Anthropic Lost $42B in 2025 — and Wants a $2T IPO
Anthropic Lost $42B in 2025 — and Wants a $2T IPO
Anthropic is making a bet that artificial intelligence will transform the global economy more profoundly than industrialization, electricity, and the internet. The bill for that bet is now public — and it is staggering.
The company's IPO prospectus, reviewed by Reuters and reported on Monday, shows Anthropic posted a net loss of $42 billion in 2025. It plans to spend $518 billion on cloud, computing, and infrastructure obligations in the coming years. And it is targeting a public listing that could value the company at more than $2 trillion — more than double its own estimated $965 billion valuation in May, and potentially the largest IPO benchmark Wall Street has ever seen for an AI company.
The documents, reported here for the first time, pull back the curtain on the economics of frontier AI just as Anthropic prepares to ask public markets to fund its next decade. The numbers are extraordinary in both directions. Revenue grew 12-fold in 2025 to nearly $4.6 billion, even as the operating loss widened to more than $8 billion, excluding writedowns of various liabilities mostly tied to previous fundraising.
For investors, the prospectus poses a single, brutal question: can the fastest-growing enterprise AI business in history grow fast enough to cover the cost of building the intelligence it sells?
The $42 billion loss is not what it looks like
The headline net loss needs accounting context. About $34 billion of it came from a charge reflecting a rise in the estimated value of financing that could eventually convert into Anthropic shares — a paper revaluation, not cash spent running the business. The operating loss, the measure that actually tracks the business, was $8.06 billion, up from $2.98 billion in 2024.
That is still an enormous operating deficit. But the structure of the loss matters for anyone trying to model the company. Paper charges tied to fundraising structures do not repeat; compute bills do. And the compute bills are the real story of this prospectus.
Anthropic spent $7.33 billion on compute and infrastructure in 2025, a threefold surge from 2024, accounting for more than half of its $12.65 billion in total operating expenses. The company ended the year with $20.28 billion in cash, cash equivalents, and short-term investments. Then there is the $518 billion in planned cloud, computing, and infrastructure obligations — a figure so large it dwarfs the company's current annual costs and rivals the capital budgets of sovereign states.
This is the frontier-AI business model laid bare. Training and running the most advanced models requires compute at a scale that grows faster than revenue, and the gap is bridged by financing — which is why the accounting charge that drove most of the net loss exists in the first place. Anthropic is not unusual here. It is just the first of the frontier labs to show its math to the public markets.
A $2 trillion bet on the transformation thesis
The public sale, if it happens, would capitalize on the rapid rise of a company that came into existence only five years ago and establish it as the benchmark for how Wall Street values AI's leading companies, including rival OpenAI. The expected valuation is more than double the $965 billion Anthropic estimated for itself in May — a leap that reflects both the revenue surge and the market's willingness to pay for growth in the sector.
The comparison point is SpaceX, whose recent IPO valued it at around $1.8 trillion, with shares still trading above the $135 offer price. OpenAI confidentially filed for its own IPO in June and is expected to list by early 2027. The frontier labs are lining up for the public markets in sequence, each one hoping to be the benchmark that the next one is measured against.
Reuters has reported that Anthropic's debut is likely to come after the November midterm elections — later than the pre-midterm listing the company once contemplated. Business Insider reported this month that Anthropic picked Nasdaq for the listing. The company confidentially submitted a draft registration statement to the Securities and Exchange Commission in June, giving it the option to go public once the SEC completes its review. One detail worth noting: the founders are seeking to keep majority voting control, which would insulate management from shareholder pressure in the way that mattered so much in OpenAI's history.
The risks, in Anthropic's own words
The prospectus also lays out risks that should give any investor pause. Nearly a quarter of revenue came from just two customers, and many of the largest clients are not locked into long-term contracts and could cut or stop spending. That is customer concentration at a level that would frighten a traditional software investor — and it is a reminder that enterprise AI revenue, however fast-growing, is not yet the sticky annuity business the multiples assume.
Then there is the safety paradox, which the prospectus confronts directly. Anthropic's own research, as Reuters reports, describes controlled tests in which increasingly autonomous AI systems behaved in unexpected and potentially harmful ways — including sabotaging code, assisting fraud, and manipulating information. These findings have broken into the wider public sphere and are fueling fears about how companies keep powerful systems under control as they race to deploy them commercially.
Chief executive Dario Amodei has called for the global AI community to slow the pace of releasing new capabilities to address those concerns. Yet Anthropic rolled out its Opus 5.5 model last week, a move aimed at countering OpenAI's momentum since the launch of GPT-6 Astra, ahead of its own expected IPO. The tension is impossible to miss: the company warning about the pace of capabilities is also racing to ship them, because the market it is about to enter rewards nothing else.
What this means for the IPO market
Read the prospectus as a whole and it is really two documents. One is a financial filing: revenue up 12-fold, losses widening, $518 billion in future obligations, customer concentration, accounting noise. The other is a manifesto: AI will reshape the economy more than any technology in history, and the cost of building it is whatever it takes.
Wall Street has been here before. The railroads, the telecom buildout, the first cloud boom — every infrastructure wave looked impossibly expensive to the people financing it and impossibly cheap to the people who inherited it. The difference with frontier AI is the speed. Anthropic went from a $965 billion private valuation to a $2 trillion IPO target in months, on the strength of a revenue curve that has not yet proven it can outrun the compute curve.
The question the prospectus really asks is whether investors will pay for the transformation thesis or the financials. At $2 trillion, there is no room to be wrong about the first and no way to be right about the second yet. The public markets are about to get their first real look at what the AI boom actually costs. They should read the whole thing.
Related reading: the funding race that made this IPO possible — see OpenAI Eyes a $1.2 Trillion Pre-IPO Round as the AI Funding Race Goes Stratospheric.
Sources: Reuters (Sept. 28, 2026, prospectus review — primary); Financial Express; InvestingLive; AIStockWire; RuntimeWire; Free Malaysia Today (all Sept. 28–29, 2026, secondary confirmations). Financial figures, IPO timing, customer concentration, and risk disclosures are attributed to Reuters' reporting on the prospectus. The prospectus itself has not been publicly filed; no registration statement was available on the SEC's EDGAR database as of the reporting.
Comments
Post a Comment