Cerebras Systems filed updated paperwork for a Nasdaq IPO on 4 May 2026, planning to sell 28 million shares at a price between $115 and $125 per share. At the top of that range, the offering would raise approximately $3.5 billion and value the company at up to $26.6 billion on a fully diluted basis. The AI chip specialist, which builds wafer-scale engines as an alternative to NVIDIA's GPU architecture, reported fourth-quarter revenue of $510 million — a 76 per cent increase year-on-year — and net income of $87.9 million.
The IPO comes at a pivotal moment for the AI hardware market. NVIDIA's CEO Jensen Huang recently acknowledged that the company's AI accelerator market share in China has dropped to effectively zero due to US export restrictions, with domestic competitors like Huawei and Cambricon filling the gap. Meanwhile, Cerebras has secured a major deal with OpenAI worth over $20 billion through 2028, positioning its wafer-scale chips as a credible alternative for training and inference workloads. The company's architecture takes a fundamentally different approach to AI compute: instead of networking thousands of individual GPUs together, Cerebras builds a single chip the size of an entire silicon wafer, eliminating the interconnect bottlenecks that plague traditional GPU clusters.
For context engineers, the Cerebras IPO signals that the AI chip market is genuinely diversifying beyond NVIDIA's near-monopoly. Google has its TPU 8t and 8i chips announced at Cloud Next, Amazon is scaling its Trainium accelerators, and now Cerebras is going public with proven revenue and a marquee customer in OpenAI. The practical implication is that the compute layer underpinning AI development is becoming more competitive, which should drive down costs and increase availability for the cloud platforms developers build on. If Cerebras succeeds at its target valuation, it would be among the largest tech IPOs of 2026 — a testament to just how much capital is flowing into AI infrastructure.
