Skip to content
Sıfırıncı Dakika
Latest

Anthropic's IPO filing: Compute money flows to Google, Broadcom and SpaceX

AIAnalysis1 min read

In brief

Anthropic's SEC filing revealed how its more than $750 billion in compute deals are distributed. Google's TPUs take the largest share at roughly $400 billion, while Amazon and Microsoft trail Google, Broadcom and SpaceX despite their investments.

  • Over $750 billion in compute deals
  • About $400 billion in capacity on Google TPUs
  • $161.2 billion equipment lease with Broadcom
  • Up to $120.5 billion from SpaceX through 2029
  • At least $110 billion with AWS over 10 years
  • $44.9 billion commitment on Azure

Where the compute money goes

Anthropic's IPO prospectus filed with the U.S. Securities and Exchange Commission lays out how much the company will pay which suppliers for AI compute. The filing details more than $750 billion in computing deals, including firm commitments and estimated future spending.

Google and Broadcom on top

The biggest share goes to Google-designed TPUs (tensor processing units). Anthropic has lined up roughly $400 billion in computing capacity, mostly over the next seven years. That includes $114.5 billion directly to Google and $161.2 billion in equipment leases with chipmaker Broadcom, the largest single arrangement in the filing.

SpaceX and Amazon

Elon Musk's SpaceX, through its AI subsidiary formerly known as xAI, will supply up to $120.5 billion in capacity through 2029. Amazon committed in April to spend at least $110 billion with AWS over 10 years, and AWS Trainium is one of four chip families Anthropic uses.

Microsoft further behind

Microsoft's piece is smaller: at least $44.9 billion on Azure from November 2026 through May 2033, up from the $30 billion deal announced in November 2025. Nscale (about $44.6 billion) and Lambda (about $41.3 billion) have deals in a similar range.

Is the demand real?

Anthropic's second-quarter revenue hit $11.6 billion and it posted its first operating profit of $974 million. It still reported a $66.3 billion net loss, mostly from accounting charges tied to its soaring valuation. The company says customer demand exceeds available compute.

Why it matters

A concrete picture showing the AI race is driven by chips and data-center supply, not just models. Worth reading for anyone curious about which tech giants actually profit from the boom.

Sources

Related stories