Anthropic reportedly signs $35 billion cloud deal with Nvidia-backed Lambda

Anthropic reportedly commits $35 billion to Lambda

Anthropic, the OpenAI rival behind the Claude model family, has reportedly signed one of the largest cloud contracts in the AI industry's short history — and not with Amazon, Google or Microsoft. According to a dev.to post citing Reuters reporting based on a person familiar with the matter, Anthropic has agreed to a deal worth roughly $35 billion over five years with Lambda, a comparatively small provider that specializes in GPU infrastructure and counts Nvidia among its backers.

The size alone would draw attention. The more consequential detail is who is selling the compute: a specialist AI cloud operator rather than one of the three hyperscalers that have functioned as the default route to frontier-scale computing power.

What the reported deal includes

Per the dev.to summary of the Reuters report:

• A commitment of about $35 billion spread across five years.
• Access to tens of thousands of Nvidia GPUs, including H100s, H200s and the forthcoming Blackwell generation.
• Infrastructure designed specifically for training large models, rather than general-purpose cloud capacity.

The terms have not been publicly confirmed by the companies involved, and the figures rest on a single unnamed source, so the details should be read as reported rather than final.

The logic of looking beyond the big three

Anthropic is no stranger to hyperscaler money. The company has already taken billions of dollars in investment and cloud credits from both Google and Amazon — support that fueled growth but created an uncomfortable dependency, since its main infrastructure suppliers are also developing competing models.

As described in the reporting, the Lambda arrangement is a hedge against that dependency. It spreads Anthropic's compute sourcing across a vendor with no rival model of its own, and it secures a large pool of Nvidia GPUs at a moment when chip supply is tight and allocation decisions carry competitive weight. There is also a technical case: a cloud built around dense GPU clusters and fast interconnects between them can shorten training runs and cut costs for models at Claude's scale, compared with generalist platforms that also serve web hosting, streaming and conventional enterprise workloads.

Nvidia's quiet win

The second storyline belongs to Nvidia. The chipmaker has invested in Lambda and a cohort of similar GPU-focused cloud operators, and the reported Anthropic contract shows how that strategy compounds: a flagship customer signs with a Nvidia-backed provider, and that provider must then buy large volumes of Nvidia's most advanced chips to fulfill the agreement. Nvidia gains predictable demand while cultivating challengers to Amazon, Google and Microsoft — all three of which are building their own custom AI silicon and have clear incentives to depend less on Nvidia over time.

Why it matters

If the reported terms are accurate, the deal undermines a core assumption of the AI boom: that any lab racing toward the frontier has no practical alternative to AWS, Google Cloud and Azure. Hyperscalers have spent years investing in AI startups in exchange for platform commitments, an arrangement that reinforced their dominance. A company of Anthropic's stature signing with Lambda suggests leading labs now have the leverage to help build a different market — one where specialized AI clouds can rise to the top tier on the strength of a single anchor contract, and where Nvidia shapes not only which chips get used but which providers get to grow.

For now, the report stands on thin sourcing — one unnamed source relayed through a dev.to post summarizing Reuters — and the companies have not confirmed it. But even as a signal, it points to a shift in how AI compute gets bought: away from renting capacity from rivals, and toward labs securing direct, predictable access to the hardware itself.