OpenAI Unveils Jalapeño, Its First Custom AI Chip
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- OpenAI revealed Jalapeño, its first custom AI processor — an ASIC designed for inference workloads like powering ChatGPT and Codex — built in partnership with Broadcom.
- Jalapeño is built specifically for inference (processing user requests), not training, and arrives nine months after OpenAI announced its Broadcom partnership to reduce Nvidia GPU dependence.
- Broadcom CEO Hock Tan told Reuters Jalapeño matches the performance of Nvidia's Blackwell chips and Google's Tensor Processing Units.
- Microsoft, Meta, and Amazon have each launched custom AI chips recently, but all still trail Nvidia on overall performance, per the report.
- OpenAI calls Jalapeño the "first step in a multi-generation compute platform," with deployment expected by the end of 2026.
- Early testing shows Jalapeño delivers performance per watt "substantially better than current state-of-the-art," according to OpenAI.
Why it matters: Jalapeño gives OpenAI a path to reduce its dependence on Nvidia's supply-constrained GPUs while matching Blackwell-class inference performance, with deployment targeted by end of 2026. Microsoft, Meta, and Amazon are pursuing the same custom-chip strategy but still trail Nvidia, making OpenAI's Broadcom-built ASIC a direct shot at the AI compute bottleneck.
