illinois’ new virtual power plant framework in the Clean and Reliable Grid Act. No single regulator currently has full visibility into the aggregate compensation picture across all three tiers. This is precisely the kind of regulatory seam that sophisticated developers exploit before it gets closed.

The Market Overhaul Changes the Calculus

Rosner’s framework was designed in part to relieve pressure on PJM’s capacity market by reducing how much new generation PJM needs to procure for data centers. But PJM’s white paper released May 6, 2026 makes clear that the capacity market itself is in structural crisis regardless. Capacity prices have spiked over 1,000% across the last two auctions, driven primarily by data center demand. PJM has warned of an electricity shortfall as early as 2027. And the cycle of price spikes triggering political intervention — governors capping capacity prices, emergency backstop procurements, ad hoc settlements — has created what PJM CEO David Mills calls a “credibility trap.” Prices must rise to attract new supply, but those same price spikes trigger interventions that undermine investor confidence. As a result, investment is delayed and the shortage deepens.

PJM’s three proposed reform pathways — stabilizing the existing capacity market with better hedging tools, rationing reliability across customer classes, or shifting toward an energy-and-ancillary-services model — each intersect with Rosner’s framework in ways that remain unreconciled. The reliability rationing pathway would formalize at market scale exactly what Rosner’s Non-Firm Contract Demand service already contemplates for individual co-location arrangements: some customers accepting curtailment risk in exchange for lower costs. But doing this systematically, across 67 million customers in 13 states, is a fundamentally different political and regulatory undertaking than doing it for individual data center agreements.

Compounding this, American Electric Power — one of the utilities that warned most loudly about co-location cost-shifting, estimating up to $140 million in annual transmission cost shifts to other ratepayers — is now considering pulling out of PJM altogether. Its CEO has said the stakeholder process gives him no confidence that these issues will be resolved anytime soon. If AEP exits, the cost allocation assumptions underlying Rosner’s framework shift materially, because the remaining load-serving entities absorb a larger share of fixed transmission costs.

The State Layer Is Moving Fast — But Unevenly

At the state level, the large load tariff movement has exploded in scale. As of March 2026, there are 77 large load tariffs pending or in place across 36 states, with 29 approved in 2025 alone — nearly double the total approved in the prior six years combined. The ComEd and PECO model of filing Transmission Security Agreements directly at FERC to lock in data centers’ contributions to the transmission revenue requirement is being replicated by Dominion in Virginia, Georgia Power, PPL in Pennsylvania, and Xcel across four states.

But most of what’s proliferating remains retail-level protection only — minimum contract terms, upfront deposits, exit fees — without the FERC-filed transmission layer that gives the Illinois and Pennsylvania structures their teeth. Transmission costs are allocated through complex regional mechanisms that even well-designed state tariffs cannot fully capture. Commissioner Chang has explicitly warned that FERC lacks a framework for assessing whether these bilateral agreements adequately protect customers, and has called on the Commission to act proactively before the patchwork calcifies into something permanent and unworkable.

The Bottom Line

Rosner’s concurrences sketched a coherent and accessible vision: pair load with generation, let physics reduce transmission costs, make co-located loads pay their fair share, and move fast enough to matter. It remains a sound framework as far as it goes. But PJM’s market overhaul announcement, the unresolved arbitrage questions across demand response and storage compensation, the potential exit of a major transmission owner, and the uneven proliferation of state-level protections all point to the same conclusion: the pieces are not yet adding up to a coherent whole.

Wholesale electricity markets, as PJM’s CEO recently wrote, are extraordinary institutions whose most essential infrastructure is not a price curve or a performance obligation. Generators, utilities, investors, and consumers must all believe, at a basic level, that the rules are fair, stable, and the product of a credible process. Rosner’s framework is a necessary contribution to rebuilding that legitimacy. But it is one piece of a much larger puzzle that no single commissioner, order, or white paper can solve alone.

As Rosner himself wrote in December: our country deserves no less than durable solutions. The harder question, five months on, is whether the institutions capable of delivering them are moving fast enough to matter before the shortfall arrives.

Not that the current grid doesn’t work. It works. It just works for a world that no longer exists. A world where load grew predictably at 1-2% a year, where you could plan a substation five years out and still be on time, where the biggest disruption to a utility’s week was a thunderstorm in July. That world is gone. AI data centers are coming online consuming hundreds of megawatts. EV charging is hitting distribution networks that were designed for refrigerators and televisions. Industrial electrification is accelerating faster than any forecast from three years ago predicted. And the transmission queue? It’s measured in decades, not years.

But here’s what makes this especially frustrating: we’re not even using the grid we have very well. The average transmission asset in this country operates at roughly 40-50% utilization. We have wires in the ground, transformers on pads, substations sitting in fields running at half capacity for most of the year. And simultaneously, we have utilities telling industrial customers they can’t get new service for three years. We have data center developers buying land and threatening to power them off-grid. We have manufacturers who want to electrify their operations and can’t get an expansion of their existing service. The constraint isn’t always physical. A lot of the time, the constraint is that we haven’t built the intelligence layer that lets us use what we already have more effectively.

I’ve spent my career watching smart, well-intentioned people try to solve this problem by doing more of what we already do — building bigger plants, pulling longer lines, filing more interconnection requests. And I understand the instinct. It’s what we know. But the instinct is expensive and in 36 Governor races around the country candidates are hostile to that approach. You cannot build your way out of a structural timing mismatch by building slower infrastructure. And you cannot solve a utilization problem by adding capacity that sits idle half the time. You solve a speed problem with speed. You solve a utilization problem with flexibility.

That’s why I believe what Torus is doing is one of the most important things happening in American energy right now.

In 2025, Torus scaled the world’s first mesh energy infrastructure. Not a pilot. Not a demonstration project. Not a press release. A real, operating system — modular, behind the meter, combining flywheel inertia with advanced battery storage, protection, controls, and intelligent orchestration — deployed across utility, industrial, commercial, and data center environments at genuine scale. By the end of the year, Torus surpassed 2 gigawatt-hours of facility-managed power in operation. Their systems delivered 99.99% uptime across customer sites. They responded to more than 360 unplanned demand response events — not simulated, not in a lab, but real grid stress events where utilities called on customer-sited infrastructure and it showed up.

That last point deserves to be repeated. Three hundred and sixty unplanned demand response events. That means utilities trusted this infrastructure enough to lean on it when the grid was under pressure. That’s not a vendor relationship. That’s a demand flexibility asset being dispatched in real time to keep the system in balance. Every one of those events is a moment where, instead of building a new peaker plant that runs 50 hours a year, the grid leaned on distributed intelligence at the facility level and got what it needed. That is how you improve grid utilization. That is how you defer capital investment. That is how you make the economics of the energy transition actually work for ratepayers.

Subscribe now

I’ve seen a lot of battery storage companies come through in my time at the Department of Energy and in the private sector. Most of them solve one problem pretty well. They either do energy shifting, or they do backup power, or they do demand response. What makes Torus different is that they’ve engineered a system that does all of it — and does it with the flywheel component providing the kind of instantaneous inertia response that pure battery systems simply cannot match. The grid needs inertia. As we retire synchronous generation and add more inverter-based resources, inertia becomes the hidden variable that determines whether a frequency event cascades or gets absorbed. But it’s the combination — flywheel inertia for the instantaneous response, battery storage for the sustained shift, and orchestration software that coordinates both across a portfolio of sites — that gives utilities a demand flexibility resource they can actually count on. Not just for a single facility. For a network of facilities acting in concert, shaping load in ways that improve utilization of every wire and transformer between the meter and the bulk system.

Now here’s the commercial reality, because this is where most clean energy companies stumble. Technology without a manufacturing strategy is just a prototype forever. I’ve watched too many promising companies spend years building beautiful systems in low volumes at high costs, never reaching the price point where the market can actually absorb them. Torus opened GigaOne — a 540,000-square-foot manufacturing and assembly facility near Salt Lake City — with 60% of production equipment already installed and test parts already running. They are not waiting to figure out manufacturing after they win more customers. They’re building the supply chain now so that when the orders come — and they are coming — they can fulfill them at a cost structure that makes the economics work for everyone.

Share

That’s the discipline that separates companies that change industries from companies that get acquired for their patents.

And the utility partnerships are real. Torus expanded to 500 megawatts of contracted demand response capacity in 2025. That is not a memorandum of understanding that lives in a drawer. That is regulated utilities, with fiduciary obligations to their ratepayers, signing contracts that say they will count on this infrastructure as part of how they manage grid reliability and utilization. When a utility signs a 500-megawatt demand flexibility contract with a distributed infrastructure company, it means their engineers and planners looked at the technology and said: this is real, this works, and we need it — not as a backup, but as a core part of how we operate the system. Demand flexibility at that scale changes the planning calculus. It means you can defer a substation upgrade. It means you can absorb a new industrial load without a three-year queue. It means the grid you have today can serve more of the economy than it could yesterday, without a single new mile of wire.

The broader market context matters here too. Every week there’s another announcement of a hyperscaler signing a 500-megawatt or gigawatt power purchase agreement. Every week there’s another industrial company announcing an electrification roadmap. Every week the gap between load growth and grid capacity gets a little wider. But that gap is not just a generation problem or a transmission problem. It is fundamentally a utilization and flexibility problem. We have assets on the system that are underused. We have loads that could shift if given the right signal and the right infrastructure. We have facilities that could be grid assets instead of just grid consumers. Torus is the platform that makes that transformation possible.

I also want to note something that often gets overlooked: supply chain. Torus’ strategic partnership with Ultion is deliberately aimed at strengthening domestic supply chains for energy storage and power electronics. We cannot build energy independence on components that come from a single foreign source. We learned that lesson in semiconductors. We cannot afford to learn it again in grid infrastructure. Companies that are thinking about where their materials come from, and building partnerships that reduce that exposure, are the ones that will still be operating reliably in fifteen years.

Subscribe now

The grid is not going to be fixed by waiting for transmission. It’s not going to be fixed by another round of utility-scale solar with no storage. It’s going to be fixed by putting intelligent, resilient, dispatchable infrastructure at the facility level — at the load — so that the grid has demand flexibility exactly where the demand is, and so that every asset on the system gets used closer to its full potential.

That is what Torus is building. And in 2025, they proved it works.

The energy transition has a utilization problem. The energy transition has a flexibility problem. Torus is the answer to both. Pay attention.

The world installed 647 GW of new solar capacity in 2025. China alone added 315 GW, lifting its cumulative solar past one terawatt. Battery cell prices fell below $80/kWh for the second consecutive year. EV sales hit 21 million units, one in four cars sold globally. China’s NDRC issued a mandate to integrate ~900 GW of distributed solar and wind by 2030. The constraint on the transition is no longer technology cost or manufacturing volume. It is grid integration, storage deployment, and execution speed.

What often gets missed in the China and Europe headlines is how fast the rest of the world is moving. Africa recorded its fastest year of solar growth ever in 2025, with installations jumping 54% to 4.5 GW. Chinese customs data suggests over three times more solar equipment was imported than showed up in official utility-scale figures, the rest going into rooftop systems that are harder to track. South Africa, Nigeria, and Egypt led the continent, but eight countries crossed 100 MW for the first time, double the number from 2024. In South America, Brazil crossed a milestone where solar generation exceeded fossil fuel generation for the first time in 2025, while Chile now generates 38% of its electricity from wind and solar combined. Wood Mackenzie estimates the region will add 160 GW of solar capacity through 2034.

The 2026 picture is dominated by the closure of the Strait of Hormuz.

After disruptions in 2022 and now in 2026, non-OECD countries are rethinking the viability of LNG for power generation. The Middle East supply shock of early 2026 shut in roughly 20% of global LNG supply from Qatar and the UAE. Pakistan had declared in 2022 it would build no new LNG plants, and by early 2026 solar covered 25% of its generation mix, saving an estimated $12 billion in fuel costs over four years. Vietnam suspended new LNG plant approvals as solar panels arrived at prices that made landed LNG economics look silly. China itself is on track for an 11% drop in LNG imports in 2026, has cancelled terminal expansions, and is redirecting capital to domestic renewables and nuclear. This is not a cyclical demand dip. It is demand destruction that compounds. Countries that take LNG out of their power generation forecast do not come back to gas when prices normalize. Memories are long in the infrastructure industry. In this scenario, non-OECD LNG-to-power generation falls 60–90 TWh in 2026.

Coal has already rebounded in 2026, but you are already starting to get air quality complaints. ASEAN coal demand is growing around 5%, China’s coal generation has partially recovered from a lower 2025, India continues expanding thermal capacity, and the US has forced coal retirements to slow until 2030. Globally, coal-fired generation likely adds 40–80 TWh in 2026, reversing some of last year’s progress. But there is a reason to believe this ceiling holds. Once populations in Jakarta, Hanoi, Manila, and Delhi got used to clean air, they can’t go back. Middle-class populations have air quality apps on their phones and governments are facing measurable political costs from PM2.5 readings. Coal’s temporary return is not invisible. It shows up in the sky. And batteries filled by solar is now cheap enough that the alternative is a real choice, not an aspirational one. The biggest move is that China and now India are committed to new nuclear power. The coal rebound peaks in 2026. It will resume its decline in 2027.

The big breakthrough is that batteries and EVs have crossed out of marginal economics. Stationary storage at ~$70/kWh has come down 45% in just the last two years. China’s target of 180 GW of new energy storage by 2027 is funded through market mechanisms rather than mandates. The energy storage has to be placed where it can earn revenue through arbitrage and ancillary services. China learned the lessons that mandates cause sub-optimal placement and reset on the right model. The practical effect is that China’s red-zone grid connection bans — areas in Shandong, Hebei, and Henan that blocked new distributed solar because local networks couldn’t handle the midday voltage surge — begin to lift as storage buffers are deployed. If the 500+ GW of distributed solar can reduce its curtailment in half (2%) you can recover 50–80 TWh of electricity that was previously being wasted. Meanwhile, Chinese EVs are now cheaper than gasoline powered cars. Importantly, around the world utility companies are making sure to operate EV charging as a grid asset by scheduling it when the grid is least in demand or wholesale power is trading at ultralow prices. Smart charging shifts EV load to midday solar surplus windows. Vehicle-to-grid programs are already running at scale in China and the UK, allow EVs to discharge back into networks during evening peaks. The world is on track to deploy 200 million EVs by the end of 2030, representing a distributed storage and demand-flexibility resource that back up the entire grid on a daily basis — grid operators know this will require them to change they way they manage their grids.

Under these assumptions, clean electricity additions in 2026 total about 1,000 TWh. Demand growth, moderated by GDP falling below 2% and slowing industrial load, will probably come in closer to 900 TWh. The fossil fuel stack ends up roughly flat with the coal rebound and the LNG destruction roughly cancelling. The key risk is battery storage execution. The distributed solar will accelerate around the world as countries copy the rapid deployment of Pakistan, Syria, Lebanon, Hungary, and others. Solar panel and battery manufacturing are not the constraint. The question is whether the regulatory plumbing moves at the speed the hardware already permits.

2025 proved low-carbon electricity can power global economic growth. 2026 tests whether execution can match what the installed base already makes possible. The LNG exit in non-OECD markets is structural. The coal rebound is real but bounded by air quality politics and nuclear is scaling in China and India. Grid batteries and EVs are introducing demand flexibility needed to increase the utilization of the grid we have already paid for. The electrons are being generated. The question for 2026 is whether the infrastructure lets them flow.

To understand why this matters, you need to understand what China has already built.

Subscribe now

China ended 2024 with 887 gigawatts of total installed solar — adding 277 gigawatts in a single year, more than twice the entire cumulative solar capacity of the United States. Distributed solar accounted for roughly 38% of new installations, tens of gigawatts embedded on rooftops, factories, and farms across the eastern provinces. By May 2025, China’s total PV capacity crossed one terawatt. The United States is at roughly a tenth of that.

China didn’t issue this grid mandate because distributed solar is a future bet. They issued it because distributed solar is already everywhere, and their grid — built for three decades to move coal power west to east — cannot handle what’s already on it. The curtailment problem is real. Governments in Shandong, Hebei, and Henan have been designating areas as “yellow” and “red” zones restricting new distributed solar connections. Residential solar additions dropped 23% year-over-year in the first quarter of 2024. That’s not a market signal. That’s a grid bottleneck signal.

And here is what doesn’t get said enough: you cannot solve the distributed solar integration problem without solving the battery storage problem. These two things are inseparable. Anyone who tells you otherwise hasn’t spent enough time standing next to a substation at 2pm on a sunny afternoon watching the voltage spike.

The physics are simple. Distributed solar generates power in the middle of the day whether the grid needs it or not. A few hundred kilowatts on one rooftop is manageable. Hundreds of gigawatts across millions of rooftops, all generating simultaneously, flooding the distribution network at the same time every day — that is a stability problem of a completely different order. The grid doesn’t just need to accept that power. It needs somewhere to put it. Local storage at or near the point of generation is the only answer that doesn’t require an entirely new transmission backbone to carry the surplus somewhere else.

China figured this out, tried to mandate it, and ran into the implementation problem I’ve watched play out everywhere. From 2022 onward, provinces required new solar and wind projects to bundle co-located storage at ratios of 5% to 20% of generation capacity. The policy drove real deployment — China ended 2024 with 73.8 gigawatts of new-type energy storage installed, a 130% year-on-year increase and about 40% of global capacity. But it created a quality problem. When storage is a condition of grid connection, developers treat it as a tax, not an asset. They buy the cheapest batteries available. The systems sit underutilized. You end up with gigawatts of storage that isn’t doing the grid flexibility job you needed. China scrapped that mandate in February 2025 for exactly those reasons.

Share

But they didn’t give up on storage. A September 2025 national action plan set a target of 180 gigawatts of new energy storage by 2027 — roughly equivalent to the entire world’s installed battery storage capacity today. This time, the policy emphasis shifted to market mechanisms: energy arbitrage, ancillary services, frequency regulation. The goal is to make storage economically self-justifying rather than administratively required. That’s the right instinct. Mandated storage that owners treat as a cost center will always underperform. Storage that earns revenue by providing real grid services will be operated well and scaled intelligently.

This is what every utility needs to internalize right now. The NDRC grid mandate is not just a solar story. It is a solar-plus-storage story, and the storage piece is load-bearing. You cannot run a distribution network built for one-way flow if you are asking it to absorb gigawatts of intermittent generation from millions of nodes. The math doesn’t work without local storage buffering the imbalances. Every utility planning its distributed solar integration strategy without a corresponding battery deployment strategy is planning to fail, just on a longer timeline.

The good news is that battery prices have collapsed. Stationary storage costs fell to roughly $70 per kilowatt-hour in 2025 — a 45% drop from 2024 alone. The economics that made storage feel like an expensive burden two years ago are not the economics of today. At current prices, paired distributed solar and storage is increasingly competitive as a dispatchable asset. That changes the business case entirely.

Subscribe now

China’s grid redesign will only succeed if the storage layer is built correctly alongside it. The same is true for any country serious about distributed solar at scale. The NDRC order is the architecture change. The batteries are the operating system. You cannot run one without the other — and the utilities and regulators that figure that out first will have a structural advantage that compounds for decades. The question for U.S. utilities is whether they treat batteries as an afterthought or as the foundational infrastructure investment that makes everything else work. China is learning that lesson in real time, at terawatt scale. We don’t have to repeat the same mistake. But we need to decide soon.

The starting point is electricity. New solar and wind power now costs at least 50% less than new fossil fuel generation, and in 2025 solar, wind, and nuclear grew fast enough globally to meet all new electricity demand growth without any increase in fossil fuel generation — the first time this has happened outside a pandemic-induced demand collapse. Cities that haven’t yet switched their municipal electricity to 100% renewable via long-term power purchase agreements are leaving money on the table. Melbourne has already done it for all municipal operations, locking in below-market rates while cutting emissions. The C40 network’s collective purchasing power means that cities acting together can secure 1,000 GWs of clean electricity with better terms, and mandate cost-effective solar installations on rooftops, parking lots, and infrastructure as a condition of planning permission. France’s national parking lot solar law could unlock over 8 gigawatts of solar deployment on parking lots that were otherwise doing nothing.

Subscribe now

Buildings are the next frontier, and heat pumps are the critical technology. Space and water heating account for roughly 45% of energy demand in buildings, and the overwhelming majority of that is currently fossil-fueled. Heat pumps running on clean electricity are three to four times more efficient than gas boilers, and in most urban markets they are now cheaper to run, particularly as gas prices become more volatile. The good news is that cities don’t need to force early retirement of functioning equipment to make rapid progress. Commercial HVAC systems last 15 to 20 years, meaning roughly 5 to 7% of the stock turns over naturally every year. From now to 2030, approximately 20 to 28% of all commercial heating equipment will be replaced anyway — no stranded assets, no political controversy, just normal end-of-life turnover. A simple mandate requiring that every replacement unit be a heat pump would, through physics and procurement alone, deliver major CO₂ reductions by 2030. Cities have direct authority to make this happen through building codes: banning fossil fuel heating in new construction, requiring heat pump installation as a condition of major renovation permits, and running neighborhood-level rollout programs that reduce per-unit installation costs by coordinating street-by-street. Amsterdam is already moving 550,000 homes off fossil gas using this model.

Transport is where city regulatory authority is most direct and most underused. Municipal bus, refuse, and maintenance fleets can be mandated to go fully electric immediately — electric buses already have lower total lifetime costs than diesel in most markets, and the guaranteed demand created by city fleet procurement pulls down prices for everyone else. For private fleets, the transition can be accelerated without stranding assets: cities requiring ride-hailing and taxi operators to switch to EVs as a condition of licence renewal create a wave of high-mileage used internal combustion vehicles that flow into second-hand markets in less-regulated cities and regions, clearing the way for electrification in the city without stranded costs. An electric vehicle used for full-time ride-hailing reduces CO₂ emissions by nearly 40 kilograms per day — roughly three times the benefit of electrifying a privately owned vehicle — because commercial drivers cover so many more kilometres. Low-emission zones and zero-emission city cores extend the pressure to private vehicles without requiring direct subsidy of private purchases.

Subscribe now

Virtual power plants represent perhaps the most underutilised tool in the urban energy toolkit. By aggregating distributed energy resources — home and commercial batteries, EV chargers, smart thermostats, heat pumps, and flexible commercial loads — cities can coordinate these assets to behave like a conventional power plant, providing grid services and displacing fossil-fuel peaker plants that run only a few hundred hours a year. The US Department of Energy projects that scaling virtual power plants could address 20% of peak electricity demand nationally and save around $10 billion per year in grid costs. Crucially, they can be deployed in months rather than years. Leaning into demand flexibility can reduce bills for everyone using the grid by 5% for everyone. Cities can drive this by requiring virtual-power-plant-ready smart panels, bidirectional EV charger wiring, and grid-connected battery storage as standard in all new buildings and major renovations — creating a massive enrolled asset base automatically, at near-zero marginal cost, simply by updating procurement specifications.

And then there is the hidden efficiency gain that most cities have barely considered. Inside every commercial HVAC system, biological material — biofilm, mold, organic fouling — accumulates deep inside the heat transfer coils over time, progressively increasing pressure drop, forcing fans to work harder, and quietly degrading heat transfer efficiency in ways that never show up on an energy bill line item. Blue Box Air, a Dallas-based company, has developed a patented enzyme-infused foam that penetrates to any coil depth while the system keeps running — no shutdown, no chemicals, no disruption to occupants — removing biological fouling and restoring coils to near-original specifications. Across more than 45,000 coils treated in over 1,000 buildings, the results are consistent: pressure drop reductions of 44 to 69%. If this technology was rolled out to all HVAC coils that are already cleaned each year we would have energy savings of ~30% depending on how severely fouled the system was before treatment. A commercial real estate portfolio of 455 units across 42 buildings saved more than 8.8 million kilowatt hours annually; a single healthcare facility cut its pressure drop by 55% and saved $50,000 in fan energy alone. For C40 cities, this one measure along could reduce electricity consumption by almost 5%. Crucially, Blue Box also extends equipment life — which keeps systems running efficiently until natural end-of-life replacement, when a heat pump mandate can take effect cleanly, without forcing premature retirement of anything. For C40 cities, it is an immediately deployable, self-financing intervention and a natural complement to the replacement-cycle strategy.

Share

The combination of these levers — clean electricity, heat pumps at natural replacement, EV fleet mandates with used-vehicle pathways, virtual power plants, and Blue Box coil restoration — is more than sufficient to reach the 50% target. Research on cities deploying batteries, heat pumps, and EVs together shows that meeting the fossil fuel reduce targets would save 22 to 37% of energy costs; the transition pays for itself. What cities need is to reduce red tape and focus on deployment. They need coordination, procurement discipline, and the political will to use the authority they already have. The 50% fossil fuel reduction target by 2030 is not an aspiration. So stop over subsidizing and lets just get it done.

This wasn’t irrational. When the telecom boom hit Africa in the 2000s, diesel was the path of least resistance — grids were weak or nonexistent, solar was expensive, and batteries hadn’t matured. The same story played out across Asia. In Indonesia’s vast archipelago of 17,000 islands, extending transmission lines to remote communities was impractical, locking in over 5,800 megawatts of diesel capacity and more than two billion dollars in annual fuel imports. In Bangladesh, operators in flood-prone rural districts faced exactly the same calculus. Operators built what worked, and what worked was diesel.

Subscribe now

Once you build a diesel-dependent system, you inherit interlocking dependencies that are hard to escape. Converting a tower to solar requires significant capital — panels, inverters, battery storage, integration systems. For operators carrying heavy debt loads and serving thin-margin rural markets, that upfront number was a blocker even when the long-term math was favorable. Capital for renewable projects in Africa has historically run to 25% interest rates versus under 10% in advanced economies, making projects that pencil out in Europe or the US simply unviable on the continent. Standardized financing structures didn’t exist — no leases, no energy-as-a-service agreements, no blended finance instruments to transfer conversion risk away from the operator. Smaller operators had no path in at all. Regulatory frameworks governing private energy generation were absent or hostile across much of sub-Saharan Africa, and Indonesia’s state utility monopoly created similar legal friction for private solar developers at tower sites. Fuel subsidies masked diesel’s true cost for years: when Nigeria removed its subsidies in 2023, prices surged 200% in a single year, leaving operators facing a $400 million annual fuel bill. Even chronic diesel theft — which regularly knocked towers offline in northern Nigeria and the DRC, disrupting mobile money transactions and emergency calls — wasn’t enough on its own to break entrenched procurement habits.

Several things shifted simultaneously. Solar panel prices fell more than 80% since my 2012 prediction, making solar the cheapest option in many markets — cheaper than extending the grid. New energy-as-a-service models emerged, letting operators consume solar without owning the infrastructure: a third-party provider installs and maintains the system, charging a fixed fee that replaces unpredictable diesel costs with a stable monthly payment and removes upfront capital risk entirely. In Bangladesh, solar-plus-battery deployments now cut diesel runtime by up to 50%. In Brazil’s Amazon basin — where vast geography and absent grids create the same problem as rural Africa — Alcoa and TIM Brasil brought 4G connectivity to remote communities in Pará entirely through solar-powered Flextower units, with no grid connection required. Indonesia launched a formal de-dieselization program in 2022 targeting 5,200 diesel plants, with estimated savings of two billion dollars annually in avoided fuel imports. Across Latin America more broadly, solar-hybrid deployments are cutting diesel use by up to 70%, with renewable-powered sites projected to grow at over 13% annually through 2030.

Geopolitical shocks removed the last excuses. The Middle East conflict tightened global fuel markets and pushed diesel prices higher across import-dependent economies in Africa and Asia alike. For operators who had long deferred the solar decision, the business case became impossible to ignore. MTN’s South Sudan operations cut fuel spending by 30% after switching to solar. Airtel Africa, working with ENGIE Energy Access, cut diesel use by more than half at sites in Zambia and the DRC. The numbers speak for themselves.

Share

None of this happened by accident. The Rockefeller Foundation, working through the Global Energy Alliance for People and Planet — co-founded with the IKEA Foundation and Bezos Earth Fund in 2021 — has been the most consequential actor in reengineering conditions for solar investment. The strategy was deliberate: use philanthropic capital not to subsidize end projects but to fix the system — regulatory frameworks, financing structures, government delivery capacity — so that private capital can flow at scale. In the past eighteen months, the Foundation and GEAPP scaled their joint Mission 300 commitment to over $100 million, funding technical assistance to government delivery units, updating electricity regulation acts in countries like Zambia, training fellows embedded in African governments to fast-track project pipelines, and advancing electrification efforts across nearly two dozen countries. The IFC’s $45 million blended-finance investment in IPT PowerTech will modernize over 2,200 off-grid tower sites in Ethiopia, Liberia, and Sierra Leone — a financing template now being replicated across the continent.

The significance extends far beyond cheaper phone bills. Nigeria’s telecom regulator now encourages operators to build solar minigrids around tower sites, positioning them as anchor energy clients that also power nearby homes, schools, and businesses. If that vision takes hold, those 500,000 African towers — part of a global fleet of five million — become distributed energy infrastructure for a continent where 600 million people still lack reliable electricity. For a teacher in northern Kenya named Martin Imwatok, it is already personal: “When these towers go off, business and life stop.”

What this global story illustrates is that cheap technology alone does not produce a transition. Solar got cheap. The transition didn’t happen. What was missing was the connective tissue: financing instruments, regulatory frameworks, government capacity, and the risk-absorption that allows private capital to move. That is what institutions like the Rockefeller Foundation have been quietly building. The latest fuel price shocks have demonstrated, with brutal economic clarity, why it was worth building. The sun has always been there. What was missing was everything else.

Subscribe now

What’s your perspective on blended finance models for energy transition in emerging markets? I’d welcome your thoughts in the comments.

Then two things happened almost simultaneously. California diesel hit $7.50 per gallon. And Tesla rolled the first Semi off its new high-volume production line in Sparks, Nevada on April 29, 2026. Now the goal feels within reach.

California has always paid a premium for diesel. But prices have jumped nearly 50% since earlier this year on Middle East supply disruptions. A typical Class 8 diesel truck averaging 6.5 miles per gallon and logging 100,000 miles per year now burns through $115,000 in fuel annually — roughly $1.15 per mile. The Tesla Semi’s energy cost runs 15–25 cents per mile at typical commercial electricity rates, putting annual fuel savings at $75,000–$90,000 per truck. Fleet managers have been running the numbers for years; this time the numbers are running away from diesel.

The sticker price of a Tesla Semi — approximately $300,000 for the 500-mile variant — stops many conversations before they start. A new Freightliner Cascadia diesel runs roughly $175,000. That $115,000 gap is real. But total cost of ownership is not purchase price. It is purchase price plus fuel plus maintenance over the working life of the asset. On that basis, Bernstein Research formally concluded this spring that the Tesla Semi carries a 3% total cost of ownership advantage over the Freightliner Cascadia diesel in today’s fuel environment. Sounds small? In trucking, where full-year operating margins run 2–5%, a 3% structural cost edge doubles your profit.

Subscribe now

The five-year fuel math tells the story plainly. Diesel at $7.50 per gallon costs approximately $577,000 over five years at 100,000 miles per year. The Tesla Semi at $0.20 per mile costs approximately $100,000 — a $477,000 saving before a single wrench is turned. Then add maintenance. The Semi eliminates entire failure categories that consume diesel budgets: turbochargers, EGR valves, DEF systems, oil changes, and transmission rebuilds. Maintenance savings of 30% or more versus diesel add another $15,000–$20,000 annually. Total five-year operating savings run $550,000–$600,000 against a $115,000 purchase price premium. Breakeven arrives in roughly 18–24 months.

What makes the economics especially compelling is who is actually buying Class 8 trucks in California. Not every Class 8 tractor is a long-haul sleeper cab crossing the Rockies. Roughly 35–40% are day cabs — no sleeper berth, because the driver returns to a home base every night. These trucks run local and regional routes: warehouse to retail, distribution center loops, port pickups. Their daily mileage is predictable, their routes are bounded, and they return to a depot every night where charging infrastructure can be installed once and used indefinitely. For this segment, the charging problem that dominates long-haul conversations simply does not exist.

Then add drayage. Approximately 33,500 Class 7 and 8 drayage trucks service California’s seaports and railyards, moving containers from ships and rail yards to warehouses on trips that typically run under 150 miles and are completed within the same day. These trucks make two to five runs daily, return to base each night, and operate on some of the most congested, emissions-burdened corridors in the state. California has set a goal of 100% zero-emission drayage operations by 2035. Given CARB compliance deadlines and $7.50 diesel, many drayage operators are rapidly making a present-tense business decision.

The regulatory target is 30% of annual Class 8 tractor sales in California by 2030 — not 30% of the operating fleet, which would be an entirely different order of magnitude. California sells roughly 10,000–12,000 Class 8 tractors per year, so 30% means approximately 3,600 ZEV tractors annually by 2030. Tesla’s Nevada factory is designed for 50,000 units per year at full capacity. Tesla could meet the California target as early as 2027. Nationwide, approximately 200,000 Class 8 trucks are sold every year. California, with its incentive stack and the highest diesel prices in the continental United States, is precisely the market where demand will concentrate.

The policy architecture supporting this transition is more durable than many expected. California scaled up solar and battery storage — technologies now scaling in Texas and around the country. Fleet operators have taken notice: 90% of the 1,067 ZEV tractor purchase vouchers currently reserved through CARB have been allocated to the Tesla Semi. Traditional OEMs — Daimler, Volvo, Paccar — have largely ceded this moment, pulling back from aggressive California ZEV commitments just as the economics turned favorable. That retreat has handed Tesla a category-defining product, a loyal fleet customer base, and a regulatory environment designed to accelerate exactly the transition Tesla is positioned to lead.

Share

Caveats are required. Tesla’s production ramp is an execution challenge that history gives us reason to watch carefully. The $7.50 diesel price reflects an acute supply shock that could ease. Depot charging infrastructure is buildable but not free. But the structural forces — California’s fuel tax premium, its regulatory framework, its dedication to technology commercialization, and a competitively priced electric tractor with megawatt charging capability — do not depend on $7.50 diesel. The economics work at $5.50. They work at $5.00.

For fleet operators running local and regional California routes — warehouse to retail, port drayage, distribution center loops — the calculation has shifted. They are now figuring out how quickly they can get allocation. California set out in 2020 to demonstrate that the hardest vehicle segment to electrify could be transformed by ambitious regulation, layered fiscal policy, and market competition. With the Semi now in volume production and diesel prices that make the business case self-evident, that demonstration is underway in earnest. The slow start is over. The finish line is closer than it looks.

Subscribe now

Data sources: EIA (California diesel prices), Bernstein Research (TCO analysis), CARB (HVIP voucher data and drayage registry), ICCT (operating cost modeling), Tesla (Semi pricing and production), ACT Research (day cab share).

After Carbon War Room, I wrote Creating Climate Wealth in 2013 and co-founded Generate Capital with Scott Jacobs and Matan Friedman in 2014. Both projects made the same argument: the energy transition isn’t a tech problem — it’s a business model and capital problem. Still my running fight with Bill Gates and Vinod Khosla.

We get into the book tour, the Josh Hawley hearing where I called myself more accessible than a ham sandwich (and then dressed as one for Halloween), the doubters who whispered “vanity project” while Generate was getting off the ground, and why I finally hired a leadership coach.

Part 3 is the LPO years. Send your questions — Jamie was there too so she'll keep me honest.

Hope you enjoy it.

Listen on your favorite podcast player

Over $156 billion in data center projects were blocked or stalled by local opposition in 2025 alone. More than 188 organized opposition groups have formed across 24 states. A nationwide poll found that only 44% of Americans would welcome a data center near where they live — less popular, remarkably, than gas plants. Utilities are seeking $31 billion in rate hikes nationwide. And unlike groceries or housing — where governors can point vaguely at supply chains or global markets and keep walking — electricity is different. Governors regulate the utilities. They approved the deals. They are now on the hook, politically, for what comes next.

North Carolina is the sharpest version of this problem. Duke Energy’s 2025 Integrated Resource Plan projects load across its North Carolina systems growing between 16% and 60% over the next fifteen years — against just 7% growth in the previous two decades. Data centers account for 30% of Duke’s economic development pipeline by project count but 80% of projected energy demand. Meanwhile, residential bills have already risen 22% since 2020, nearly two-thirds of it driven by natural gas costs. Duke is now seeking an additional 13.5 to 13.9 percent rate hike by January 2027, while posting $5 billion in profit and running the largest capital expenditure plan of any utility in America. The utility’s own projections show bills rising another 40 to 70 percent over the next fifteen years.

What made this worse was deliberate actions by Duke. Over the past legislative session, Duke and Republican legislators enacted changes to North Carolina energy law estimated to cost ratepayers $23 billion through 2050. The package stripped the Governor’s appointment majority to the Utilities Commission, allowed Duke to bill customers for gas plants before they generate power, shifted fuel-cost volatility from industrial users to households, and retroactively legalized fuel charges a state appeals court had ruled improperly collected. Then the restructured NCUC paused Duke’s planned procurement of 770-plus megawatts of solar. Paused the solar. Approved the gas cost recovery. Let the rate hike proceed. These are not accidents of administrative procedure. They are choices with identifiable authors and predictable beneficiaries — and the bill is landing on ordinary households.

Governor Stein has responded correctly. His Energy Policy Task Force, created through Executive Order 23, produced an interim report with nine concrete recommendations: large load interruptible tariffs, bring-your-own-capacity options, load flexibility programs, grid-enhancing technologies, demand-side resources. These are the tools other states are already deploying. The question is whether he will fight to use them — and whether he understands the economic argument that makes that fight winnable.

That argument was supplied in March 2026 by the Brattle Group, in an analysis commissioned by GridLab and the Utilize Coalition. The finding: improved grid utilization could save customers of vertically integrated U.S. utilities between $110 billion and $170 billion over the next ten years. From infrastructure already built. And you can improve the number with new wires! The power grid runs at roughly 50% capacity for the majority of the year — a system built for extreme peak loads sitting idle for thousands of hours. The Brattle analysis found that a 10% boost to system utilization translates to a 3.4% decline in customer rates by 2030. In a status quo scenario with the same load growth and no utilization improvement, rates rise 1.4% instead. The utilization-first approach doesn’t just beat the alternative. It inverts the trajectory entirely. The state-level evidence is unambiguous: Texas, Nebraska, and New Mexico grew load by 15% or more from 2019 to 2024 and saw rates decline. California and Hawaii saw declining loads alongside sharp rate increases. States that grew load while managing utilization saw rates fall. States that didn’t, didn’t.

The single highest-leverage policy Governor Stein can pursue flows directly from this analysis: require data centers to co-locate battery storage at the 7,200 MWs of solar already deployed around the state as a condition of grid interconnection. Duke earns its return on equity on every dollar of physical infrastructure, which means the utility profits from overbuilding, not from using what exists. The Brattle report modeled both paths. In the status quo, new load triggers new infrastructure, new customers don’t cover full incremental costs, and other ratepayers fill the gap. In the utilization-first approach, half the new transmission-level load connects without imposing new capacity costs — through self-supply or peak-hour flexibility — while a distributed energy resource portfolio offsets distribution-level demand at $50 per kilowatt-year. The result is a 3.4% rate reduction rather than an increase. Applied to North Carolina specifically, requiring co-located batteries at new data center interconnections could achieve a 5% reduction in electricity rates for all Duke customers. Not a 5% reduction in the rate of increase. An actual reduction. The cost to the data center is less than 1% of facility construction. The most sophisticated operators already have backup generation — they just haven’t been asked to integrate it with the grid.

The community opposition to data centers is not irrational NIMBYism. It is a rational response to a policy failure: utilities are still being allowed to ratchet up investment in ways that transfer risk to ratepayers while rewards flow to shareholders. That failure now has political consequences. In Tucson, residents packed meetings until their city council unanimously walked away from a major data center deal. In Maine, a moratorium paused all new construction — which was vetoed by a Governor desperate to court data centers. These are not fringe outcomes. They are what happens when voters figure out what happened to their bills.

Minnesota shows what the alternative looks like. Google’s data center agreements with Minnesota’s utilities are projected to deliver $1.7 to $1.9 billion in ratepayer savings over fifteen years. That is what it looks like when a technology company is required to serve the grid rather than merely draw from it. It should be the rule, not the exception — and Governor Stein has the authority to make it the rule in North Carolina.

The numbers are now available to make this case with authority: $170 billion in potential savings from better grid utilization, a 3.4% rate decrease from a 10% utilization improvement, a 5% reduction achievable in North Carolina from requiring co-located batteries. Against those numbers, Duke’s request for a 13.5% rate hike looks like exactly what it is: a choice, not a necessity. Governors who move first on this will get credit for solving a cost-of-living crisis. Those who don’t will own it.

The North Carolina Energy Policy Task Force Interim Report was released February 15, 2026. The Brattle Group’s “Untapped Grid” analysis was published March 2026. Duke Energy’s proposed rate increase is pending before the North Carolina Utilities Commission.

The economics are blunt. A single mile of new gas distribution line can cost $500,000 to over $1 million to install, with maintenance costs compounding over decades. Meanwhile, the homes connected to that infrastructure are increasingly running appliances — heat pumps, induction stoves, heat pump water heaters — that don’t need gas at all. For utilities staring down aging pipeline networks and aggressive state climate targets, the arithmetic is starting to point in one direction.

In April 2026, Xcel Energy put that arithmetic to the test with a $10 million Neighborhood Residential Retrofit Program in Denver and Aurora, Colorado — the largest non-pipeline alternative evaluation investment by any utility in the country. Rather than extending gas lines in two target neighborhoods, Xcel is funding whole-home electrification for approximately 2,200 households, then comparing the total cost against what continued gas infrastructure maintenance would have required. A matched control group ensures the comparison is rigorous. The findings could reshape infrastructure investment decisions not just for Xcel, but for utilities nationwide.

Zero Homes, a Denver-based technology company, is operating the program end-to-end — managing homeowner enrollment, home assessments, project design, installation coordination, and rebate processing through a software platform validated by the U.S. Department of Energy. The company brings prior experience running Green Homes Chicago, a large-scale city-led electrification initiative alongside the City of Chicago and ComEd, giving it a track record in managing complex utility programs at neighborhood scale. For income-qualified homeowners in Denver’s Valverde neighborhood, the program covers 100% of project costs. Market-rate homeowners in Aurora can access up to $20,000 in incentives, stackable with Colorado state rebates for total savings of up to $22,500 per home.

Retrofit programs like this are compelling, but the economics get even clearer in new construction. Every gas line buried under a new subdivision is a cost recovered over decades from customers and a stranded asset risk if electrification accelerates. Utilities and municipal planners in states from California to Colorado to New York are beginning to design new developments as all-electric from the ground up — the gas utility never connects at all. Developers install cold climate heat pumps, heat pump water heaters, induction-ready panels, and EV charging infrastructure in every home. The utility builds out electrical distribution capacity instead of gas infrastructure, often at comparable or lower upfront cost with no ongoing maintenance liability.

The technology stack needed to support this transition is no longer a limiting factor. Cold climate heat pumps from Mitsubishi, Bosch, and Carrier now operate efficiently to -15°F, with broad installer networks nationwide. Heat pump water heaters from Rheem and AO Smith are 2 to 4 times more efficient than gas units and available through standard distribution channels. Smart thermostats from Ecobee and Google Nest integrate with utility dispatch systems, allowing heat pumps and water heaters to shift load to off-peak hours automatically — reducing grid strain and customer bills simultaneously. Smart panels from Span.io enable whole-home energy management and EV charging priority from a single interface. Home battery systems from Tesla, Sunnova, and Enphase allow households to store grid or solar energy, reducing peak demand charges and providing resilience during outages. And when these batteries and smart appliances are aggregated across a neighborhood, they form what the industry calls a virtual power plant — a dispatchable grid resource that utilities like Sunrun and Tesla Energy are already operating at scale today. AI-powered home energy modeling platforms further optimize where program dollars are spent, prioritizing upgrades that deliver the highest savings per household.

The longer arc of this transition is ambitious. An all-electric neighborhood of 500 homes with heat pump water heaters, home batteries, and smart thermostats represents meaningful dispatchable capacity if the utility can orchestrate it. Rather than passive consumers of electricity, these homes become active participants in a grid that increasingly needs flexibility more than it needs new generation capacity. The economics of gas infrastructure avoidance are just the beginning of the value case.

For the utility industry watching Xcel’s Colorado experiment, the data produced over the coming years may prove more valuable than the $10 million investment itself. Gas distribution networks represent hundreds of billions of dollars in utility assets across the country, many of them aging and facing rising maintenance costs. A credible, data-backed pathway to avoid rebuilding those networks — while meeting state climate mandates — would transform how utilities allocate capital for decades to come. The study in Denver and Aurora is designed to produce exactly that evidence. The infrastructure question those 2,200 Colorado homes will help answer is one the entire country is waiting on.

This isn’t really a story about five batteries. It’s a story about what happens when batteries become so cheap that the economics of grid infrastructure flip — and what that means for the hundreds of distribution utilities, cooperatives, and municipal systems across PJM that are about to face the same math.

For decades, electricity economics in PJM have revolved around a handful of critical hours. Coincident peak pricing — where a utility’s transmission costs for the entire year are determined largely by how much load it carries during the five highest-demand hours — is the mechanism that makes grid infrastructure so expensive to serve at the margins. Build enough capacity to cover those hours, and you pay for it all year. For small distribution utilities, that dynamic is brutal. They can’t negotiate transmission tariffs. They simply absorb costs that grow every year as data centers and electrification push demand higher.

That’s what makes the Lightshift deployment so interesting. A transmission-connected storage project can cost tens of millions of dollars and take several years to clear PJM’s interconnection queue. A distribution-connected 5 MW battery can be deployed this year, at a fraction of the cost, and begin shaving coincident peak exposure immediately. “The number the transmission utility has to solve for, then, is lower,” Lightshift’s chief commercial officer told Utility Dive. That one sentence describes a tectonic shift, not a project.

Now layer in the policy context, energy storage is being mandated across the PJM.

Virginia, Maryland, Illinois, and New Jersey have all enacted legally binding storage mandates with deadlines converging on 2030, and they’re all operating inside the same PJM footprint, under the same capacity pricing pressure, served by the same overburdened interconnection queue. Virginia’s Clean Economy Act requires Dominion to deploy 1.2 GW of storage by end of 2030. Governor Spanberger just signed legislation targeting 20.78 GW long-term, with a new FAST Act allowing batteries to co-locate at existing solar sites and new siting rules eliminating local permitting hurdles for those projects. Maryland set a path to 1.5 GW by 2030 and 3 GW by 2033, then in 2025 added a competitive procurement for up to 1,600 MW of transmission storage and a direct mandate for 150 MW of distribution-connected projects — nearly identical in architecture to what Lightshift is building in Virginia. PJM had described Maryland’s resource adequacy situation as “dire.” New Jersey is targeting 2 GW by 2030, with competitive auctions already underway for the first gigawatt.

Then there’s Illinois. PJM’s largest state by load just enacted the most aggressive near-term battery mandate in the entire region. Governor Pritzker signed the Clean and Reliable Grid Affordability Act in January 2026, directing the Illinois Power Agency to procure 3 GW of grid-scale storage by 2030, financed through 20-year contracts designed to give developers the revenue certainty to move fast. The Illinois Power Agency projects the law will save consumers $13.4 billion over two decades by suppressing capacity market prices. The timing is no coincidence: Illinois capacity costs had risen 22% year-over-year, and roughly 3 GW of fossil generation is headed for retirement by the same deadline.

Add it up: over 7 GW of mandated storage deployment by 2030, across four states, in one market. At current costs, that’s a procurement wave well north of $10 billion. And PJM itself has been told it needs at least 16 GW of storage by 2032 just to maintain reliability (PJM only has 400MW operating now). The gap between where the grid is and where state mandates require it to be is the investment opportunity.

The problem is the queue. PJM’s interconnection process has made transmission connected batteries wait five years. A new resource entering the queue today wouldn’t receive an interconnection agreement until 2029 at best. PJM needs capacity added faster than that. Batteries sited on the distribution grid can bypass the queue entirely.

Which brings us back to what Lightshift is actually building. The company calls its model “VPP+” — aggregating distribution-connected batteries across multiple utilities into portfolios that can reach transmission-scale, while deploying at behind-the-meter speed. They’ve already executed this playbook with 14 projects through Massachusetts Municipal Wholesale Electric in ISO New England. Virginia is the PJM proof of concept. The math their CCO offered is worth sitting with: in some geographies, ten 30 MW distribution-connected projects make more sense than one 300 MW transmission facility. Same aggregate capacity, a fraction of the interconnection risk, a timeline measured in months rather than years. And the Blue Ridge Power Agency model — a nonprofit wholesaler coordinating procurement across multiple small members and standardizing deal terms — shows how to compress transaction costs enough to make 5 MW projects economically viable. That’s a replicable blueprint for any public power wholesaler sitting inside a mandate state with a congested queue.

The deeper disruption is what all of this does to the grid’s existing cost logic. Transmission charges in PJM are allocated heavily on coincident peak demand. If a small battery shaves a utility’s load during just a few of those hours, the savings are earned all year long. The battery doesn’t need to run year-round. It just needs to show up at the right moments. That asymmetry is what makes distribution-connected storage so compelling, and what makes it threatening to the incumbents who profit from current pricing constructs. As more utilities opt out of coincident peak exposure, the remaining unhedged load bears more of the cost. The collective action problem — everyone benefits, no one wants to move first — gets resolved by mandates, falling battery costs, and projects exactly like this one.

The Virginia cooperatives that moved this week are small. Their loads are rural. Their peak exposures are modest by PJM standards. And yet they moved first, with a deployment architecture that the rest of the region’s mandated buildout is going to need.

This is James’s fourth time on Energy Empire. Each time it gets darker. This time he walked us through why the initial shock has been absorbed but the real pain is still in the pipeline, why energy independence gave the US, Israel, Russia, and China each a permission structure to act more aggressively on the world stage, and why Europe — which didn’t want this war — is now building a post-war security architecture for the Strait without the country that started it.

On the ground, 70% of American farmers can’t afford fertilizer. Europe has six weeks of jet fuel left. The US can’t deliver weapons its NATO allies already paid for. And James makes the case that this war may bring us to peak oil demand — not because of any climate agreement, but because the world is making sure this never happens again.

My favorite quote from the episode: “You can say from London or Paris or Berlin or Brussels, I love America. But right now, America has to deal with America. So Europe has to hold America’s beer for a while.”

We’re dropping this on the day the ceasefire expires and the day before Earth Day. I got the Earth Day award two years ago. James thinks Trump deserves it this year.

Listen on your favorite podcast player

Over $156 billion in data center projects were blocked or stalled by local opposition in 2025 alone. More than 188 organized opposition groups have formed across 24 states. In Maine, a first-in-the-nation moratorium paused all new data center construction. In Tucson, residents packed meeting after meeting until their council unanimously walked away from a major deal. A nationwide poll found that only 44% of Americans would welcome a data center near where they live — less popular than gas plants.

This backlash is bipartisan, and is dominated by impact on electricity bills. Utilities are seeking $31 billion in rate hikes nationwide, energy costs were already a ballot issue last fall, and unlike groceries or housing, most governors can’t blame “the market” and move on. They regulate the utilities. They approved the deals. They are now on the hook — politically — for what comes next. Deploy Action showed that the tools already exist. Governors are choosing not to use them.

Utility demand forecasts have historically overshot actual electricity growth by 8% over five years and 23% over ten. The current forecasting frenzy is driven largely by speculative data center pipeline — land purchases and letters of intent, not operational facilities. In PJM alone, consumers are already on the hook for $16.6 billion in capacity costs tied to forecasts that industry experts widely expect to be wrong. This rhymes uncomfortably with the 1999–2002 IT buildout that left hundreds of power plants stranded and investors burned.

Subscribe now

At the same time, some meaningful new load is coming — and the question isn’t whether to serve it, but how to serve it without socializing the risk onto ordinary ratepayers. The chip constraint makes this more urgent, not less. NVIDIA wait times have stretched to 2028; the high-speed memory these chips require is booked through 2027. Utilities are being asked to build billions in new generation to serve data centers that cannot say, with any precision, when they will actually come online. Ratepayers are funding a speculative bet stacked on another speculative bet. Better public data on actual delivery timelines — not just investment announcements — is the minimum the public is owed.

The solutions are not theoretical. They are deployed, proven, and cheaper than the alternative.

Data centers should sign interruptible service agreements — like PG&E’s T-Flex — as a condition of grid interconnection. Texas has mandated the same. In exchange for priority access and faster connection timelines, the data centers accept curtailment during peak grid stress. This doesn’t harm operations: the most sophisticated data center operators already have this flexibility built in. Many have put in back-up generators, they just haven’t been asked to use it.

Data centers can also invest in distributed batteries, smart water heaters, and managed EVs in the homes and businesses on their local circuit — turning their neighbors from opponents into beneficiaries. The cost is less than 1% of building a new facility. Virtual power plants assembled from these assets can shift consumption to off-peak hours within months, without building anything new. Peak demand falls, bills fall with it, and community opposition loses its sharpest edge.

Governors have to force utilities to plan smarter. New software can break through data silos and complete grid analysis in days instead of months — stress-testing proposals against actual project commitments rather than speculative land purchases, before a single ratepayer dollar is committed. Batteries were the single largest source of new US generating capacity in 2025; they deploy fast and scale modularly to match data center build cycles. And the PJM model of socializing large loads’ capacity costs onto all ratepayers should be named for what it is: a subsidy to large customers at the public’s expense. PG&E estimates that spreading fixed infrastructure costs across a broader, more flexible load base — rather than building new infrastructure to serve each new megawatt — could reduce electricity bills by 20% by 2030. That is what basic economic discipline looks like in a sector that has avoided it for too long.

The model is already working in practice. In Minnesota, Google’s data center projects in both the north and south of the state have filed Energy Savings Agreements with the Public Utilities Commission. Together, they are projected to deliver $1.7 to $1.9 billion in ratepayer savings over fifteen years — $1.1 billion through Xcel Energy and between $600 million and $800 million through Minnesota Power. This is what it looks like when a major technology company offers to serve the grid rather than just draw from it, and when regulators hold the line. It should be the rule, not the exception.

Subscribe now

Michigan offers a cautionary example of what happens when utilities resist these tools. State regulators recently issued Consumers Energy a formal prospective warning for failing to model virtual power plants — aggregations of home batteries, electric vehicle batteries, and smart thermostats that can collectively reduce grid stress and displace the need for new gas generation. The technology is not experimental: utilities in other states are already running VPP programs that compensate customers directly, cutting electricity costs while adding flexible capacity to the grid. The commission found that these approaches can be 90% cheaper than business as usual. The Michigan Public Service Commission has made the stakes clear: model VPPs in your long-range plans, or risk having rate increases denied. That is the kind of regulatory pressure that moves utilities. It should not take a formal warning to get there.

The public opposition to data centers is not irrational NIMBYism. It is a rational response to a policy failure: governors were unaware as electric utility companies ratcheted up investment that transferred risk to ratepayers while letting rewards flow elsewhere. That failure now has political consequences, and the fix is straightforward. Mandate interruptible tariffs for all large new loads as a condition of interconnection. Require utilities to deploy demand flexibility investments while carefully planning new generation. End the socialization of speculative capacity costs onto residential ratepayers. None of this requires new technology or more time. Governors who move first will get credit for solving a cost-of-living crisis. Those who don’t will own it.

I had invited him to give a lunch talk at a Generate Capital gathering in San Antonio. He showed up — then a 20-year WSJ energy reporter, two-time Pulitzer finalist — and told a room of solar developers that the “small is beautiful” era was over, that the fossil fuel industry was coming at us bare-knuckled, and that the clean energy industry had brought a fan to a gunfight. He kept thinking about that talk. Years later, he quit journalism to go work for a solar company.

On this week’s Energy Empire, we talk about how Russell broke the Deepwater Horizon story and exposed PG&E’s role in the Camp Fire, what he thinks the clean energy industry keeps getting wrong about itself, what we can learn from Landman, and why — even at a solar company — he goes weeks at work without ever mentioning carbon.

Listen to the full episode

For years, we’ve asked the wrong question about Electric Vehicle adoption: How do we persuade people to buy them? The assumption was that consumers needed to be subsidized—through tax credits or mandates.

But that’s never been the real driver.

The real driver is much simpler: deliver a better car for less money.

When fuel prices spike—driven today by instability around the Strait of Hormuz—families don’t suddenly become climate activists. They become cost-conscious. The question shifts from “Do I like EVs?” to “Why am I spending hundreds of dollars a month just to drive to work?”

That shift is what actually moves markets. It’s practical electrification.

When gasoline prices rise, consumers don’t optimize for making a statement—they optimize for savings and convenience. And that means the first winners won’t be luxury electric vehicles. They’ll be the practical end of the market:

-Used hybrids tighten first.
-Used EV inventory follows.
-New hybrids and plug-in hybrids surge.
-Affordable EVs with solid range and manageable payments gain share.

In other words, people don’t switch categories—they get a higher mileage car or switch fuels.

And the market is already moving. Shopping activity for hybrids and EVs is rising. Used EV inventory is tightening. Vehicles that reduce fuel costs are outperforming those that don’t. Not because consumers changed their beliefs—but because they’re responding rationally to higher gasoline expenses.

Even the industry’s biggest players are being pulled in that direction. Tesla has regained the global lead in pure battery-electric sales from BYD. The next wave of more affordable vehicles—like Rivian’s R2—matter more than any high-end model at a time when folks are looking to switch.

At the same time, a critical barrier is quietly disappearing. The U.S. charging network is scaling rapidly, with more than 192,000 public charging ports nationwide and roughly 1,000 new ones installed each week. The constraint is no longer “Can I charge?” It’s increasingly “Is everyone putting in a charging station?”

That’s why high gasoline prices are such a powerful accelerant. Tax credits and incentives don’t create durable demand. A monthly fuel bill does. Once consumers experience lower operating costs, that preference tends to stick.

To be clear, price spikes alone don’t create energy transitions. They reveal when a better alternative is already viable. And today, for millions of drivers—especially commuters, two-car households, and fleet operators—that alternative is ready.

This is also why the conversation shouldn’t be limited to fully electric vehicles. If gasoline remains expensive, consumers will move toward anything that reduces costs including more Ethanol in their fuel. The administration announced an increase to E15 for everyone and more subsidies for E85 stations in their first term.

In 2008, oil magnate T. Boone Pickens struck a chord with the country when he presented “The Pickens Plan” that called for investing in clean electricity and switching from oil to natural gas for heavy trucks. Tesla is finally shipping Semis at scale this year.

If the United States matched European fuel economy standards, the impact would be massive because U.S. vehicles are roughly 30% less efficient, we’re effectively burning 30–40 billion extra gallons of gasoline each year—waste that translates directly into higher costs for households. At $5 per gallon, that’s $150–200 billion annually flowing out of Americans’ pockets, alongside a reduction of roughly 2–3 million barrels of oil demand per day and 300–350 million metric tons of CO₂ emissions. Clearly the technology exists in Europe and China—we already know how to build more efficient vehicles.

What’s happening now is bigger than a product shift. People are starting to realize that we are going to have high prices for as long as oil markets are disrupted, which is now looking like years.

Gasoline ties household budgets to global volatility—geopolitics, refinery outages, and supply shocks. Electrification, by contrast, shifts transportation onto a more stable, increasingly domestic energy system. Every mile driven on electricity is a step away from that volatility.

We often talk about electrification as a climate strategy. And it is. But in moments like this, it’s also a consumer protection strategy, an energy security strategy, and a resilience strategy.

With the global fuel disruptions, electric vehicles are sold out in countries across Southeast Asia and many are now predicting over 30 million EVs will by sold in 2027, with growth led in India, Indonesia, Vietnam, Philippines, Brazil, and other emerging markets with heavy oil imports. High fuel prices are pushing Chinese EV exports into something more durable: mass adoption driven by affordability.

Because that’s how energy transitions actually happen.

They don’t begin with belief, policy white papers, or moral persuasion—they begin when the price sign at the gas station turns into a daily, unavoidable shock and the economics hit you in the face.

Vishal Kapadia lived this from the other side. As Head of Energy for Walmart, he watched non-weather-related outages go up 20% in four years. He was buying diesel generators and building microgrids just to keep the lights on across thousands of stores. Then he left to run LineVision, where his sensors unlock 30 to 40% more capacity on existing transmission lines.

So why aren't we deploying this everywhere? On this week's Energy Empire, we talk about the slide rule problem, why Chris Wright is suddenly talking about grid utilization, Virginia's first-in-the-nation grid utilization bill, and what I call the aspirin problem — everyone is in pain, the fix is sitting right here, and somehow we still need a five-year study before anyone will take it.

Listen now

A third of the world’s helium passes through the Strait of Hormuz. Your kid’s birthday balloons. The fabs that make our semiconductors. Gone. I didn’t see that one coming either.

James Gutman came back on Energy Empire — more somber than I’ve ever seen him. We got into the full picture: the humanitarian crisis on the ground, how alliances are being redrawn, what happens to the petrodollar, China’s clean tech play, and the one question nobody in Washington wants to answer: how do you reopen the Strait without boots on the ground?

Arnab Pal filled in for Jamie. He was great.

Listen now

This is not a “glass half full / glass half empty” situation. If 11–12 million barrels per day of liquids supply is offline, that supply is offline. That is not a matter of interpretation. It is a physical reality.

Once you accept that, the world has only two options: draw inventories or destroy demand. There is no third option. And while many people point to “global inventories” as a source of comfort, not every barrel in storage is truly available. A meaningful share of observable stocks are operational minimums — pipeline fill, refinery working inventory, tank bottoms, and strategic reserves you do not want to breach unless you are in a true emergency.

So the real question is not the headline number for barrels in storage. The real question is: how many usable barrels exist before the system starts to fail? That is where this stops being an abstract market conversation and becomes something much more serious.

A prolonged Hormuz closure is not mainly about whether wealthy countries can tolerate higher prices. It is about how long the global system can keep functioning before shortages begin to appear.

And when that happens, the pain will not be distributed evenly. North America will be relatively insulated. Europe will pay up. Higher-income buyers in Asia will scramble and survive. But many lower-income countries do not have that option. They do not “adjust” in some elegant macroeconomic sense. They simply lose access to fuel — diesel for transport, LPG for households, LNG for power, and fuel oil for industry.

That is why this should not be framed as just another inflation story or a clever debate about market resilience.

It is a humanitarian story.

People in wealthy countries may experience this as higher prices. People in poorer countries may experience it as not having enough energy at all.

A prolonged Hormuz closure is, in plain terms, a global fuel rationing event. And when fuel gets rationed, poor people lose access first — to hospitals, fertilizer, cooking fuels, water pumping, lighting, transportation, refrigeration, and basic economic life.

We should talk about it that way.