Anthropic S-1/A: Q3 revenue $17.0B, $895B commitments, and a warning it may 'imminently' hit its AI-acceleration threshold
Amendment No. 1 to Anthropic's IPO filing adds Q3 2026 numbers, about $895 billion of compute commitments and a new risk factor on AI-accelerated research.
· 7 min read
Written by AI from the sources linked below. How we use AI
On this page
Anthropic's amended IPO filing, Amendment No. 1 to its Form S-1 dated 2 November 2026, adds the company's third-quarter numbers and a new warning in its risk factors: the company says its own estimates indicate it "may imminently meet" the Responsible Scaling Policy's definition of "dramatically accelerated progress attributable to AI". That is the company's wording for the point at which AI is speeding up AI research enough to matter for safety. The same filing shows revenue of $17.0 billion in the quarter, an operating loss that has nearly closed, and about $895 billion of compute and infrastructure commitments.
Key points
- Anthropic filed Amendment No. 1 to its S-1 on 2 November 2026; it plans to list Class A shares on the Nasdaq Global Select Market under the symbol ANT, and the filing still leaves the price range blank.
- Q3 2026 revenue was $16.956 billion, up about 47% from $11.555 billion in Q2, after a roughly 2.4x jump from Q1 to Q2.
- The new risk factor says Anthropic "may imminently meet" its Responsible Scaling Policy threshold for "dramatically accelerated progress attributable to AI", but has not designated any current model as exceeding it.
- Purchase commitments and leases total about $895.4 billion, of which about $447.5 billion is expected to be incurred before 31 December 2030.
- Anthropic says it does not expect revenue to keep growing at comparable rates and that it cannot assure investors it will keep growing at all.
What happened
The amendment adds the quarter that ended 30 September 2026 to the numbers investors saw in the first filing, which we covered in our report on the original S-1. The first filing covered the first half of 2026, with $16.3 billion of revenue. The new filing's quarterly table gives revenue of $4.731 billion for Q1 and $11.555 billion for Q2, which add up to the same $16.3 billion, and $16.956 billion for Q3. For the nine months to 30 September, revenue was $33.2 billion, which the filing puts at 1,206% year-over-year growth.
Costs grew too. Compute and infrastructure expense was $9.863 billion in Q3, up from $5.558 billion in Q2 (about 77%). The filing says $5.0 billion of that was compute bought on short-term commitments, which "currently costs significantly more" than compute under multi-year deals. Operating loss for the quarter was $567 million, or 3% of revenue, against an operating income of $974 million in Q2. Net loss was $23.3 billion in Q3, but almost all of that is not operating: a $21.9 billion line for the change in fair value of convertible notes, derivative liabilities and SAFEs (a type of investment contract), which moves with the company's valuation.
The filing also says a price cut mattered. Between Q2 and Q3, higher usage was partly offset by a decrease of about $600 million from price reductions introduced with new model versions and contractual discounts. Customers with at least $100,000 of trailing-twelve-month revenue rose from about 400 in March 2025 to more than 11,900 in September 2026.
The RSI warning, in the filing's words
Most coverage of the amendment, including TechCrunch's report, focuses on a passage in the risk factors. We checked it against the filing itself. It says Anthropic's increasing use of AI to accelerate its own research and development may make its models advance faster than its ability to build and test safety measures, and that "we have in the past not identified, and may in the future not identify, a significant acceleration in our rate of development until after it has occurred."
It then says: "our internal estimates and our expectations regarding future models, rather than our current models, indicate that we may imminently meet our Responsible Scaling Policy's definition of 'dramatically accelerated progress attributable to AI.'" And it adds that Anthropic has not designated any current model as exceeding the threshold and "continue[s] to assess the underlying data". It also says the company is "evaluating the implications" and will publish further reports as its policy sets out.
Two points are worth keeping apart. First, TechCrunch and others describe this as "recursive self-improvement" (RSI). The filing's own term is the policy threshold, and the filing says the company does not expect to safely use Claude "to fully autonomously design and build successive generations of frontier models without meaningful human supervision or understanding". It is "considering updates" to its Responsible Scaling Policy "to keep humans in control of frontier model development." Second, the statement is about internal estimates of future models, not a claim that a model has crossed the line.
Background
Anthropic's revenue was $41 million in 2023, $386 million in 2024 and $4.6 billion in 2025, according to the filing, which gives year-over-year growth of 841% in 2024 and 1,088% in 2025. The first S-1 landed in late October. Since then, the Wall Street Journal has reported, citing sources, that institutional investors told Anthropic its valuation should be closer to $1.5 trillion to $1.6 trillion than the $2 trillion that had been estimated; we have not seen that report ourselves, and the filing gives no price range.
The numbers
| US$ billion | |
|---|---|
| Q1 25 | 0.376 |
| Q2 25 | 0.787 |
| Q3 25 | 1.382 |
| Q4 25 | 2.040 |
| Q1 26 | 4.731 |
| Q2 26 | 11.555 |
| Q3 26 | 16.956 |
Source: Anthropic Amendment No. 1 to Form S-1, quarterly statements of operations (SEC, 2 November 2026)
Growth is still steep but has slowed: Q2 was 2.4 times Q1, while Q3 was 1.47 times Q2. The Register's report on the filing makes the same point, noting that compute costs grew 77% over the same quarter while revenue grew about 46%. Our own calculation from the filing's figures gives 46.7% for revenue and 77.5% for compute.
| US$ billion | |
|---|---|
| Q1 25 | 0.944 |
| Q2 25 | 1.390 |
| Q3 25 | 2.047 |
| Q4 25 | 2.945 |
| Q1 26 | 3.387 |
| Q2 26 | 5.558 |
| Q3 26 | 9.863 |
Source: Anthropic Amendment No. 1 to Form S-1, quarterly statements of operations (SEC, 2 November 2026)
As a share of revenue, compute fell from 251% in Q1 2025 to 48% in Q2 2026, then rose to 58% in Q3 2026 as more committed capacity came online. The filing says compute is typically bought 12 to 18 months ahead, so these costs are fixed in the near term.
The commitments are the other big figure. The filing says Anthropic has entered purchase commitments and leases, including leases not yet started, with an aggregate committed spend of about $895.4 billion, almost all for compute and infrastructure. About $447.5 billion is expected before the end of 2030, most of the rest between 2031 and 2045, and about $69.8 billion falls due in the twelve months after 30 September 2026. The Register adds its own breakdown (roughly $320 billion of existing contractual obligations and $451 billion of leases that have not started, and an estimate that only about $213 billion of the total allows Anthropic to avoid payments); those splits are the outlet's reading, and we have not checked them against the filing's notes. It also adds Anthropic's valuation of unwritten "blank checks" at $150 billion to $175 billion to reach a total "north of $1 trillion". Treat that as the outlet's estimate, not a figure Anthropic reports.
What it means
For developers, the most direct signal is price and capacity. The filing says new model versions came with price reductions that cost about $600 million of quarterly revenue, and that short-term compute costs much more than contracted compute. Whether that translates into tighter limits or cheaper tiers for API users is not something the filing says.
For the industry, the filing shows how much of the AI buildout depends on lenders. It says some compute and infrastructure partners "depend on third-party financing or credit support to acquire chips and develop capacity", and that contracted capacity could be delayed, repriced, reduced or not delivered if that financing tightens. The cloud providers are also a distribution channel: sales through cloud marketplaces were about 44% of revenue in the nine months to September 2026, down from 48% in the same period a year earlier and from 47% for all of 2025.
For safety policy, the notable thing is a company putting a measurable threshold, defined in its own policy, into a securities filing. Investors can now hold it to what it said. The Register's reading is more sceptical, calling much of the language ordinary protective drafting for an IPO. TechCrunch notes the same tension: a prospectus is both disclosure and marketing.
What we don't know yet
- The IPO price range and share count; both are blank in this amendment.
- Whether and when Anthropic will designate a model as meeting the "dramatically accelerated progress" threshold, and what its further reports will contain.
- What the planned changes to the Responsible Scaling Policy will say.
- The final valuation, which depends on investor demand and has only been reported by outlets citing unnamed sources.
What to watch next
Watch for a further amendment with the price range, for any Responsible Scaling Policy update or capability report from Anthropic, and for the Q4 numbers once it reports as a public company. Also watch whether the financing risk the filing describes shows up in delayed data center deliveries.
Sources
- Anthropic, PBC, Amendment No. 1 to Form S-1, SEC, 2 November 2026 (primary source)
- TechCrunch: Anthropic warns that it's very close to RSI capability, aka self-building AI
- The Register: Anthropic's compute tab could top $1T as Q3 revenue nears $17B
- CNBC: Anthropic reports $17 billion in Q3 revenue, purchase commitments of $895.4 billion in revised S-1
Written by
The official account of the people who run audiotale.net: announcements, conversation starters and the newsroom's drafts. Posts written with AI say so. Every other account here is a real person.
Read next
Anthropic S-1: $16.3B H1 2026 revenue, $80.7B net loss, Nasdaq symbol ANT
Anthropic's public IPO filing shows revenue of $16.29 billion in six months, a $693 million operating loss, $112 billion in cash and a governance structure that gives founders and a trust control.
audiotale.net team8 min read
Anthropic launches Claude Duet, a proactive background agent app, as a US research preview
Claude Duet sends morning briefs, works on its own virtual machine and asks before it acts; it is open by waitlist to US Pro and Max users.
audiotale.net team6 min read
Anthropic's draft federal AI safety bill: what is reported about court-gated model blocks
Politico reports Anthropic is pitching Congress a bill letting a judge halt an AI model only for imminent risks of 50+ deaths or $1 billion in damage.
audiotale.net team7 min read