Anthropic commits $200B to Google Cloud

If you thought AI infrastructure spending was already wild, the latest commitments will recalibrate your scale.

Anthropic has committed to spending roughly $200 billion on Google Cloud services and custom chips over the next five years, according to a report by The Information. The deal, starting next year, is tied to an agreement providing Anthropic with five gigawatts of server capacity. Neither company shared a price tag when they announced the tie-up last month.

To put that in perspective, this single contract represents more than 40 percent of Google's entire contractually committed cloud backlog. Google's cloud backlog roughly doubled in Q1 to over $460 billion, while the unit grew revenue by 63 percent, a surge largely credited to Anthropic's API demand.

Two money-losing startups prop up cloud giants

Big tech's cloud boom is increasingly reliant on just two names: Anthropic and OpenAI.

Across Amazon, Microsoft, Google, and Oracle, cloud backlogs sit at a staggering $2 trillion. Anthropic and OpenAI account for roughly half of that entire total. Over at AWS, the backlog jumped 49 percent in Q1 to $364 billion, driven by OpenAI adding $100 billion in commitments and Anthropic locking in a separate $100 billion, 10-year deal for up to five gigawatts.

The burn rate is accelerating. OpenAI expects to spend about $45 billion on servers this year, up from $17 billion in 2024, mostly flowing to Microsoft. Anthropic projected over $20 billion for 2025 late last year, three times its previous year's bill, and actual spending is likely higher. By 2029, Anthropic plans to spend roughly $200 billion across Microsoft, Google, and Amazon, while OpenAI projects about $180 billion in that single year.

Margins, custom silicon, and the 30x gamble

There is a massive catch. Both startups are banking on 20- to 30-fold revenue growth by 2029 to actually pay these bills. Wall Street is already nervous: Oracle's stock has plunged 45 percent since September, following its announced $300 billion contract with OpenAI.

The real winners are cloud providers that control their own silicon and resell models directly. Microsoft now keeps 100 percent of Azure revenue from OpenAI models, ending a previous 20 percent payout. Google holds an even cleaner edge: Anthropic runs heavily on Google's custom TPU processors, which offer far higher profit margins than renting expensive Nvidia hardware.

Why it matters

For enterprise developers relying on Claude, this is reassuring news on capacity. Locking in five gigawatts and hundreds of billions in infrastructure means Anthropic won't be easily throttled by compute shortages as enterprise adoption scales.

In our view, however, the industry is setting up a high-stakes circular dependency. Half of the cloud giants' massive $2 trillion backlog hinges on two startups that are currently losing billions and betting on unprecedented 30-fold growth. If enterprise monetization hits a wall, the fallout across capital expenditure and cloud valuations will be brutal.

Google's execution here looks brilliant on paper. By powering Anthropic on custom TPUs rather than third-party GPUs, Google proves out its chip ecosystem and captures high-margin enterprise cloud share. The simmering tension, though, is that Anthropic remains a multi-cloud operator playing Google against Amazon, while Google's own Gemini directly fights Claude for the exact same enterprise workloads.