On a flat stretch of Nueces County, Texas, where a bitcoin miner once planned to hum through the night, Anthropic has placed a $35 billion bet — and Nvidia is standing on all three sides of it.
The AI company signed the cloud computing agreement with Lambda on Monday, Aug. 31, according to people familiar with the terms, securing roughly 350 megawatts of GPU capacity at a data center under development by Hut 8, a former cryptocurrency miner now reinventing itself as an AI landlord. Neither Anthropic, Lambda, Nvidia nor Hut 8 has confirmed the figure on the record. Hut 8 shares climbed as much as 4% in premarket trading the following morning anyway.
What makes it unusual is not the size. It is that one company occupies three seats at the table. Nvidia manufactures the chips Lambda will install. Nvidia is an equity investor in Lambda, having joined the cloud provider's $480 million Series D in 2025. And Nvidia, according to reporting on the arrangement, holds the lease on the building itself — two 15-year leases covering about 704 megawatts of IT capacity at Hut 8's 1-gigawatt Beacon Point campus, with a combined base-term contract value of $19.6 billion. Anthropic's deal draws on roughly half that footprint.
The Deal Behind the Deal
Hut 8 disclosed the leases in July 2026, identifying the tenant only as a "high-investment-grade technology customer." Nvidia was later named in press reports. At the time, chief executive Asher Genoot framed the second lease as validation of strategy, not financial engineering.
"The real test of our power-first approach is what our partners are willing to commit against it," Genoot said. "Our tenant at Beacon Point chose to double its footprint at the site, the strongest validation an asset can receive."
The Lambda agreement lands on top of a separate roughly $45 billion, six-year commitment Anthropic made days earlier to rent about 460 megawatts from Nscale's West Virginia campus — another Nvidia-backed provider, running Nvidia's forthcoming Vera Rubin silicon. Together that is roughly $80 billion in new infrastructure obligations in about a week, atop a 2026 run that already includes a roughly $50 billion Fluidstack partnership across Texas and New York and a roughly $45 billion arrangement with SpaceX. Anthropic's disclosed compute commitments now exceed $150 billion.
Anthropic's preliminary second-quarter 2026 revenue topped $11.5 billion, up from about $787 million a year earlier — more than fourteenfold growth — with positive adjusted operating income for the quarter. Demand for Claude Code, its software-development tool, has been the loudest driver. The company submitted confidential IPO paperwork in June 2026.
Why the Structure Is Drawing Scrutiny
The concern is not that the demand is fake. It is that the loop is closed. Nvidia's capital helps finance the landlord, the cloud operator and, through a disclosed investment and technology partnership, the AI lab that ultimately pays the bill — and the money returns as chip revenue.
Mizuho chip analyst Jordan Klein has described the underlying model as one in which Nvidia GPUs function as collateral, are rented back into the AI supply chain, and are repaid through lease payments from tenants like Lambda and Anthropic — effectively, he argued, Nvidia pre-funding purchases of its own products.
Nvidia's finance chief met the criticism head-on. "We recognize the scale of this support, and we know some will call this circular financing. We see it differently," Colette Kress told analysts. Her defense rests on redeployability: "The NVIDIA Compute platform is fungible and durable and can be redeployed to support other customers." Kress said Nvidia has "invested nearly 50 billion in the Frontier AI Labs" and expects balance-sheet-supported lab demand to represent "roughly a quarter of our business next year."
The exposure is substantial: Nvidia closed its most recent quarter with $279 billion in supply obligations and $108.5 billion in guarantee obligations tied to AI cloud and data center partners.
There is also a strategic irony. "A lot of these investments are designed to help the frontier labs, especially OpenAI and Anthropic, but both of them are designing their own [custom chips]," Bank of America's Vivek Arya noted — meaning Nvidia is helping underwrite the buildout of two customers working to need it less.
What This Means for the AI Buildout
Strip away the mechanics and a simpler picture emerges: frontier labs can no longer buy compute the way ordinary companies buy cloud. Capital requirements have outgrown the buyers' credit profiles, so the supplier became the financier, and the financier became the landlord. Anthropic's own CFO, Krishna Rao, has described the bind bluntly: "If you buy too much compute, you go out of business. If you buy too little compute, you can't serve your customers."
That is the trade Anthropic is making, against revenue it has not yet raised or earned at that scale. Skeptics and defenders are both partly right: fourteenfold growth is real and independently reported, and a deal's economics are still harder to assess when the chip supplier is also the buyer's investor and the building's tenant of record.
What to Watch
Three things. First, energization: Beacon Point is not expected online until the first quarter of 2027, and Texas has paused new data center power connections amid grid-demand concerns, which could slow the ramp. Second, Lambda's own financing — the company is reportedly raising up to $3 billion at a valuation of $12 billion or more, a round that will test whether investors price the Anthropic contract as durable revenue or a pass-through. Third, and most consequential, Anthropic's IPO, which would force disclosure of what these commitments look like on a balance sheet and of what Lambda actually pays Nvidia for capacity — terms undisclosed today.
Until then, the $35 billion sits in the same category as most of the AI buildout's biggest numbers: contracted, unconfirmed, and dependent on how many tokens Claude ultimately sells.
“We recognize the scale of this support, and we know some will call this circular financing. We see it differently.”— Colette Kress, Chief Financial Officer, Nvidia