The data center boom is facing voters who care about bills, investors who care about cash, and regulators setting the rules on who pays.
Picked from 32 new articles across 13 sources · arrow keys for other days
Today in 30 seconds
30s left
In a survey experiment, rate guarantees raised support for local data centers. Solar power instead of gas made no significant difference. story 1 ↓
AI and power stocks fell after Meta went free cash flow negative on AI spending. GE Vernova dropped 10% in one day. story 2 ↓
PG&E cut $2 billion from 2027 spending as Fitch moved its outlook to negative at BBB-, the lowest investment-grade rating. story 3 ↓
Trump chose trade talks over new wind turbine tariffs, with a report due in 180 days. Steel, copper and aluminum tariffs still apply. story 4 ↓
Alabama regulators get 90 days to review each Alabama Power data center contract, or it's approved by default. The cost math stays secret. story 5 ↓
Only 77 heads of state or government are on COP31's early speakers' list, about a third fewer than at COP28's peak.
1/5Heatmap· Jael Holzman· Oct 23
Survey experiment: rate guarantees win data center support, solar power doesn't
GoodPower and the pollster Data for Progress showed Americans different hypothetical local data centers. Promises that power bills wouldn't rise, or home energy upgrades, raised support. Switching from gas to solar didn't.
Why it matters · Utilities and tech firms are deciding now whether to meet data center demand with gas or clean power. This suggests local acceptance hinges on electricity bills, not on which fuel is used.
Know this first
Context from the editor, not the article · 30 sec
1 · Data centers and power bills
AI data centers can draw hundreds of megawatts each. Utilities may build new plants and grid lines to serve them. If those costs are spread across all customers, household rates can rise, and that fear drives much local opposition.
2 · Large load tariffs
A large load tariff is a special utility rate for very big customers like data centers. It can require minimum payments, long contracts or upfront grid costs, so the new load pays for what it causes instead of households.
3 · Say versus do in polling
Plain polls ask people what they would favor. Survey experiments randomly show different groups different versions of a scenario. That isolates which feature actually changes opinion, and it often gives a different answer than asking directly.
2/5Heatmap· Robinson Meyer· Oct 22
AI and power stocks slide as investors question data center spending
After a hot inflation report and Meta's disclosure that AI spending has pushed its free cash flow negative, AI and power stocks fell sharply. Heatmap's Robinson Meyer argues the money behind the data center boom may be slowing.
Why it matters · The data center buildout has been the main driver of new US electricity demand and gas plant orders, so any slowdown in its funding changes energy forecasts and emissions.
Know this first
Context from the editor, not the article · 30 sec
1 · Why data centers drive power demand
Training and running AI models takes huge numbers of chips that draw a lot of electricity around the clock. After years of flat US demand, utilities and grid planners now forecast steep growth, much of it from data centers.
2 · Who sells the picks and shovels
Beyond chipmakers like Nvidia and cloud giants like Microsoft and Alphabet, the boom lifted power companies such as Vistra and equipment makers such as GE Vernova, whose gas turbines are booked years ahead.
3 · Free cash flow
Free cash flow is the cash a company has left from operations after paying for capital spending. When it turns negative, the company is spending more on building than its business brings in and must cover the gap with savings or borrowing.
3/5Utility Dive· Diana DiGangi· Oct 23
PG&E cuts $2 billion from 2027 spending, presses California on wildfire liability
California lawmakers failed to limit utilities' wildfire liability, and Fitch then warned it may downgrade PG&E. PG&E cut 2027 capital spending from $13.4 billion to $11.4 billion and is asking for liability caps and a permanent backstop fund.
Why it matters · In California, wildfire liability rules now limit how fast the grid can connect housing, renewables and big new loads. The cost of that risk shows up in customers' bills.
Know this first
Context from the editor, not the article · 27 sec
1 · Strict wildfire liability
Under a California doctrine called inverse condemnation, a utility whose equipment starts a fire can owe for the property damage even if it wasn't negligent. Liability from the 2018 Camp Fire pushed PG&E into bankruptcy in 2019.
2 · The Wildfire Fund
A 2019 state law, AB 1054, created a shared fund of about $21 billion, paid for by ratepayers and utility shareholders. It covers wildfire claims above a utility's insurance, so a single fire can't bankrupt the company.
3 · Why credit ratings matter
BBB- is the lowest investment-grade rating. A drop below it makes borrowing more expensive. Regulated utilities pass their interest costs on to customers, so ratepayers pay for a weaker rating.
4/5Utility Dive· Phil Neuffer· Oct 23
Trump chooses trade talks, not new tariffs, on imported wind turbines
After a 14-month national-security review of wind turbine imports, President Trump ordered trade negotiations rather than new tariffs. He also called foreign-made parts, inverters in particular, a cybersecurity risk.
Why it matters · Turbine costs drive the price of onshore wind power. Skipping a wind-specific tariff spares developers one cost hit for now, but keeps the threat of one open.
Know this first
Context from the editor, not the article · 27 sec
1 · What Section 232 is
Section 232 of the Trade Expansion Act of 1962 lets the president restrict imports after the Commerce Department finds they threaten national security. The president can respond with tariffs, quotas or negotiated agreements.
2 · US reliance on imported turbines
Much of the equipment in US wind farms, including blades, nacelles, towers and power electronics, comes from abroad. Big makers such as Vestas, Siemens Gamesa and Nordex build mostly overseas. GE Vernova is the main domestic manufacturer.
3 · The administration and wind
Since returning to office, the Trump administration has worked against wind power, especially offshore projects. It has tried to revoke permits and has offered to pay developers to give up their leases.
5/5Inside Climate News· By Dennis Pillion· Oct 23
Alabama adopts data center power contract rules, but cost math stays secret
Alabama's utility regulators approved a new review process for Alabama Power's contracts with data centers. Regulators see full contracts. The public gets redacted copies without the numbers showing whether other customers end up paying.
Why it matters · Data centers are driving new load growth at US utilities, and contract rules like these decide whether households end up paying for the plants and power lines built to serve them.
Know this first
Context from the editor, not the article · 28 sec
1 · Who regulates Alabama Power
Alabama Power, a Southern Company subsidiary, is the state's largest utility. It owns its own power plants and wires. The elected Alabama Public Service Commission (PSC) approves its rates and its special contracts with large customers.
2 · The cost-shift worry
A data center can need as much power as a small city. If a utility builds plants or lines for it and the contract doesn't fully cover those costs, the shortfall can end up in everyone else's rates.
3 · Tariffs and special contracts
A tariff is a utility's filed, regulator-approved rulebook for prices and terms. Large customers often negotiate their own contracts on top of it, and those contracts are usually kept confidential as trade secrets.
GoodPower, a clean-energy advocacy group led by CEO Leah Qusba, and 1, a progressive polling firm, have released a survey experiment on why Americans oppose AI data centers near them. The main finding: whether the facility runs on solar or gas made no measurable difference to local support. Promises about electricity bills did. Qusba walked through the results in an interview with Heatmap.
Why they ran an experiment
Earlier polls found that saying a data center would run on renewables raises its favorability. But Qusba points to a gap. Large-scale solar polls well, even with Republicans, and solar projects still get cancelled because of local protests. A direct question captures what people say. It doesn't capture how they react when a project is coming to their own town.
So the survey had two parts. The first was a standard poll, and it matched earlier results: people said solar power would make them more favorable. The second told respondents a data center was coming to their community. It varied the features of that project across 2, the versions shown to different groups. Each version was powered by either gas or renewables. Some included a binding guarantee that the data center would pay 100% of its costs, so household rates wouldn't rise. Others offered residential benefits such as rooftop solar, heat pumps or home batteries.
What moved people
Only two things raised support in the realistic setting: a guarantee of stable rates, and meaningful household or community benefits. Solar versus gas on its own was not 3. Statistically, the fuel choice had no effect. Qusba puts this down to the cost of living. She cites the figure that upwards of 65% of Americans live paycheck to paycheck.
The interview doesn't give the sample size or how much each condition raised support.
Other findings
No spillover. Linking solar to data centers didn't lower solar's favorability.
Trust. Respondents trusted consumer watchdogs most and renewable developers second. They trusted data center developers least.
AI itself. Feelings about AI were the strongest driver of opposition, ahead of political ideology. Qusba says anti-AI sentiment is spread evenly across Democrats, Republicans and independents.
The policy angle
Qusba says proposals for 4 plants have roughly doubled in the past six months, as tech companies hunting for power ease off earlier clean-energy pledges. She wants federal permitting reform to keep strong ratepayer protections and 5. She also backs "bring your own clean energy" rules, which let big customers supply their own new clean generation. Her pitch to tech firms is to pair clean power with rate guarantees. Solar won't win local approval on its own merits, but clean-energy developers are trusted more.
GoodPower, a clean-energy advocacy group, ran a survey experiment with the progressive polling firm 1. It was published on October 19 and asks why Americans oppose AI data centers in their communities, and what would change their minds. The headline finding cuts against a common industry message. Telling people a local data center would run on solar rather than gas did not significantly raise their support. Guaranteeing that their electricity rates wouldn't go up did. CEO Leah Qusba discussed the results with Heatmap's Jael Holzman.
The backdrop: a fork between gas and clean power
Qusba starts with the market. Over the last 15 years or so, Big Tech has been the world's largest buyer of decarbonized electricity. Data center growth is still adding clean energy. But the race for power is now urgent enough that, in her words, earlier clean procurement commitments "have slid a little bit." She says proposals for 2 plants have roughly doubled over the past six months. These plants burn gas in a turbine and use the waste heat to run a second, steam turbine, and they typically run for decades. She frames this as a crossroads. One path uses data center demand to fund new clean energy while protecting 3, the ordinary utility customers who share grid costs. The other locks in decades of gas.
The problem: people say one thing and do another
Earlier polls, widely covered in the press, found that saying a data center runs on renewables raises its favorability. Qusba's question is why that doesn't show up on the ground. Large-scale solar polls well even among Republicans. Yet 4 ("not in my backyard" opposition to local projects) still gets projects cut back or cancelled. Her explanation is a gap between stated preferences and how people respond when a real project is proposed nearby.
How the experiment worked
The design had two stages.
A conventional poll. Respondents were asked how they felt about data centers powered by different sources. The results matched earlier surveys: say "solar," and stated favorability goes up.
A randomized scenario. Respondents were told a data center was coming to their own community. The project's features were varied across 5, the different versions shown to different groups. The power source was either gas or renewables. Other features were added or left out. One was a rate-binding guarantee: the data center pays 100% of its costs and household rates don't rise. Another was residential benefits such as rooftop solar, heat pumps and home batteries.
Here is the logic, with made-up numbers for illustration only. Suppose one group sees "gas-powered data center" and another sees "solar-powered data center," and both are 40% supportive. Then the fuel choice isn't doing anything, whatever people said in stage one. If a third group sees "gas-powered, with a rate guarantee" and is 52% supportive, the guarantee is what moves opinion. Randomizing which version each person sees means differences between groups can be pinned on the feature itself.
Results
Qusba calls the result "pretty definitive." Only two features raised support in the realistic scenario:
a guarantee of stable electricity rates, and
meaningful household or community benefits.
The power source on its own, solar versus gas, did not reach statistical significance. Any difference was too small to tell apart from random noise, which Qusba calls a null effect. Pairing solar with rate protections did help.
The interview does not report the sample size, the size of each effect, or the confidence intervals. So it can't tell us how much a rate guarantee helps, only that it registered and the fuel choice didn't. Readers who want the magnitudes need the full Data for Progress release.
Why technology didn't matter
Qusba doesn't think people are indifferent to technology. She thinks cost crowds everything else out. She cites the figure that upwards of 65% of Americans live paycheck to paycheck, and calls affordability "the queen issue." That is her reading of the data, not something the experiment tested directly.
Secondary findings
No spillover. Some worried that tying solar to unpopular data centers would drag down solar's reputation. The survey found zero spillover in solar favorability.
Trust rankings. Respondents trusted consumer watchdogs most and renewable energy developers second. They trusted data center developers least. Qusba's advice to tech firms is to buy clean power where they can, because being linked to a trusted developer helps, even though the fuel alone doesn't win support.
AI is the main driver. Attitudes toward AI as a technology were the strongest predictor of opposition, ahead of political ideology and even concern about climate change. Qusba says anti-AI sentiment is spread evenly across Democrats, Republicans and independents.
Policy implications, as Qusba argues them
Qusba connects the findings to federal permitting reform. She wants it to keep "maximum ratepayer protections" and 6, special utility rates that make very large customers pay for the grid upgrades and generation they cause. She also backs 7, "bring your own clean energy." Under these arrangements a large customer supplies or pays for its own new clean generation, so it doesn't strain the existing grid. In her view, the survey results support both.
What to keep in mind
GoodPower is an advocacy group and Data for Progress is a progressive pollster. The finding that suits their goals is that clean energy plus rate protection is the winning package. The finding that cuts against them is that solar alone doesn't move people, and they report it anyway. That adds some credibility.
Several questions remain open. Can a rate guarantee be credibly enforced? Utility regulators, not developers, set rates. Do hypothetical scenarios predict behavior at real public hearings? Qusba says they plan to dig further into the unexpected results, especially the role of AI sentiment. If feelings about AI are the strongest driver, rate guarantees may only go so far.
Remember
What people say in polls about solar didn't predict their support when shown a realistic local project.
No measurable spillover: linking solar to data centers didn't hurt solar's own favorability.
Opinions about AI, not partisanship, were the strongest predictor of opposition to data centers.
Stocks tied to AI and data centers fell sharply this week. Robinson Meyer, Heatmap's executive editor, argues this may be the first sign that the money behind the data center boom is slowing.
What happened
Two things set off the selloff. An inflation report last week came in hotter than expected. Then on Wednesday evening, Meta disclosed it has gone 1 negative to pay for its AI buildout. Free cash flow is the cash a company has left after capital spending, so Meta is now spending more on construction than its business brings in.
The next day:
2, which rents out AI chips, fell about 10%.
3, a power generator that sells to data centers, fell about 10%.
4, maker of gas turbines, fell 10%.
Meta fell almost 7%.
Microsoft, Alphabet and Nvidia also dropped.
Over a longer window, Coreweave is down 54% in six months, and Alphabet and GE Vernova are below their summer levels. Nvidia and Microsoft are still well up.
Two loops, one wobbling
Meyer explains the boom as two "5." First, AI models got better the more data and computing power they had. Second, investors kept paying for more scale as long as the models kept improving. Together, he writes, these made the data center boom one of the biggest infrastructure buildouts in US history.
The second loop is the one now under strain. Meyer is careful here. Investors are not (yet) calling AI a dud. They are asking whether companies need to build so much so fast. He says this is what a slowdown could look like, not proof that one has started.
How fast things changed
Meyer opens with the case of 6. In 2023 the company went public with a gas power plant design that captures its carbon. At the time, the Environmental Protection Agency was drafting rules that would have required new gas plants in the 2030s to capture their emissions. Now AI demand has electricity use surging, but Net Power hasn't commercialized its plant. Meyer says few buyers would pay its premium, and the government won't require it.
The caveats
The evidence so far is stock prices and one company's cash flow statement. The piece does not point to any cancelled data center, power plant or turbine order. Still, turbine orders, power contracts and utility demand forecasts all assume the buildout keeps growing fast. If its funding tightens, those assumptions, and the emissions that come with new gas plants, shift too.
AI-linked stocks fell hard on Wall Street this week. Robinson Meyer, Heatmap's executive editor, reads the drop as the first sign that the money behind the data center boom may be slowing. Two things set it off: an inflation report last week that came in hotter than expected, and a disclosure on Wednesday evening that Meta has gone 1 negative to pay for its AI buildout. Free cash flow is the cash a company has left after paying for capital projects. Meta is now spending more on construction than its business brings in.
The day's moves
Companies close to AI or data centers all fell:
2, a cloud company that rents out AI chips, fell about 10%.
3, a power generator that sells electricity to data centers, fell about 10%.
4, which makes gas turbines, fell 10%.
Meta lost almost 7%.
Microsoft, Alphabet and Nvidia also fell during the day.
Meyer argues the decline looks bigger over a longer window. Since late June, much of what he calls "the AI portfolio" has dropped below where it started. Coreweave is down 54% over six months, and Alphabet and GE Vernova are also lower than they were in summer. The pattern is uneven, though: Nvidia and Microsoft are still well up.
Two 5
Meyer explains the boom as two trends feeding each other. He admits he didn't foresee either when he co-founded Heatmap a little over three years ago.
Models improved with scale. The more data, training and, above all, computing power AI models got, the better they performed.
Wall Street paid for the scale-up. As long as models kept improving, investors kept lending and investing more in the companies building them.
Put together, Meyer writes, these two loops made the data center boom one of the biggest infrastructure buildouts in American history. He also notes that people who had studied computing's energy use for years did not see the near-term demand surge coming either.
His point is that the second loop is the one now wobbling. He is careful about what that means. Investors are not, or not yet, calling AI a failed business. They seem to be asking whether companies need to build so much so quickly. A slowdown driven by "finance-applied friction" would look like this week's selloff, Meyer argues, though he says it guarantees nothing.
A cautionary example: 6
Meyer starts with a colleague's reporting on Net Power, which shows how fast the energy picture has changed. Net Power went public in 2023 with a design for a gas-fired power plant that would emit little or no CO2, partly by capturing and storing carbon. At that time the Environmental Protection Agency was drafting rules that would have required new gas plants built in the 2030s to capture their emissions. Net Power's technology looked like one of the best ways to meet that standard.
By 2026, AI has sent electricity demand soaring, the setting a gas-plant developer would hope for. Yet Net Power has not commercialized its plant. Meyer writes that even if it had, few buyers would pay its large premium when cheaper options exist, and the government is not going to require it. In the rush to build, buyers chose cheaper gas plants over cleaner ones. The newsletter only previews this story and does not say what happened to Net Power next.
What to watch
The energy stakes are why this matters beyond the stock market. GE Vernova's turbine order book, Vistra's power contracts and utilities' demand forecasts all assume data centers keep growing fast. If funding tightens and building slows, those assumptions weaken. That would affect how many new gas plants get built and how much CO2 the buildout adds.
The article leaves several questions open:
Is this a correction or a turn? A few weeks of falling prices after one inflation report do not make a trend, and Nvidia and Microsoft are still up. Meyer gives no threshold for when investor caution would actually slow construction.
What does it mean for power projects? The piece does not tie the selloff to any cancelled or delayed data center, power plant or turbine order. The link between stock prices and steel in the ground is inferred, not shown.
Does the first scaling loop still hold? Meyer's argument rests on the money faltering, not the technology. He doesn't address whether models are still getting better with scale.
For now, the evidence is market prices and one company's cash flow statement. It is a signal worth watching, but not yet proof that the boom is slowing.
Remember
Coreweave is down 54% in six months, while Nvidia and Microsoft are still up, so the selloff is uneven.
Meyer says the funding loop is faltering, not the technology. Investors are questioning how fast to build, not AI itself.
Net Power shows the cost of the rush: demand surged, but buyers chose cheaper gas plants over carbon-capture ones.
Pacific Gas & Electric (PG&E), California's largest investor-owned utility, used its third-quarter earnings call to press the state on wildfire liability. CEO Patti Poppe said operations are "strong and improving." She said the company's finances will stay strained until lawmakers limit its exposure to catastrophic fires.
In September, the legislature adjourned without passing liability protections in Senate Bill 492. PG&E's stock had already fallen 20% when the bill advanced without them. On Sept. 16, Fitch Ratings kept PG&E at 1, the lowest investment-grade rating, but changed its outlook to negative, which means a downgrade is possible. Fitch cited "no meaningful progress" on spreading wildfire liabilities more broadly.
Why it's hard
California courts can hold a utility liable for fire damage its equipment causes even without negligence. One large fire can create claims bigger than the company. That is the "2" Poppe means, and it pushed PG&E into bankruptcy after the 2018 Camp Fire. A shared state 3, created in 2019 and paid for by ratepayers and utility shareholders, covers claims above insurance. But the fund can be drained.
What PG&E wants
PG&E is asking for two things:
a permanent source of cash to pay claims beyond the Wildfire Fund
a hard cap on 4, meaning costs that regulators refuse to let the utility pass on to customers, no matter how claims are paid
Poppe also wants insurers to limit 5, where insurers pay homeowners and then sue the utility to recover the money. A legislative hearing on the subject is set for Oct. 29.
She put a price on doing nothing. Debt PG&E has issued as sub-investment grade since 2017 will cost customers an extra $4 billion in interest over its life, she said.
The spending cut
On Sept. 2, PG&E cut 2027 capital spending from $13.4 billion to $11.4 billion, about 15%. The delayed work includes connecting new housing, connecting new renewable projects, technology upgrades, and large loads beyond a first 1.8-gigawatt 6. For scale, PG&E estimates its plan to put lines underground against wildfire risk at $16.6 billion.
Data centers
Poppe said the 1.8 GW of data centers promised by 2030 is still in the plan. The total pipeline is 12.87 GW, but only 174 MW has been connected since 2024. Housing and renewables are absorbing the cuts while the data center tranche is protected.
What's open
The Oct. 29 hearing is a discussion, not a vote. A Fitch downgrade would push PG&E's borrowing costs higher still. The article doesn't cover the usual objection to liability caps, which is that they weaken a utility's incentive to prevent fires.
What happened
Patti Poppe is CEO of Pacific Gas & Electric, the largest investor-owned utility in California. On PG&E's third-quarter earnings call on Thursday, she said the company's day-to-day operations are "strong and improving every day." Its finances, she said, will stay under pressure until California changes how it assigns liability for wildfires. She called the credit rating "a major issue that must be solved."
The pressure has been building since summer. California lawmakers advanced Senate Bill 492 without language shielding utilities from wildfire liability, and PG&E's stock fell 20%. Negotiations ran into a special session, but the bill never got a final vote before the legislature adjourned in September. On Sept. 16, Fitch Ratings kept PG&E Corp. and its utility subsidiary, PG&E Co., at 1, the lowest investment-grade rating. It changed their outlook from stable to negative, which signals that a downgrade is possible. Fitch said the session "ended with no meaningful progress towards a more durable solution to socialize wildfire liabilities."
Why wildfire risk is hard to finance
The issue is 2: the small chance of a very large loss. California courts apply a doctrine called 3. Under it, a utility whose equipment ignites a fire can be liable for the property damage even without negligence. One catastrophic fire can produce claims bigger than the company. After the 2018 Camp Fire, PG&E filed for bankruptcy.
The state's main response so far is the 4. A 2019 law, AB 1054, set it up at about $21 billion, paid for by ratepayers and utility shareholders, to pay claims above a utility's insurance. Credit agencies care whether a backstop like this is permanent. A fund that a few bad fire seasons could drain does not remove the tail risk. It only delays it.
What PG&E is asking for
According to PG&E's earnings presentation, the company wants two things from lawmakers:
A permanent source of liquidity beyond the Wildfire Fund. This would be cash that is reliably available to pay claims, not a fund that can run out.
A maximum 5 independent of funding source. A disallowance is a regulator's ruling that the utility cannot recover a cost from customers, which leaves shareholders to pay it. PG&E wants a hard cap on that amount, whether the claims are paid from the fund or from somewhere else.
Poppe also wants insurers involved. Without legislation, she said, "insurance still needs to join in discussions… to make sure that claims are limited and our customers are not the infinite backstop for insured properties." She is referring to 6. When insurers pay homeowners after a fire, they can sue the utility that caused it to recover what they paid, and those recoveries end up in utility costs and rates. A joint legislative hearing on utilities, energy and insurance on Oct. 29 is set to discuss subrogation. Poppe said talks have been "productive."
The cost of a weak rating
Poppe put a number on it. She said all the debt PG&E has issued as sub-investment grade since 2017 will cost customers an extra $4 billion in interest over the life of that financing. She described this as a cost "that can be immediately remediated with good long-term policy." She also argued "it's very realistic that this business be investment-grade." The article does not say which ratings or which bonds the $4 billion covers.
The $2 billion capital cut
On Sept. 2, PG&E began a strategic review and cut its 2027 capital plan from $13.4 billion to $11.4 billion, about 15%. Poppe said at the time that the liability framework "create[s] financing risks that drive higher costs, affect customer affordability, and limit investment in the energy system."
The work being delayed or deferred includes:
connecting new housing projects
interconnecting new renewable generation
technology upgrades
large loads beyond an initial 1.8-gigawatt (GW) 7
Poppe said $11.4 billion still leaves money for "critical safety issues, our critical compliance issues" and for projects already under way that are meant to lower rates. A separate figure shows how big the safety program is: PG&E's plan to put power lines underground to reduce wildfire risk is estimated at $16.6 billion.
Data centers
Poppe said the cut does not affect the 1.8 GW of data center projects PG&E has said will be online by 2030. The full data center pipeline is 12.87 GW, up slightly from 12.71 GW in June. PG&E has connected 174 MW of data centers since 2024. That is under 10% of the 1.8 GW committed and about 1.4% of the pipeline. Pipelines like this usually include many speculative requests, so the 1.8 GW figure is the one to watch.
The cut also shows who loses when capital is limited: housing hookups and renewable interconnections are deferred, while the data center tranche is protected. The article does not say how many housing or renewable projects are affected, or for how long.
What's still open
The legislature. The Oct. 29 hearing is a discussion, not a vote. Nothing in the article suggests a bill is close.
Fitch's next move. A downgrade from BBB- would push PG&E below investment grade at Fitch and make Poppe's interest-cost argument worse.
Who pays. Each option shifts costs among ratepayers, shareholders, insurers and fire victims. PG&E argues that capping its exposure would lower customer costs through cheaper borrowing. Critics of liability limits generally argue that limits weaken a utility's incentive to prevent fires. The article doesn't cover that side.
Deferral costs. The article gives no estimate of what delaying renewable and housing connections will cost California's clean-energy and housing goals.
Remember
PG&E puts the extra interest on its junk-rated debt since 2017 at $4 billion, a cost customers pay.
The capital cut defers housing and renewable connections but protects 1.8 GW of data centers due by 2030.
Only 174 MW of a 12.87 GW data center pipeline has been connected since 2024.
Wind turbines at sunset in Fenton, Iowa. Iowa is one of the biggest US wind-power states.
On Thursday, President Donald Trump decided against new tariffs on imported wind turbines and parts. He ordered trade negotiations instead.
The decision closes a 1 investigation that the Commerce Department opened in August 2025. Section 232 is a 1962 law that lets the president restrict imports judged a national-security threat. Trump directed Commerce Secretary Howard Lutnick and 2 Jamieson Greer to open or continue talks with trading partners and to report on progress within 180 days, which is around April 2027.
What the probe concluded
According to the 3, the U.S. "is reliant" on imported turbines and components, and those imports have grown over the past decade. It also says foreign-made parts "present unique cybersecurity risks." It singles out 4, the power electronics that turn a generator's output into grid-ready alternating current, and says "the vast majority" are made abroad. The reporting does not describe a specific threat or give import figures.
The order also calls wind "unreliable" and "burdensome" on the grid and cites the investigation's findings. That is the administration's characterization. The article provides no supporting data.
What actually changes
There is no new tariff aimed at wind. Some turbines and parts already pay tariffs on steel, copper and aluminum, and those stay in place. Trump also said he would consider "alternative remedies" depending on how the talks go, so the 180-day report could bring tariffs back.
Context
Most of this administration's Section 232 cases have ended in tariffs, including those on drones, furniture and some semiconductors. There is one recent exception: in July, a probe into commercial aircraft and jet engines also led to negotiations rather than tariffs.
The decision doesn't signal a friendlier stance on wind. The administration has tried to revoke offshore wind permits and has offered to pay developers to give up offshore leases. The proclamation does not say which countries are involved in the talks or what the U.S. wants from them. Until the report arrives, developers face uncertain equipment costs even without a new tariff.
On Thursday, Oct. 22, President Donald Trump signed a 1 that closes out a 2 investigation into imported wind turbines and their components. Section 232 is a 1962 trade law that lets the president restrict imports the Commerce Department judges a threat to national security. Trump did not impose a new tariff on wind equipment. He told Commerce Secretary Howard Lutnick and 3 Jamieson Greer to open or continue talks with trading partners about the imports, and to report back on progress within 180 days.
What the investigation found
Commerce opened the probe in August 2025, so it ran about 14 months. The proclamation repeats several of its conclusions:
Dependence. The U.S. "is reliant" on imported turbines and parts, and those imports have grown over the past decade. The article gives no import volumes or values.
Cybersecurity. Foreign-made turbines and parts "present unique cybersecurity risks." The order names 4 and says "the vast majority" are made outside the U.S. Inverters are the power electronics that convert a generator's variable electrical output into grid-ready alternating current. Because they control how the machine feeds the grid, and are often networked for remote monitoring, they are the most plausible route for a cyberattack. The proclamation, as reported, does not describe a specific threat.
What changes and what doesn't
No tariff specific to wind turbines takes effect. Tariffs already in place on steel, copper and aluminum still hit some turbines and parts, and the order leaves them alone. Towers are mostly steel, and generators and cabling use a lot of copper, so those metal tariffs already add to project costs. The article does not say how much. The proclamation also keeps tougher action available: Trump said he would consider "alternative remedies" depending on how the talks go. In practice the 180-day report is a checkpoint at which tariffs could return.
Where this fits
Most of the administration's Section 232 cases have ended in tariffs. The article lists drones and drone components, furniture, and some semiconductors. There is one recent precedent for choosing talks first: in July, a Section 232 probe into commercial aircraft, jet engines and parts ended the same way.
What sets the wind decision apart is the policy around it. The administration has tried to revoke offshore wind permits and has offered to pay developers to give up offshore leases. Against that record, passing on a tariff looks less like support for wind and more like a holding position that keeps the option open. The article does not say why the administration chose this route. Possible reasons include the effect on domestic power prices, US makers' own reliance on imported parts, or bargaining leverage with Europe. All of these are guesses the reporting does not confirm.
What's still open
Which partners, and what deal? The proclamation does not say which countries the talks cover or what concessions the U.S. wants. Supply-chain security terms, local-content rules and limits on Chinese components are all possible. None is confirmed.
Inverter rules. If the cybersecurity concern turns into procurement rules or certification requirements for inverters, the effect on projects could be bigger than a tariff. The order does not set any such rules.
Project economics. Developers can't count on stable equipment costs until at least the 180-day report, which is due around April 2027. That uncertainty affects turbine orders and power-purchase pricing in the meantime, even with no new tariff in place.
Remember
Existing steel, copper and aluminum tariffs still raise turbine costs. Only a wind-specific tariff was avoided.
The 180-day report, due around April 2027, could bring tariffs or other remedies.
Calling inverters a cybersecurity risk could lead to supply-chain rules beyond tariffs.
Alabama now has a dedicated process for approving Alabama Power's electricity contracts with data centers. The state's largest utility had been treating these huge, power-hungry facilities like any other large industrial customer. The Alabama 1 (PSC), the elected body that regulates the utility, voted 3-0 earlier this month to approve the framework and issued its final order this week. Alabama Power filed a matching 2 (its official rulebook of prices and terms) on Wednesday, and the rules take effect November 23.
Most observers agree the rules are better than what Alabama had before. Several advocacy groups say they still don't let the public check who pays.
How the review works
Each new data center contract goes unredacted to the PSC and to the Alabama Attorney General's office, which is the state's legal 3.
Those agencies have 90 days to confirm that the contract serves the public interest and that the data center pays the full cost of its electricity, including new infrastructure and fuel.
The public then gets a redacted copy. It must name the location, the maximum power capacity, and the company and its parent.
Alabama Power says data centers already pay their full cost of service, and that state law requires it.
What stays hidden
The inputs and findings of the 4 stay confidential. That test is the analysis meant to show that other customers aren't subsidizing the data center. "The numbers that matter most are still secret," said Monika Gerhart, who heads an association that filed comments asking for more disclosure. She says outsiders can't verify that families aren't paying a data center's bills, or that fuel-free resources like solar "got a fair shot" when the utility planned how to serve the load.
Claire Douglass of Manifest Energy, a nonprofit that promotes grid modernization, put it more bluntly: "We want to see the math, we want to see the receipts."
The default-approval clause
The PSC must publicly vote on each contract within 90 days. But under a "5" provision, a contract counts as approved if the commission doesn't act in time. Advocates worry this lets commissioners skip public votes on controversial projects. Contracts approved this way would only be announced at the next meeting. The PSC says it intends to vote on every contract.
Jaclyn Brass of the Southern Environmental Law Center called the process "better than what we had before." She said much more is needed to keep customers from footing the bill for the data center build-out.
Alabama has new rules for how its largest utility signs electricity contracts with data centers. The Alabama 1 (PSC) is the elected body that sets rates for Alabama Power, a Southern Company subsidiary that owns its own power plants and lines. Earlier this month it voted 3-0 to approve a data center contract framework, and it issued the final order this week. On Wednesday Alabama Power filed a new 2 (the regulator-approved rulebook of prices and terms) to put the process in place. It takes effect November 23.
Until now, Alabama treated data centers like any other large industrial operation. Nearly everyone quoted in the article agrees the new process is an improvement. The argument is over whether it lets anyone outside government confirm that households aren't paying for the build-out.
The problem being solved
A large data center can draw as much power as a sizable town, and it can arrive faster than a utility normally plans for. Serving it may require new generation, transmission lines and substations, plus the fuel to run the plants. If the contract price doesn't cover all of that, a 3 like Alabama Power can recover the shortfall through rates paid by everyone else. That cost-shift risk is what regulators across the US are now trying to manage. It's hard to police because the contracts are negotiated privately and usually treated as commercially confidential.
Alabama Power's position is that this is already handled. "Alabama Power will continue to require large load data center customers to pay the full cost to serve their energy needs, so costs are not shifted to existing customers," the company said. It added that state law sets the same standard.
How the new process works
Full disclosure to regulators. Each contract for a new data center goes unredacted to the PSC and to the Alabama Attorney General's office. Under Alabama law, the AG is the state's designated advocate for ratepayers.
A 90-day review. The two agencies check that the contract is "consistent with the public interest." They also check that the data center will cover every cost of serving it, including new infrastructure and fuel.
A public vote. The PSC must vote publicly to approve or reject each contract within 90 days of receiving the application.
A redacted public copy. The public version must state the facility's location, its maximum power capacity, and the company behind it, including the parent company. Other terms can be withheld.
What critics say is missing
The main complaint is about the 4. That is the analysis meant to show whether the contract leaves other customers better or worse off. Its method, inputs and findings stay confidential. The article doesn't describe how the test is built.
Monika Gerhart, who heads an association that filed comments in the PSC proceeding, said the final rule delivered "a good share" of the transparency her group sought. (The article's text cuts off the association's name.) But she said "the numbers that matter most are still secret." Without them, she argues, "nobody outside the Commission can check that Alabama families and businesses aren't paying a data center's bills." Nor can anyone check that "resources with no fuel bill, like solar, got a fair shot," meaning whether the utility fairly considered cheaper ways to supply the new load.
Claire Douglass of Manifest Energy, a nonprofit that pushes for grid modernization, efficiency and lower costs, made the governance case: "We want to see the math, we want to see the receipts. Just saying, 'Trust us, this is going to benefit families,' isn't really a good governance model."
Daniel Tait of Energy Alabama welcomed the PSC's list of what it will review before voting. He said he was disappointed that the details of what will be required are not public.
The deemed-approval loophole
The rule's own timeline has a fallback. Under a "5" provision, a contract counts as approved if the commission takes no action within the review period. Advocates fear this gives commissioners a way to avoid a recorded public vote on a contentious project. Contracts approved this way would only be announced at the next commission meeting.
A PSC spokesman said the commission weighed many public comments before keeping the provision. The final process, he said, "retains that 'deemed approved' provision while making clear the Commission intends to vote on each contract within the 90-day review period." He added that the provision doesn't change the public interest standard. Even so, the commission has stated an intention to vote rather than accepted a binding duty to.
How to read it
On paper, Alabama now checks each contract against a cost-recovery standard, gives the 6 the full contract, and sets a deadline for decisions. Those are real changes from treating data centers as ordinary industrial load. But the public can verify the outcome only to the extent the regulators publish their reasoning. Under these rules they don't publish the cost analysis.
Jaclyn Brass of the Southern Environmental Law Center summed up the critics' view: "The process that was adopted is better than what we had before. But much more needs to be done to ensure that customers aren't going to foot the bills for the data center build-out." She also called for more accountability in how Alabama Power is regulated generally.
Open questions
The article doesn't say how many data center contracts are pending or expected. It doesn't give the load involved or explain how the rate impact test is calculated. Contracts start flowing through the new process in late November. Whether the commission actually votes on each one, or lets some pass by default, will be the first visible test of the rules.
Remember
Regulators and the Attorney General see full contracts; the public sees location, capacity and owner, but not the cost analysis.
A contract the commission doesn't vote on within 90 days is approved by default, despite the commission's stated plan to vote.
Critics accept the rules as progress; their objection is that the cost-shift claims can't be independently checked.
Grid impact. The order calls wind energy "unreliable" and "burdensome" on the domestic grid, and cites the investigation as its source. That is the administration's position. Grid operators generally treat variable output as something to manage with forecasting, reserves and storage. The article offers no data either way.
