Broadcom
OpenAI’s Jalapeño Chip Just Rewrote Inference Economics
The AI Prism Admin
OpenAI published the first benchmarks for Jalapeño, its custom inference chip built with Broadcom — up to 1.9x throughput per kilowatt and 3.6x lower latency than a current Nvidia Blackwell system. Here's what the numbers mean for token prices, agents, and Nvidia's moat.
