// blog · analysis · frontier-models2026-05-215 min read

The $380B watermark — what Anthropic's valuation actually prices

Anthropic's $380B post-money round is the price of a moat that has to keep widening. Public-market investors are starting to do the arithmetic — and the answer doesn't favor closed-lab IPO multiples.

The arithmetic is uncomfortable

Anthropic closed Series G at $380B post-money on roughly $14B annualized revenue — a ~27× multiple. That's premium relative to a SaaS comp and modest relative to OpenAI's last round. Both numbers depend on the closed-lab moat holding against an open-weight tier that costs 90% less per token.

CNBC's read of the IPO prep work is that public-market investors are no longer willing to take the moat thesis on faith. They're pricing it.

The 60% number

Chinese open-weight models — DeepSeek, Qwen, Kimi, GLM — went from roughly 1% of OpenRouter usage in mid-2024 to more than 60% in May 2026. That's a 60× share shift in 18 months, driven by a 5–20× price gap and a near-elimination of the capability gap on most evaluation suites.

If you priced Anthropic at $380B assuming the closed-lab tier holds production share, the OpenRouter data is the early-warning signal that the assumption is wrong.

What the moat actually is

The honest case for the multiple isn't "closed labs hold capability lead." It's three more specific claims:

The first two are decay-curves measured in years, not decades. The third is a real moat — but it depends on the capability ceiling still being where the buyers actually want to spend.

The forward read

If the OpenRouter share shift is the leading indicator, the closed-lab IPO window narrows from "Q3/Q4 2026" to "before the next earnings cycle confirms the production-share data." Anthropic and OpenAI both have to file before the IPO market prices in what the OpenRouter numbers already show. That's a tighter window than the public messaging suggests.

The structural read isn't "closed labs lose." It's "closed labs need to be valued like premium-tier SaaS, not like category-defining monopolies." $380B was the latter. The next public-market data point reveals whether the markets agree.

CNBC — cheap AI could derail IPOs → · Air Street — State of AI May 2026 → · Crunchbase — Q1 2026 venture record →