Carbon Capture Blinks, Batteries Step Up · Morning Gist
Friday, October 235 stories · 16 min
Carbon Capture Blinks, Batteries Step Up
AI's hunger for power is changing who builds what: dirty gas engines win over carbon capture, home batteries cover peaks, and grid rules get contested.
Picked from 37 new articles across 13 sources · arrow keys for other days
Today in 30 seconds
30s left
Net Power, the carbon-capture pioneer, bought 123 MW of reciprocating gas engines to sell unabated power to data centers fast. story 1 ↓
Home batteries fed California's grid during the October 6 heat wave; Sunrun says 1.2 million US units could deliver 8 GW. story 2 ↓
Ex-EIA chief Guy Caruso blames shortages in PJM's deregulated market, not utility greed, for rising power bills, rebutting Jigar Shah. story 3 ↓
Indonesia's PLN plans wood co-firing at 52 coal plants; its biomass supply target already slipped to 2031, and forests are at risk. story 4 ↓
Data centers are offering grid flexibility for faster interconnection; on-site batteries are easiest, flexing compute hardest, and "flexible" remains undefined. story 5 ↓
The Justice Department is moving to revoke already-issued offshore wind permits; a Democratic co-sponsor says that could sink the Senate permitting bill.
1/5Heatmap· Emily Pontecorvo· Oct 22
Net Power drops zero-emission gas plants to sell fast, unabated power to data centers
Net Power was founded to build gas plants with near-zero emissions. It has now dropped that technology and bought reciprocating gas engines to power a data center, because buyers now pay for speed, not low carbon.
Why it matters · It shows that, at least for now, AI power demand is outweighing U.S. carbon-capture subsidies, even after the 45Q tax credit was expanded.
Know this first
Context from the editor, not the article · 27 sec
1 · Carbon capture on gas plants
Carbon capture and storage (CCS) separates CO2 from a plant's exhaust and buries it underground. On gas plants, the exhaust is mostly nitrogen and only a few percent CO2, which makes capture costly and energy-hungry.
2 · The 45Q tax credit
45Q is a U.S. federal tax credit paid per tonne of CO2 captured and then stored or used. Trump's 2025 One Big Beautiful Bill Act kept it and expanded it while cutting other clean-energy credits.
3 · Data centers want speed to power
AI data centers need hundreds of megawatts quickly. Waits for grid connections and for large gas turbines stretch for years, so developers increasingly build their own generation on site, known as co-location.
2/5Grist· Matt Simon· Oct 22
Home batteries supplied California grid power during October heat wave
During California's October 6 heat wave, Tesla and Sunrun home batteries sent stored solar power to the grid at peak hours. The nonprofit Common Charge says that output equalled the demand of every home in San Francisco and San Jose.
Why it matters · Cheap ways to meet a few hours of peak demand decide whether utilities build gas plants, and home batteries are one of the few options that can grow quickly.
Know this first
Context from the editor, not the article · 31 sec
1 · Why peaks are expensive
Grids must be sized for the highest few hours of the year, usually hot late afternoons when air conditioning runs and solar output fades. Meeting those hours often means gas "peaker" plants that sit idle most of the year, or costly imported power.
2 · Virtual power plants
Software can combine thousands of home batteries into one resource that a utility or grid operator calls on, a setup known as a virtual power plant. Owners are paid to let it discharge their batteries when the grid is stretched.
3 · Power versus energy
Gigawatts (GW) measure the rate at which electricity flows. Gigawatt-hours measure how much is delivered in total. A typical home battery stores roughly 10–15 kWh and can supply a few kilowatts for a few hours.
3/5Utility Dive· Guy Caruso· Oct 22
Former energy statistics chief disputes Jigar Shah's "utility greed" claim
In a Utility Dive op-ed, Guy Caruso, former head of the US government's energy statistics agency, argues that rising power bills come from too little new supply in deregulated markets, not "utility greed."
Why it matters · Who gets to build power plants, whether regulated utilities or merchant generators, is now a live policy fight in PJM states, and both camps claim bill-payers' interests.
Know this first
Context from the editor, not the article · 29 sec
1 · Two ways to run power
In vertically integrated states, one regulated utility owns generation, transmission and delivery, and charges rates set by regulators. In deregulated or restructured states, utilities mostly deliver power, and independent companies own the plants and sell into wholesale markets.
2 · How utilities earn money
Regulated utilities recover costs plus a return on equity (ROE) on approved investments. State regulators set that return in rate cases, which are public proceedings where consumer advocates and other parties can challenge costs.
3 · PJM and capacity auctions
PJM Interconnection runs the grid and wholesale market for 13 states plus Washington, DC. Its capacity auctions pay plants years ahead to be available at peak. Clearing prices have jumped sharply in recent auctions as data-center demand grows.
4/5Climate Home News· Chloé Farand· Oct 23
Indonesia's plan to burn wood in coal plants risks forests and delays coal's closure
Two Indonesian think-tank directors argue that the state utility's plan to burn wood alongside coal at 52 power plants would drive deforestation, add emissions and give old coal plants a reason to keep running.
Why it matters · Co-firing biomass is often counted as renewable energy in national climate plans. If the wood comes from cleared forest, both the "renewable" share and the emissions cuts may be accounting fictions.
Know this first
Context from the editor, not the article · 29 sec
1 · What biomass co-firing is
A coal plant replaces a small share of its coal with wood chips, pellets or crop waste. At low blends, plants need little or no new equipment. That makes co-firing a cheap way to report some generation as renewable.
2 · The carbon-neutral biomass assumption
Burning wood is often counted as zero-carbon because regrowing trees reabsorb the CO2. That holds only if the land stays forested. Clearing old forest releases centuries of stored carbon, which plantations take decades to recapture, if they ever do.
3 · Indonesia's coal dependence
5/5Latitude Media· Lisa Martine Jenkins· Oct 21
Data centers trade grid flexibility for faster grid connections
Grid connections now take years, and delays cost data centers about $3 million per megawatt per month. Many are offering to cut grid draw at peaks in exchange for faster hookups, mostly by adding on-site batteries.
Why it matters · Whether large new loads can flex decides how much grid capacity, gas backup and battery storage the data center buildout actually needs.
Know this first
Context from the editor, not the article · 26 sec
1 · Interconnection queues
Before a large load or generator can connect, the utility studies whether wires and supply can handle it. Demand from AI data centers has backed these queues up, so in busy markets connection can take many years.
2 · Why flexibility helps
Grids are sized for the few hours of highest demand. A load that agrees to reduce its draw during those hours needs less new capacity built for it, so utilities can connect it sooner.
3 · Meter location
Behind-the-meter equipment sits on the customer's side of the utility meter and serves the site directly. Front-of-meter equipment sits on the grid side but is paired with the site.
Net Power was founded to build natural gas power plants with essentially zero emissions. It has now given up on that goal. Last month it said it had acquired 10 Wärtsilä natural gas 1, 123 megawatts in all. It plans to install them next to a data center on the site once meant for its first commercial clean plant. No data center customer has been named.
The original idea
Net Power's technology was 2, which burns gas in pure oxygen instead of air. Air is mostly nitrogen, which dilutes the exhaust and makes CO2 expensive to separate. Burning in pure oxygen produces a nearly pure CO2 stream: some of it drives the turbine, and the rest can be captured cheaply. The company's demonstration site in La Porte, Texas, never showed that the plant produced more energy than it used.
Two pivots in a year
In November 2025, CEO Danny Rice said the first 200-megawatt plant would cost about $1.7 billion and could not come online before 2030 or 2031. Net Power switched to ordinary gas turbines with 3, which scrubs CO2 from normal exhaust. It ordered two 30-MW Siemens turbines for 2028 and signed a letter of intent with Entropy, a Canadian maker of capture equipment. By year's end it had suspended all oxy-combustion work.
In August, Rice said buyers now want only speed, reliability and scale, and that carbon has dropped off the list. Net Power would build 4, meaning gas without capture, and hoped to add capture "in the future."
The engines it bought are among the dirtiest gas options. An Earthjustice estimate found a 1-gigawatt reciprocating-engine plant would emit 1.5 times the CO2 of a modern 5 plant, which reuses exhaust heat to run a second turbine. It would also emit twice the nitrogen oxides and fine particulates.
Why capture isn't selling
Only one proposed U.S. gas plant with capture has a buyer: Illinois's Broadwing Energy Center, which signed a deal with Google in October 2025. Columbia's Jack Andreasen Cavanaugh says demand for unabated gas is "massive." He says the 6 tax credit, which pays per tonne of CO2 stored and was expanded in 2025, still isn't enough to get capture deployed at scale. The Global CCS Institute counts more than 266 U.S. capture projects in development, held back by permitting delays, policy uncertainty and rising capital costs.
A smaller bet continues
Arbor Energy, a startup founded by former SpaceX engineers, is still pursuing oxy-combustion. It is starting with a 1-MW demo that is weeks from running and plans a 25-MW unit by 2028; neither will store carbon. Asked whether anyone will pay extra for clean gas power today, COO Alicia Harmon said, "Honestly, my answer is no." She claims the cost gap will close by the early 2030s.
Net Power was founded to build natural gas power plants with essentially no emissions. It now plans to build some of the dirtiest gas generation available. Last month the company said it had acquired 10 Wärtsilä natural gas 1, 123 megawatts in all. It plans to co-locate them with a data center on the site once meant for its first commercial clean plant, meaning the engines will sit next to the data center and feed it directly. No data center customer has been named. Heatmap's Emily Pontecorvo argues this is more than one company's change of course. It shows how the AI boom has weakened demand for clean firm power, and that U.S. carbon-capture policy still isn't enough to make the technology viable in the power sector.
What Net Power was trying to build
Net Power's core technology was 2, which burns natural gas in pure oxygen instead of air. Air is mostly nitrogen, so a normal gas plant's exhaust is a dilute mix, and pulling CO2 out of it is expensive. Burning in pure oxygen leaves an exhaust of nearly pure CO2 and water. Net Power's design sent some of that CO2 back through the cycle as the working fluid that drives the turbine, and captured the rest more cheaply than a conventional plant could. (The design is widely known as the Allam cycle, though the article doesn't use that name.)
The company tested the system at a demonstration site in La Porte, Texas. It never showed that the plant could produce more energy than it consumed. Making pure oxygen and compressing CO2 take a lot of power, and the company's name refers to clearing exactly that bar.
Pivot one: conventional turbines plus capture
By November 2025 the economics had stopped working. CEO Danny Rice told investors that the first 200-megawatt oxy-combustion plant would cost about $1.7 billion. That works out to roughly $8,500 per kilowatt, by our arithmetic. The plant could come online in 2030 or 2031 at the earliest, costs were still climbing with inflation, and the company could not realistically finance it before testing was finished. Net Power said it would "broaden the scope of its business" and build conventional gas plants with 3, meaning equipment that scrubs CO2 from ordinary exhaust. It signed a letter of intent with Entropy, a Canadian capture-equipment maker, and ordered two 30-megawatt Siemens gas turbines for delivery in 2028. "Speed is everything," Rice said at the time. By the end of 2025 the company had suspended all oxy-combustion development and testing, according to its annual filing.
Pivot two: no capture at all
Six months later the company changed its plan again. On an August earnings call, Rice said buyers now care about three things: speed, reliability and scale. Carbon emissions had dropped off the list. Turbines with capture had not turned out to be the "fastest-to-market and most cost-competitive" option he had predicted. "Natural gas power generation, co-located with customer load, and deployable on a fast timeline, is how we expect to meet that demand today," he said. He added that the company hoped to add capture later. "We believe the clean piece is important," he said, but it "will only come at scale once society's basic energy pillars of reliable, affordable power are fully in place."
Reciprocating engines are large piston engines, similar in principle to a car engine but scaled up. They come in modular units, here about 12 MW each. As general background, they are quicker to buy and install than big turbines, whose order backlogs run for years. The cost is emissions. An Earthjustice report estimated that a 1-gigawatt reciprocating-engine plant would emit 1.5 times the CO2 per year of a 4 plant, the efficient design most modern gas plants use, which recovers exhaust heat to run a second, steam turbine. It would also emit about twice the nitrogen oxides and fine particulate matter.
Why carbon capture on gas isn't selling
The demand is thin. Only one proposed U.S. gas plant with capture has announced a buyer for its power: the Broadwing Energy Center in Illinois, developed by Low Carbon Infrastructure, which signed a power purchase agreement with Google in October 2025. "There is a massive demand for 5 right now, and there is much less of a demand for carbon capture," said Jack Andreasen Cavanaugh of Columbia University's Center on Global Energy Policy. He said 6 (the largest cloud and AI companies) are still talking with capture companies, but the cost and logistics of capturing, transporting and storing the CO2 remain obstacles. More broadly, he said, the U.S. has never put up enough public money to get CCS deployed at scale.
The Global CCS Institute's annual status report, published this week, counts more than 266 U.S. carbon capture, transport or storage projects in development. It says permitting delays and policy uncertainty are holding them back, and that inflation and higher capital costs have weakened project economics, "reducing the effectiveness of existing policies, such as tax and investment credits."
Someone else is still trying oxy-combustion
Arbor Energy, a startup founded by former SpaceX engineers, is still developing oxy-combustion, and is taking a smaller, more incremental path than Net Power did. Last week it ran a 7 at high speed for more than two hours. This is a key component that uses CO2 squeezed into a dense, fluid-like state. Arbor uses 3D-printed turbines and builds smaller, modular units. A 1-megawatt demonstration in California is "weeks away" from operating, according to COO Alicia Harmon, and a 25-megawatt demo is planned for 2028. Neither demo will store any carbon.
Arbor has also changed course because of data centers. It originally planned to burn biomass and bury the CO2, which would remove carbon from the air. It switched because demand for power outran demand for carbon removal, and because natural gas is easier to come by than biomass. Asked whether anyone will pay a green premium for carbon-capture gas power today, Harmon said: "Honestly, my answer is no." She says Arbor is already "very close" to cost parity with combined-cycle plants once the 8 credit is counted, and expects to close the gap entirely by the early 2030s. These are her claims, not demonstrated results.
What's left open
Net Power declined to comment. It has not named a customer for the engines, said when they would run, or explained how capture could be "added later" to equipment that wasn't designed for it. The bigger open question is whether expanded 45Q, falling costs or buyer pressure can ever make capture competitive while speed to power is the only thing buyers pay for.
Remember
Net Power never showed its oxy-combustion plant could produce net power; its first 200 MW plant was costed at $1.7 billion.
Only one proposed U.S. gas-with-capture plant, Broadwing in Illinois, has a buyer for its power.
Reciprocating-engine plants emit an estimated 1.5 times the CO2 of combined-cycle plants and twice the nitrogen oxides.
On October 6, during a California heat wave, home batteries installed by Tesla and 1 sent stored solar power into the grid at the evening peak. Sunrun installs and leases rooftop solar and batteries. 2, an energy nonprofit, says the batteries supplied enough power for every home in San Francisco and San Jose, about 1.8 million people, during peak hours.
Why it helps
Utilities have to meet a few hours of peak demand, usually late afternoon when air conditioning comes on and solar output fades. Traditionally they burn more natural gas or import power from other states. A home battery fills up on midday solar and releases it in the evening. When a utility calls on many batteries at once, called a 3, they act like one power plant.
Sunrun says it ran 1,300 dispatch events last year and more than 800 so far this year. It always leaves some charge in each battery, so most owners never notice, and the house can still run on its own power during an outage. Common Charge's Mary Rafferty argues that people without batteries benefit too, because local stored power is cheaper than new gas plants or imports. The article puts no dollar figure on those savings.
Cars and buses next
4 lets some electric cars send power back to the grid. A driver who charged at work can sell some of the remaining charge during the evening peak and recharge overnight. School buses are promising candidates: they have big batteries and predictable schedules. Sunrun and others are running V2G pilots.
The numbers in proportion
Sunrun says it and Tesla have 1.2 million batteries installed across the US, able to deliver 8 5. It calls that almost 11 times the heat-wave contribution, which implies roughly 0.7 GW on October 6. California's peak demand on a hot day is typically 40–50 GW (general knowledge, not from the article), so the batteries covered a few percent at most. That is useful, because the tightest hours are the most expensive to supply. It is not yet a remade grid.
The "every home" comparison counts only household demand, not businesses. The 8 GW is a rate of delivery, and the article doesn't say how long it could be sustained. There is plenty of room to grow: the US has 86 million single-family homes.
The article doesn't say how much owners are paid, which program called the dispatch, or how much frequent cycling wears batteries down. Its sources, an advocacy group and a battery company, have a stake in the story.
What happened
On October 6, during a late heat wave, California's grid pulled power from thousands of home batteries. 1, a nonprofit focused on energy, says batteries installed by Tesla and 2 supplied enough electricity during peak hours to cover every home in San Francisco and San Jose, two cities with about 1.8 million residents between them. Sunrun installs and leases rooftop solar and batteries.
Grist's Matt Simon treats the event as a preview of a grid where households both buy power and sell it. Most of the article's sources have a stake in that outcome: an advocacy nonprofit and a company that sells home batteries. The numbers below come from them and are worth reading with that in mind.
The problem being solved
Simon describes the traditional model as rigid. A utility forecasts when demand will spike, typically late afternoon in summer when people get home and turn on the air conditioning, and covers the spike by burning more natural gas. It may also import electricity from other states. Either way it pays for capacity that only matters for a small number of hours each year.
Solar complicates this. Rooftop panels produce the most power at midday and then fade just as the evening peak arrives. Early solar owners exported their surplus and were paid for it, but only until sunset. Batteries close that gap. A home battery fills up on cheap midday solar and then discharges in the evening, either to run the house or to export power to the grid.
How a dispatch works
A 3 is a period when the operator tells enrolled batteries to discharge into the grid. Sunrun says it ran 1,300 of them last year and more than 800 so far this year. According to Sunrun spokesperson Wyatt Semanek, homeowners usually don't notice. The company always holds back some charge, so the house can still run as an "island" on its own power if the grid fails.
As an illustration (these numbers are not from the article), picture 100,000 enrolled batteries that each export 5 kilowatts for three hours. Together they behave like a 500-megawatt power plant for those three hours. Each home gives up 15 kilowatt-hours at most, and less if it keeps a reserve. Simon notes that the more batteries are enrolled, the less any one of them has to give.
Simon argues that people without batteries benefit too. Common Charge's executive director, Mary Rafferty, says local stored power is cheaper than building gas plants or importing electricity, so all ratepayers save over time. The article gives no dollar figure for those savings.
Electric vehicles as grid batteries
4 lets some EVs send power back to the grid as well as draw from it. Sunrun is one of several companies running V2G pilots. Simon gives an example: a driver charges at work, comes home in the late afternoon with plenty of charge left, and lets the utility take some during the evening peak. The car then recharges overnight, after demand has fallen.
School buses look especially promising. Their batteries are large and their schedules are predictable, so a utility knows when the power will be available. Simon's point is that the goal is not just more backup batteries but a smoother demand curve.
The numbers, in proportion
Sunrun says it and Tesla have 1.2 million batteries installed across the US, able to dispatch 8 5 to the grid. Sunrun calls that almost 11 times what was delivered during the heat wave. Working backward, the October 6 contribution was about 0.7–0.75 GW. The article does not give that figure directly.
For scale, using general knowledge rather than the article: California's statewide peak demand on a hot day is typically 40–50 GW. A contribution of about 0.7 GW is therefore a few percent at most. That matters, because a few percent at the tightest hour is what operators otherwise cover with the most expensive plants. But it is a long way from remaking the grid. The comparison to "every home in San Francisco and San Jose" counts residential load only. It leaves out commercial and industrial demand, which is much of what peaks.
The 8 GW figure is power capacity, a rate of delivery, and the article doesn't say for how long it could be sustained. It is also spread across the whole country, not concentrated where any one grid needs it. The headroom is large, though. The US has about 86 million single-family homes, so 1.2 million batteries reach only a small fraction of them.
What's left out
The article does not explain how owners are paid or how much. It doesn't say which program called the October 6 dispatch, how many batteries took part, or how long they ran. It doesn't address battery wear from frequent cycling, or whether utility payments will stay generous enough to keep households enrolled. V2G is still at the pilot stage, and the article gives no data on its performance.
Simon also mentions plug-in solar panels that renters can mount on a balcony. Like the V2G material, this is a direction of travel rather than a measured result.
The October 6 event shows that a large fleet of distributed batteries can be called on reliably and without disrupting the homes involved. Whether it grows into a major grid resource depends on enrollment, payment levels and economics that the article doesn't measure.
Remember
Working back from Sunrun's figures, home batteries delivered roughly 0.7 GW on October 6, a few percent of California's typical hot-day peak.
The 8 GW claimed for 1.2 million US batteries is a rate of delivery; how long it can be sustained isn't stated.
Payment levels, battery wear and V2G performance data are all missing from the article.
Guy Caruso, a former head of the U.S. 1 (EIA), the federal energy-statistics agency, has written a Utility Dive op-ed answering Jigar Shah. Shah, a clean-energy financier who once ran the Department of Energy's loan office, blamed "utility greed" for rising power bills in an August Newsweek column. Caruso agrees that data centers alone didn't raise everyone's bill. He says the greed explanation is wrong and points policymakers at the wrong fix.
Why he says profits aren't the cause
Returns are regulated. Utilities earn a return only on investments that state regulators approve as prudent. Large projects go through 2, public reviews where consumer advocates can challenge costs. An authorized 3 is a chance to earn that return, not a guarantee.
Returns are falling. A report from the Alliance for Innovation and Infrastructure finds that both the ROEs utilities requested and the ROEs regulators granted have trended down for two decades. The op-ed gives no figures.
The increases are regional. A Charles River Associates analysis found that recent residential increases are concentrated in California, the Northeast and parts of the Mid-Atlantic. Most states saw stable or below-average rates.
What he blames instead
Caruso points to deregulated markets, especially 4, the grid operator for 13 eastern states plus Washington, DC. In PJM states such as Pennsylvania and New Jersey, utilities don't own power plants. Supply comes from independent generators, paid partly through 5 that buy availability years in advance. He says customers there face volatile prices and shortages because new plants are slow to arrive. Public auction results back up the volatility: PJM's capacity price rose roughly tenfold between the 2024/25 and 2025/26 delivery years.
His fix is to let regulated utilities build generation in those states again. Regulators would still vet each project, and any market revenue a utility plant earned would lower customer bills. He also argues that unregulated generators have earned record profits during the shortage without building enough new capacity. He doesn't name companies or cite figures.
How to weigh it
The factual points about ROE and PJM prices hold up. The causal claim is weaker:
California, one of the high-increase regions he lists, is mostly served by regulated utilities, and its wildfire-related grid spending is a utility cost.
The Northeast had high rates before it deregulated.
Plants built by utilities shift cost-overrun risk back onto customers.
Caruso and Shah agree more supply is needed. They disagree on who should build it and who should carry the risk.
The dispute
Guy Caruso spent more than 40 years in energy markets and policy and once ran the U.S. 1 (EIA), the federal agency that collects energy statistics. In a Utility Dive op-ed, he replies to an August Newsweek column by Jigar Shah. Shah is a clean-energy financier who led the Department of Energy's Loan Programs Office, and his column blamed "utility greed" for rising electricity bills. Caruso agrees with one part of it: data centers alone did not "jack up" every household's bill. He argues the rest gets the cause wrong. In his account, bills are rising where supply is short, and supply is shortest in deregulated markets that leave plant-building to independent companies.
This is an opinion piece. The op-ed doesn't lay out Shah's full argument, so this summary covers only the parts Caruso quotes or answers.
Caruso's case that utility profits aren't the problem
Utility returns are regulated. Under cost-of-service regulation, utilities earn a return only on investments that state regulators judge prudent, useful and in customers' interest. Big capital projects go through 2, public proceedings where consumer advocates and other outside parties (called 3) can challenge costs. Caruso also points out that an authorized 4 is a chance to earn that return, not a promise of it.
Returns have been falling. He cites Ratemaking for a Reliable Grid, a report from the Alliance for Innovation and Infrastructure, a policy group. It finds that both the ROEs utilities requested and the ROEs regulators granted have trended down over two decades. The op-ed gives no figures. Industry trackers have generally shown average authorized ROEs drifting from above 10% in the mid-2000s to the mid-9% range more recently, which fits the direction he describes.
The increases are regional. A Charles River Associates analysis from earlier this year found that recent residential rate increases are concentrated in California, the Northeast and parts of the Mid-Atlantic. Most states have seen stable rates or rates below the national average. Caruso's point is that a national greed story would not produce such a regional pattern.
The alternative diagnosis: shortages in deregulated markets
Caruso's broader claim, which he says comes from reviewing the data, is that deregulated regions have done worse on affordability than vertically integrated ones. His main example is 5, the grid operator and wholesale market covering 13 Mid-Atlantic and Midwestern states plus Washington, DC. In PJM states such as Pennsylvania and New Jersey, utilities generally don't own power plants. Supply comes from unregulated generators, and capacity is bought through forward auctions that pay plants years ahead to be available at peak demand.
He says this setup leaves customers exposed to volatile capacity prices and ongoing supply shortages while new generation struggles to get built. The article gives no numbers. Public auction results show the scale. PJM's capacity price rose from about $29 per megawatt-day for 2024/25 to about $270 for 2025/26, and then hit a negotiated price cap of roughly $329 for 2026/27.
His proposed fix
Caruso wants regulated utilities in PJM states to be allowed to build generation again. He argues this would keep oversight in place, because regulators would still decide whether each plant is prudent and whether its costs can be recovered. He also calls it a hedge for customers. If a utility-owned plant earns money in the PJM market, those earnings reduce what customers pay. An 6 keeps market and scarcity-pricing profits for its shareholders and has no obligation to serve local customers.
He then turns Shah's question about incentives back on the merchant generators. Caruso writes that publicly traded generators, which are not regulated the way utilities are, have made record profits in recent years without building the new capacity demand requires. He says some announced new plants only after shortages were obvious. He doesn't name companies or cite profit figures.
He dismisses the argument that better use of the existing grid can close the gap. Utilities already use demand response, efficiency programs, advanced grid management and 7, he says. Meeting demand at today's scale still needs new generation, transmission, substations and distribution upgrades.
How to weigh it
Caruso's main factual points are well supported:
Utility returns are set in public proceedings.
Authorized ROEs have not been climbing.
PJM's capacity prices really have jumped.
The weaker link is the step from those facts to "deregulation causes high bills." Several points cut against it:
California doesn't fit. It is on his list of high-increase regions, but its large investor-owned utilities are rate-regulated. Much of their recent increase has been linked to wildfire-related grid spending, which is a utility capital cost, the kind of cost Shah's critique is about.
High rates came before restructuring. The Northeast had high rates before it deregulated, and high rates were part of why it deregulated. Comparing regions doesn't untangle fuel mix, geography and history from market design.
Rising rates have many sources. Rate increases also come from distribution and transmission spending, the costs of utilities that are still regulated even in deregulated states. Caruso's argument doesn't fully address that.
Utility-built plants carry their own risk. If a regulated plant runs over budget, customers absorb the overrun. Markets were meant to shift that risk onto investors.
Sources aren't characterized. The op-ed doesn't say who funds the Alliance report or the Charles River Associates analysis, or whether they were commissioned by industry. Readers should check before treating either as neutral.
Both sides agree that more supply is needed. They disagree on who should build it and who should carry the risk.
Remember
Average authorized utility returns on equity have drifted down for two decades, which undercuts a simple greed explanation.
PJM capacity prices rose about tenfold between the 2024/25 and 2025/26 auctions, the strongest evidence for Caruso's shortage argument.
Letting utilities build plants adds regulatory oversight but shifts cost-overrun risk back onto ratepayers.
Coal generates most of Indonesia's electricity, and PLN, the state utility, runs much of the fleet. Under its Paris Agreement pledge, its Nationally Determined Contribution, Indonesia targets 23% renewables in its energy mix.
Image: Climate Home News
Indonesia's state electricity company, 1, plans to mix wood into the coal burned at 52 power plants. This is called 2, and PLN presents it as progress toward clean energy. In an opinion piece for Climate Home News, two directors of the Jakarta think tank 3 (Center of Economic and Law Studies) argue the opposite: the plan would clear forests, could add emissions, and would give old coal plants a reason to stay open.
The plan
The co-firing target is in PLN's national electricity supply plan. It is tied to Indonesia's 4, its Paris Agreement pledge, which aims for 23% renewables in the energy mix. Most plants mix in only 1% to 5% wood, so they don't need new equipment. PLN wanted 10.2 million tonnes of biomass a year by 2025. Because it cannot source enough, that target has slipped to 2031.
Where the wood comes from
A 2024 NGO assessment estimated that the mandate needs more than 10 million tonnes of wood pellets a year. That is an area of forest equal to 3.27 million football fields. It put possible deforestation as high as 2.1 million hectares a year, an upper-bound figure whose assumptions the article doesn't give.
PLN says the wood will come from "5", land set aside for fast-growing fuel trees. But the assessment found that of the 1.2 million hectares designated for these plantations, more than 400,000 hectares are still undisturbed tropical forest.
The emissions math
Burning wood counts as carbon-neutral only if the trees grow back on the same land. Clearing old forest releases carbon that took centuries to build up. The Centre for Research on Energy and Clean Air (6) estimated that a 10% wood blend across all 52 plants could produce up to 26.5 million tonnes of 7 a year, once land clearing and pellet production are counted. The article doesn't compare that with the coal emissions the wood would replace.
Who benefits
Existing timber permit holders can convert their land to energy plantations without applying for new forest land. Research by the advocacy group 8 found the push is led by firms with deforestation records, including APP Group and Wilmar International. Japan and South Korea have bought more than 99% of Indonesia's pellet exports since 2021, counting them toward their own clean-energy targets.
The authors' bottom line
The authors accept biomass made from farm or municipal waste. But they say leftovers can't supply a programme this large. They also argue that a token wood blend extends coal plants' lives, though the article gives no example of a delayed retirement. They want funders, Japan and South Korea to stop counting co-firing as a climate win.
Indonesia presents 1 as a step toward clean power. Co-firing means mixing wood or other plant matter into the coal a power plant burns. In an opinion piece for Climate Home News, Muhammad Zulfikar Rakhmat and Bhima Yudhistira Adhinegara of the Center of Economic and Law Studies (2), a Jakarta think tank, argue that the plan will do more climate harm than good. Their case rests on three claims. It will drive deforestation. Counted honestly, it may add emissions. And it gives ageing coal plants a green excuse to keep running.
What 3 is planning
PLN, Indonesia's state electricity company, wants to co-fire biomass at 52 coal plants. The target is written into its national electricity supply business plan, known in Indonesia as the RUPTL, and tied to the country's 4. That is the climate pledge Indonesia filed under the Paris Agreement, and it aims for renewables to reach 23% of the energy mix.
The supply side is already struggling. PLN planned to secure 10.2 million tonnes of biomass a year by 2025. That deadline has moved to 2031 because the utility cannot line up enough wood.
The blend at most plants is small, typically 1% to 5% of the fuel load. That is low enough that plants don't have to change any machinery. The authors see that as the core problem: the plants carry on burning coal as before, with a little wood added.
Why the wood demand matters
Supply studies model higher blends than plants use today, so the numbers below describe what the full programme could require, not current practice.
A 2024 assessment by a group of NGOs (the article doesn't name them) found:
Fuel needed: more than 10 million tonnes of wood pellets a year to meet the mandate. That is an area of forest equivalent to 3.27 million football fields.
Deforestation: potentially up to 2.1 million hectares a year. This is an upper-bound estimate, and the article doesn't give the assumptions behind it.
Forest at risk: cutting coal use by 10% through co-firing could threaten 10 million hectares of untouched forest in the areas where wood is collected.
Biggest plants alone: switching 10% of coal to wood at Indonesia's largest plants could clear forest roughly 35 times the size of Jakarta.
PLN and its suppliers say the wood will come from "energy plantation forests", land set aside to grow fast-growing trees for fuel. The same assessment found that 1.2 million hectares are already designated for these plantations, and more than 400,000 hectares of that, about a third, is still undisturbed tropical forest. The authors conclude that the claim the wood comes from already-degraded land does not hold up.
The carbon accounting
Wood is usually treated as a clean fuel on the logic that trees grow back and reabsorb the carbon released when they burn. The logic fails when natural forest is cleared to make room for plantations. The old forest's carbon, built up over decades or centuries, goes into the atmosphere at once. A young plantation recaptures only part of it, and slowly.
The Centre for Research on Energy and Clean Air (5), an independent research group, estimated the full-chain cost. Co-firing 10% wood across all 52 plants could produce up to 26.5 million tonnes of 6 a year, counting land clearing and pellet manufacturing. The article doesn't set that against the coal emissions co-firing would avoid, so you can't tell from it alone whether the net effect is positive or negative. The authors' claim rests on CREA's inclusion of land-use emissions, which standard power-sector accounting usually leaves out.
Who supplies the wood, and who buys it
Government rules make entry easy, the authors say. Companies that already hold industrial timber plantation permits can convert them into 7 without applying for new forest land. Indroyono Soesilo, who chairs the Indonesian Forestry Entrepreneurs Association (APHI), has publicly asked for tax breaks and relief from state forestry levies in return for making the switch.
Research by 8, an Indonesian advocacy group, found the expansion into energy plantations is led by large firms with deforestation records: APP Group, Korindo Group, Barito Pacific and Wilmar International. Much of the output leaves the country. Since 2021, Japan and South Korea have bought more than 99% of Indonesia's wood pellet exports, which they use toward their own clean-energy targets. So two other countries' renewable statistics also count Indonesian forest wood as renewable fuel.
The lock-in argument
The authors call the biggest risk the effect on the coal fleet itself. A low wood blend needs no shutdown or upgrade. On paper, though, it makes a plant look greener, which can justify running it past its planned retirement. Every extra year a coal plant runs, they argue, delays solar and wind. This argument is qualitative. The article gives no example of a plant whose retirement has actually been pushed back.
What the authors would accept
The authors don't reject all biomass. Generating power from farm residues or municipal waste can make sense, they say. Their objection is to scale: a programme of 10 million tonnes a year can't run on leftovers, which is why forests get pulled into the supply chain. They call on the countries and banks funding Indonesia's energy transition, plus Japan and South Korea as pellet buyers, to stop counting co-firing as a climate win. Their conclusion: real progress means closing coal plants.
What stays open
This is an advocacy piece, and its numbers come mostly from NGO and think-tank assessments, not independent peer review. Several gaps would matter to a careful reader:
No net-emissions comparison against the coal displaced.
No breakdown of how the 2.1-million-hectare upper bound was derived.
No evidence yet of retirements actually delayed.
No word on whether residues could cover today's 1–5% blends, as opposed to the 10% scenarios.
The supply data is harder to dismiss. PLN's own target slipped six years for lack of wood, and a third of the land set aside for fuel plantations is still natural forest.
Remember
Today's blends of 1–5% wood need no new equipment, so the plants go on burning mostly coal.
More than 400,000 of the 1.2 million hectares set aside for fuel plantations is still undisturbed tropical forest.
Japan and South Korea buy over 99% of Indonesia's wood pellet exports and count them toward their own clean-energy targets.
A data center waiting for a grid connection is expensive. Some industry estimates put lost revenue at about $3 million per megawatt per month, and in busy markets the connection queue runs for years. So operators are starting to offer "flexibility": a promise to draw less from the grid when it is stressed, in return for being connected sooner. Experts at Latitude Media's first Flex Summit last week said the hard part is working out how much flexibility, and what kind.
Batteries first, compute last
Flexibility can come from inside the facility, by adjusting cooling, power distribution or the computing workload itself. It can also come from on-site energy such as batteries or generators that cover demand during grid peaks.
Jake Elder of 1 (EIP), a climate-focused investment firm, said most deals he has seen rely on on-site power. The reason is organisational. A battery "doesn't require any real conversation with the end customer." Nobody has to touch cooling systems or renegotiate 2 (SLAs), the uptime guarantees owed to the customers whose workloads run there. Peter Hans Hirschboeck of impactECI said flexing compute is "probably the hardest" option, a point he said the energy industry doesn't grasp.
The leading public example is Aligned Data Centers. In 2025 it agreed with Portland General Electric to add a 31 MW, 62 MWh battery, which is about two hours at full output, in exchange for faster 3. Elder said several similar private deals exist.
Nobody agrees what "flexible" means
Seyed Madaeni, CEO of Verse, said the lack of a standard definition is the main obstacle. Should a site cut its load for two hours or eight? How fast must it respond? Each answer changes how much equipment the site has to buy, and each site faces a different constraint: a shortage of generating capacity in one place, a transmission bottleneck in another.
One EIP analysis, which Elder called "just one indicative study," modelled a territory with 200 hours of grid stress a year. Workload flexibility could cover about half of those hours and batteries most of the rest, leaving roughly one event a year that needs dispatchable backup such as gas. He warned it gets harder from there. The first peak hours to be shaved are short and rare, and each further hour means longer events.
The outlook
The panelists expect most data centers coming online by 2030 to use a mix of software, batteries and backup generation. Madaeni said the alternative is going fully off-grid. Until utilities and regulators set rules on duration, response time and location, every deal stays a one-off negotiation.
Data centers have long sold one thing above all: reliable compute. That is changing because getting a grid connection is now so slow. Some industry estimates put the revenue lost to a delayed data center at about $3 million per megawatt per month. In popular markets the queue to connect can run for years. At those numbers, a 100 MW campus that sits idle for a year gives up billions. Operators are therefore starting to accept "flexibility": a promise to draw less from the grid when it is stressed, in return for being connected sooner.
The real questions are how much flexibility and what kind. Speakers at Latitude Media's first Flex Summit last week took those on. Jake Elder, SVP of research and innovation at 1 (EIP, a climate-focused investment firm), named a basic obstacle: people from the energy world underestimate how complex data centers are, and data center people underestimate the grid. Conversations between the two, he said, often "miss the mark" because each side misjudges who it is talking to.
Two places to find flexibility
Flexibility can come from inside a facility or from outside it. Inside, an operator could in principle modulate cooling, power distribution, or the computing workload itself, for example by slowing or pausing jobs. Outside, it can add on-site energy such as batteries or generators to cover its own demand during peaks while its draw from the grid falls. The startup Emerald AI, which works on compute flexibility, calls this being "a good grid citizen." The payoff comes in two parts: faster 2 first, then a more stable grid as more big loads connect.
These options are not equally easy. Peter Hans Hirschboeck, founder of impactECI and an adviser to Emerald AI, said that of all the types, "the energy industry doesn't understand that compute is probably the hardest."
Why batteries are winning so far
Elder agreed. He said most projects he has seen trade flexibility for speed use on-site power, either 3 (on the customer's side of the utility meter) or front-of-meter (on the grid side but tied to the site). The reason is organisational, not technical. On-site power "doesn't require any real conversation with the end customer," he said. Engineers don't have to get comfortable adjusting cooling. Nobody has to renegotiate 4 (SLAs), the contracts that guarantee customers uptime and performance for their workloads. "You can just kind of turnkey transact."
The best-known public example is Aligned Data Centers. In 2025 it signed a deal with Portland General Electric to put a 31 MW, 62 MWh battery next to its facility in exchange for faster interconnection. That works out to about two hours of output at full power. Elder said several private bilateral deals of the same kind exist.
5 (VPPs), which are networks of distributed batteries, thermostats and other devices that a utility can dispatch together, are also being added to some data centers' capacity plans. The article notes this does something different. It shows the grid has room for the new load, rather than flexing the facility's own draw from day to day.
No shared definition
The biggest practical problem, according to Seyed Madaeni, co-founder and CEO of Verse, is that "flexible" has no standard meaning. How long must 6 (a requested cut in power use) last? How fast must the site respond? Do location standards apply? Without answers, flexibility is hard to write into grid planning or into the interconnection study for a single site. "Do I cut the load for two hours, four hours, six hours, eight hours? Because that impacts the amount of 7 that we need to bring in," he said. Every site differs, too. One location may face a shortage of generating capacity, another a transmission bottleneck.
How the mix might work out
Elder cited one EIP analysis as an illustration. Take a service territory with 200 hours of grid stress a year, with events of varying length and depth. Workload flexibility could cover about half of those hours, and batteries most of the rest. That would leave "maybe one outlier event per year" that needs something dispatchable, such as fossil generation. He was careful to call it "just one indicative study" that depends heavily on site and market.
He added a warning. As flexible resources "clip the far left side of the 8," the task gets harder. A load duration curve ranks every hour of the year from highest demand to lowest. The few top hours on the left are cheap to shave because they are short and rare. Each further step down means longer, more frequent events, which short batteries and brief workload pauses cover less well.
Hirschboeck said he weighs options on four things:
Cost, including lost revenue
Capability, meaning short versus long duration
Benefit
Second-order effects, such as pollution from gas generators
Where this leaves things
The panelists broadly agreed that most data centers coming online by 2030 will use at least one form of flexibility, and often several. Nima Amirzadeh, COO of LōD Technologies, described a stack: "a software layer, a battery layer, and a generation behind the meter, or some sort of backup generation." Madaeni put the alternative bluntly. Data centers that refuse to flex will be pushed off-grid entirely.
Much remains open. The article gives no figures on how much compute flexibility has actually been delivered in practice. It also doesn't say how the 200-hour EIP scenario was built. The rules are the central gap. "There is absolutely a path for operations here," Madaeni said. "What's not clear, which creates uncertainty, is what are the rules around this." Until regulators and utilities set standards for curtailment duration, response time and location, every deal is a one-off negotiation. That favours batteries and generators over the harder work of flexing compute.
Remember
Most speed-for-flexibility deals so far use on-site batteries or generators, because they avoid renegotiating customer uptime contracts.
The Aligned–Portland General Electric deal paired a two-hour, 31 MW battery with a faster grid connection.
Flexibility gets harder the deeper it goes: shaving rare peaks is cheap, while covering longer, frequent events is not.
