AI-assisted research analysis by FutureKnowledge. Sources checked 21 October 2026. This is not investment advice.
Reviewed by FutureKnowledge: figures, dates and attributions were checked against the linked sources before publication.
Anthropic’s draft IPO prospectus shows first-half 2026 revenue of $16.3 billion and a net loss of $80.7 billion, according to reports by The Wall Street Journal and Reuters, which both reviewed the confidential draft. Most of that loss is an accounting charge, but the document also shows how much the maker of the Claude models has promised to pay for computing power over the next decade.
What the Anthropic draft S-1 shows
Anthropic, which is organised as a public benefit corporation, announced on 1 June 2026 that it had confidentially submitted a draft registration statement on Form S-1 to the US Securities and Exchange Commission. The share count and price have not been set, and the company said the offering would depend on the SEC’s review and market conditions. (Verdict reported it at the time.) The draft is not public, so the figures below come from news organisations that have seen it. Anthropic declined to comment to the Journal and did not respond to Reuters.
- Revenue: $16.3 billion in the six months to 30 June 2026, up from about $1.1 billion a year earlier, according to the Journal’s reporting as summarised by Implicator.ai and Stocktwits on Yahoo Finance. Roughly 15 times higher.
- Net loss: $80.7 billion, against about $13 billion in the first half of 2025. Reuters, republished by Global Banking & Finance Review, says about $72 billion of this is a non-cash charge reflecting a higher estimated fair value of convertible notes that could turn into Anthropic shares.
- Computing costs: nearly $9 billion on compute and infrastructure in the half, up from $2.3 billion a year earlier and above the roughly $7.3 billion spent in all of 2025.
- Customer concentration: about one-fifth of first-half revenue came from two unnamed customers, TradingKey and Implicator report.
Why the loss is so large, and what it does not mean
The headline loss is mainly the result of how convertible financing is accounted for. When a company’s estimated value rises, notes that can convert into its shares become more valuable, and that change is booked as a loss. Reuters says the same effect drove most of Anthropic’s reported $42 billion net loss for 2025. Taking the $72 billion charge out leaves about $8.7 billion. Implicator cautions that this residual should not be read as an operating loss or cash burn.
The cash picture looks better than the headline. The roughly $9 billion of first-half compute spending is a little over half of revenue for the period. Separately, Reuters reported that Anthropic told investors at an investor meeting on 14 October that third-quarter revenue rose 45% from the previous quarter to $16.8 billion, after $11.6 billion in the second quarter. The company described the quarter as profitable on an adjusted basis. Neither Reuters nor the Journal reporting explains how that adjusted measure is defined, so it should not be confused with GAAP profit.
The compute commitments are the bigger number
The figure that may matter most for the AI industry is the size of Anthropic’s long-term purchase commitments. According to the Journal-based reports, the draft lists about $450 billion of cloud agreements over seven to ten years, most of them non-cancelable. These include about $111 billion with Google Cloud, $110 billion with Amazon Web Services and $31 billion with Microsoft. There is also up to $86 billion to rent Nvidia-based capacity from Nscale and Lambda through 2033 and 2034, and more than $160 billion of lease obligations for Google chips through a Broadcom arrangement. Total commitments are put above $600 billion, with talks that could add about $150 billion.
The sources do not fully agree. Reuters puts the commitment at “at least $518 billion” over a decade. On our reading, the individual line items in these reports also do not add neatly to the stated totals. Until a public S-1 is filed, these numbers should be treated as approximate. The draft also reportedly says that Amazon, Google and Microsoft increasingly see Anthropic as a competitor and have not supplied all the capacity it wanted.
How it affects markets, India and businesses
Reuters has framed the Anthropic IPO as a test for the wider AI trade: a weak reception could weigh on AI-linked stocks ahead of a possible OpenAI listing next year. Reuters also reported that some investors were discussing a valuation of about $1.5 trillion for a share sale expected around November, below earlier talk of $2 trillion. Part of the reason given was the slower sequential growth in the latest quarter. No price range has been set.
For India the links are indirect but real. Global sentiment towards AI hardware and software moves foreign flows into Indian technology shares, as seen in Nifty IT’s best week since 2008 and the chip-stock losing streak that followed.
Who is more or less exposed
- More exposed: cloud and data-centre suppliers whose order books lean on a few AI labs; newer GPU-rental operators that depend on long contracts; and investors in pre-IPO vehicles priced near the top of the valuation range.
- Less exposed: diversified hyperscalers, whose AI contracts are a part rather than the whole of their revenue, and enterprise users with contracts across several model providers.
Our earlier report on Anthropic’s 5% public benefit share pledge covers the governance side of the listing.
Scenarios to watch
These are conditional scenarios, not predictions.
| If… | Then a likely consequence would be… |
|---|---|
| The public S-1 confirms revenue growth and shows positive operating cash flow | Investor focus could shift from losses to the size and timing of compute payments |
| Sequential growth slows further in the fourth quarter | Valuation talk could settle nearer the lower figures reported by Reuters |
| Commitments are confirmed near $600 billion or more | Questions about the financing of AI infrastructure and supplier concentration could sharpen |
| The listing timetable slips beyond the expected November window | AI-linked stocks may stay sensitive to each new leak or report |
What to watch next
- The public S-1 filing, which must appear at least 15 days before an investor roadshow, and the audited figures it contains.
- A price range, and whether the valuation lands nearer $1.5 trillion or $2 trillion.
- How Anthropic defines “adjusted” profit and how it reconciles that with GAAP results.
Bottom line
The Anthropic draft prospectus shows a business whose revenue grew about 15-fold in a year, a headline loss dominated by an accounting charge, and compute commitments that run into the hundreds of billions of dollars. The figures are still second-hand and some do not reconcile; the public filing will be the first chance to check them.
Method note: This analysis compared reporting by Reuters (via Global Banking & Finance Review), Wall Street Journal-based summaries by Implicator.ai and Stocktwits (on Yahoo Finance), TradingKey’s analysis and Verdict’s June report, against Anthropic’s own 1 June 2026 announcement of its confidential S-1 submission. Where totals for compute commitments differ, both figures are given. FutureKnowledge’s research tool is built on a model made by Anthropic; all figures here are attributed to independent reports, not to the tool.