The money was yesterday’s story. The compute supply chain is today’s.
Anthropic’s Series H closed at $65B on a $965B post-money valuation, figures that were covered when the round was announced on May 28. What wasn’t mapped in detail was the infrastructure architecture embedded in the deal: a confirmed 5 gigawatt compute agreement with Amazon, and reported agreements for an additional 5 GW through Google and Broadcom’s next-generation TPU capacity, plus access to SpaceX’s Colossus 1 and Colossus 2 GPU clusters.
The Amazon deal is confirmed. Anthropic’s own announcement states the company has “signed a new agreement with Amazon that will deepen our existing partnership and secure up to 5 gigawatts (GW) of capacity” for training and deploying Claude. That’s not a letter of intent. It’s a signed capacity reservation at a scale that would rank among the largest single compute agreements ever disclosed.
The Google and SpaceX components are different in verification status. Anthropic reportedly secured agreements for up to 5 GW of next-generation TPU capacity from Google and Broadcom, according to reporting by Axios and The Wall Street Journal. Anthropic is also reported to have secured GPU capacity access within SpaceX’s Colossus 1 and Colossus 2 clusters, per the same reporting. Neither has been independently confirmed from primary sources.
The chipmaker layer adds a third dimension. Micron, Samsung, and SK hynix joined the round as strategic infrastructure partners, according to Anthropic, with their participation described as aimed at securing memory and logic supply. Yahoo Finance corroborates their investor participation, though the “strategic infrastructure partners” characterization is Anthropic’s framing rather than an independently verified term of art.
The capital connection matters for context. The round includes $15B of previously committed investments from hyperscalers, including $5B from Amazon, confirmed directly in Anthropic’s announcement. What’s notable is that Amazon and Google aren’t just investors, they’re also the providers of the confirmed and reported compute agreements. Capital and infrastructure access are flowing through the same relationships simultaneously.
This is the third major AI compute infrastructure agreement anchored by a hyperscaler commitment this quarter, following CoreWeave’s GPU loan facility and the string of power capacity announcements tracked in TJS’s coverage of the gigawatt race. The pattern: frontier labs are locking in supply before the compute market tightens further, and the capital structure of their funding rounds is the mechanism they’re using to do it.
Verification
Partial Amazon 5 GW confirmed via Anthropic official announcement; Google/SpaceX capacity reported by Axios and WSJ, sources not independently accessible Chipmaker 'strategic infrastructure partner' framing is Anthropic's characterization; T3 corroboration only for investor participationThe catch is
the verification asymmetry. Amazon’s 5 GW is on the record. Google’s 5 GW and SpaceX’s cluster access are attributed to Axios and WSJ, credible sources, but remain unconfirmed from primary disclosures. That distinction matters for enterprise teams modeling Claude availability risk and for investors pricing infrastructure moat depth into Anthropic’s valuation.
Watch the Amazon earnings call and any AWS infrastructure capacity disclosures for the first hard confirmation of when the 5 GW agreement phases into active deployment.