OpenAI partners with Cerebras to add 750MW of high-speed AI compute, reducing inference latency and making ChatGPT faster for real-time AI workloads.