$380B: The Sticker Price on AI's Most Uncomfortable IPO

Anthropic is weighing an IPO. The harder question is what comes after the opening bell.

The AI safety company behind Claude confirmed this week what markets have anticipated since its $30 billion Series G in February: it is considering going public as soon as October, with Goldman Sachs, JPMorgan, and Morgan Stanley in preliminary discussions for lead roles on what could be a $60 billion raise, according to Bloomberg. That valuation sits at $380 billion post-money, one of the largest ever sought by a private company.

But the roadshow brief will have to explain something the pitch deck cannot change: Anthropic is simultaneously fighting the United States government in federal court. In February, the Pentagon designated Anthropic a supply chain risk after the company refused to remove contractual restrictions on using Claude for domestic surveillance and fully autonomous weapons. A federal judge in San Francisco blocked that designation last week. The legal fight continues, and the timing of an IPO beside it is, at minimum, notable.

The governance structure that makes Anthropic distinctive also makes the IPO complicated. Anthropic is a Delaware Public Benefit Corporation, with a Long-Term Benefit Trust holding a special class of shares that gives it escalating board-election rights — eventually a majority of the board, as the trust's founding document describes. The trust was designed with five voting members; Anthropic announced in January 2026 that two had concluded their terms, leaving three current trustees: Chair Neil Buddy Shah, Richard Fontaine, and Mariano-Florentino Cuéllar. Those three are required to ensure the company balances profit with its AI safety mission. That structure is nearly untested at public market scale. The SEC will want to know how it works when a major institutional investor with a 14 percent stake, Google, wants something the trust does not.

The financial picture is where the IPO brief gets genuinely complicated. Anthropic reported $5 billion in GAAP revenue from 2023 through December 2025, according to a court filing by CFO Krishna Rao. Its annualized run-rate revenue sat at $14 billion as of February 12, and climbed to $19 billion by month-end, Anthropic said in its Series G announcement. The gap between those two numbers is not a sign of trouble. It reflects how Anthropic counts revenue: it annualizes the last 28 days of consumption-based sales, multiplies monthly subscriptions by 12, and adds them together — a formula that inflates the headline figure relative to cash collected. Reuters Breakingviews first reported the distinction.

The accounting cuts both ways. Unlike OpenAI, which reports hyperscaler distribution revenue net of payments to partners, Anthropic books the full gross amount — including the chunk it immediately pays back to Amazon Web Services and Google Cloud. Bank of America estimated in March that Anthropic could owe cloud providers $6.4 billion in revenue-sharing payments in 2026, up from $1.9 billion in 2025. At a $380 billion valuation, that cost is not trivial. The SEC will almost certainly ask Anthropic to reconcile its revenue presentation against OpenAI's when both file, which will make the comparison harder to avoid.

The Pentagon dispute has already shown up in enterprise contracts. One customer paused a $15 million contract discussion after the supply chain risk label, Chief Commercial Officer Paul Smith told Reuters. Two financial-services companies refused to finalize agreements worth a combined $80 million unless they secured broad cancellation rights. Those are not existential losses for a company with 500 customers spending more than $1 million annually and eight of the Fortune 10 as Claude users, per the Series G announcement. But they are real data points for what a government blacklist actually costs when it bleeds into enterprise sales cycles.

Anthropic's investor base is a cross-section of the AI industry's fault lines. Amazon is the largest external investor, with $8 billion committed across multiple rounds. Microsoft has invested $5 billion separately and Anthropic has committed $30 billion in Azure spending. Google holds roughly 14 percent. Nvidia and Microsoft co-invested up to $15 billion in November 2025 as part of a partnership in which Anthropic committed to buying $30 billion in Azure computing capacity running on Nvidia AI systems. The Series G itself was led by GIC and Coatue Management, with D.E. Shaw, Dragoneer, Founders Fund, and ICONIQ as co-leads. Managing that list of interests in the boardroom is a different skill than managing them in the market.

The OpenAI comparison will be unavoidable. OpenAI is targeting a $750 billion to $830 billion valuation for its own listing, per Business Insider, after closing a $110 billion funding round in February. Both companies are burning cash at a scale that makes traditional SaaS metrics inadequate. Both are navigating the same hyperscaler dependency. Both are trying to convince public markets that a mission-driven structure — OpenAI's capped-profit entity, Anthropic's PBC — can survive the obligations of a quarterly earnings calendar. The difference is that Anthropic is doing it while actively suing the Department of Defense.

What happens next is fairly predictable: the S-1 will arrive, the LTBT governance section will get lawyered into something passable, and the roadshow will emphasize the $19 billion run-rate and the Fortune 500 adoption story. The harder question — whether a public Anthropic can actually maintain the constitutional constraints its founders wrote into the charter when the next DoD contract, or the next political moment, tests them — is the one that will sit underneath the stock price for years.