Anthropic just bought one point seven gigawatts of gas generators. Not the electricity. The machines. When the grid can't deliver, apparently you go shopping for engines. It's Monday, November 9th, and this is Lantern AI Daily Briefing. On the rundown: that three point four billion dollar generator deal, OpenAI turning commitments into cash at a one point two five trillion dollar valuation, Shield AI paying five point three billion for a radar maker, and Democrats finding out that wanting AI rules is the easy part.
Reuters, relaying a Financial Times scoop, via ET Enterprise AI:
Anthropic has struck a deal worth about $3.4 billion to secure gas-fired power generators for data centers, the Financial Times reported on Sunday. The AI startup bought commitments for 1.7 gigawatts of heavy-duty generators in late September from OnePWR Solutions, a San Antonio, Texas-based equipment supplier, the report said.
So the deal itself dates to late September; the FT only surfaced it Sunday. And it's not an acquisition. Anthropic didn't buy OnePWR and no staff moved over, but it did get the engineering and construction capability to actually install the things. The Context's Hacker News roundup adds the FT's detail that these were generator orders OnePWR had placed with Caterpillar, and that Anthropic first planned to simply buy power from OnePWR before taking over the orders. The FT calls it the largest known single purchase of power-generation equipment by an AI lab. The caveat: Reuters couldn't verify the report, and neither company responded. MLQ's math puts it around two thousand dollars per kilowatt. That's the going rate for not waiting on a utility. And Anthropic isn't alone. The New York Times reports xAI installed eight converted aircraft turbines in Mississippi for Colossus 2, because a used engine ships in roughly twenty-four months versus three to eight years for a new turbine. For continuity: Anthropic's amended S-1 on November 2nd listed about eight hundred ninety-five billion dollars in purchase and lease commitments, and the Times reported last week that it expects more than three and a half gigawatts by the end of this year and aims to double that in 2027. Read together, power looks like the gating item now, not chips. This is the infrastructure story I'd rank above most model launches. And these are gas-fired, which lands right on top of last week's ballot fights over local air, water and electricity. Owning the generators doesn't get you the permit.
AI Daily Insights, summarizing Bloomberg's reporting:
OpenAI had received about $11 billion in new funding by November 6, U.S. time, according to people familiar with the matter cited by Bloomberg. This was the first close of a round valuing the company at $1.25 trillion. The total could reach $20 billion, with the remainder expected to close over the coming weeks.
We first covered these talks on Friday, October 23rd. What's new is money actually arriving, and a lower bar than OpenAI started with. Its original ambition was at least thirty billion dollars at about one point four trillion, before those goals came down amid a decline in AI-related stocks. OpenAI declined to comment. So: smaller round, lower price, and it still pulled in eleven billion dollars in cash. That's this market in one sentence. For scale, back in March OpenAI announced a hundred twenty-two billion dollars in committed capital at an eight hundred fifty-two billion post-money. Committed and received are different stages, and this week was the received kind. It matters for the other trillion-dollar candidate. Reuters reported Thursday that prospective investors questioned whether Anthropic could start its roadshow this week, with some treating the previously discussed one point five trillion valuation as the high end. The FT reported Friday that a pre-Thanksgiving listing was still planned, raising eighty to a hundred billion dollars at a range of one point two five to one point seven five trillion. None of that is final. OpenAI's mark now sits at the very bottom of Anthropic's reported range. Every banker on that roadshow will bring that up. If OpenAI had to give ground to get cash, I wouldn't bet on Anthropic pricing at the top.
Colin Campbell, writing in Axios:
Defense contractor Shield AI is raising $2.5 billion to acquire radar provider Echodyne Corp. for $5.3 billion.
Announced Thursday, with the new Series H valuing Shield at twenty-eight point eight billion dollars post-money and closing expected in early 2027. Ultrathink's breakdown of the release shows the capital stack, which is the interesting part. The one billion dollar Series H is co-led by Disruptive and Innovation X Capital, another billion is structured equity from funds managed by Blackstone, and five hundred million is debt with JPMorgan Chase as sole arranger. And the markup. Bloomberg reported Shield raised two billion at a twelve point seven billion valuation in March. Twenty-eight point eight is more than double that. Of the five point three billion, four point eight is paid and issued at closing, and up to five hundred million more depends on Echodyne hitting performance targets. What Shield's buying is sensing. Echodyne's solid-state radars, which it calls MESA, are built into hundreds of counter-drone and security systems in more than forty countries, and its new plant can produce about thirty thousand radars a year. Echodyne's CEO, Eben Frankenberg, says autonomous systems can only act on what they can sense. A rare autonomy company admitting the software isn't the whole product. Two things left undisclosed: the cash-versus-stock split, and Echodyne's revenue. Both companies say they expect triple-digit revenue growth this fiscal year. Regulators still get their say before that early 2027 close.
Jared Perlo, writing for NBC News:
Democrats across the country who swept into power amid growing voter skepticism about AI and data centers are now reckoning with knotty discussions about how to tackle AI regulation and which technological impacts deserve their attention first, according to AI experts, strategists and legislators themselves.
This picks up right where last Tuesday's results left us. The appetite is real. Sacha Haworth of the Tech Oversight Project points to fifty-six percent of voters saying AI concerns were crucial or important to their vote. And incoming Representative Scott Wiener says every House Democrat he's talking to wants AI regulation. The priorities scatter, though. Sherrod Brown is focused on data centers. Senator-elect Peggy Flanagan committed to restricting AI in insurance decisions. Jon Ossoff talked about extinction risk in his victory speech. And Samuel Hammond of the Foundation for American Innovation highlights the distance between Bernie Sanders's proposal to ban artificial superintelligence and routine permitting updates for power plants. Alex Bores, who wrote New York's RAISE Act, calls that variety an asset, with Congress working several fronts at once. I'll believe it when one of those fronts survives a veto. Trump still calls AI a golden goose. My standing complaint applies here. Data centers, insurance algorithms and superintelligence are three separate problems, and filing them all under AI regulation helps nobody draft a bill. For builders, the practical read is unchanged since last week: with preemption off the table, state rules are where compliance actually happens.
Quick favor before we go. If this briefing saves you a morning of scrolling, rate it or leave a short review in your podcast app. It helps other engineers and investors find the show.
On the calendar: the rest of OpenAI's round is expected over the coming weeks, and we'll be watching whether Anthropic's roadshow actually gets underway this week. Back tomorrow with whatever moves. You've been listening to Lantern AI Daily Briefing, part of the Lantern Podcasts family. Talk to you Tuesday.