OpenAI and Broadcom unveil 'Jalapeño,' a custom LLM-inference chip co-designed in nine months

OpenAI·high signal

OpenAI announced Jalapeño, its first Intelligence Processor, co-developed with Broadcom from design to tape-out in roughly nine months — one of the fastest high-performance ASIC cycles ever, partly accelerated by OpenAI's own models. Engineering samples are already running ML workloads (including GPT-5.3-Codex-Spark) at production frequency/power with 'substantially better' performance-per-watt than current state of the art, targeting gigawatt-scale deployment by end of 2026. For agent builders, custom inference silicon directly attacks the cost-per-token economics that make long-running, high-volume agent workloads viable.