Enterprise DNA
News AI News

OpenAI models assist chip design

OpenAI published an account of using its models to advance chip design, reporting quantified results including area reductions and PPA improvements..

Enterprise DNA |
OpenAI models assist chip design

AI Pulse · AI Trends Pulse

The play

Track AI-assisted chip design as a specialized productivity signal, but ask vendors for independently validated area and performance results before setting budgets.

OpenAI says its models have helped improve chip designs, including its Jalapeño inference chip and hardware developed with AMD. In Jalapeño’s vector and matrix units, the company estimates model-proposed changes reduced standard-cell area by about 13% and 10%, respectively. In one matrix multiplier redesign, cell area fell about 15% and latency dropped from two clock cycles to one. OpenAI says those Jalapeño optimizations were proven during silicon bring-up.

With AMD, OpenAI reports up to 14% improvement in power, performance and area compared with heavily optimized baselines on flagship GPU IP. AMD has included some model-written changes in a design sent for manufacturing and added others to files for future versions. Those are reported results from OpenAI’s collaboration, not proof that every chip-design task will see similar gains. For teams, the example underlines the value of tracking what models do and what needs checking.

The work also has limits. At about 112,000 tokens per run, all five models scored higher when asked to design from written requirements than when improving existing designs. OpenAI notes that the models could omit features, and says tighter area limits and physical timing checks would make the test more representative of production work. For a business, the useful takeaway is that AI may contribute to specialized engineering tasks, but results depend on how the work is tested and verified. OpenAI’s account details both the reported improvements and the evaluation caveats.

More like this?

Free daily email

Get this every morning.

This brief is one item from today's AI Pulse, the short daily read we run for ourselves on what is actually happening in AI. Subscribe free and it lands in your inbox each morning.

Free daily email

Subscribe to the daily AI Pulse

One short read every morning on what is actually happening in AI. Free.

One email a day. Unsubscribe any time.

Working With Claude field guide cover

Free Resource

Put what you just read to work

The free 32-page Working With Claude guide: the full ecosystem, Claude Code, and how to roll it out across a business.

Add your name (optional)

No spam. Unsubscribe any time.