Anthropic’s compute and infrastructure expense was 58% of its revenue in the third quarter of 2026, down from 148% a year earlier, according to the percentage-of-revenue table in the amended Form S-1 it filed with the SEC on Monday, 2 November. In dollars, that is about 58 cents of compute per dollar of revenue, against $1.48 a year before, by our arithmetic.
The ratio rose from 48% in the second quarter. The filing attributes the rise to the delivery of additional compute, including $5.0 billion under short-term commitments that it says costs significantly more than compute bought under multi-year deals.
The key insight: Anthropic’s amended S-1 shows compute and infrastructure expense at 58% of revenue in the third quarter, down from 148% a year earlier but up from 48% in the second quarter, which the filing attributes to $5.0 billion of short-term compute.
The filing says compute and infrastructure expenses as a percentage of revenue “decreased from 251% for the three months ended March 31, 2025 to 48% for the three months ended June 30, 2026, as revenuegrowth outpaced the growth of our committed compute cost base, and increased to 58% for the three months ended September 30, 2026, reflecting the delivery of additional compute”.
The quarterly table puts the ratio at 177% in the second quarter of 2025, 148% in the third, 144% in the fourth, 72% in the first quarter of 2026 and 48% in the second.
Anthropic’s compute and infrastructure expense as a percentage of revenue, first quarter of 2025 to third quarter of 2026 (unaudited). Source: Anthropic PBC, Form S-1/A filed with the SEC on 2 November 2026.
Why the Ratio Rose in the Third Quarter
The filing says the third quarter included “$5.0 billion of compute under short-term commitments, which currently costs significantly more than comparable compute procured under our multi-year commitments”, and that “this reduced revenue per dollar of compute compared to the preceding quarter”.
Adjusted Operating Income, a non-GAAP measure, rose to $2.5 billion from $2.2 billion, while Adjusted Operating Margin “decreased from 19% for the three months ended June 30, 2026 to 15% for the three months ended September 30, 2026, primarily driven by the $5.0 billion of short-term compute noted above”, the filing says.
For the first nine months of 2026, the filing says short-term compute came to $6.5 billion.
Revenue Grew Faster Than Compute
The filing says revenue “grew 1,127% year over year for the three months ended September 30, 2026, compared to our compute and infrastructure expenses that grew 382% during the same period”. Revenue was $16,956 million in the quarter and compute and infrastructure expense $9,863 million, against $1,382 million and $2,047 million a year earlier. That is about 12.3 times the revenue for about 4.8 times the compute, by our arithmetic.
The $2.0 Billion Donation-Matching Charge
The operating loss of $567 million also included “$2.0 billion of non-cash charges related to our donation matching program”, the filing says. For the nine months to 30 September, it says, “expenses related to our donation matching program were $2.1 billion, representing approximately 6% of our total operating expenses for that period”.
Business Pill · Gross Margin
Business Pill: Gross Margin — what is left of each revenue dollar after the direct cost of delivering it.
As we read the quarterly table, the ratio fell for five straight quarters as revenue grew faster than a compute base Anthropic says it commits 12 to 18 months ahead, then rose once extra short-term compute arrived.
As we read the margin lines, the same $5.0 billion explains both the higher compute share and the fall in Adjusted Operating Margin to 15% from 19%, according to the filing.
Anthropic’s amended Form S-1
“This reduced revenue per dollar of compute compared to the preceding quarter”
Three Implications
FOR IPO INVESTORS The filing shows compute and infrastructure expense at 58% of revenue in the third quarter, against 48% in the second, and Adjusted Operating Margin at 15% against 19%.
FOR COMPUTE SUPPLIERS The filing says $5.0 billion of third-quarter compute was under short-term commitments and currently costs significantly more than compute bought under multi-year commitments.
FOR AI SUBSCRIBERS The filing says revenue grew 1,127% year over year in the quarter while compute and infrastructure expenses grew 382%.
The framework says: “The new job of finance is to manage the ratio between the two โ work produced per token burned.”
As we read it, Anthropic’s compute-to-revenue ratio is one company-level version of that ratio: 58 cents of compute per dollar of revenue in the third quarter, by our arithmetic. The framework is the owner’s argument and does not discuss this filing.
The ratios come from unaudited quarterly figures. The filing does not say how much more the short-term compute costs per unit than multi-year compute, and it does not split compute cost between serving customers, training models and internal use. The cents-per-dollar figures and the 12.3-times and 4.8-times comparisons are our arithmetic.
This Week In The AI Supercycle
Anthropic’s Reality Check, The Capital Squeeze, & The New World Order
This week’s issue: for two years the AI build-out was priced as a growthstory. This week it started being priced as a credit story.
Anthropic’s amended S-1 shows compute and infrastructure expense at 58% of revenue in the third quarter of 2026, down from 148% a year earlier but up from 48% in the second quarter. The filing attributes the rise to $5.0 billion of short-term compute, which also pulled Adjusted Operating Margin to 15% from 19%.
A note on sourcing. All figures and quotations are from Anthropic’s amended Form S-1 filed with the SEC on 2 November 2026; quarterly figures are unaudited. The cents-per-dollar figures and the 12.3-times and 4.8-times comparisons are our arithmetic.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
