Cerebras can trade like a new AI-chip public company and still read like an old customer-concentration case.
That was the paper's position when the IPO close reopened customer questions, and it was the older frame when Cerebras's IPO made revenue the hard question. The frame remains useful Tuesday. The accessible market page shows CBRS with a $46.84 billion market cap and $509.99 million in trailing-12-month revenue after a June 1 close at $213.28, down 10 percent. [2] The accessible S-1 breakdown says the more important denominator is who supplies the revenue and backlog. [1]
Mostly Metrics' analysis of the filing says Cerebras had $510 million in fiscal 2025 revenue and $24.6 billion in remaining performance obligations. It also says MBZUAI accounted for 62 percent of 2025 revenue, G42 accounted for 85 percent of 2024 revenue, and OpenAI accounted for much of the backlog. [1]
Those are not footnotes. They are the company. The optimistic version is that Cerebras has landed the exact strategic customers a non-Nvidia accelerator company needs: sovereign AI buyers, Gulf capital, and OpenAI-scale demand. The skeptical version is that public-market investors are buying a revenue base whose durability depends on a small set of counterparties and one very large AI cycle.
The divergence matters because the IPO story rewards mood. X turns Cerebras into either proof that Nvidia has a real challenger or proof that AI listings have outrun sense. Market coverage can flatten the company into a ticker and a price move. The filing frame is narrower and better: how much revenue, how much backlog, how many customers, and how much of the next year depends on one buyer's budget.
There is a source caveat. SEC pages were blocked in the research stack, so this article uses an accessible filing analysis and an accessible market-data page rather than quoting the S-1 text directly. That weakens any claim about legal wording, but it does not weaken the main editorial point: the public-market argument is not complete until concentration sits beside valuation.
Cerebras may yet prove that concentrated early demand is the normal shape of a new accelerator market. But that is an argument, not a receipt. The receipt is the customer list.
-- THEO KAPLAN, San Francisco