OpenAI Says Its First Chip Jalapeño Beats Nvidia's GB300 on Inference

OpenAI announced its first inference chip, Jalapeño, outperforms Nvidia's GB200 and GB300 on AI workload per watt and latency on public benchmarks, with deployment expected in small volumes by late 2026 and meaningful scale in 2027. The chip aims to reduce OpenAI's inference costs, challenging Nvidia's pricing power in AI hardware but full system economics and large-scale deployment remain to be proven.

This Wire brief sits within Fusion42's coverage of AI Infrastructure and Semiconductors, and 5 sources have reported it between 25 Aug 2026 and 26 Aug 2026.