OpenAI has described its new AI chip, Jalapeño, in a blog post, saying it completes tasks more efficiently and returns responses faster than other AI systems. In a briefing with reporters, OpenAI hardware vice president Richard Ho said the chip offers the “best of both worlds” by combining lower latency with higher throughput, characteristics he said AI systems usually have to trade off against one another.
First introduced in June, Jalapeño is an Application-Specific Integrated Circuit (ASIC) developed in partnership with Broadcom. It is designed for AI inference — the process of running a trained model to complete a task or deploy an agent.
Why it matters
Inference performance affects how quickly AI systems respond and how much work they can handle. A chip aimed at improving both latency and throughput could influence how OpenAI deploys its models.
Who should care
Organizations running AI models at scale and those following AI hardware developments may take an interest in a purpose-built inference chip and OpenAI’s collaboration with Broadcom.