🤖 OpenAI’s Jalapeño chip is coming for Nvidia

🤖 OpenAI’s Jalapeño chip is coming for Nvidia OpenAI has published the first real performance results for its custom AI inference chip, Jalapeño and the numbers are crazy. Against Nvidia’s GB200 and GB300, OpenAI says Jalapeño delivers 1.5–1.9× more AI work per watt and 1.7–3.6× lower end-to-end latency across GPT-OSS 120B, DeepSeek R1 and Kimi K2.5 1T. For highly interactive workloads, exactly the kind needed for AI agents, OpenAI claims 2.1–4.1× higher performance. The chip is rated at just 700W, compared with 1,200W for GB200 and 1,400W for GB300. OpenAI says its measured sustained power stayed at or below 550W in the tested workloads. OpenAI plans to start deploying Jalapeño inside its own compute infrastructure by the end of 2026, with Gen 2 already in development and Gen 3 taking shape. Source. @aipost 🏴