Anthropic Signs a $35 Billion Compute Deal With Lambda — Its Biggest Cloud Contract Yet

Anthropic has signed a $35 billion, six-year cloud computing agreement with Lambda, the Nvidia-backed “neocloud” provider based in San Jose, California — the latest and largest in a string of massive compute contracts the AI developer has clinched amid fierce competition for advanced AI infrastructure.

The Wall Street Journal was first to report the deal on August 31, and AFP confirmed the terms on September 1 through a source familiar with the matter. The contract is one of the largest single cloud-computing commitments ever signed by an AI lab, and it lands at a moment when Anthropic and rival OpenAI are locked in a two-front war: one for customers and revenue, and one for the physical capacity — chips, power, and data centers — that makes frontier AI possible at all.

The structure: a three-way chain built on Nvidia

The details of the arrangement, as reported by the WSJ and AFP, reveal a supply chain that is almost baroque in its construction:

• Anthropic signs a six-year deal worth $35 billion to secure cloud computing capacity.
• Lambda — which rents out GPU capacity and is backed by Nvidia — is the counterparty Anthropic pays.
• Hut 8, a bitcoin miner that has pivoted into data center development, is building the facility, including one in Texas that Anthropic will tap for computing power.
• Nvidia — which has invested in both Anthropic and Lambda — will lease the data center from Hut 8, and Lambda will then pay Nvidia an undisclosed amount to access it.

In other words, the money and the hardware flow through four companies before a single Claude token is served. Nvidia appears at two points in the chain simultaneously: as the chip supplier and landlord via the Hut 8 lease, and as an investor in both the buyer (Anthropic) and the seller (Lambda) of the capacity. According to the reporting, the capacity will be used to meet growing demand for Claude, Anthropic’s flagship AI product line.

A month of mega-deals: $135 billion and counting

The Lambda contract does not stand alone. According to AFP, Anthropic has now signed compute deals worth at least $135 billion in 2026 alone. Just days earlier, the company confirmed a similar six-year agreement worth $45 billion with Nscale, a London-based cloud provider that will give Anthropic capacity at a facility in West Virginia.

The pace is staggering by any historical standard of infrastructure procurement. Earlier this year, reporting indicated Anthropic had agreed to pay Google roughly $200 billion for TPU capacity — a figure that, combined with the Nscale and Lambda deals, sketches the true scale of what it costs to compete at the frontier. Anthropic’s April agreement with Google and Broadcom alone covered multiple gigawatts of next-generation TPU capacity.

For context on how quickly the numbers have escalated: Anthropic’s entire Series F in early 2025 was around $3.5 billion. The company now signs individual compute contracts ten times that size.

Why neoclouds, and why Lambda?

The rise of the “neocloud” — companies like Lambda, CoreWeave, Nscale, and Crusoe that buy Nvidia GPUs at scale and rent them out — is one of the defining structural shifts of the AI buildout. Hyperscalers like Microsoft, Google, and Amazon cannot expand fast enough to absorb frontier-lab demand, and the Big Three’s own AI divisions compete with their cloud customers. Neoclouds occupy the gap: capacity-hungry, Nvidia-aligned, and willing to sign long-term contracts that banks will finance.

Lambda has spent 2026 positioning itself exactly for deals like this one. In May, the company closed a $1 billion secured credit facility. In August, it raised another $1 billion in private short-dated debt — arranged by JPMorgan — to buy Nvidia GPUs it will lease to Microsoft, alongside a $926 million senior secured term loan to fund GB300 GPUs for a deployment contracted to Nvidia itself. It is reportedly in talks for a $3 billion pre-IPO round at a valuation above $12 billion. Every one of those transactions was a step in the same direction: assembling the balance sheet to sign enormous, multi-year capacity contracts with investment-grade AI labs.

The Anthropic deal is the payoff. A six-year, $35 billion revenue commitment transforms Lambda from an interesting GPU reseller into a company with contracted cash flows on a scale that public markets can underwrite — and it does so weeks before an anticipated IPO filing.

The compute arms race, and the IPO shadow over it

Both AFP and Reuters frame the deal against the same backdrop: Anthropic and OpenAI are “barreling toward initial public offerings” while racing to win customers and grow revenue. Compute is the constraint that binds everything else. Demand for chips and advanced computing power has skyrocketed over the past two years, and that demand is now cruncing global supply chains well beyond AI — AFP notes that Apple has already raised prices on Macs and iPads due to memory and chip costs, and is expected to raise prices on the new iPhones it unveils in September.

The macro signal is hard to miss: frontier AI has become a capital-intensity game where the winners are determined less by model architecture than by who can contract, finance, and energize the most silicon. Anthropic’s $135 billion in 2026 commitments is a bet that Claude’s revenue growth will outrun the cost of the capacity underneath it. OpenAI, with its own web of mega-deals, is making the same bet.

There are real risks embedded in the structure. These contracts concentrate enormous counterparty exposure on relatively young companies — Lambda’s ability to deliver depends on Hut 8’s construction schedule, Nvidia’s allocation priorities, and debt markets staying open. A six-year deal spans multiple GPU generations; capacity that looks ample in 2026 may be the wrong silicon in 2029. And the circularity of Nvidia investing in both sides of the transaction is efficient for the ecosystem but raises questions about how demand for Nvidia’s chips is being manufactured as much as met.

What to watch

• Lambda’s S-1: with $35 billion of contracted revenue from a single customer, the anticipated IPO filing becomes one of the most interesting reads of the year — what margins survive a deal this size?
• Nscale’s West Virginia buildout: the $45 billion sibling deal depends on capacity that is still under construction.
• Anthropic’s revenue disclosures: the company needs Claude’s growth to justify nine-figure annual capacity costs. Any IPO prospectus will have to reconcile those two curves.
• Neocloud debt markets: over $400 billion in AI-related debt has been raised globally in 2026. The Lambda-Anthropic contract is precisely the kind of asset banks can lend against — which means the deal’s ripple effects will show up in credit markets, not just tech headlines.

For now, the machine is running: Nvidia sells the chips, Hut 8 builds the halls, Lambda borrows against the contract, and Anthropic gets the gigawatts it needs to keep Claude growing. Thirty-five billion dollars over six years is the new unit of measurement in the AI infrastructure race — and by the pace of the last few weeks, it will not hold the record for long.