Anthropic opens its books for Wall Street, showing record revenue and mounting obligations

Anthropic filed its S-1 with the SEC, becoming the first major AI lab to publicly file an IPO prospectus. The company plans to list on the Nasdaq under the ticker "ANT," with Goldman Sachs and Morgan Stanley leading the offering. Wall Street expects a valuation of roughly $2 trillion, according to Yahoo Finance, well above Anthropic's most recent private valuation of about $965 billion, which already exceeded rival OpenAI's roughly $852 billion.
The revenue picture is dramatic. Anthropic reported $11.6 billion in revenue for Q2 2026, up from $376 million in Q1 2025, a 1,368 percent year-over-year increase. First-half 2026 revenue totaled $16.3 billion, more than triple the company's full-year 2025 revenue of $4.6 billion. In Q2 2026 Anthropic posted its first positive GAAP operating profit, at $974 million. Compute costs rose 300 percent over the same period, but that growth trailed revenue growth.
On a per-person basis, revenue per employee is $6.6 million with about 5,300 staff, higher than any Magnificent 7 company except Nvidia. Chip stocks including Nvidia, Broadcom and AMD rallied after the filing became public, which suggests investors read the disclosure as confirmation of sustained demand for AI compute rather than as a warning sign.
A few revenue characteristics stand out. Eighty percent of revenue comes from per-token billing based on actual usage, with the remaining 20 percent from fixed subscriptions. Government contracts account for only about one percent of revenue. Roughly 5 to 7 percent comes from customers who violate usage policies, primarily through model distillation, a practice Anthropic has already been cracking down on. Net revenue retention sits at roughly 900 percent, meaning existing customers spend nine times what they did a year ago.
Customer concentration is moderate but worth watching. More than 11,900 customers pay over $100,000 annually, and about 250 spend more than $10 million. No single customer accounts for more than ten percent of revenue, but five customers together make up 25 percent. Claude Enterprise users who also use Claude Code are the largest spenders, paying about 20 times more than other customers.
Weighing against the revenue is a very large balance-sheet obligation. Compute capacity contracts with Broadcom, Google, Amazon, Microsoft and other providers add up to roughly $780 to $810 billion, and those costs are more or less fixed and locked in long-term while revenue remains variable and usage-dependent. That mismatch is the central tension in the filing: the company must keep growing quickly to cover future bills. To gain flexibility and lower costs over time, Anthropic is building its own data centers under the name "Anthropic Direct."
In practice, the 80/20 usage-versus-subscription split is a useful lens for product readers. Usage-based billing scales with customer value but makes revenue sensitive to token prices, model efficiency gains and competitive pressure, since cheaper models cut bills as well as costs. High net revenue retention far above the levels typical of enterprise software suggests early deployments are expanding rather than being replaced. At the same time, a large share of that retention may reflect customers moving more workloads onto Claude rather than price increases, which is a healthier signal, though it also means the base is still small enough for growth rates to swing sharply.
Anthropic did not include Q3 figures in the S-1 even though the company already has them. A plausible reading, and this is inference rather than stated fact, is that growth slowed noticeably in the quarter. OpenRouter data shows demand for Anthropic models climbing again after the October release of the 5.5 series, especially Opus 5.5, though OpenRouter captures only a small slice of total token demand and skews toward open-weight models.
The ~$2 trillion expected valuation against roughly $16.3 billion in first-half revenue implies investors are pricing in years of continued rapid expansion, and the fixed compute commitments are the mechanism that both enables and constrains that expansion. Comparable infrastructure-heavy businesses have historically traded at far lower multiples of revenue, so the offering will test how much of AI's growth story public markets are willing to capitalize up front.
Why it matters: Developers and product teams now get an unusually detailed look at the economics behind the Claude API, including the usage-based billing model, heavy spending by enterprise and coding customers, and the scale of committed compute. If the listing proceeds near expectations, Anthropic's capital position and pricing power could shift, which in my inference may affect model pricing, capacity and roadmap cadence for teams building on Claude. The S-1 also sets a disclosure benchmark other labs will likely be measured against.
Based on reporting from the original publisher. Visit the source for full context and later updates.
Publisher excerpt
Anthropic is the first major AI lab to file an S-1 with the SEC, planning to list on the Nasdaq under the ticker "ANT." The numbers are strong, with $11.6 billion in quarterly revenue, a first-ever operating profit, and existing customers spending nine times more year over year. But the company also carries roughly $800 billion in long-term compute commitments. The article Anthropic opens its books for Wall Street, showing record revenue and mounting obligations appeared first on The Decoder .