Announced 20 Oct 2026 · Sources checked
What the draft shows
Anthropic took in $16.3 billion of revenue in the first six months of 2026 and recorded a net loss of $80.7 billion, according to a draft of its initial public offering prospectus. The Wall Street Journal reported the figures on 20 October, and Reuters said it had reviewed the draft too. The revenue compares with $1.1 billion in the first half of 2025, so sales grew almost fifteen-fold in a year.
The prospectus is a Form S-1, the document a company files with the US Securities and Exchange Commission before listing shares. Anthropic has not made it public. AI Lookout found no Anthropic S-1 on the SEC's EDGAR database when it checked on 21 October, which fits a confidential submission. That means every number in this article comes from journalists who have seen a draft, not from a published filing, and the figures can still change before the public version appears.
The new draft extends an earlier version that covered only 2024 and 2025. Reuters first reported that version in late September, when it showed a net loss of about $42 billion for 2025 on revenue of nearly $4.6 billion. Reuters says Anthropic is expected to go public in November. Anthropic did not respond to requests for comment from Reuters or TechCrunch.
Why the loss is so large
Most of the $80.7 billion is not cash the company spent. Reuters says the loss includes a $72 billion accounting charge reflecting a rise in the estimated value of bonds that could eventually convert into Anthropic shares, "rather than cash costs tied to its operations." Convertible notes are recorded as a liability; as investors value Anthropic more highly, the notes are worth more, and the increase is booked as a loss under US accounting rules.
Take the $72 billion out and roughly $8.7 billion of loss remains for the half year. That is AI Lookout's own subtraction from the reported figures, not a number from the draft, and it does not match any specific line in the filing. A year earlier, according to Reuters, the first-half net loss was $13 billion.
The same pattern showed up in 2025. Reuters reported in September that the near-$42 billion annual loss included a roughly $34 billion charge of the same kind, and that excluding it Anthropic posted an operating loss of about $8.1 billion on $4.6 billion of revenue. Separately, Anthropic told prospective investors on 14 October that third-quarter revenue reached $16.8 billion and that it was profitable on an adjusted basis, a measure that strips out items such as these charges. Third-quarter revenue alone was therefore higher than the whole first half.
Computing costs and cloud commitments
The draft shows how quickly computing costs are rising. Anthropic spent nearly $9 billion on computing capacity and infrastructure in the first half of 2026, Reuters says, up from $2.3 billion a year earlier. Investing.com's summary of the Journal's report puts that against $7.3 billion for the whole of 2025, a figure that matches the $7.33 billion Reuters reported from the earlier draft.
The longer-term promises are much larger. Reuters says Anthropic has committed to spending at least $518 billion over a decade on cloud services, computing and infrastructure. According to the Journal's account, as summarised by Investing.com, about $450 billion of that is cloud computing agreements, including $111 billion with Google Cloud, $110 billion with Amazon Web Services and $31 billion with Microsoft. The draft also says Anthropic could spend $750 billion in the coming years once deals still under negotiation are counted.
Some of those deals have surfaced in recent weeks. Reuters and Bloomberg reported that Anthropic is negotiating a second roughly $45 billion compute deal with Nscale in West Virginia, and Anthropic has signed a 20-year lease at a new Nova Scotia campus. The draft gives those moves context: Anthropic is buying capacity wherever it can find it.
Tension with the cloud providers
The filing suggests that buying enough computing power has become a source of friction with the three big cloud companies. TechCrunch, citing the Journal, writes that Anthropic is having trouble securing enough capacity from AWS, Google and Microsoft, which increasingly see it as a competitor rather than a customer. According to the Journal's report, Anthropic says in the draft that it has asked for more capacity than the hyperscalers were willing to supply.
That is why Anthropic is leasing its own data centres and working with third-party operators. TechCrunch says the filing flags a risk there too: some of those operators have a "limited operating history," which could make them less reliable than the established cloud providers.
The relationship was already circular. Amazon and Google have invested heavily in Anthropic while selling it the computing it runs on, and in the earlier draft Reuters found that nearly half of Anthropic's 2025 revenue came through Amazon's and Google's cloud marketplaces. The new figures show Anthropic still depends on those companies for most of its committed capacity, even as it competes with them for customers.
Customer concentration and risk factors
Revenue is also concentrated. About 20% of first-half revenue came from two customers, according to the Journal, which did not name them. In 2025 the two largest customers accounted for nearly a quarter of revenue, Reuters reported from the earlier draft, and Anthropic warned then that many of its largest clients were not locked into long-term contracts and could cut or stop spending.
The risk factors go beyond what a typical software company lists. According to the reports, the draft discusses problems with model alignment and the possibility of catastrophic incidents, and warns that damages from a serious AI-related security event could exceed Anthropic's insurance coverage. TechCrunch describes this as building on catastrophic-risk disclosures already reported from the earlier draft.
Reuters says investors are weighing the figures closely. Some were disappointed by slowing quarter-on-quarter growth after the 14 October investor meeting, it reports, and some are now considering a lower valuation of around $1.5 trillion. Investing.com's summary of the Journal's report says the listing could value Anthropic at $2 trillion. Neither figure comes from the prospectus itself.
Pay, pledges and governance
The draft also covers executive pay. Dario Amodei, the chief executive, and Daniela Amodei, the president, each drew a base salary of about $655,000 in 2025, according to the Journal's report as summarised by Investing.com. Both have pledged 80% of their Anthropic equity to philanthropy, and the filing describes the 5% public benefit fund Anthropic announced on 18 October, which still needs stockholder approval.
Anthropic is a public benefit corporation, which lets its board weigh a stated public mission alongside shareholder returns. How that structure and the new fund will work once public shareholders are involved is one of the questions investors are likely to examine when the full filing is released.
What is still unconfirmed
None of these numbers can yet be checked against the filing. The Journal and Reuters have seen a draft; the version Anthropic eventually publishes may restate figures, add quarters or change the risk factors. The reports also differ in how they frame computing costs. Reuters compares first-half spending with the same period of 2025, while the Journal compares it with full-year 2025, and the $450 billion, $518 billion and $750 billion commitment figures measure different things.
AI Lookout has not seen the document and relies on Reuters' wire story, Investing.com's summary of the Journal's paywalled report, TechCrunch and Reuters' September coverage. Other AI companies' numbers make a useful comparison: SpaceX, which owns xAI, recently reported that its AI segment turned an operating profit in the third quarter, but its figures come from a public company's results, not a draft prospectus.
Why it matters
These are the first 2026 figures from Anthropic's own filing, and they show both sides of the business at once. Revenue grew almost fifteen-fold in a year, but the cost of the computing behind it is rising quickly, and the headline loss is dominated by an accounting charge that grows as the company's valuation climbs.
Anthropic's listing is widely seen as a test of whether public investors will fund frontier AI on the terms private investors have. Reuters notes that a cooler reception could weigh on the wider AI market ahead of OpenAI's potential listing next year. When the public S-1 appears, the details to watch are the adjusted profit measure, how the convertible notes are treated at listing, the identity or type of the two largest customers, and how firm the cloud commitments really are.
Common questions
Has Anthropic published its S-1?
No. The figures come from a draft seen by the Wall Street Journal and Reuters. AI Lookout found no Anthropic S-1 on SEC EDGAR on 21 October 2026.
Why did Anthropic lose $80.7 billion?
About $72 billion is an accounting charge from the rising value of convertible notes, not cash spent on operations, according to Reuters.
When is Anthropic expected to go public?
Reuters says the listing is expected in November 2026; Anthropic has not announced a date.
Sources & further reading
How this story was made
Written by Kristian Kostov with AI assistance and checked against the linked sources. Company performance claims are attributed to the company. Analysis reflects AiLookout’s interpretation; we have not independently tested the products discussed. Cover photography is illustrative and does not depict the specific announcement or product.
Our editorial standards





