ChipsAug 25, 2026 · 23:487 sources

OpenAI Jalapeño Delivers 1.9x More Work Per Watt in First Custom Inference Chip

From huggingnews.com

Image: huggingnews.com
AI;DR

OpenAI will begin deploying its first custom inference chip, Jalapeño, within its compute infrastructure by the end of 2026. Built by Broadcom on TSMC N3P, the 700W chip delivers 1.5 to 1.9 times more AI work per watt and 1.7 to 3.6 times lower end-to-end latency than Nvidia's GB300 system. The hardware was developed using AI-assisted design via Codex and is designed exclusively for inference rather than model training.

Read the sourceSign in to suggest

Sources

+ 6 more
  1. SUPPORT@edludlow“OAI's Richard Ho is on Bloomberg Tech today”x.com
  2. SUPPORT@cpmou2022“2.1–4.1× higher performance on highly interactive workloads”x.com
  3. SUPPORT@wallstengine“OpenAI says a second-generation chip is already approaching tape-out, while work on a third generation has begun.”x.com
  4. SUPPORT@wallstengine“HBM4 with 15.4TB/s of memory bandwidth”x.com
  5. SUPPORT@deitaone“However, Jalapeno wasn’t tested against Nvidia’s newer Vera Rubin chips.”x.com
  6. SUPPORT@stocksavvyshay“Jalapeno reportedly delivered more AI work per watt and faster response times as inference becomes a larger share of AI compute.”x.com

One ranked edition, every morning

The stories that matter today. Free, cancel anytime.

Telegram @aidr_today