Cerebras Surges 68% on Debut, Then Reality Bites: Inside the Biggest IPO of 2026

When Cerebras Systems priced its initial public offering at $185 per share on May 13 — above every forecasted range — Wall Street braced for a spectacle. It got one. The AI chipmaker’s debut on the Nasdaq the following day was the kind of first-day pop that makes institutional investors feel clever and retail traders feel vindicated: shares opened at $350, surged to an intraday high near $385, and closed at $311.07, a gain of 68.2% from the IPO price. The $5.55 billion raised made it the largest public offering of the year so far. Then, on day two, the music slowed. Shares dropped roughly 10%, settling around $279.72, as the market began asking the questions that never get answered during a roadshow frenzy.

A Different Kind of Chip Company

Cerebras is not trying to be Nvidia. Its bet — one it has been making since its founding in 2016 — is that the economics of AI inference will eventually swing toward specialized, wafer-scale silicon rather than clusters of general-purpose GPUs. The company’s Wafer-Scale Engine is a physically massive chip, roughly the size of a dinner plate, that integrates an entire silicon wafer into a single processor. The architecture eliminates the inter-chip communication bottlenecks that slow down traditional GPU clusters for certain AI workloads, enabling inference speeds that Cerebras claims are dramatically faster for the right use cases.

That speed story landed with at least one very important buyer. OpenAI has committed more than $20 billion over three years — from 2026 through 2028 — to rent 750 megawatts of compute capacity directly from Cerebras. The partnership, which became public in April, is the anchor of Cerebras’s entire investment thesis. It also explains why the company’s $24.6 billion backlog dwarfs its $510 million in 2025 revenue: the bulk of that commitment hasn’t been invoiced yet.

From $23 Billion to $86 Billion in Three Months

The trajectory of Cerebras’s valuation is dizzying even by 2026 standards. In February, the company’s most recent private funding round valued it at $23 billion. The IPO pricing at $185 per share implied a market capitalization of roughly $48.8 billion. After the first-day surge to $311, the outstanding-share market cap approached $70 billion. On a fully diluted basis — accounting for restricted stock units, options, and warrants — the number climbed closer to $86 billion. That is a nearly fourfold jump in valuation in under three months, driven almost entirely by investor excitement rather than new financial disclosures.

The demand metrics were extraordinary. During the roadshow, orders exceeded available shares by more than 20 times, forcing underwriters to repeatedly lift the price range. Cerebras had originally filed in May targeting a range of $115 to $125 per share, bumped it to $150–$160 after the oversubscription became clear, and ultimately priced at $185 when demand still wasn’t satisfied.

The OpenAI Concentration Problem

Cerebras bulls have a compelling narrative: AI inference demand is growing faster than GPU supply, Cerebras hardware is demonstrably faster on certain workloads, and the company has already landed the most important customer in the industry. Bears have a different read of the same facts.

The OpenAI relationship is both Cerebras’s greatest asset and its largest single risk factor. A company that derives the overwhelming majority of its forward backlog from one customer — even a customer as well-capitalized and strategically committed as OpenAI — is effectively a contractor dependent on a single contract renewal. If OpenAI develops its own in-house silicon, shifts inference workloads to Nvidia’s next-generation hardware, or restructures the compute agreement as part of its ongoing legal and financial reorganizations, Cerebras’s revenue profile changes dramatically.

OpenAI has been actively diversifying its hardware relationships, running parallel deals with Nvidia, Microsoft Azure, and its own internal chip efforts under Project Sage. The Cerebras relationship appears to be specific to the high-throughput code-generation workload — a critical use case, but not the entirety of OpenAI’s compute needs.

The Market’s Two-Day Verdict

The first-day surge and second-day retreat is a pattern investors have seen before with highly anticipated AI hardware listings. The initial pop reflects genuine scarcity — investors who couldn’t get IPO allocations buying in the secondary market — and the day-two correction reflects some of those same investors taking profits once the lock-up-free window closes.

Analysts who follow semiconductor IPOs note that Cerebras’s post-trading valuation still prices in near-flawless execution of its backlog. At ~$280 per share as of May 16, the stock trades at roughly 55 times trailing revenue. By comparison, Nvidia — the undisputed leader in AI infrastructure — trades at around 35 times trailing revenue despite its dominant market position. For Cerebras to justify its current multiple, it would need to demonstrate that the OpenAI relationship is durable, that it can win additional customers at scale, and that its Wafer-Scale Engine architecture remains competitive as Nvidia ships its next-generation Vera Rubin and Blackwell Ultra products.

None of those questions will be answered this week. The company’s first earnings call as a public company — expected in August — is when investors will learn whether the backlog is converting to recognized revenue on schedule.

What Comes Next

The Cerebras IPO is the most visible signal yet that the AI infrastructure market has entered a phase where the public markets are willing to price in scenarios that would have seemed implausible two years ago. The company is not yet profitable, has a customer concentration problem, and faces competition from one of the best-resourced semiconductor companies in history. It also has the fastest-growing AI inference backlog of any publicly traded hardware company.

For the broader tech IPO market, the debut matters as a sentiment indicator. A clean Cerebras listing — however volatile — keeps the window open for other AI infrastructure companies considering public markets in the second half of 2026. A sustained post-IPO slide, by contrast, would cool enthusiasm for the roughly dozen AI chip and infrastructure companies watching from the sidelines.

The stock’s trajectory over the next 90 days will say more about investor appetite for AI hardware risk than any single earnings report. Day one said: the appetite is enormous. Day two said: but not infinite.