Cerebras will begin trading on the public market Thursday in what is expected to be the largest AI-focused IPO of 2026. The offering marks the first of a wave of AI chip company listings and is likely to draw significant investor interest given the sector’s rapid expansion.

The company’s flagship wafer-scale engine spans an entire semiconductor wafer—approximately the size of an iPad—enabling a single-processor design with integrated high-speed SRAM. This architecture delivers greater memory density and faster data throughput than multi-chip GPU assemblies.

Cerebras enters a market dominated by Nvidia and AMD in AI training and inference, aiming to capture share by offering superior single-chip performance. Its unique design could appeal to hyperscale data centers seeking to accelerate large-scale AI workloads.

Building wafer-scale chips involves high complexity and cost, as any fault affects the entire wafer rather than individual die. Cerebras claims a fault-tolerant routing architecture to mitigate yield losses, but production challenges and expense remain significant hurdles.