Anthropic has agreed to pay $35 billion over several years for computing capacity from Lambda, an Nvidia-backed cloud provider, in a deal that anchors a new data-center campus in Nueces County, Texas. The agreement, signed August 31, is structured as a multiyear lease and offtake arrangement rather than a lump-sum payment, and it puts Nvidia at the center of the transaction in three separate roles at once.
The Texas facility, developed by Hut 8 — a company that converted from cryptocurrency mining into AI infrastructure — will initially deliver about 350 megawatts of capacity toward a campus designed to scale to roughly 700 megawatts. Nvidia supplies the GPUs that will fill it, holds an equity stake in Lambda, and separately holds the lease on the data center itself — meaning a large share of the $35 billion Anthropic is committing effectively flows back toward Nvidia's own balance sheet.
Circular financing, again
That structure has revived a criticism that has followed several recent AI infrastructure deals: that chipmakers are financing the very customers, and the very facilities, that generate demand for their own hardware. Nvidia's chief financial officer, Colette Kress, addressed the concern directly.
We recognize the scale of this support, and we know some will call this circular financing. We see it differently.
Colette Kress, Nvidia CFO
Nvidia's argument, echoed by analysts covering the deal, is that the infrastructure itself is redeployable — the data centers and chips retain value even if any single tenant relationship changes. Bank of America analysts framed the underlying dynamic more bluntly: frontier AI labs like Anthropic and OpenAI cannot yet finance compute at this scale on their own, leaving Nvidia's balance sheet as the backstop, even as both companies simultaneously work on custom chips that could someday reduce their dependence on Nvidia hardware.
The Lambda deal is one of several large infrastructure commitments Anthropic has made in recent weeks as it races to secure computing capacity ahead of demand. The company's annualized revenue run rate reached roughly $74.3 billion as of July 22, putting it ahead of OpenAI's reported $41.3 billion by the same measure — a growth trajectory the company is citing to justify locking in compute years in advance.
Whether that revenue can comfortably service tens of billions of dollars in lease and offtake obligations is the question hanging over the deal, and one that will only be answered as the Nueces County campus comes online and Anthropic's spending commitments start coming due.
