Anthropic at $900 Billion: Think Twice Before Buying the IPO

Before the hard questions — the company is real, the growth is real, and the mission matters more now than when Anthropic was founded. But at $900 billion, you deserve the full picture before October.

First, What Is Genuinely Extraordinary

Salesforce took 20 years to reach $30 billion in annual revenue. Anthropic did it in under three years from zero. Dario Amodei described the growth himself as "crazy" — they planned for 10x annual growth and saw 80x.

Claude Code — the company's agentic coding product — became the fastest-growing enterprise software product in history. It went from launch in May 2025 to $1 billion in annualised revenue by November 2025. Eight of the Fortune 10 are customers. Over 1,000 companies are spending more than $1 million per year.

And the efficiency number that does not get enough attention: Anthropic has approximately 5,000 employees generating $44 billion in revenue. That is $8.8 million per employee. Nothing in enterprise software history comes close.

Question 1: Why Are They Selling Shares Now?

When a company raises money or sells shares, the first question is always: why now, and who benefits?

There is also a timing calculation that has nothing to do with operational readiness. A recession, a major AI safety incident, a Chinese lab releasing a model that matches Claude at a fraction of the compute cost — any of these could close the window. October 2026 was chosen because the window is open, not because the business is optimally positioned for public scrutiny.

Question 2: Is the $900 Billion Real?

The most sophisticated critique of this valuation has a sharp edge: "This is being valued like a future monopoly before the market structure even exists."

The bull case: if Anthropic hits $100 billion in annualised revenue by end of 2026 — which its own trajectory suggests is plausible — the $900 billion valuation implies roughly 9x forward revenue. At that point it looks reasonable. Anthropic also generates $2.10 of revenue per compute dollar spent, versus OpenAI's $1.60.

The deeper issue: we do not actually know what kind of company Anthropic is. If it is a software company, 20x revenue is aggressive but defensible. If it is an infrastructure company — burning capital on compute, locked into $15 billion annual hardware commitments — then the multiple that applies is closer to a data centre operator than a software firm. Data centre operators do not trade at 20x. They trade at 4x.

Question 3: Will It Actually Be Profitable?

Anthropic is projecting its first profitable quarter in Q2 2026. That is a real milestone. But the most important detail about that profitability has received almost no coverage this week.

The gross margin story is also more complicated than the headline suggests. In 2024, Anthropic's gross margin was negative 94% — spending nearly twice what it earned serving customers. In 2025 it improved to approximately 40% — but that was already 10 percentage points below the internal target, because inference costs came in 23% higher than anticipated. The 2028 target of 77% assumes one of the most aggressive margin expansion trajectories ever embedded in a private technology valuation.

Question 4: What Happens to Safety When Quarterly Earnings Start?

This is the question that started the whole company. And the one that gets the least airtime in the funding coverage this week.

Anthropic was founded on the belief that AI safety and commercial success could coexist. The IPO puts that thesis under the most intense test it has ever faced. Public shareholders demand growth every 90 days, on earnings calls, in analyst notes, in the instant feedback of a stock price. Safety frameworks slow deployment. Slower deployment can mean losing market share.

Some academics argue directly: "You cannot simultaneously optimise for public-market hypergrowth and maximum caution."

Anthropic's governance structure — including its Long-Term Benefit Trust and public benefit corporation status — was designed to protect the safety mission from exactly this pressure. But the practical reality of being a public company has a way of eroding commitments that seemed ironclad in private. None of these structures have been tested against sustained quarterly earnings pressure. We simply do not know yet.

The Five Structural Risks (Stated Plainly)