
Wafer-scale AI chips and inference
cerebras.ai (opens in a new tab)Cerebras Systems builds the largest AI chip in the world, 56 times the size of a GPU, and the systems around it. That architecture delivers training and inference speeds more than ten times faster than GPU-based hyperscale cloud inference, which changes what AI applications can do: real-time iteration becomes possible, and agentic systems can spend more computation per request without the latency becoming unusable. Customers span leading model labs, global enterprises and AI-native startups. OpenAI has committed to a multi-year partnership deploying 750 megawatts of capacity to move key workloads onto high-speed inference. The team publishes and open-sources its research.





