Skip to content
AI daily

6 min · 4 stories

AI Data Centers Lost the Midterms and Wall Street Noticed

By Robert Hattala

In this issue

Funny week. The AI industry spent Tuesday learning that a lot of regular folks do not want a data center down the road, then spent Friday learning that the money might not be as bottomless as everybody assumed. Four stories, and they all rhyme.

01 of 04

Voters went 40 for 41 against data centers

On November 3, there were 41 local ballot measures about data centers in 10 states: Ohio, Illinois, California, Florida, Kansas, Maine, Nevada, New Jersey, North Dakota, and Wisconsin. In 40 of them, the result went against data center development. That count comes from Duane Morris Government Strategies' State Capital Lobbyist tracker, which lists every measure and the vote.

Some of these were not close. Granville, Ohio banned data centers 82 to 18. Alhambra, California passed Measure DB, prohibiting them, 82 to 18. Edgewater, Florida put a ban in its charter 77 to 23. Wells and Peru in Maine passed 180-day moratoriums. Boulder City, Nevada voted down Question 1, which would have opened up the Eldorado Valley for data centers, 71 to 29. Wilmington, Ohio shot down four separate rezonings for data center land, each about 80 to 20.

The lone exception was Harwood, North Dakota, a town of under 800 people near Fargo. Voters there rejected an initiative that would have required extra review before the city approved industrial projects, data centers included. It lost 76 to 24.

Meta noticed. Meta president Dina Powell McCormick told the WSJ Tech conference that the results are "going to make it a lot harder for companies to build data centers," and admitted the industry has "not done as good a job as we needed to explain the benefits." Meta has data centers in 23 states, with five planned at a gigawatt or more, one of them in Texas. (TechCrunch)

Here is where I land on it. Nobody in Granville voted against chatbots. They voted against a building that pulls as much power as a small city and doesn't hire many people once it's built. The industry told everybody this was about national competitiveness and never told the guy next door what was in it for him. Forty out of forty-one is not a messaging problem. That is a verdict, and the companies that start paying for their own power and water up front are the ones that are going to get built.

02 of 04

Crusoe hands its Wyoming megaproject to Google

Crusoe confirmed Friday that Google acquired its data center campus site in southeast Wyoming. Crusoe and Blackstone-backed Tallgrass Energy started the project last year and talked about scaling it to 10 gigawatts, powered by a mix of fuel cells, grid power, and natural gas turbines, with carbon capture through Tallgrass's CO2 sequestration hub. (Crusoe announcement)

Crusoe's release frames it as a handoff: they did site selection, approvals, and the power plan, and now Google takes it from there. CEO Chase Lochmiller said the work "set up the Wyoming project for construction and eventual operation." The less cheerful version, per TechCrunch, is that Google had tried and failed to land as a customer, and reportedly raised concerns about cost and construction timeline under Crusoe's management. Bloomberg first reported back in June that Crusoe was being pushed aside. Neither side disclosed the price. Google says it plans to invest $17 billion in the campus.

What I want to know is how many more of these we see. Crusoe is the company building Abilene for Oracle and OpenAI, and it raised money at a $30.9 billion valuation in September. If even Crusoe ends up selling a site to the customer instead of building for it, that tells you the real bottleneck is not land or ideas. It is turbines, transformers, and the ability to hit a schedule. The hyperscalers with the deepest pockets are going to end up owning the dirt.

03 of 04

AI spending stalled while OpenAI and Anthropic valuations shrank

Ramp's AI Index, built from corporate card and bill pay data, shows token spend grew only 1.5 percent across August and September combined. In July it grew 42 percent in a single month. October bounced back some, up 8.9 percent from September 23 to October 23. (TechCrunch)

Ramp lead economist Ara Kharazian says it is not a demand problem. Usage is up. It just got cheaper. OpenAI cut API prices in half for GPT-6 Sol and GPT-6 Luna in September, and Anthropic cut pricing on Fable 5.1 and Opus 5.5, which a recent Anthropic filing said knocked about $600 million off revenue growth. Companies are also getting smarter: blocking expensive models on long jobs, setting employee budgets, and routing work to cheap lightweight models. His line: "You can use more AI, but use it for less." The median company spends $12 to $15 per employee per month. The top 10 percent spend $600 to $1,000.

Same day, Bloomberg reported OpenAI has raised $11 billion of a round that could reach about $20 billion, at a $1.25 trillion valuation. That valuation is down from the $1.4 trillion it was pitching in September. Bloomberg also reported Anthropic's potential IPO valuation slid from about $2 trillion in early October to $1.2 to $1.3 trillion. (TechCrunch)

I think this is healthy, and I say that as somebody who uses this stuff all day. Prices dropping while usage climbs is exactly what is supposed to happen to a useful technology. The problem is only for anybody who priced these companies like the bill would keep doubling. Anthropic shedding $700 billion or more of paper value in a month is a wild sentence to type, and it is still a trillion-dollar company. Everybody calm down.

04 of 04

A judge says xAI has to face the NAACP over its turbines

US District Judge Debra M. Brown of the Northern District of Mississippi denied xAI's motion to dismiss a Clean Air Act lawsuit from the NAACP over unpermitted gas turbines powering its power plant at 2875 Stanton Road in Southaven, Mississippi. The NAACP filed in April and is represented by the Southern Environmental Law Center and Earthjustice. (Memphis Flyer)

xAI made two arguments. First, that the trailer-mounted turbines were mobile sources, so they did not need air permits. The court said they are not mobile sources under federal rules. Second, that citizen lawsuits under the Clean Air Act are unconstitutional because they step on the executive branch. The court rejected that too, saying citizen plaintiffs are not exercising the federal government's executive power. Back in October, the same court also blocked the Justice Department's attempt to get the case dismissed. SELC litigation director Kym Meyer said the ruling makes clear the right to sue "cannot be tossed aside simply because a massive corporation finds it inconvenient." (Engadget)

Put this one next to the ballot results and the picture gets real clear. Calling a power plant "mobile" because it sits on a trailer is the kind of move that makes people vote for moratoriums. The citizen suit argument was the bigger deal, though. If that had worked, neighbors would have lost one of the only tools they have when regulators look the other way. I am glad it didn't.

Thursdays

Get the digest on Thursdays

Each Thursday: the shop notes and AI daily issues from the last seven days, and one product. Unsubscribe from any issue.