OpenAI CEO Sam Altman says "we made a chip and it is fast" about OpenAI's first custom processor

OpenAI chief executive Sam Altman announced that the company has produced its first custom processor, a chip named Jalapeño built for AI inference—the phase where trained models respond to user requests. The company said the chip, developed with AI-assisted design, showed better efficiency and lower latency than comparison systems in tests on three large models. OpenAI plans to put Jalapeño into its data centers by the end of 2026. The move could reduce its reliance on Nvidia, though OpenAI says it will continue using Nvidia chips for training and inference. The announcement matters because custom hardware may cut costs and improve speed for services like ChatGPT and AI agents.