GoKawiilOpenAI says Jalapeño AI chip built for internal use, hints at wider access later
OpenAI's Head of Hardware Richard Ho told Tom's Hardware that the company's new Jalapeño inference ASIC, developed with Broadcom, is designed primarily to meet OpenAI's own growing compute demand. Ho said the chip 'could' be used by others eventually, but that satisfying OpenAI's internal needs remains the priority for the foreseeable future. At Hot Chips 2026, OpenAI shared benchmarks showing Jalapeño running GPT-OSS, DeepSeek R1 and Kimi K2.5 against Nvidia's GB200 and GB300 accelerators.
GoKawiil's interpretation of the reporting above, not reported fact.
Ho's comments suggest OpenAI is positioning Jalapeño first as a way to reduce reliance on Nvidia GPUs rather than as a commercial product, which could limit near-term competitive pressure on Nvidia despite the public benchmarking. The willingness to eventually open the chip to outside users hints OpenAI may see hardware as a future revenue stream once its own compute backlog eases. The decision to publish comparative benchmarks, unusual for purpose-built ASICs, signals OpenAI wants to be seen as a credible hardware competitor even while insisting its scope is internal.
- Jalapeño is currently intended mainly for OpenAI's own AI compute needs, per hardware chief Richard Ho.
- Ho did not rule out broader availability of the chip in the future, saying 'you could use it for anybody.'
- OpenAI presented Hot Chips 2026 benchmarks pitting Jalapeño against Nvidia's GB200 and GB300 accelerators.
Source: tomshardware.com — Jake Roach, 2026-09-28
Published there as: “OpenAI's custom Jalapeno AI inference ASIC is for OpenAI’s internal use, but company leaves the door open to broader rollout”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.