Cerebras IPO to Test Its $29 Billion Valuation Amid Execution Risks

Cerebras Systems, a challenger to Nvidia’s dominance in AI computing, has filed for an initial public offering that puts the market for private technology valuations on notice. The company, which designs massive, wafer-sized processors to accelerate AI workloads, saw its valuation climb to $29.4 billion in secondary markets, a 28% premium to its last private funding round and a steep 58 times its 2025 revenue of $510 million.

“If you believe that the AI inference market is infinitely sized, which I do, this is going to be the chip that is the standard for inference,” Darian Shirazi, managing partner at Gradient, said. Shirazi, who purchased shares on the secondary market, believes Cerebras could become a $100 billion company.

The IPO filing reveals a company with a massive $24.6 billion in remaining performance obligations, primarily from a landmark deal to provide OpenAI with 750 megawatts of AI inference capacity between 2026 and 2028. That deal, worth up to $20 billion, is contingent on Cerebras executing an unprecedented operational scale-up. The company’s revenue grew 76% to $510 million in 2025, a fraction of its order book.

However, the valuation and execution risks are substantial. At 58 times trailing sales, Cerebras is priced significantly higher than Nvidia, which trades at 23 times sales while holding a dominant and secure position in the data center market. “For companies that have gone public at a price-to-sales ratio of over 40... on average they have underperformed,” said Jay Ritter, director of the IPO Initiative at the University of Florida, noting such companies saw an average return of minus 44.8% between 1987 and 2024.

Technology vs. Production Reality

Cerebras’s core innovation is its Wafer-Scale Engine, a chip the size of an entire silicon wafer that the company claims can be up to 15 times faster for AI inference than leading GPU solutions. By placing more cores and memory on a single piece of silicon, it reduces data transfer bottlenecks and power consumption. This technology attracted not only OpenAI but also Amazon Web Services, which will integrate the CS-3 system into its cloud platform.

The challenge lies in manufacturing. Cerebras is entirely dependent on Taiwan Semiconductor Manufacturing Co. (TSMC) to produce these unique chips, and no other foundry has the capability. Yields on wafer-scale chips are inherently lower, as a single defect can ruin the entire unit. This single-source dependency and manufacturing complexity represent a critical risk as Cerebras attempts to scale from $510 million in revenue to fulfilling a $20 billion order.

A Market Test for AI Hype

The Cerebras IPO is set against a backdrop of intense investor enthusiasm for AI infrastructure. Companies like Anthropic and OpenAI are engaged in a multi-hundred-billion-dollar scramble for computing capacity, striking deals with cloud providers and chipmakers. This demand has fueled a surge in private market valuations, with Cerebras’s own valuation jumping 184% from $8.1 billion to $23 billion in just over four months between its Series G and H rounds.

The public offering will force a collision between this private-market momentum and public-market fundamentals. If the IPO succeeds, it could validate the high valuations across the private AI sector and encourage more listings. If it falters, it may signal that public investors are unwilling to underwrite the significant execution risks tied to even the most promising AI technologies, potentially cooling the secondary market. Investors will be watching closely to see if Cerebras can grow into its 58x sales multiple or if it will join the list of high-flyers that failed to meet lofty expectations.

This article is for informational purposes only and does not constitute investment advice.