Batteries Got Safer; Permits Might Get Faster

The finding

EPRI puts the annual failure rate of stationary storage systems at 0.04% to 0.14%. The report compares this with the roughly 0.1% annual failure rate of oil-filled transformers, which every utility already runs in large numbers and treats as a normal, manageable risk. The article does not say exactly what one "failure" is for this rate, for example whether it is counted per site or per unit.

Why it is getting better

Lakshmi Srinivasan, a senior technical leader in EPRI's energy storage program and a principal investigator on the report, told Utility Dive that systems built in the past five years appear significantly more reliable than earlier ones. "There are things we know in 2026 that were not necessarily understood in 2019 or 2020, when the older systems were being put in place," she said.

A caveat that is our reading, not the article's: global storage deployment has grown very fast. Some of the per-GWh drop may come from the denominator growing faster than the number of incidents. The article does not give absolute incident counts, so the two effects can't be separated here. The trend to newer systems being safer holds either way.

Outcomes of the failures that did happen

The report's data support that point. Indoor facilities fail more often than outdoor ones and suffer more damage on average. EPRI says two measures can limit spread indoors: propagation prevention inside each battery unit, and fire walls or similar barriers between units. Most modern projects instead use outdoor, containerised units spaced apart.

Where the risk sits now

Facilities commissioned before 2021 make up only 3% of global storage capacity. They account for a disproportionate share of the risk. Srinivasan suggests their owners run thorough risk assessments and fix problems where needed.

The APS explosion shows the main concern for these older sites. Srinivasan says explosion risk is "a different risk than fire and a bigger one from a personnel and equipment standpoint." The pressure wave can hit firefighters who open a door, which a slow-burning fire does not do. Retrofits range in difficulty:

• Low difficulty: adding flammable-gas detectors inside battery enclosures, so a dangerous build-up is spotted early.
• Higher difficulty: new ventilation systems to clear the gas, or structural changes that direct pressure waves upward and away from first responders.

The bigger retrofits usually mean modifying the battery unit's enclosure. That can void the manufacturer's warranty or interfere with other parts of the design. In some cases, Srinivasan says, retiring an outdated, higher-risk system early may make more sense than retrofitting it.

The permitting problem

Safer hardware has not made permitting easy. Srinivasan says NFPA 855 has limits that push most developers to apply for exemptions, for example to allow higher energy densities than the standard's default limits. Many local authorities don't have the in-house expertise to judge those requests.

New York has responded with a fire-code change that took effect in January. Local authorities can now make storage developers pay for independent technical reviews of exemption applications. That moves the cost of expert review onto the applicant rather than leaving small fire departments to decide alone. (An earlier version of the article mentioned a new NFPA standard due this fall. Utility Dive removed it. NFPA 855 remains the governing standard.)

What stays open

The article gives rates and percentages but not the underlying counts, the GWh denominator by year, or how EPRI's database finds incidents. Public databases built from reports can miss small events. It also doesn't say how many pre-2021 sites have actually been assessed or retrofitted, or whether other states will copy New York's approach. The overall picture is clear, though: failures are now rare, they cluster in older and indoor designs, and the remaining work is mostly on those older sites and on the permitting process.

Last month a bipartisan group of senators introduced the Bipartisan American Affordability and Jobs Act (BAAJA), a permitting reform meant to speed up federal approval of energy projects of every kind. One of its biggest changes is to the role courts play in federal environmental reviews. The bill would limit who can challenge a federal environmental decision, which courts can hear the challenge, what plaintiffs can argue, and what remedies they get if they win.

What NEPA does, and where its teeth come from

The National Environmental Policy Act (NEPA) is an information-gathering law. When the federal government makes certain decisions, such as issuing a permit or writing a rule, it has to consult the agencies with relevant expertise and invite public comment. If the decision could significantly affect the environment, it also has to do a broad assessment of those effects, including a comparison with alternatives.

Other laws require narrower reviews. A Clean Air Act analysis asks whether a project meets federal pollution limits. A NEPA analysis asks what that pollution does to nearby communities and wildlife, and whether putting the project somewhere else could meet the same goals with less risk.

The courts have already narrowed NEPA. Last year the Supreme Court gave agencies more discretion over how much to analyze a project's upstream and downstream effects. Experts expected that to make injunctions and vacated permits much harder to win.

What the bill changes

BAAJA goes much further. If a community or other stakeholder sues under NEPA and wins, the court could no longer enjoin the project or vacate the decision. It would have one remedy left: ordering the agency to address the gap in its analysis that the lawsuit exposed. Construction could continue in the meantime.

An illustrative example: an agency approves a transmission line without studying its effects on a protected bird population. A group sues and wins. Today the judge could pause construction until the study is done. Under the bill, the judge could only order the study. By the time it's finished, the line might already be built.

Littleton says this removes the reason to do NEPA properly. Right now agencies and developers know that a sloppy review could get a decision suspended, a permit vacated or construction halted. It's hard to say how much that risk shapes agency behavior, but without it, nobody has a reason to invest in the analysis. Agencies are under-resourced, he noted: "If at the end of the day you know a court is not going to stop you, even if you completely ignore the statute, you're going to take the action as quickly as you can, consistent with your other obligations." The worst case becomes a court telling the agency to go back and do more paperwork.

According to Mechanick, a court might find an APA violation if:

• a decision's environmental costs plainly outweighed its benefits;
• a less damaging option offered the same benefits; or
• the agency didn't analyze a relevant part of the decision.

If the Interior Department adopted a new oil and gas leasing plan without considering effects on local wildlife, a court could call that arbitrary and capricious. The same goes if the department studied those effects, found significant risks and ignored them with no explanation. Littleton didn't dispute this. He suggested NEPA review could become the way agencies show they've complied with the APA rather than a goal in its own right.

A tougher rule for wetlands permits

Under BAAJA, a court that finds a flaw in the Corps' analysis couldn't vacate a Section 404 permit, and the bill specifically blocks vacatur for APA challenges too. The one exception is when the permitted activities "would present an imminent and substantial danger to human health or the environment."

Mechanick calls this a "timing rule." What a court can do depends on how far along the project is: either the agency still has time to act before harm arrives, or it doesn't. "I don't understand why you should have to wait until the harm is imminent to enjoin the agency," he said, "but I suppose the idea is that this is a lighter touch remedy until it's really a do-or-die moment."

What's still open

The article doesn't measure how much NEPA litigation actually delays energy projects. It doesn't say whether the bill's limits on who can sue and where would change outcomes on their own. It also doesn't address the bill's chances of passing. The main open question is the one Littleton raises: whether agencies will keep doing thorough reviews once courts can no longer stop them, or whether the APA ends up doing the work NEPA used to do.

FERC approved one smaller piece: a large load registry. This is a central database, run by PJM, of every existing and proposed load of 50 megawatts or more.

The problem PJM was trying to solve

IRAS worked like this:

Why FERC said no

FERC gave two main grounds. First, the plan intruded on state authority over retail service and shutoffs. Second, PJM hadn't justified treating new large loads differently from other new demand.

FERC also rejected PJM's proposed BYOC rules, including two features aggregators had liked because they widened who could take part:

Mary Yang, a former PJM regulatory counsel who is now COO of 804 Advisory, says the order doesn't make clear whether FERC is rejecting these features outright or just wants better justification. A redesign could keep similar benefits in a different form.

What a refiling would need

FERC invited PJM to refile, with conditions. A new IRAS must:

1. Stay out of state authority over retail shutoffs and rates.
2. Provide concrete evidence for singling out large loads.
3. Require contracted capacity to be physically deliverable within the constrained zone where the load connects.

The commission didn't throw out the idea of cutting data centers first. In a concurring opinion, Commissioner David Rosner wrote that "in an emergency, data centers not serving critical functions should be shut off before ordinary customers," such as homes, hospitals and schools. In his view, though, setting that order is a job for state regulators.

Who wins and who waits

IRAS had opponents across the board: merchant generators, state utility regulators and the AI companies themselves. Data centers' main objection was curtailment authority. With states and utilities enforcing it, a data center could bring new capacity and still be told to curtail. Supporters, including Pennsylvania Governor Josh Shapiro, argued the plan would stop the cost and strain of connecting data centers from landing on everyday consumers.

For BYOC deals, Yang calls the ruling a setback but not a dealbreaker. Many deals were already drafted but unsigned, waiting for the rules to become clear. That uncertainty "is going to keep dragging," she says, because rewriting IRAS probably isn't PJM's top priority. She says she is hearing anecdotally that some developers have decided to go to "a different region that's got more certainty."

Aggregators seem less worried. Most stakeholders expect states to create their own IRAS-style regimes, and the newly approved registry gives states the data to design them. An earlier FERC order had already sent back PJM's plan for a one-time central "backstop" auction to procure new supply. Aggregators saw that delay as useful extra time to close deals.

For generators, the effect depends on the deal. Some may want to sign bilateral contracts quickly to avoid the uncertain auction. Others may wait to see whether the auction pays better.

What's still open

The article doesn't say when PJM might refile or what a state-run version would look like. It also leaves open how deliverability rules would limit distributed capacity in constrained zones. Other grid operators are watching PJM, so the outcome will likely shape how data center load is handled well beyond its 13 states.

California's electricity prices are now double the US average and the second highest of any state. Eighty percent of Californians say lowering utility bills should be a top priority for lawmakers, and the issue is now part of the race for governor. Xavier Becerra, who leads statewide polls, has proposed a rate freeze and two hours of free power at midday, when solar generation peaks.

In an op-ed for Utility Dive, Ric O'Connell, executive director of the grid-research group GridLab, and Michael O'Boyle, senior director for policy and strategy at the think tank Energy Innovation, argue that these steps would bring relief but leave the causes of rising bills in place. They also make a political point: clean-energy policies are not driving the increases, but California's failure to control bills gives fossil fuel interests an argument against electrification. Electrification only saves households money if electricity is affordable.

What is driving costs

• Distribution, meaning local poles, wires and substations, is now the largest part of customer bills. These costs doubled over the two decades and jumped in 2019, after catastrophic wildfires pushed utilities into heavy spending to reduce future fire risk.
• Transmission, the high-voltage long-distance lines, saw spending triple over the same period. It still makes up only about 10% of the bill.
• Generation, the cost of producing power, rose less steeply: about 40% over 10 years.

In other words, the wires are pushing bills up, not the power plants. The authors say similar utilities in other states show the same rise in distribution and transmission costs. The difference is size: wildfire spending has made California IOUs' distribution costs grow faster than most peers' since 2019.

The roadmap

1. Regulation that can be followed. The authors say utility regulators everywhere are stretched thin. Batteries, customer-owned resources and data-center demand make their core job of setting "just and reasonable" rates harder. California has added more on top: the CPUC also oversees research programs, public-purpose programs, mandates for specific technologies, resource planning, transportation electrification and wildfire planning.

4. A budget for utilities. US utilities earn a return on the capital they invest, so their incentive is to build. That is the capital bias. The authors propose setting utility budgets, with higher returns when service and affordability improve and limits on profits when they don't. In industry terms, this is a form of performance-based regulation, though the op-ed doesn't use the phrase.

Use the grid you have

The authors' last point is about underused equipment. They say California's grid utilization, roughly how much of its capacity is used on average, has fallen to about 48% from a high of about 60%, a decline they say most of the country shares. When fixed grid costs are spread over relatively fewer kilowatt-hours, rates rise. They argue that making customer demand more flexible can delay or avoid new grid investment. "Flexible grid connections," which connect large new loads such as data centers on terms that let their use be curtailed during peaks, could bring in new customers who use existing capacity more fully.

What the piece doesn't settle

This is a summary of a policy roadmap written by advocates, not a cost study. It gives no estimate of how much each measure would save. It doesn't say how the 48% utilization figure is measured. It also doesn't address what happens when costs move to taxpayers or to fixed charges: they don't disappear, they land on different people. Moving utilities to performance-based budgets is, by the authors' own description, a major long-term project. The authors' plan is staged: slow rate increases first, then cut rates over time. How long either stage would take is left open.

What was promised

In 2023 the Department of Energy (DOE) chose seven regional hubs to share $7 billion. A hub is a group of hydrogen producers, users and researchers in one region. Pennsylvania anchored two of them, which also covered Delaware, New Jersey, West Virginia and Ohio.

Both had support from both parties, including Democratic Gov. Josh Shapiro and Republican Sen. Dave McCormick.

Trouble began before Trump

Underneath that fight is a physics problem. Electrolysis means passing an electric current through water to split it into hydrogen and oxygen. If a solar panel feeds an electrolyzer, the article says, about 30% of the panel's energy is lost in that step alone. Compressing or liquefying the hydrogen, moving it, and turning it back into electricity raises the total loss to about 70%. To illustrate: 100 kilowatt-hours of solar power ends up as roughly 30 kilowatt-hours of electricity again. Hydrogen is also the smallest element and leaks easily, even from sealed steel tanks.

The US makes about 10 million metric tons of hydrogen a year, almost all of it from oil and gas. Experts have warned that clean hydrogen, even with subsidies, will never compete for buses, cars or heating buildings. Where it matters is as a chemical feedstock, a raw material for industry. Ammonia, mainly used for fertilizer, accounts for about 2% of global energy use and 1.5% of greenhouse gas emissions. Green ammonia could also help clean up shipping, which produces another 3% of emissions.

MACH2 did make a show of testing hydrogen buses with SEPTA, Philadelphia's transit agency. But its application named local oil refineries as its main early customers. Community and environmental justice groups objected, citing flammability. They also pointed out that leaked hydrogen acts as an indirect greenhouse gas, because it makes methane, ozone and water vapor stronger warmers.

Funding saved on paper, money not spent

The Trump administration cut much of the Inflation Reduction Act, the 2022 climate law. Most hydrogen hubs survived on paper. The California and Pacific Northwest hubs lost their funding early and are now in court. In April, Energy Secretary Chris Wright said the remaining hubs would keep their awards. Experts describe a "will they, won't they" pattern from week to week.

Three changes make that promise thin:

1. Shorter deadlines. Projects must now start construction by December 31, 2027 to qualify for 45V, instead of 2032. The article calls that close to impossible.
2. No spending. DOE's July apportionment plan, its schedule for releasing appropriated money, shows no further hub spending this year. Each hub has drawn only about a fifth of its Phase One money, which Tom Torres of the Ohio River Valley Institute calls "a drop in the bucket."
3. Proposed cancellation. The 2027 President's Budget Request would "permanently cancel" unspent clean hydrogen money. It would move funds to "baseload power" upgrades, power plants that run around the clock, and put $1.2 billion into seven AI supercomputers at Argonne and Oak Ridge national labs.

Each hub lost what it was built on

MACH2 needed new solar and offshore wind to run its electrolyzers. Some solar is still being built. Offshore wind on the East Coast has largely collapsed. Atlantic Shores and Invenergy left their contracts in 2025. Early-stage projects ended as the administration paid companies to give up their leases. In April, New Jersey's Board of Public Utilities stopped building the grid connections for offshore wind.

ARCH2 existed mainly so gas producers could keep operating under tighter emissions rules. The administration is now moving to scrap power-plant emissions rules altogether, which removes much of that purpose.

Supply without demand

Anika Juhn of the Institute for Energy Economics and Financial Analysis argues the original design had a flaw. It paid people to produce hydrogen but did little to make anyone want to buy it. "The hydrogen story is really an ammonia story, in the U.S. at least," she says. Without help for farmers paying more for clean fertilizer, there is no domestic market, so any US green ammonia would likely be exported.

Steel shows the same gap. Hilary Lewis of Industrious Labs notes that most US steel still comes from coal-fired blast furnaces. Direct reduced iron (DRI) furnaces are cleaner. They run on methane today but could switch to hydrogen. Hyundai's $5.8 billion mill in Louisiana is one candidate. Lewis says Pennsylvania needs two things: clean hydrogen and new furnaces. It has neither.

Where the partners went

Torres says former partners are moving to data centers. Adams Fork Energy in West Virginia started as a coal-to-gas project. It joined ARCH2 with a plan for a large ammonia plant and was later dropped. This year it began building gas plants to power a data center. At least four other former ARCH2 partners have done the same: Hog Lick Aggregates, EQT, Fidelis New Energy and CNX Resources. "It's all gas plants and data centers now," Torres says.

Open questions

The article doesn't say how much MACH2's partners have actually committed, whether anyone will claim 45Q for ARCH2, or how the lawsuits over the West Coast hubs might affect these two. DOE did not respond to requests for comment.