Cerebras Systems made its public market debut on Thursday, raising $5.5 billion in what became the largest IPO of the year. Its shares shot as high as $385 immediately after opening before trading was briefly halted for volatility. It closed the day up 68% at $311, giving the company a market capitalization of nearly $70 billion. Less than a year ago, it was valued at $8.1 billion in a private financing round.
The offer was priced at $185 per share on Wednesday evening, raised multiple times during bookbuilding from an initial target of $4.8 billion, as demand consistently outran supply.
What Cerebras Builds
Cerebras makes chips the size of a dinner plate, literally. Its wafer-scale engine uses entire sheets of silicon, producing a chip 58 times larger than the Nvidia GPUs that dominate the AI market. The design eliminates the need to link multiple smaller chips together, removing a key bottleneck that slows inference, the process of running a trained AI model to generate responses for users.
That focus on inference, rather than training, is precisely what made Cerebras a laggard in the early AI cycle and what makes it compelling now. The industry has shifted. When the priority was training large models, Nvidia's architecture won. As the priority has moved to serving those models to millions of users at speed, Cerebras' hardware becomes strategically relevant in a way it simply was not before.
The Deals That Made the IPO Possible
Cerebras arrived at its listing with two anchor relationships that provided the institutional credibility the offering needed. First, a $20 billion deal with OpenAI will see the AI lab deploy 750 megawatts of Cerebras chips coupled with an agreement to co-design future hardware. Second, a partnership with Amazon Web Services opens access to thousands of corporate clients, a critical diversification from a heavily concentrated revenue base. In 2025, two UAE-based customers, G42 and MBZUAI, together accounted for 86% of revenues. The AWS relationship is the clearest path to resolving that concentration risk.
Additional partnerships with Meta, Mistral, Cognition, and Windsurf round out a customer base that is now materially broader than the financials currently reflect.
The Numbers Behind the Hype
Cerebras reported a $146 million operating loss on $510 million in revenue in 2025, with revenue up 76% year-on-year. It is not yet profitable. At $70 billion market cap against $510 million in revenue, the valuation multiple is extreme by any conventional measure, but it is being applied to a company whose revenue trajectory and strategic positioning are both moving fast in the same direction.
Nvidia still controls roughly 80% of the AI chip market. Cerebras is not displacing that. What it is doing is carving out a defensible position in inference infrastructure at exactly the moment hyperscalers are competing aggressively on the speed and cost of serving AI at scale. The OpenAI and AWS deals suggest that position is real, not theoretical.
Why It Matters
The market's response to Cerebras doubling on debut, with Nvidia and Broadcom both hitting all-time highs on the same day, is the clearest signal yet of where investor appetite sits ahead of the larger listings coming this year. OpenAI and Anthropic are both targeting public markets as soon as year-end. SpaceX's roadshow is already being planned for June.
Cerebras was a hard business to build: a decade in the making, several near-death moments, and years of being ahead of a market that did not yet care. The IPO does not resolve its execution risks. Revenue concentration, operating losses, and Nvidia's entrenched dominance are all real. But the debut tells you something important about what the public market is willing to pay for a credible position in AI infrastructure with strong hyperscaler relationships. That read matters for everything that follows.


