Anthropic is meeting investors ahead of a listing it wants to hold in September or early October, and the questions it is fielding are now on record. The Decoder reported Tuesday, citing original reporting by The Wall Street Journal, that the company is carrying a $965 billion valuation into those conversations. Exact pricing and the final timeline have not been disclosed.
The figure sitting under that valuation is annualised revenue of more than $47 billion, which The Decoder said Anthropic reported in May 2026. Divide one by the other and the multiple lands near 20 times annualised revenue. That is our arithmetic, not a figure either outlet published, and it is the ratio a public market will interrogate line by line.
The valuation is not the most useful disclosure here. The objection list is. A company running preliminary meetings before an IPO is publishing its own risk section early, because the doubts raised in a private room in August are the doubts that come back in the roadshow in September. Three came up.
Start with price pressure out of China. Investors named Kimi K3 and Qwen3.8, per The Decoder’s account, which turns a vague anxiety about low-cost Chinese labs into two specific releases buyers are already pricing against Claude. Leadership brushed the threat aside, saying Anthropic builds nothing but frontier-grade systems and that demand concentrates on whatever is smartest. Treat that as the company’s stated position rather than a settled matter. It rests on an assumption about market structure: that the quality gap holds and that customers refuse to trade down. What happens if they do trade down was not addressed.
Washington came up second. Investors raised friction between Anthropic and the Trump administration, and The Decoder’s account of those meetings carries no company answer on that theme. An objection left hanging in a private room is still a disclosure, because it returns with a prospectus attached.
Land and power made the third theme. Americans are fighting new data centre builds, and Anthropic’s plan for the sour public mood around AI is to foreground health and biology work. That aims at national sentiment. A rezoning hearing, a water allocation, and a substation fight are local, and they are rarely settled by what a model can do in a lab.
This is where the timing gets awkward. AI Insiders reported yesterday that Anthropic signed a $9.1 billion data centre deal with Bitcoin miner Riot Platforms, a 20-year lease covering 191 megawatts at a Texas site. Investors are flagging construction resistance as a risk in the same week the company commits to a decade-scale build. Both things are true at once, and that is the tension worth naming: Anthropic’s capital plan requires exactly the thing its own investors say the public may not let it have.
The operating picture underneath the doubts is stronger than the doubts suggest. The Decoder credited Claude Code with pulling Anthropic ahead of rivals, and reported fresh compute contracts signed with SpaceX and with Google. It also noted repeated service outages this year. That detail reads as a footnote in a private meeting; it reads as a reliability disclosure in a filing.
For anyone negotiating Anthropic capacity or enterprise terms, the window before a listing is the leverage window. A company chasing clean revenue growth ahead of a September pricing has reasons to be flexible that a company answering to a share price in November does not. When the risk factors publish, check whether these three items appear at all, and whether the order matches what the private meetings surfaced first.
This account comes from The Decoder’s August 11, 2026 coverage of a story originally reported by The Wall Street Journal.
