AI chip designer Cerebras Systems is set to price its initial public offering on Wednesday, May 13, at a revised range of $150 to $160 per share following a surge in demand that oversubscribed the deal more than 20 times and forced the company to raise its target range twice in the space of a week.
At the top of the new range, Cerebras would raise approximately $4.8 billion and achieve a fully diluted market capitalisation of nearly $48.8 billion, making it the largest IPO globally in the past 12 months and the biggest pure-play AI hardware listing of 2026. Trading is expected to begin on the Nasdaq Global Select Market under the ticker CBRS on Thursday, May 14.
The deal began with a price range of $115 to $125 per share when the company filed its amended S-1 on May 4. Bloomberg first reported plans to lift the range to $125 to $135, reflecting initial bookbuilding momentum, before Reuters reported Sunday that Cerebras was considering a further increase to $150 to $160 after investor orders surpassed $10 billion against $3.5 billion of shares on offer. CNBC confirmed Monday that Nasdaq expects the IPO to proceed on May 14. Morgan Stanley, Citigroup, Barclays, and UBS are running the offering.
Cerebras reported 2025 revenue of $510 million, representing 76% year-on-year growth, with a net income of $238 million based on a 47% net margin. The company’s primary product is the CS-3 AI supercomputer, built around its Wafer-Scale Engine, a chip physically several times larger than Nvidia’s Blackwell B200 GPU and designed specifically for AI inference workloads rather than the general-purpose computing that conventional GPUs handle. The company’s founding argument is that the GPU was never built for AI and that a chip purpose-built for inference can deliver meaningfully faster performance at lower energy cost per token.
The commercial momentum behind that argument is substantial. OpenAI committed to purchasing more than $20 billion worth of Cerebras compute capacity across a multi-year agreement and invested alongside the deal. Amazon Web Services announced in March that Cerebras chips would be available through its data centre infrastructure, providing an enterprise distribution channel that the company’s previous revenue base, which was 86% concentrated in two UAE entities in 2025, had notably lacked. Those two customer additions transformed the IPO narrative from a concentrated hardware vendor with regulatory baggage into a platform play with hyperscaler validation.
Morningstar senior equity analyst Brian Colello offered a grounded assessment of where the excitement meets the risk. “It’s remarkable that the Business was valued at $8 billion in October… although large deals with OpenAI and Amazon Web Services certainly help the cause,” he said, before cautioning that Cerebras competes with rivals including Nvidia, AMD, Google, AWS, Microsoft, Meta, Broadcom, Marvell, and MediaTek. The competitive set is not merely large, it includes organisations spending tens of billions annually on R&D. Nvidia alone spent more than $18 billion in its most recent financial year.
At a valuation of nearly $49 billion against $510 million in 2025 revenue, the implied price-to-sales multiple of roughly 96 times trailing revenue positions Cerebras as one of the most expensive AI listings in recent memory. The bull case rests on OpenAI and AWS contract lock-in, a $24.6 billion remaining performance obligation balance, and the structural tailwind of the AI inference market expanding rapidly as model deployment replaces model training as the dominant compute workload. The risk is customer concentration, operating losses at the core business level before non-operating accounting effects, and the asymmetric difficulty of competing against Nvidia’s CUDA ecosystem, which remains the default infrastructure for most enterprise AI deployments.