Cerebras Systems priced its initial public offering at $185 a share on Wednesday night, $25 above the top of a range it had already raised twice, and shares were expected to begin trading on the Nasdaq Global Select Market at 9:30 a.m. Eastern under the ticker CBRS. [1] The company sold 30 million shares for $5.55 billion in proceeds, valuing it at $56.4 billion on a fully diluted basis — the largest US stock market debut so far this year. [2] The first trade typically prints 30 to 90 minutes after the bell while underwriters balance the book.

The paper's Wednesday account put the priced number at "above $160" — the marketed range. The real number is higher, and the higher number changes the math on a single line in the S-1: 86 percent of 2025 revenue was concentrated in two United Arab Emirates entities, 62 percent in MBZUAI and 24 percent in G42. The IPO did not dilute that concentration. It put it on a public ticker.

A second number now matters more than the price. G42's Series F-2 preferred shares — issued in February 2026 — carry a conversion feature that lets G42 buy roughly 23 million additional shares at $14.66 each. At $14.66, those rights were a private-market footnote. At $185, every $1 of share-price appreciation widens the gap between G42's strike and the public-market clearing price by another $4.3 million in unrealized value across the right. The better CBRS trades on Thursday, the more dilutive the F-2 becomes.

CEO Andrew Feldman, who is not selling shares in the offering, holds a stake worth roughly $1.9 billion at the IPO price. [2] OpenAI, the largest customer Cerebras disclosed in its prospectus, holds warrants on 33.4 million shares as part of a multi-year deal valued in the $10 billion to $20 billion range covering 750 megawatts of inference capacity through 2028. [3] The customer is also an in-kind shareholder, the supplier is also a strategic counterparty, and the trial that will decide whether OpenAI remains structurally what it is today is in closing arguments today in Oakland.

Morgan Stanley, Citigroup, Barclays, UBS, Mizuho and TD Cowen ran the book. [1] Needham, Wedbush, TD Cowen and Rosenblatt sit in the co-manager block and are likely first to publish initiation coverage; the lead book-runners are under a 25-day quiet period before they can publish notes. The first analytical question the sell side has to answer is not whether the demand was real — orders ran 20 times oversubscribed — but whether the named UAE share is a risk factor or a pricing risk. Those are different words for different conclusions.

The S-1 had three structural features that the paper has flagged before: the customer concentration line, the OpenAI warrant overhang, and the CFIUS-cleared G42 stake restructured into non-voting shares. [4] None of those have changed. What changed Wednesday night was the denominator. At $26.6 billion of valuation — the company's original midpoint a week ago — the F-2 strike was theoretical. At $56.4 billion, it is a balance-sheet line item for any analyst building a fully diluted model.

The framing on X around the print has split predictably. The flow-trader accounts on platforms that priced the company privately at $102 to $107 a share two months ago are watching what happens to the secondary discount. The analytical Substack stack — Futurum, AdValorem, Semicon Alpha — is writing the customer-concentration risk note that the lead book-runners cannot. [5] The first Wall Street initiation that names MBZUAI by percentage of revenue, rather than burying it in a risk-factor recitation, will set the analytical floor for the stock. The cynical read is that it will not happen for 25 days. The structural read is that the longer it takes, the more the F-2 strike matters.

There is a connecting line to today's other AI-state-power stories. Project Glasswing's 90-day public-report clock is on day 37 with no artifact. The Trump delegation to Beijing includes Jensen Huang as a last-minute addition. Closing arguments in Musk v. Altman are this morning, the same hours CBRS opens. Each of those four stories names a different mechanism. Cerebras names the one that has a ticker symbol attached.

The cleanest test is liquidity. If CBRS opens above $200, the F-2 right widens further into the money and the dilution question becomes part of every secondary-offering calculation Cerebras runs for the next two years. If it opens below par, the lead book-runners have a price-stabilization mandate to defend, and the demand story becomes a different sentence by Friday. Either way, the customer concentration moves from a confidential filing to a market-disclosed input. The S-1's most uncomfortable paragraph just acquired daily volume.

-- THEO KAPLAN, San Francisco