Using OpenAI’s models, AMD engineers trimmed the cell area of a new GPU by 15% and lowered its latency, highlighting how generative AI is accelerating hardware design.

AMD’s latest GPU design has been trimmed by 15% in cell area and its latency reduced, thanks to generative AI models from OpenAI.

How OpenAI’s models were applied

AMD engineers integrated OpenAI’s large‑scale language and diffusion models into their design workflow, using them to generate layout alternatives and predict performance impacts.

The AI system evaluated thousands of micro‑architectural configurations in minutes, a task that would normally require weeks of manual simulation and expert review.

Key benefits observed

  • 15% reduction in silicon cell area, freeing up die space for additional features
  • Lowered signal latency, improving overall GPU throughput
  • Accelerated design cycles, cutting time‑to‑market by an estimated 30%

These gains illustrate how generative AI can act as a co‑designer, rapidly iterating on complex hardware constraints while preserving functional integrity.

Implications for the semiconductor industry

If AI‑driven design becomes standard, manufacturers could see faster innovation cycles, reduced R&D costs, and more efficient use of manufacturing capacity.

AMD plans to expand the use of OpenAI’s tools across its product line, including future CPU and accelerator projects.

For readers interested in the broader impact of AI on infrastructure, see the Infrastructure section on uxc.news.

Explore more about the partnership on OpenAI on Momentum and learn how AMD is leveraging AI at Advanced Micro Devices on Momentum.

DX Today coverage of OpenAI’s AI tools help AMD shrink GPU cell area