Anthropic’s Reported $35 Billion Lambda Deal Pushes AI Compute Commitments Higher
A reported 350-megawatt Texas cloud agreement would add another enormous infrastructure obligation to Anthropic’s expansion, but the parties have not confirmed the terms.
Anthropic has reportedly agreed to spend $35 billion on cloud capacity from Nvidia-backed Lambda, adding another extraordinary long-term obligation to the race for artificial-intelligence compute. Reuters, citing a source familiar with the matter and an earlier Wall Street Journal report, said the capacity would come from a roughly 350-megawatt data centre under development in Nueces County, Texas.
None of Anthropic, Lambda, Nvidia or the site developer, Hut 8, confirmed the agreement in the initial reporting. The size, lease structure and delivery timetable should therefore be treated as reported rather than final public facts. Even with that qualification, the deal illustrates how leading AI laboratories are locking in power, buildings and accelerators years before the associated revenue is certain.
A chain of interdependent contracts
The Texas project connects several parts of the AI capital stack. Hut 8, once primarily known for crypto mining, is developing the physical campus. Lambda would provide the cloud service. Nvidia is an investor in Lambda and, according to the Journal's reporting, would hold the lease on the site. Anthropic would become the ultimate user of the computing capacity.
That chain distributes financing and operating roles, but it also creates interdependence. The developer needs a creditworthy lease. The cloud provider needs chips and customer commitments. The chip supplier benefits when capacity is built around its hardware. The AI company obtains compute without owning every layer of the data centre.
Anthropic's demand is not hypothetical: products such as Claude and Claude Code require substantial training and inference capacity. Yet a $35 billion commitment is large relative to the revenue of any private AI company. It would sit alongside a reported $45 billion commitment for capacity from Nscale in West Virginia and other cloud relationships. The total makes future revenue growth, utilisation and financing costs central to the company's economics.
The 350-megawatt figure is also significant. A site of that scale can draw power comparable to a substantial industrial complex. Bringing it online requires grid connections, cooling equipment, transmission upgrades and reliable access to advanced chips. Construction delays or power constraints can turn a contracted asset into an expensive bottleneck.
Circular financing becomes harder to untangle
The AI boom increasingly links vendors, investors, lessors and customers. Nvidia can benefit at several points: through chip sales, an investment in Lambda and a lease structure that supports deployment. Anthropic gains access to scarce computing infrastructure, while the financing chain may reduce the upfront cash it would otherwise need.
These arrangements are not automatically circular or uneconomic. Long-term commitments are common when infrastructure takes years to build. An anchor tenant can make a project financeable, and a supplier can rationally support partners that expand its market. The concern is transparency. Investors need to understand who bears construction risk, who guarantees payments, whether obligations are conditional and how much demand depends on capital supplied by the same ecosystem.
The lack of public confirmation leaves those questions unanswered. A headline contract value may represent total payments over many years rather than cash transferred at signing. It may include options or capacity that is never fully used. Without the contract, the number cannot be treated as an immediate expense or as guaranteed revenue for Lambda.
Why it matters
Compute commitments are becoming the balance-sheet equivalent of the AI model race. The winning laboratory will not simply have the best software; it will need enough power and hardware at a cost that its products can support.
For Anthropic, securing capacity protects growth but raises the hurdle for an eventual public listing. Prospective investors will have to compare contracted infrastructure payments with recurring revenue and margins. For Lambda and Hut 8, the deal could validate a shift from crypto-era computing to AI infrastructure. For Nvidia, it reinforces demand but adds scrutiny of financial links across its ecosystem. Utilities and local communities, meanwhile, will confront the power, water and construction requirements of a campus whose economics depend on AI demand years from now.
Sources: Reuters, The Wall Street Journal