OpenAI released the most credible AI chip benchmark data of this year on August 25, 2026. The next night, Jensen Huang sat down on Nvidia's Q2 fiscal 2027 earnings call and told investors not to worry. Both things can be true — and the tension between them is now the most important unresolved question in AI infrastructure.
On the InferenceX benchmark platform developed by semiconductor research firm SemiAnalysis — whose engineers visited OpenAI's labs to verify the runs — Jalapeño delivered up to 1.9 times more AI work per watt and up to 3.6 times lower end-to-end latency than Nvidia's GB200 and GB300 systems across three open-weight models. On interactive workloads — the low-latency, conversational traffic that represents most of ChatGPT's actual usage — the advantage widened to 2.1 to 4.1 times faster response times. These results mark the first time an AI lab's purpose-built inference chip has been independently assessed against Nvidia's production silicon in a public benchmark.
To read more, click here.