OpenAI Taps Cerebras for Faster AI Inference

OpenAI is teaming up with Cerebras to add 750 megawatts of ultra low-latency compute power. Cerebras makes special chips that are basically one giant processor, which speeds up AI responses by removing the usual bottlenecks. This means when you ask a hard question or run an AI agent, you’ll get answers faster — making interactions feel more natural and real-time. OpenAI says this is about building a flexible compute portfolio, matching the right hardware to the right job. Cerebras’s CEO compares the shift to broadband transforming the internet. The new capacity will roll out in phases through 2028. Why it matters: This partnership signals that even the biggest AI labs are hitting hardware limits with traditional chips, and the next leap in user experience may come from specialized silicon, not just bigger models.