Broadcom and OpenAI unveil custom-built Jalapeño inference processor — OpenAI's first chip is a massive reticle-sized ASIC built in an ultra-fast nine-month development cycle

Broadcom and OpenAI have unveiled Jalapeño, OpenAI's first custom-built inference ASIC. Developed in a rapid nine-month cycle, this reticle-sized processor is designed to handle ChatGPT and other services, allegedly surpassing leading-edge performance-per-watt metrics. OpenAI aims to build its full stack to increase efficiency and intelligence delivery. While industry experts state that ASICs are less flexible than Nvidia GPUs, they are more cost-effective and specifically designed for AI tasks to ensure broader access to advanced AI.