Cloud backlog is mostly two customers

Our read: the cloud backlog boom is mostly two customers, not broad demand. Anthropic and OpenAI account for ~$971B, about 41% of ~$2,352B in combined backlog at $MSFT $GOOGL $AMZN $ORCL (our calc).

Combined backlog was ~$602B in mid-2024. It has risen ~$1.75T since, and the two labs' named deals are about 55% of that rise (our calc; Google's mid-2024 figure is total backlog, as it gave no Cloud split before Q1'26; on total backlog at both ends the result is the same, ~55%).

Anthropic alone is ~$252B across three clouds: $111B Google Cloud, $110B AWS, $31B Microsoft (WSJ-reported, draft S-1). That is ~22% of Google Cloud backlog, ~22% of AWS's and ~5% of $MSFT commercial RPO (our calc). $AMZN's own 10-Q says the AWS-Anthropic commitment grew "by more than $100.0 billion over 10.0 years."

OpenAI's named deals add ~$719B: ~$281B at $MSFT (its CFO put OpenAI at ~45% of the $625B commercial RPO as of 12/31/25; $MSFT's total here is 6/30/26), $138B at AWS, and $300B at $ORCL (WSJ-reported; Oracle doesn't name the customer) (our calc).

The stretch: Anthropic's ~$450B in cloud commitments over 7–10 years is ~$45–64B a year, 1.4–2.0x its ~$32.6B H1 revenue run-rate (WSJ-reported, draft S-1; our calc).

Caveats: backlog definitions differ by company and fiscal quarter ends don't line up, so the combined total is indicative. Lab and backlog figures are from different dates, and WSJ says Anthropic's commitments are "mostly" noncancelable, not all.

What would prove us wrong

Set before publication. We report back in public when one trips.

• Q3'26 filings show combined backlog still rising while the two labs' named deals fall below 40% of the rise since mid-2024. Open · next check: Q3 '26 filings (late Oct; $ORCL in Dec)

Method and sources

Figures marked "our calc" are our own arithmetic on the cited primary sources (company filings, earnings releases, public datasets). Every figure was checked against its source before publication.

Original post on X → Download data note

General research, the same for every reader. Not investment advice.