Cerebras Priced Its IPO at $185. The Market Said $385. Nothing Means Anything Anymore.

The Cerebras pricing committee sat down Wednesday night, looked at the AI chip market, looked at investor demand, looked at their $510 million in revenue and their new-found profitability, and announced a share price of $185. Careful. Measured. Even a touch conservative — they'd already raised the range twice in the days before.

The market opened Thursday morning, squinted at those same numbers, spotted the word "AI" on the prospectus, and replied: $385.

That's a 108% pop on the first trade. Not over a quarter. Not over a year. In the time it took most people to get through their morning coffee. Welcome to the first big tech IPO of 2026. Please keep all assumptions about rational markets tucked safely inside your carry-on.

The IPO That Wasn't (Until It Very Much Was)

To understand how surreal Thursday was, you need to remember where Cerebras was a year ago: absolutely nowhere near public markets.

The company had first filed to go public in 2024. Cerebras, for the uninitiated, makes chips — giant, purpose-built AI chips designed to challenge Nvidia at the inference layer. Inference is the part of the AI pipeline where models actually do things — answer your questions, write your code, tell you the restaurant doesn't take reservations. Training gets all the glamour. Inference is where the compute bills live.

Cerebras built a chip from scratch for this. Physically massive, architecturally bold, and — critically — backed heavily by Group 42, an Abu Dhabi-based AI firm that accounted for almost all of Cerebras's revenue. This caught the attention of CFIUS — the Committee on Foreign Investment in the United States — which proceeded to review the IPO at a pace that could charitably be described as "glacial." The IPO was quietly shelved.

It was, objectively, a rough place to be. Your entire revenue base is one foreign-linked customer, the government is not sure how it feels about that, and every week someone else is announcing a new chip to challenge Nvidia while you sit in regulatory limbo.

The Comeback Arc Nobody Wrote a Medium Post About (Yet)

Then something genuinely interesting happened: Cerebras fixed it.

Not fixed it in a Silicon Valley "we pivoted our narrative" way. Fixed it in an actual "the numbers got dramatically better and we diversified the customer base" way — which, as a strategy, remains deeply underrated in this industry.

By the time Cerebras refiled in earnest this spring, it had revenue of $510 million in 2025 — up 76% year-over-year — from a handful of real customers rather than one concentrated foreign bet. More startlingly, it had swung from losing nearly half a billion dollars the year before to posting $237.8 million in net income. That's not a modest improvement. That's a different company.

New customers included OpenAI, Amazon Web Services, and Saudi Arabia's Mohamed bin Zayed University of Artificial Intelligence — which, as customer lists go, reads like a who's-who of "entities with large ongoing AI compute budgets." This is how you get CFIUS comfortable. You stop being "the chip company with the Abu Dhabi problem" and start being "the chip company that counts OpenAI and AWS as customers."

About That OpenAI Relationship

TechCrunch described the relationship with OpenAI as a "complicated circular-deal relationship," which is the tech press's polite way of saying: buckle up, there's a lot going on here.

OpenAI uses Cerebras chips for inference. OpenAI also uses Azure, which runs on Nvidia. OpenAI also has its own chip ambitions. OpenAI is simultaneously a customer of, competitor to, and strategic partner with approximately everyone in the AI supply chain at this point. I've written before about how hard it is to pin down Anthropic's valuation — pinning down OpenAI's actual vendor relationships is a similar exercise in creative ambiguity.

Does the OpenAI-Cerebras relationship matter for the long term? Yes, almost certainly. Is it clear exactly how it matters? Absolutely not. But "OpenAI is a customer" is the kind of sentence that makes retail investors double-click the buy button before they finish reading it, so here we are.

What $385 Actually Means

Let me be direct about what happened Thursday morning: retail investors smashed the buy button as fast as humanly possible, and the stock went parabolic before institutional investors had finished their second cup of coffee.

At the $185 IPO price, Cerebras entered trading with a fully-diluted valuation of $56.4 billion. At $385, where it opened, you're looking at something north of $117 billion — briefly, before it cooled to trade above $330 by midday. CEO Andrew Feldman's stake at the IPO price was worth nearly $1.9 billion. At $385, you do the math.

The question everyone immediately starts asking about a 108% first-day pop is: is this justified? And the answer, as with most of 2026's AI funding spectacles, is: it depends entirely on whether you believe the AI inference market becomes as large and entrenched as the bulls say.

If Cerebras captures a meaningful slice of global AI inference compute — and that market continues to explode — then $56B might look cheap in three years. If Nvidia iterates its inference stack faster than Cerebras can scale, the story gets much shorter. This is not a company where you shrug and say "seems fine either way." It's a large bet on a specific chip architecture winning in a market that's only a few years old.

For what it's worth, I've watched a lot of AI companies try to turn hype into durable revenue, and Cerebras at least has actual revenue, a path to profit it has already walked once, and customers who are buying for infrastructure reasons rather than experimentation budgets. That's more than most.

The Part Where I Admit This Is Actually Impressive

I spend most of my professional existence chronicling the absurdities of the AI funding machine — the $150M rounds for companies that have been alive 18 months, the valuations that seem to have been calculated by someone who lost a bet — so I want to be honest when something earns its moment.

Cerebras earned this one. A year ago, it was a company stuck in national security review purgatory, entirely dependent on a single foreign-linked customer, watching its IPO plans evaporate. It rebuilt. It diversified its revenue. It got profitable. It got CFIUS clearance. And then it went public and the market rewarded it spectacularly.

Is the stock pricing in a lot of future optimism? Yes. But the company actually did the hard things: it shipped real chips to real customers, turned a massive loss into a massive profit, and survived the kind of regulatory scrutiny that would have quietly ended most startups.

The 108% pop is the market being irrational. The existence of Cerebras as a public company at all — after everything that happened — that part is genuinely earned.

Now let's see if the stock is still at $330 in six months. I'll be here either way, with snacks.