I think OpenAI just made AI chip design impossible to dismiss.
OpenAI says AMD saw up to 14% gains on flagship GPU IP, while Jalapeno blocks fell 13% and 10% in area.
OpenAI says GPT-4o rewrote adders and multipliers in Jalapeno to save area and cut one multiplier from two cycles to one.
OpenAI says GPT-6.2 Astra led 9,881 RISC-V runs at 112,000 tokens and scored higher designing from specs than editing old RTL.
Synopsys said at its 2026 Investor Day in New York it will embed OpenAI intelligence directly in EDA workflows, per EE Times.
OpenAI admits cores built from specs could skip counters and need tighter area and timing checks to match production.
I still buy the result, so if you build chips, test model rewrites now before rivals do.





