Anthropic’s S-1 filing describes vast future commitments for the computing power behind its AI models. The Register calculates that listed agreements could add up to between $782bn and $807bn, though the contracts vary in certainty and some payments depend on conditions.
Anthropic Watch analysis
What happened
The Register reports that Anthropic entered into new or amended compute and infrastructure agreements carrying about $417bn in future minimum payments after 31 December 2025, then added agreements with roughly $245bn in further minimum payments after 30 June 2026. Some of those later obligations are conditional.
Adding up the agreements disclosed in the filing, The Register puts the potential total at $782bn to $807bn. Its breakdown includes at least $110bn for Amazon, at least $114.5bn for Google, and approximately $161.2bn for Broadcom. The report says some major cloud-provider contracts require payment regardless of usage, while other arrangements have different conditions or cancellation terms.
Why it matters
These figures show the scale of the infrastructure bet behind Anthropic’s growth plans. The company’s filing says its ability to develop and deploy frontier models depends on access to increasing amounts of specialised computing power, and that meeting the resulting financial obligations depends on significant revenue growth.
The headline total is not a single, unconditional bill due tomorrow. The agreements stretch across years, differ in their terms, and include conditional commitments. But the disclosed sums make compute access and financing central to the company’s financial story, not a footnote to its models.
Our read
AI growth is being built on contracts for extraordinary amounts of computing capacity. That may be a rational wager on demand, but it leaves investors with a clear question: can revenue keep pace with the infrastructure commitments? The Register’s calculation is based on the filing’s listed agreements, and the varying terms matter as much as the eye-catching total.
What to watch
- How Anthropic’s revenue develops against its expanding compute obligations.
- Which conditional agreements become firm commitments, and when.
- How the company explains its capacity needs and financing plans in future filings.
Discussion spark: Are commitments of this scale a sensible bet on sustained AI demand, or do they leave too much riding on revenue growth keeping pace?
Sources and evidence
- Humanity can't afford not to spend $800B on AI, Anthropic insists (26 October 2026, 23:39 UTC)
- From Wall Street to Silicon Valley, Anthropic IPO poised to have massive 'trickle-down effect' – CNBC (26 October 2026, 23:39 UTC)
Anthropic Watch is independently operated by WittyWires. It is not affiliated with, endorsed by, or operated by Anthropic.