Cerebras Systems enters 2026 with a backlog that dwarfs its current size. The AI chipmaker said it exited 2025 with $24.6 billion in remaining performance obligations, and a significant share of that was tied to OpenAI, making the company one of the most closely watched artificial intelligence IPO stories of the year.
That is why Cbrs stock is being searched now: investors are waiting for the company’s first earnings report after going public, looking for proof that the contracts can turn into revenue. Cerebras revenue rose 75.7% in 2025 to $510 million, while revenue from its cloud and other services segment climbed 93.6% to $151.6 million, a sign that the company is pushing beyond one-time hardware sales and into recurring AI access and inference.
The OpenAI relationship is the center of the story. In January 2026, OpenAI agreed to add 750 megawatts of Cerebras high-speed compute capacity to its platform in a deal valued at over $10 billion through 2028. later reported that OpenAI could spend more than $20 billion over three years on servers powered by Cerebras chips, and that the newer commitments reportedly doubled the earlier agreement. The numbers matter because they suggest the company has already lined up demand at a scale far larger than its 2025 revenue base.
But the deal is not as simple as a headline number. OpenAI received warrants that can convert into a minority stake in Cerebras, with the share count depending on how much OpenAI spends. That means the ownership outcome is still elastic, and the eventual dilution is not fixed today. OpenAI also agreed to provide about $1 billion in loan financing to help Cerebras scale manufacturing and build data centers for AI workloads, showing that the customer relationship reaches into both demand and funding.
There is also a second thread investors are watching. Amazon has partnered with Cerebras to make fast AI inference services available through Amazon Bedrock in AWS data centers, but parts of that relationship still require definitive agreements. That leaves one of the company’s most promising distribution channels only partly locked in, even as inference becomes an increasingly important use case in AI workloads.
Cerebras is trying to prove that its wafer-scale engine can keep more computing and communication on one massive processor, reducing the data-movement bottleneck in complex AI workloads and offering an alternative to Nvidia’s chips. The market is already pricing in a great deal of that future; Cerebras was trading at 90.2 times trailing-12-month sales. The next earnings report will show whether the company can start converting that OpenAI-linked backlog, and whether the business is becoming more than a promise attached to a very expensive stock.
