Cerebras IPO Prices at $150–$160 Ahead of Thursday Nasdaq Debut

Cerebras Systems is pricing its IPO today at $150 to $160 per share, ahead of its Thursday debut on Nasdaq under the ticker CBRS. At the top of that range, the AI chip company would raise up to $4.8 billion and carry a fully diluted valuation of $48.8 billion — placing it among the most valuable tech IPOs of the decade.

The offering was oversubscribed more than 20 times, forcing Cerebras to raise its share count from 28 million to 30 million just to meet demand. Two weeks ago, its target range was $115 to $125 per share. That's how fast the market appetite grew.

A Long Road to Wall Street

Cerebras has been trying to go public for two years. In 2024, the company filed confidentially, but a national security review halted everything. The concern: its early investors included G42, a major Abu Dhabi AI company with reported ties to China's tech sector. That review wrapped up earlier this year, clearing the path for an April 2026 S-1 filing.

The delay turned out to be fortuitous. AI infrastructure spending exploded in the interim, and Cerebras arrived at its IPO roadshow with $510 million in 2025 revenue — and a jaw-dropping 47% net margin, rare for a hardware startup.

The Chip That's Bigger Than Your Dinner Plate

What makes Cerebras unusual is its core product: the CS-3, a wafer-scale engine (WSE — a chip the literal size of a dinner plate) with 4 trillion transistors. That's roughly 57 times larger than a conventional GPU die.

The architecture isn't just a novelty. Because there are no chip-to-chip communication bottlenecks, the CS-3 can run AI model inference — the process of actually using a trained AI model to generate answers — at speeds that outpace Nvidia GPU clusters for certain workloads. Cerebras describes itself as "the market leader in high-speed AI inference," and the claim has enough real-world support that major customers are paying up.

For context: when you ask an AI assistant a question, inference is what happens. The faster inference is, the more responsive your AI feels. For enterprises running AI at scale, inference speed translates directly to cost savings and user experience.

OpenAI's $10 Billion Bet

The most striking line in Cerebras's S-1 is a multi-year agreement with OpenAI valued at more than $10 billion. OpenAI didn't just sign a contract — it made a loan of $1 billion to Cerebras in December, secured by warrants to buy over 33 million shares.

Sam Altman, Greg Brockman (OpenAI's president), former OpenAI chief scientist Ilya Sutskever, and OpenAI board member Adam D'Angelo are all listed as angel investors. That's essentially the original OpenAI leadership team betting on Cerebras as the hardware backbone for next-generation AI inference.

Amazon is also in the picture. AWS announced in March 2026 that it would deploy a combination of Amazon Trainium and Cerebras CS-3 chips in its data centers, making them available through Amazon Bedrock — the service that lets enterprises access foundational AI models. That deal gives Cerebras a distribution channel that reaches every AWS enterprise customer.

Why This IPO Matters for the AI Industry

Cerebras is the first pure-play AI inference chip company to go public. Nvidia dominates AI hardware today, but its strength is in training — teaching AI models to get smart. Inference, the far more frequent task of using those models in production, is a different game.

As AI moves from labs into every product and workflow, inference demand is growing much faster than training demand. Every query, every document summary, every AI-generated customer response is an inference event. The company that can run those faster and cheaper wins.

Cerebras is betting its wafer-scale approach is the answer. The market, at 20x oversubscription, seems to agree — at least for now.

What to Watch After the IPO

Cerebras shares will begin trading Thursday under CBRS on Nasdaq. Final pricing happens Wednesday evening (May 13).

Watch for: the first-day trading pop (or drop), which will signal how much of the demand was speculative versus fundamental. Also watch for any analyst commentary on Cerebras vs. Nvidia pricing dynamics — Cerebras prices its inference services competitively, but Nvidia has both scale and an incumbent advantage in enterprise AI stacks.

Longer term: Cerebras has promised to expand its product line beyond the CS-3 wafer-scale engine, with next-generation architectures targeting both inference and training. Whether it can hold its performance edge as Nvidia's Blackwell and Vera Rubin chips deploy at scale will be the real story of 2027.

The Bottom Line

Cerebras's IPO is a milestone for the AI chip industry — proof that the market believes specialized inference hardware deserves its own category, separate from the GPU oligopoly. With $510 million in revenue, a deep relationship with OpenAI, and AWS distribution, the fundamentals are real. The $48.8 billion valuation is aggressive, but in a market where AI infrastructure is treated as essential infrastructure, aggressive valuations have a way of growing into themselves. If you're following the AI chip space, Thursday's Nasdaq debut is the one to watch.