Deep Dive — Congress has the data center numbers, still no bill

Share
Deep Dive — Congress has the data center numbers, still no bill

A yearlong Senate investigation into seven of the biggest data center developers in the country concluded this week that the public case for the AI buildout does not survive the companies' own paperwork — and it landed at the exact moment Congress needs it, because the one federal bill written to make data centers pay for their own power fell three votes short of advancing in the Senate, despite passing the House 417-3. The report, led by the offices of Senators Elizabeth Warren, Chris Van Hollen and Richard Blumenthal and first shared with TIME on October 8, is not a law and will not be one this year. What it offers instead is something rarer in this debate: numbers pulled directly from Amazon, Google, Meta, Microsoft, CoreWeave, Digital Realty and Equinix, plus their own answers, on the record, about who pays for the grid the boom is forcing everyone to rebuild.

How the fight got here

The report arrives late to a fight that has moved fast all year. Community pushback was already forcing concessions before Senate staff finished their questions: Amazon said in early October that it had stopped using NDAs with county officials, with Wired reporting that the company itself linked the change to local backlash that was producing project moratoriums. Reporting by the Wall Street Journal in August described how Amazon and Gilroy, California quietly negotiated a $2 billion data center project that never required a public meeting or a vote — the kind of deal the Senate report argues NDAs have enabled elsewhere. Data centers have since become a campaign-trail issue for both parties, and President Trump has warned publicly that opposition could drive the buildout overseas. What changed in Washington this week is not public opinion; it is that Senate staff can now answer the industry's talking points with the industry's own figures.

What the paperwork says

The jobs claim is the load-bearing argument data center developers make to win incentives, and it is the first thing the report takes apart. When Senate staff asked the seven companies for comprehensive information on permanent employment, several refused to provide it. The ones that answered described roughly one permanent worker per megawatt of power demand. The arithmetic is stark enough to state plainly: a 100-megawatt facility can consume about as much electricity as 100,000 homes while employing around 100 people once construction crews leave. Construction employment is real but temporary; the incentive case has always rested on what remains, and what remains is thin.

The second finding is about where the subsidy actually hides. Public fights over data centers center on property tax breaks, but the report concludes the most lucrative incentive is the quieter one: sales-tax exemptions on computer equipment. That matters because of the hardware mix — GPUs account for an estimated 39 percent of spending at an average gigawatt-scale AI data center. Tax the building and you collect on concrete; exempt the machines and you forgo the revenue on the single largest line item in the structure.

The transparency section is where the report names names. All four big tech companies in the investigation — Amazon, Google, Meta and Microsoft — have routinely sought nondisclosure agreements while developing these projects, and some acknowledged asking government officials to sign them. The report's argument is not that NDAs are illegitimate when negotiations involve commercially sensitive terms; it is that they have been used to wall off scrutiny of deals involving tax dollars, utility rates and public infrastructure. The responses split the group: Microsoft told investigators it will stop seeking NDAs with local governments, though it will keep using them with state agencies, public utility commissions and utilities; Amazon announced a similar policy, consistent with Amazon ends data-center NDAs, pledges $1B to host towns; Google and Meta declined to commit to anything.

The grid question nobody will answer

The heart of the report is the cost-allocation fight, and the finding is blunt: none of the seven companies would agree to a standard requiring them to pay for new power infrastructure that would not have been needed if not for their data centers. The companies' counter-position, which they stated to investigators, is not unreasonable on its face — they say they pay the direct costs of serving their own facilities, and that larger investments like new power plants and transmission lines benefit other customers too, so they should not automatically be assigned to the company that helped drive the need.

Louisiana is where that abstraction gets a dollar figure. The report points to a power plant in Richland Parish that the local utility, Entergy, is seeking to buy, where analysts argue the purchase was driven primarily by Meta's planned $50 billion data center — a campus expected to draw 4,500 megawatts, roughly four times the peak electricity demand of the entire city of New Orleans. Estimates cited in the report indicate the purchase could raise electric bills for the average Entergy customer by $8 to $13 a month. Meta has disputed that its project is responsible for the costs. Both things can be true: the plant serves a system wider than one customer, and one customer's load is the reason the plant is being bought at all. That ambiguity is not a bug in utility regulation — it is the central question the field has never settled, and it is now arriving at state commissions at gigawatt scale.

The bill that 417-3 couldn't save

For all the report's detail, the most sobering data point came last month, before the report existed: the House passed the Ratepayer Protection Act 417-3, a vote so lopsided it would normally end the story. As we covered in The House votes 417-3 to make data centers pay for their own power, the bill would direct states to consider standards requiring large-load customers to cover the incremental costs of the generation, transmission and distribution infrastructure needed to serve them. It then stalled in the Senate, where Democrats argued it was toothless precisely because it only told states to consider such standards rather than requiring costs be pushed to the companies. The final vote to advance it was 57-43 — short of the 60 needed, and short because the bill's friends in the House had written something its friends in the Senate considered too weak.

That failure reframes what the report is for. When Congress reconvenes after the midterm elections, the investigation gives lawmakers a fresh set of findings to work with as they consider federal standards for large electricity users, transparency around data center deals, and public subsidies for the industry. But most authority over electricity rates still sits with state regulators, which is why the real action this year played out in places like Virginia ends by-right approval for big data centers, as states rewrite how these projects get approved at all. A Senate staff report changes the granularity of the argument in Washington; it does not move a single rate case.

What skeptics say

The obvious caveat is in the byline: the investigation was led by three Democratic senators, and Warren has already called for a national moratorium on new AI data centers until developers agree to cover the full costs. A report whose lead author wants a freeze is written to support that conclusion. TIME itself notes the findings "largely match reporting from various news outlets over the past year," which cuts both ways — the numbers are corroborated, but they are not, individually, a bombshell.

The industry's rejoinder is also on the record in the report rather than in a press release: the companies say they pay their direct serving costs, they bring construction jobs and a tax base to host communities, and grid investments do broadly benefit the customers on the other end of the wire. President Trump has warned that opposition to data centers could push the AI buildout overseas, and public opinion gives that argument traction — about half of Americans say data center construction is bad for the country, according to an Economist/YouGov poll taken August 28-31, which also means half do not. A companion analysis in the investigation, covered by Bloomberg, adds that state and local tax breaks are draining public revenue while the qualification thresholds for claiming them sit low enough that projects clear them with little effort.

The deeper skeptical point is about leverage rather than facts. Congress already had near-unanimity and still produced no law; a staff report cannot fix a jurisdiction problem. The countable facts entering the record matter because they change what state commissions and congressional committees argue about — not because they settle who pays.

What to watch

The first test is whether the report produces a fresh Senate vote on the Ratepayer Protection Act when Congress returns, this time with the floor holding findings of its own. The second is whether Google's and Meta's refusal to drop local NDAs becomes the next transparency flashpoint, now that Microsoft and Amazon have set a baseline. The third is state-level: Entergy's Richland Parish purchase and Meta's disputed cost attribution will be litigated at the Louisiana commission long before Washington acts, and how that ruling assigns costs will be the template other states copy — or reject.

Should a data center be required by default to cover the grid infrastructure its load creates? Tell us in the comments.

Read more

Only 4.5% of US consumers pay for AI, and the top 1% spends $900

Only 4.5% of US consumers pay for AI, and the top 1% spends $900

Consumer AI has near-half household reach and a paying base the size of a rounding error — and today's numbers on who actually spends put a fine point on it. Only 4.5% of US consumers pay for AI, and the top 1% of them spends $900 a month. Andreessen Horowitz published the seventh edition of its Top 100 generative AI apps ranking, and for the first time the firm tracked observed spending on US consumer cards alongside traffic and downloads. The picture: nearly half of US adults now use AI, abo

Google's Gemini 4 Carbon reportedly matches Opus 5.5 on coding

Google's Gemini 4 Carbon reportedly matches Opus 5.5 on coding

Google hasn't launched Gemini 4 Argon yet, but internal documents suggest a faster follow-up is already in testing — and early impressions put it level with Anthropic's best coding model. Google is testing a Gemini 4 variant called "Carbon" that employees say performs on par with Anthropic's Opus 5.5 on programming tasks. Business Insider, citing internal documents, screenshots, and chats, reports that Google deployed Carbon on Jetski — its internal coding platform — over the past few days, wi

Open Source Radar — October 10: reviewers, skills, and 3D maps

Open Source Radar — October 10: reviewers, skills, and 3D maps

Today's trending board skews practical: tooling that fixes code, teaches agents a framework, and rebuilds 3D scenes from video. Here's what's worth your stars. alibaba/open-code-review (Go, 45.6k stars) — Alibaba open-sourced the code review tool it runs internally, and it shows: a hybrid of deterministic static-analysis pipelines for the known bug classes (null dereferences, thread-safety, XSS, SQL injection) plus an LLM agent that writes precise, line-level comments. It talks to OpenAI- or A

WorldArena 2.0 puts world models to the test on real robots

WorldArena 2.0 puts world models to the test on real robots

The world-model field has argued for two years about video quality. This week the first full results landed for a benchmark that asks the harder question: can a robot actually use the prediction? WorldArena 2.0's global challenge has closed its leaderboard, and for the first time world models are being graded on physical robots instead of plausible video. The benchmark — designed by a Tsinghua-led consortium with PKU, CMU, Stanford, Princeton and others — extends its 1.0 video scoring along th