AI is now helping design the very chips that will power the next generation of AI!
@OpenAI just revealed some impressive results from its work with AMD. Its models achieved up to 14% improvements in power, performance, and area (PPA) on heavily optimized GPU designs, compressing roughly a year's worth of engineering progress into less than five months! Some of these AI-generated improvements have already been incorporated into AMD chips sent for manufacturing.
OpenAI's models also optimized its own "Jalapeño" inference chip, reducing the area of key compute blocks by 10–13%. Meanwhile, GPT-6.2 Astra is showing remarkable progress in designing entire RISC-V processor cores directly from written specifications.
And OpenAI is now partnering with Synopsys to train a specialized model for chip design.
Think about the feedback loop here: More powerful AI designs better chips, which enable even more powerful AI. This is where things start getting really interesting!