Anthropic's confidential IPO prospectus shows the company behind Claude betting that AI could drive economic change bigger than industrialization, electricity and the Internet. But that ambition comes with a huge bill: Anthropic recorded a net loss of about $42 billion in 2025 and expects to spend up to about $518 billion on cloud, compute and infrastructure obligations in coming years, per documents reviewed by Reuters.

Has Anthropic actually filed for an IPO?

Yes — but precisely stated. Anthropic confirmed it confidentially submitted a draft Form S-1 to the SEC for an IPO. The filing lets the company move toward a listing after SEC review, but Anthropic stresses the IPO timing still depends on market conditions and other factors; the number of shares and the offering price are undecided. Reuters says it reviewed a roughly 300-page prospectus revealing many previously undisclosed financial figures.

A $42 billion loss — but read the number carefully

The most eye-catching figure is a $42 billion net loss in 2025. But putting just that in a headline is easily misleading. Reuters says about $34 billion of the loss was a non-cash accounting charge tied to the estimated value of financing that could convert into shares — not cash spent on the business.

Strip that out and Anthropic still posts an operating loss of over $8 billion — up from around $3 billion in 2024. Put simply: ~$42B net loss, of which ~$34B is a financing-related write-down, with the operating loss over $8 billion. $8 billion is still a very large loss, but it reflects the business more clearly than the $42 billion figure.

Meanwhile, revenue grew 12-fold

The other side of the story is very different. Anthropic's revenue grew about 12-fold in 2025 to nearly $4.6 billion. Growth kept accelerating: in February 2026 Anthropic said run-rate revenue reached about $14 billion; by May, announcing its Series H, run-rate revenue had passed $47 billion. This is the central paradox of today's frontier AI: revenue can grow extremely fast — but so does the scale of compute demand.

$518 billion: a figure more notable than the loss

The prospectus Reuters reviewed says Anthropic has about $518 billion in spending obligations tied to cloud, computing and infrastructure in coming years, and the company cannot cancel about 80% of it. Importantly, don't read this as 'Anthropic will spend $518 billion immediately' — these are future commitments spread over years.

But its scale shows a clear reality: frontier AI isn't just a model problem. Behind Claude are data centers, accelerators (TPU/GPU/Trainium), power, networking, cloud capacity, inference and training — all needing enormous capital.

Amazon and Google are both partners and risks

Another notable part is how much Anthropic depends on Big Tech. Per Reuters, a large share of 2025 revenue reached customers through cloud partners including Amazon and Google (the prospectus also shows high customer concentration). The relationship is unusual: Amazon and Google can simultaneously be investor, compute supplier, distribution channel and AI rival.

Anthropic also warns in the filing that if third-party compute access is restricted, repriced or terminated, the business could be affected. It's one of the company's key structural risks.

How powerful an AI is Anthropic imagining?

This is what makes the prospectus differ from a typical tech IPO filing. Anthropic uses the term 'Powerful AI' for systems that could surpass top human experts across nearly every cognitive domain. Per what Reuters saw, such systems could coordinate with other AIs, control lab equipment, robots and machines, run continuously for weeks with little human intervention, drive scientific research and drug discovery, and perform other complex intellectual tasks.

In other words, Anthropic isn't selling investors a simple story that 'Claude will become a better chatbot'. The vision is much broader: AI as a capability layer able to affect nearly the entire economy.

Yet nearly a third of the prospectus is about… risk

This may be the most unusual detail. Reuters says nearly a third of the roughly 300-page filing is devoted to risk factors — more than the pages describing Anthropic's own operations. And the risks aren't just competition, revenue, legal or chip supply: Anthropic also warns advanced AI could pose 'catastrophic or existential' risks to humanity if not developed or deployed appropriately.

The filing also mentions the possibility that advanced systems could resist shutdown, hide or manipulate information, or behave deceptively or coercively. These are risk disclosures and scenarios Anthropic raises, not a claim those outcomes will certainly happen.

A very Anthropic paradox

What makes the prospectus special is that the company says two things at once: AI can create huge economic value, and AI powerful enough can also create extremely serious risks. Yet Anthropic's proposed solution requires the company itself to keep building more powerful models. Reuters describes this as the central paradox in the IPO story: to make AI safer, Anthropic argues it must be at the frontier of capability; but the more capability grows, the larger the potential risks it warns about.

Seven founders could retain outsized control

The prospectus reveals an unusual governance structure. Per Reuters, a Founder LLC tied to Anthropic's seven co-founders would use Class F shares to jointly hold 50.1% of voting power over corporate matters, including board elections. That means Anthropic could become a public company while founders keep very strong control — and the filing itself warns this structure could lead to decisions not fully aligned with shareholders' short-, medium- or long-term financial interests.

Is Anthropic really valued at $2 trillion?

Not yet — a point to be careful not to get wrong. Reuters discusses the possibility that an IPO could target a valuation around $2 trillion — more than double Anthropic's own ~$965 billion estimate in May. But that's not a finalized IPO valuation; Anthropic officially says the number of shares and offering price are still undetermined.

Recent official valuation markers confirmable from the company: about $380 billion after Series G in February 2026, and about $965 billion after Series H in May 2026. So a headline like 'Anthropic IPO at a $2 trillion valuation' is premature; more accurate is 'Anthropic is preparing for an IPO that could target a $2 trillion valuation'.

When is the IPO?

No certain date. Reuters says earlier plans aimed for a public filing in late September and a roadshow around mid-October, but the timeline appears to have slipped; some parties are weighing a deal after the US midterm elections. Anthropic hasn't announced an official IPO date. So as of September 30, 2026, the precise status is: Anthropic is preparing for an IPO, but the final timing and terms aren't locked.

Why is this filing more notable than a typical tech IPO?

Because Anthropic is essentially asking investors to bet on three things at once. First: growth speed — revenue is rising very fast. Second: the ability to raise and spend enormous capital — frontier AI needs compute at the hundreds-of-billions scale. Third: a future that hasn't existed — if 'Powerful AI' becomes real, Anthropic argues it could change nearly every part of the economy, yet the company itself devotes much of the prospectus to why that future could carry serious risks.

That's what makes this IPO filing unusual. Anthropic isn't only telling investors 'believe we'll make a lot of money', but also 'believe the technology we're building is important enough to change the economy — and powerful enough that we must be extremely careful with it'.

30-second recap

  • Has Anthropic IPO'd? — No; it filed a confidential draft S-1.
  • 2025 revenue? — About $4.6 billion, up roughly 12x.
  • How big is the loss? — Net loss ~$42B, of which ~$34B is a non-cash financing write-down; operating loss over $8B.
  • Compute/infra commitments? — About $518B in coming years (~80% non-cancelable), per the prospectus Reuters saw.
  • Recent run-rate revenue? — Anthropic said it passed $47B in May 2026.
  • Is the $2T valuation set? — No; it's a potential figure discussed; price and share count undetermined.
  • Notable risks? — Nearly a third of the prospectus covers risk factors, including serious risks Anthropic says advanced AI could pose.