Anthropic Files for Historic IPO After Losing $80.7 Billion

Anthropic lost $80.7 billion in the first half of 2026 — and it is still heading for what could be one of the biggest IPOs in history. The company disclosed the loss in a Form S-1 filed Monday with the Securities and Exchange Commission, saying it plans to list on the Nasdaq under the ticker symbol "ANT."

The five-year-old lab, built around its Claude model family, confidentially filed with the SEC in June. That gave regulators time to review the financials before the numbers went public. Now the numbers are here, and they cut both ways: explosive revenue growth alongside staggering accounting losses.

The loss figure needs unpacking. Anthropic said the total largely reflects accounting charges tied to changes in the value of convertible notes, "resulting from an increase in our enterprise valuation during each of the respective periods." The company lost $12.9 billion over the same stretch in 2025 and about $41.9 billion for all of 2025, widened from an $8.3 billion loss in 2024. In other words, the bigger the company got, the bigger the paper loss — an artifact of the debt instruments, not cash burn alone.

How fast is Anthropic actually growing?

Revenue tells the bullish half of the story. Anthropic reported $16.29 billion in revenue for the six months ended June 30 — up 1,304% from $1.16 billion a year earlier. Full-year 2025 revenue jumped 1,089% to $4.59 billion.

The growth is concentrated in consumption-based deals. That revenue category made up $13.1 billion, or 80% of total revenue, in the first half of 2026, versus $3.2 billion, or 20%, from subscriptions. "We expect consumption-based revenue to continue to represent the substantial majority of total revenue for the foreseeable future," the filing says. A growing number of enterprise customers sign non-cancelable annual or multi-year contracts.

The enterprise base runs deep. As of Sept. 30, Anthropic counted more than 11,900 customers with $100,000 or more in trailing twelve-month revenue, more than 1,900 customers above $1 million, and roughly 250 customers above $10 million. "The speed and depth of our enterprise customer adoption are powerful evidence of Claude's position at the frontier of AI," the filing says.

No single customer represented more than 10% of revenue for the nine months ended September 30 — better diversification than Nvidia, which had five clients contributing at least 10% of accounts receivable as of July. Still, five Anthropic customers accounted for about a quarter of revenue, and cloud marketplace customers made up 48% of revenue at the end of 2025.

Who runs Anthropic — and who controls it?

CEO Dario Amodei, 43, co-founded the company in 2021 after leaving OpenAI with a small group of researchers. He grew up in San Francisco, holds a biophysics doctorate from Princeton, worked as a senior research scientist at Google Brain and spent more than four years at OpenAI, where he led development of GPT-2 and GPT-3. His sister, Daniela Amodei, is president, co-founder and the second-highest paid executive.

The S-1 discloses the family ties in detail. Daniela Amodei's husband, Holden Karnofsky, has worked at Anthropic since January 2025 and received just over $430,000 in total compensation, excluding options and restricted stock. Karnofsky previously co-founded GiveWell and was co-CEO of Open Philanthropy, since rebranded to Coefficient Giving, which advises Good Ventures, the charity established by Facebook co-founder Dustin Moskovitz and Cari Tuna.

Control will stay with the founders. The Amodei siblings and five other co-founders — Tom Brown, Jack Clark, Jared Kaplan, Sam McCandlish and Christopher Olah — are sole members of Founder LLC, which will hold 50.1% of voting power. The group owns about 1.7 billion Class A shares. Anthropic said its governance structure "has no direct public company precedent" and may limit unsolicited merger proposals, proxy contests or director removals — potentially depriving stockholders of premium sale opportunities.

Pay is disclosed too. Dario Amodei earned $17.96 million in 2025, mostly in stock and options, on a $654,167 base salary. Daniela Amodei earned $16.4 million on the same base. Both saw their base salaries raised to $1.38 million effective July 1. CFO Krishna Rao's base salary climbed from $706,250 in 2025 to $1.25 million.

Where does the money go — and what could break?

Compute is the giant line item. Anthropic agreed to pay SpaceX around $3.1 billion per month for access to NVIDIA GPU-powered compute once all promised capacity is delivered, with total commitments up to $120.5 billion extending into 2029. SpaceX's third-quarter filing last week showed a customer — Anthropic — paid just over $5 billion, or 34.5% of SpaceX's $14.58 billion quarterly revenue.

Microsoft looms even larger. After investing $5 billion last November in a round valuing Anthropic near $350 billion, Microsoft now has commitments from Anthropic of at least $44.9 billion in spending between November 2026 and May 2033. Anthropic separately plans to spend $30 billion on Azure and contract up to a gigawatt of additional compute. Microsoft, which booked a $3.2 billion gain on its stake after Anthropic's May raise, integrates Claude into GitHub Copilot, the Foundry model garden and Copilot Cowork — while advertising its own in-house models' advantages over Anthropic alternatives. "Amazon, Google, and Microsoft each offer AI services and develop and deploy competing AI models through their respective cloud platforms," Anthropic noted. Former Microsoft CFO Christopher Liddell joined Anthropic's board in February.

The S-1 lists more than 90 pages of risk factors. Anthropic flagged U.S. data center moratoriums, which "could delay or prevent the construction or expansion of facilities on which we depend for compute," and said it is building data centers overseas to reduce concentration risk. It called Taiwan's chip manufacturing concentration at TSMC a vulnerability. It warned that safety "judgement calls" carry existential stakes: as models grow more capable, "the potential severity of harm from misaligned or misused models increases correspondingly," and "there can be no assurance" its safeguards will prove adequate.

The competitive set is narrow and stark. OpenAI and Google are the only named competitors, alongside open-weight models and well-capitalized tech companies racing toward the frontier.

What do investors make of it?

The offering lands in choppy water. SpaceX raised $85.7 billion in June in the biggest IPO on record, and its shares finished last week slightly below their June opening price. AI infrastructure stocks are wilting: CoreWeave and Nebius each lost over 30% of their value in October as credit risk rises across the data center space.

Byron Deeter of Bessemer Venture Partners said his firm's Anthropic stake is its largest investment by cost basis in its 100-plus year history. "We were fortunate to partner early and aggressively," Deeter said on CNBC's "Squawk on the Street."

MIT Sloan professor Sinan Aral framed the valuation math on "Squawk Box": the $1.5 to $2 trillion range implies a 22-to-30-times revenue multiple. "The question is, where is the enduring value?" Aral said. "I think the enduring value is going to come from companies that can demonstrate enterprise profitability, productivity, and actual value add, rather than just providing intelligence."

Anthropic closed its last private round in May, raising $65 billion at a $965 billion valuation led by Altimeter Capital, Dragoneer, Greenoaks and Sequoia — nearly triple its $380 billion February valuation, and above OpenAI's $852 billion mark for the first time. Goldman Sachs holds lead left on the IPO, with Morgan Stanley beside it and JPMorgan, Citigroup, Barclays and roughly 20 other banks behind.

The proceeds are earmarked for "general corporate purposes," including capex, working capital, hiring — and possibly acquisitions, though Anthropic said it has "no agreements or commitments for any material acquisitions or investments at this time." With 5,300 employees generating $6.6 million in revenue per employee, and more than 90% of codebase contributions in September written by Claude itself, Anthropic is betting its own model can keep the machine lean while Wall Street decides what it's worth.