Cerebras IPO Details: $3.5B Raise, Wafer-Scale Tech, and Investor Risks

AI chipmaker Cerebras Systems has filed updated IPO details with the Securities and Exchange Commission, revealing a public debut that could raise as much as $3.5 billion. The company announced plans to commence its roadshow on May 4, 2026, with pricing expected May 13 and trading to begin the following day on the Nasdaq under ticker "CBRS."

The offering will sell 28 million shares of Class A common stock priced between $115 and $125 per share. At the high end of the range, Cerebras would be valued at approximately $26.6 billion, nearly 16% higher than the $23 billion valuation it received in February's venture round. Underwriters have been granted a 30-day option to purchase an additional 4.2 million shares, which could bring total proceeds to $4.025 billion if fully exercised.

According to the official press release, Morgan Stanley, Citigroup, Barclays, and UBS Investment Bank will act as lead book-running managers. The company has applied to list on the Nasdaq Global Select Market, positioning itself alongside other high-profile AI infrastructure plays.

Cerebras' core technology differentiates it from competitors through the Wafer-Scale Engine (WSE). Rather than cutting a silicon wafer into hundreds of smaller chips, the company uses the entire wafer to create a single processor. The WSE-3 is 58 times larger than Nvidia's B200 AI chip, boasting 900,000 compute cores and 2,625 times more memory bandwidth. This architecture addresses latency issues inherent in multi-chip AI systems, since communications are thousands of times faster on-chip than across chips.

The financial picture shows explosive growth but persistent losses. In 2025, Cerebras generated $510 million in revenue, up 76% year over year, while posting an operating loss of $146 million. The company's remaining performance obligation—contractually obligated revenue not yet recognized—stands at $25 billion. Investors should note that 15% of this RPO will be recognized in 2026 and 2027, with 43% coming in 2028 and 2029.

Customer concentration represents a material risk. Two customers accounted for 86% of Cerebras' revenue last year. The company's high-profile partnerships include a $20 billion, 750-megawatt deal with OpenAI announced in January 2026, a multi-year agreement with Amazon Web Services to deploy WSE in data centers, and a deal with Meta Platforms to run inference on its Llama 4 model.

The share structure creates a significant control imbalance. Class A shares carry one vote per share and will be issued to the public. Class B shares are entitled to 20 votes per share and will be held by early investors and insiders, who will retain majority voting control. Non-voting N Class shares were issued as warrants to OpenAI and Amazon, allowing them to buy up to $1.27 billion in these shares. Co-founder and CEO Andrew Feldman is not selling shares in the IPO and will own 10.3 million shares worth up to $1.28 billion at the high end of the range.

Relatively few technology companies have gone public since central banks raised interest rates in 2022 to fight inflation, making investors less interested in unprofitable names. But with the ascent of generative AI products, investors have become rabid about betting on companies that benefit from the trend. Cerebras competitor CoreWeave, which rents out Nvidia GPUs as a cloud service, raised $1.5 billion in an IPO last year.

Cerebras sought to go public in 2024 but later withdrew the paperwork as its business model shifted away from selling hardware and toward operating a cloud service based on its own chips. The April 2026 filing marks the company's second attempt at a public debut.

The physical reality of Cerebras' approach means fewer chips to manage in a data center rack, but the engineering complexity is immense. A single WSE-3 requires specialized cooling and power infrastructure that traditional GPU clusters don't demand (which is both a selling point and a deployment headache). The latency benefits are real, but only if customers can actually integrate the massive chips into their existing workflows.

Whether the market rewards Cerebras' novel architecture remains to be seen. The $25 billion RPO provides revenue visibility, but the concentration risk and unprofitable status create genuine uncertainty. Time will tell if investors care more about the technology's potential or the bottom line.

For now, the real question isn't whether Cerebras can compete with Nvidia technically—it's whether public shareholders will tolerate the losses while insiders keep 20 votes per share. Whether users actually pay for the performance gains remains the real question.

Artūras Malašauskas is an AI Systems Integrator with 20+ years of production-grade web engineering experience. He has designed, shipped, and scaled enterprise Python/PHP systems for logistics, SaaS, and public-sector clients. For the past year, he has focused exclusively on AI integrations: deploying open-source LLMs, building generative media pipelines (image, audio, video), and engineering multi-agent workflows for real production environments. His standard: reproducibility, security, cost-efficient inference—no vaporware. He documents and evaluates emerging AI tooling, separating verified capabilities from marketing noise. Technical editor at: muza-ai.eu, ai-verslas.lt, ai-naujinos.lt Connect on LinkedIn