Anthropic Signs $US35 Billion Lambada Cloud Lease Backed by Nvidia

Key Takeaways

• Anthropic signed a $35 billion cloud computing lease with Nvidia-backed Lambada.
• Nvidia’s stronger credit rating puts its name on the underlying data centre lease, deepening the connections among investor, supplier and customer.
• The structure highlights growing concentration and counterparty risk across the rapidly expanding AI infrastructure market.

Anthropic has signed a $35 billion cloud computing lease with Lambada, a neocloud provider backed by Nvidia. The arrangement is significant not only for its size, but also for the financial links connecting the participants. Nvidia is a major investor in Anthropic and an investor in Lambada, while its processors are central to the infrastructure economics behind much of the AI market.

In an unusual structural move, Nvidia’s name, rather than Lambada’s, will appear on the lease for the data centre because Nvidia has the stronger credit rating. That effectively places the chipmaker’s balance-sheet credibility behind infrastructure intended to serve Anthropic through Lambada. The precise allocation of obligations will depend on the contracts, which have not been disclosed, but the structure illustrates how financing and commercial relationships are becoming intertwined.

Circularity does not automatically make a transaction unsound; vendors have long financed customers, provided guarantees, or supported distribution partners. Cloud providers also sign long-duration capacity agreements before customer demand is fully realized. What distinguishes this transaction is the sheer scale and the number of roles occupied by Nvidia, serving simultaneously as an investor, technology supplier, neocloud backer, and creditworthy participant in the data centre lease.

The incentives broadly line up. Anthropic obtains access to scarce computing capacity without constructing every facility itself. Lambada secures a large customer and gains support from a stronger counterparty. Nvidia helps stimulate demand for infrastructure likely to contain substantial accelerated computing capacity. Yet the same arrangement can make it harder for investors and lenders to determine where market demand ends and vendor-supported demand begins.

Is the AI infrastructure market expanding because end customers are generating durable returns, or because capital is moving through a tightly connected group of model developers, chip suppliers, cloud operators and property owners? In practice, both dynamics can coexist. The challenge is measuring how much capacity is supported by repeatable workload economics rather than expectations about future adoption.

The growth forecasts explain why companies are accepting that risk. IDC reported that global AI infrastructure spending reached $318 billion in 2025 and forecast that it would exceed $1 trillion by 2029. Cloud and shared deployments accounted for 84.1% of spending, while servers represented 98% of AI-centric infrastructure expenditure. Those figures point to a market dominated by large, capital-intensive computing installations.

Broader estimates are similarly striking. Gartner projected that worldwide AI spending would approach $1.5 trillion in 2025, encompassing hardware, software and services. S&P Global Market Intelligence, through 451 Research, estimated that the AI infrastructure market reached $337 billion in 2025 and is forecast to grow to $1.2 trillion by 2030, with inferencing becoming the principal growth driver.

That shift toward inference matters. Training frontier models demands huge bursts of computing, but inference can create a longer and potentially steadier stream of consumption as applications reach employees and customers. Capacity providers are betting that production workloads will eventually fill today’s data centres. If adoption develops more slowly, long leases, power commitments and equipment financing could leave multiple parties exposed to the same shortfall.

For enterprise buyers, the immediate concern is less about the financing label and more about resilience. Procurement teams may want visibility into which party operates the capacity, who carries the lease obligation, what happens if a provider restructures, and whether workloads can move elsewhere. Credit support from Nvidia may strengthen the arrangement, but it also reinforces the industry’s dependence on a small set of suppliers and balance sheets.

The Anthropic, Lambada and Nvidia deal therefore serves as a useful stress test for the AI boom. It shows how the sector is finding ways to fund enormous capacity commitments, even when younger infrastructure providers lack comparable credit strength. It also shows why revenue quality, contract duration, utilization rates and counterparty exposure deserve as much attention as headline spending forecasts. The buildout is real. So are the financial connections holding it together.