Cerebras relaunches US IPO: AI chip developer files again
Cerebras has filed for a US IPO again following its unsuccessful previous attempt. The AI chip developer and data center operator seeks a Nasdaq listing…
AI-processed from Bloomberg Tech; edited by Hamidun News
Cerebras Systems is returning to the public market: the AI chip developer and data center operator filed a new IPO application in the US just months after withdrawing a previous one. For the market, this is not just another startup listing riding the wave of AI interest, but a rare attempt to take public a major player in AI infrastructure that sells not models, but computational power, server systems, and access to them via the cloud. According to the new filing, the company plans to list Class A shares on Nasdaq under the ticker CBRS.
Cerebras builds its business around its own architecture for AI accelerators, server systems, and accompanying software. In the documents, the company emphasizes that it now earns not only from supplying hardware to customer data centers, but also from a cloud consumption model, where customers rent computations on a monthly or yearly basis or pay for usage based on token volume. For a business long perceived as a niche alternative to Nvidia, this is an important shift: Cerebras sells not just a chip, but an entire AI infrastructure circuit.
The financial dynamics look impressive. Cerebras' revenue grew from $24.6 million in 2022 to $78.
7 million in 2023, then to $290.3 million in 2024, and reached $510 million in 2025. Year-over-year growth for the last reporting period was 76%.
Formally, the company showed GAAP net income of $237.8 million compared to a loss of $481.6 million a year earlier.
But the picture is not so straightforward: the same documents indicate that on a non-GAAP basis, Cerebras remained unprofitable, and part of the accounting profit is related to the revaluation of liabilities. That is, the IPO story is built primarily on accelerated growth and scaled demand, not on already-proven sustainable profitability. Another key argument for the offering is new contracts.
In January 2026, Cerebras announced a multi-year agreement with OpenAI worth over $20 billion. This involves deploying 750 megawatts of high-speed AI computing, as well as joint work on future models for the next generation of Cerebras' hardware platform. In March, the company also announced a partnership with AWS, which is expected to expand global distribution of its infrastructure for inference workloads.
As of the end of 2025, the remaining contract obligation backlog reached $24.6 billion, with a significant portion of that amount tied to agreements with OpenAI. At the same time, the prospectus directly outlines the risks.
Cerebras' business remains notably dependent on a limited number of major customers. G42 provided 24% of the company's revenue in 2025 and 85% in 2024, while Mohamed bin Zayed University of Artificial Intelligence, or MBZUAI, contributed another 62% of revenue in 2025. In other words, even amid rapid growth, the income structure remains concentrated.
This is particularly important given that the previous IPO attempt was protracted due to regulatory scrutiny of Cerebras' relationships with G42 by American authorities, after which the company withdrew the previous documents in October 2025. The new filing marks the company's return to its public listing plan in an updated configuration. What does this mean?
If the IPO succeeds, investors will get one of the few pure public bets on the AI infrastructure layer outside the usual market giants. But along with that, they are buying not a calm mature business, but a company in a phase of aggressive growth, with a massive order portfolio, high capital intensity, and dependence on a few partners. For Cerebras, this is a chance to turn the hype around AI into long-term capital.
For the market, it's a test of whether it's ready to finance not just creators of models, but those who build the computational foundation for the next stage of competition in artificial intelligence.
Want to stop reading about AI and start using it?
AI News is a curated feed of AI/tech news. Hamidun Academy teaches you to use AI systematically in your work.