Cerebras Systems Inc. has set its initial public offering price at $185 per share, exceeding the previously raised range and generating $5.55 billion before any underwriter options. The Sunnyvale, California-based AI chipmaker will begin trading on the Nasdaq Global Select Market under the ticker CBRS on Thursday, May 14. The offering values the company on a fully diluted basis at approximately $56.43 billion, according to Reuters.
The IPO marks the largest debut of 2026 so far, giving public investors a direct opportunity to invest in the infrastructure powering artificial intelligence—not just training but also inference, the stage where models generate responses. Demand for the offering was robust, with investor interest exceeding available shares by more than 20 times, as reported by Reuters. Cerebras had already increased its price range to $150–$160 from an initial $115–$125, while also boosting the offering size to 30 million shares from 28 million.
Cerebras specializes in building oversized processors designed for AI workloads. Its latest Wafer-Scale Engine 3 chip is 58 times larger than a top GPU chip and can run inference tasks up to 15 times faster than leading GPU-based systems, depending on the open-source model used. The company aims to carve out a niche in a sector dominated by Nvidia.
Financially, Cerebras reported revenue of $510 million in 2025, up from $290.3 million the previous year, according to filings seen by Reuters. However, the Financial Times noted an operating loss of $146 million for 2025. At the IPO price, investors valued the company at about 111 times last year's revenue.
A key element of the bullish thesis is Cerebras' partnership with OpenAI. In January, OpenAI announced a collaboration to deploy 750 megawatts of low-latency AI compute capacity, to be rolled out in phases through 2028. OpenAI's Sachin Katti described Cerebras as a dedicated low-latency inference solution, while Cerebras CEO Andrew Feldman stated that real-time inference will transform AI.
Amazon Web Services also represents a significant channel for Cerebras. In March, AWS announced plans to integrate Trainium chips and Cerebras CS-3 systems into its data centers, offering customers access via Amazon Bedrock. David Brown, vice president of compute and machine-learning services at AWS, noted that speed remains a critical bottleneck for real-time AI workloads.
Prediction markets reflected expectations of a strong debut. On Polymarket, bets on Cerebras' closing market cap placed the $60 billion to $70 billion range at roughly 33%, while a separate market pegged a $50 billion-plus finish at 96%. However, trading volumes were modest at $99,800 and $139,600.
Earlier this year, Arm Holdings and its majority shareholder SoftBank made an early bid to acquire Cerebras ahead of the IPO, but Cerebras declined the offer, according to Bloomberg. None of the parties responded to requests for comment.
Risks remain significant. The Financial Times reported that two customers—G42 and Mohamed bin Zayed University—accounted for 86% of Cerebras' revenue last year, highlighting heavy customer concentration. Additionally, scaling up and manufacturing challenges persist for chip startups competing with larger players.
Cerebras has granted underwriters a 30-day option to purchase up to 4.5 million additional shares at the IPO price, less underwriting fees. The bookrunners include Morgan Stanley, Citigroup, Barclays, and UBS Investment Bank. The offering is expected to close on May 15, subject to standard conditions. The opening trade will test whether Wall Street is willing to pay up for an alternative AI compute play outside Nvidia and whether Cerebras can convert its large contracts into a sustainable business.
