A lab's own inference chip beats every incumbent part tested

Claim

OpenAI's Jalapeño inference chip, taped out in nine months with AI help, delivers 1.9 times more work per watt and 3.6 times lower latency than Nvidia's best, with verifiers finding it beats every Nvidia, AMD and Google chip tested and declaring the CUDA moat potentially dead.

Reported metric: 1.9x per watt

Reporting date: 2026-08-27 (not necessarily the event date)

Sources

Attribution and reporting linked by this record; links do not imply independent verification.

• Welcome to August 27, 2026 (theinnermostloop.substack.com)