Anthropic Signs $35 Billion Cloud Computing Deal, Nvidia Backs Lambda as Compute Provider
nashnova research
Anthropic signed a $35 billion compute lease with Nvidia-backed cloud provider Lambda, on top of a $45 billion deal with Nscale earlier this month — both totaling $80 billion and both underwritten by Nvidia's credit, showing that AI companies increasingly cannot secure compute without Nvidia as middleman.
How does the $35 billion actually flow?
Lambda deploys Nvidia chips inside a Texas data center developed by Hut 8, then leases the compute capacity to Anthropic.
The critical piece: Nvidia itself holds the lease on that data center, effectively pledging its own credit to backstop the entire deal.
This means → Anthropic's credit rating is too low to sign a lease of this size on its own. Nvidia's role is essentially lending its creditworthiness to its customer.
Why does Anthropic need this workaround?
Anthropic hit a compute supply bottleneck this year — product demand surged but its own capacity could not keep pace, forcing it to lock in outside resources fast.
As a sub-investment-grade company, Anthropic cannot independently sign multi-billion-dollar long-term leases.
In plain terms = it is like a tenant whose credit score is too low to rent an apartment, so a wealthy landlord co-signs the lease. Nvidia is that co-signer.
What exactly is Nvidia's "circular endorsement" model?
This is not a one-off: earlier this month Anthropic also signed a $45 billion deal with Nscale — another Nvidia-backed cloud provider — for Nvidia compute in West Virginia.
Nvidia's playbook: hold the data-center lease → let a cloud provider it has invested in move in and deploy chips → that provider re-leases compute to AI companies.
This reflects a deeper shift — Nvidia is no longer just selling chips. It is embedding itself in every link of the compute supply chain through credit guarantees and equity ties.
Is Hut 8 also building Google TPU data centers for Anthropic?
Hut 8 is developing the Texas facility for Nvidia GPUs while simultaneously building a separate batch of data centers for Anthropic running Google's TPUs — tensor processing units, Google's custom AI chips.
Google is providing financing guarantees for the TPU projects, putting it in direct competition with Nvidia's endorsement model.
This means → Anthropic is not putting all its eggs in one basket, but both paths require a chip giant's credit backing — the AI company's own bargaining power remains limited.
What does Nvidia get out of this?
In July, Nvidia announced credit support for emerging cloud providers like Lambda in exchange for a share of their cloud revenue.
But that revenue-sharing program has partially paused recent deals. Whether Lambda must share revenue from the Texas project with Nvidia remains unclear.
In plain terms = Nvidia earns hardware revenue from chip sales and takes a cut of compute rental income through its endorsements — but the "cut" mechanism is still being recalibrated, and the business model is not yet fully settled.
What does $80 billion in combined deals tell us?
The two agreements total $80 billion, both completed entirely on Nvidia's credit endorsement.
Hut 8 disclosed in July that a "high-investment-grade company" signed a 15-year lease on its 700-megawatt Texas campus, valued at $20 billion, to support Nvidia chips — in hindsight, that company is almost certainly Nvidia.
This reflects a broader reality: the AI compute arms race is no longer just about who has the best model. It is about who can mobilize credit to lock down physical infrastructure — and in that race, Nvidia is both referee and player.
市场有风险，内容仅供研究参考，不构成投资建议。