Anthropic's compute tab could top $1T as Q3 revenue nears $17B
Revised IPO filing warns 12x revenue growth is unlikely to continue
ai and ml
Anthropic's compute tab could top $1T as Q3 revenue nears $17B
Revised IPO filing warns 12x revenue growth is unlikely to continue
Anthropic has published a regulatory filing revealing nearly $900 billion in compute and infrastructure commitments, with additional estimated payments potentially pushing the total beyond $1 trillion.
The Monday filing reports that Anthropic expects to pay nearly $70 billion in contractual obligations by September 2027, with a total of roughly $447.5 billion in commitments expected to be incurred by the end of 2030. That's assuming of course its revenues continue their upward trajectory, which the prospectus warns is far from guaranteed.
"In the nine months ended September 30, 2026, our revenue was $33.2 billion, representing a year-over-year growth rate of 1,206%," the prospectus reads. "We do not expect our revenue will continue to grow at comparable rates, and we cannot assure you that it will continue to grow at all."
"Our ability to meet the substantial and growing financial obligations of these requirements to secure compute depends on our ability to generate significant revenue growth," the filing reads.
Which is to say, if Anthropic's revenues don't grow faster than its compute commitments, it could end up falling into a nasty spiral. And while Anthropic has several options, including debt and equity financing, relying on these long term could dilute existing shareholders or force the company to borrow on unfavorable terms.
Every IPO filing reads a bit like this because no company can guarantee eternal success and growth. A lot of this filing's language, as damning as it might sound, is therefore just Anthropic covering its butt.
However, even though the company is still growing steadily year-over-year, the latest filing shows revenue growth is already slowing. Between Q1 and Q2 of this year, revenues grew nearly 2.5x. Yet in Q3, Anthropic's reported revenues of nearly $17 billion represented an increase of just 46 percent from Q2. In the same period, Anthropic's compute and infrastructure costs grew 77 percent and it kept signing deals that commit it to more expenditure.
Last week, we reported that Anthropic's purchase commitments exceeded $800 billion over the next decade or so.
Anthropic now says its commitments are closer to $895 billion.
Roughly $320 billion of those commitments relate to existing contractual obligations. Another $451 billion covers leases Anthropic has signed, but that have yet to commence, with the company expecting them to begin between 2026 and 2028, although some remain subject to termination rights or other conditions.
Overall, however, Anthropic's compute spend is on track to increase substantially over the next few years. And that's not even counting the blank checks Anthropic has written to secure adequate compute, which it values at $150-$175 billion, putting Anthropic's total compute tab at somewhere north of $1 trillion.
Anthropic can't walk away from some of this spending because its deals with Microsoft, Google, Amazon, and Broadcom, which now total $437 billion, are non-cancellable. Some require payment regardless of whether Anthropic lights up a single GPU.
The AI upstart estimates that of its $895 billion in contracts, only about $213 billion worth allow it to avoid payments.
The revised S1 filing also offered insights into the company's customer base, including a major reliance on cloud service providers for distribution. The likes of Amazon, Google, and Microsoft are some of the model dev's biggest compute providers, but according to the filing, cloud marketplaces accounted for roughly 44 percent of its revenue in the nine months ended September 30.
Anthropic insists that none of its customers account for more than 10 percent of revenue. Yet reliance on hyperscalers for distribution remains a risk because they are all either actively developing competing products or provide access to models from Anthropic's rivals.
Despite all the risks in its prospectus, Anthropic believes it can become a sustainable business if it can access enough compute capacity. And by doing so, the prospectus claims AI will reshape the economy "in ways greater than the impact of industrialization, electricity, semiconductors, and the internet."
With so much at stake, how can humanity afford not to spend like a drunken sailor? ®
Originally published on The Register

