OpenAI Unveils Jalapeño, First Custom Chip With
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- OpenAI unveiled "Jalapeño," its first custom-built inference processor, developed in collaboration with Broadcom in a partnership officially announced in October.
- OpenAI's own AI models assisted in the chip's design, and TechMeme's coverage notes the design-to-manufacturing tape-out cycle took just nine months.
- Jalapeño targets inference specifically — running pre-built models in response to user commands — with OpenAI emphasizing the chip's low operating cost for real-time coding models.
- More performance-intensive pre-training will likely still rely on Nvidia hardware, meaning the chip targets inference economics rather than displacing Nvidia's training dominance.
- President Greg Brockman told OpenAI's podcast the company pursued "specific workloads that are underserved" and brought a "deep understanding of the workload" to chip design.
- OpenAI framed the move as full-stack infrastructure ownership, spanning chip architecture, kernels, memory systems, networking, scheduling, and deployment systems.
Why it matters: Jalapeño targets inference — the cost layer that scales with every ChatGPT query — not training. Early performance-per-watt gains could meaningfully cut the marginal cost of running real-time models, even as Nvidia retains its grip on the capital-intensive pre-training workloads that remain beyond the chip's stated scope.
Ask SkimNews
