Anthropic has sealed a six-year, $35 billion cloud computing agreement with Lambda, a cloud provider backed by Nvidia, according to Bloomberg. The deal is the largest reported compute contract in Anthropic’s history and one of the biggest in AI infrastructure overall.

The arrangement involves three parties working in sequence: Hut 8, a Texas-based company transitioning from Bitcoin mining to data center operations, is building the physical facility in Nueces County. Nvidia holds the lease on that data center. Lambda installs Nvidia’s chips inside it and resells the resulting compute power to Anthropic. The deal delivers approximately 350 megawatts of capacity to power Claude products.

That structure is unusual. Most cloud deals are bilateral. Here, Nvidia is embedded as a capital partner — it holds the lease rather than simply selling chips — which tells you something about how tight Nvidia’s grip on AI infrastructure has become. The company is no longer just making GPUs. It is increasingly financing and anchoring the physical infrastructure those GPUs sit inside.

A Pattern of Aggressive Compute Acquisition

This is not Anthropic’s first big compute move in 2026. Earlier this year, Anthropic signed a $25 billion deal with Amazon Web Services and a $1.8 billion compute agreement with Akamai. Add the Lambda deal and Anthropic has committed to over $60 billion in compute capacity in a single year.

That level of spending is not reckless. It reflects a genuine constraint that has shaped the company’s roadmap for the past 18 months: demand for Claude has consistently outpaced available capacity, leading to rate limits, wait times, and service degradation for enterprise customers at exactly the moments they needed Claude most.

The Lambda agreement addresses that by bringing a large, independent pool of Nvidia compute online — separate from AWS and separate from Anthropic’s own infrastructure — so the company is less exposed to any single supply chain or platform relationship.

What This Means for Business

If your business depends on Claude through the API, through AWS Bedrock, or through third-party tools built on Claude, this deal is relevant context.

More headroom, more stability. Anthropic’s chronic capacity problem has been one of the quiet frustrations of building on Claude. With 350 megawatts of dedicated capacity coming online, the infrastructure story improves materially. Reliability for production workloads should strengthen.

Pricing pressure may ease. Compute scarcity has been one driver of AI inference costs staying high. As Anthropic’s total capacity expands through multiple agreements, the unit economics of running Claude improve. That does not guarantee lower prices, but it removes one of the structural arguments for keeping prices elevated.

The Nvidia chokepoint is real. The unusual structure of this deal — Nvidia holding the lease, not just selling chips — shows how much leverage Nvidia has over AI deployment right now. For any business planning its AI stack, understanding that Nvidia sits at the center of most AI compute supply chains is important context for evaluating risk and vendor lock-in.

Anthropic is building for scale. $60 billion in compute commitments in one year sends a signal about where Anthropic expects the demand curve to go. The company is not hedging. If you are building on Claude, the infrastructure runway behind you is growing fast.

For teams evaluating AI platforms, one of the practical questions used to be whether Anthropic could keep up with demand. That concern is not gone, but it is far less urgent than it was 12 months ago.

The compute race is still running, but the participants are now separated by the size of their balance sheets as much as by their models.