Skip to main content
World Today News
  • Home
  • News
  • World
  • Sport
  • Entertainment
  • Business
  • Health
  • Technology
Menu
  • Home
  • News
  • World
  • Sport
  • Entertainment
  • Business
  • Health
  • Technology
OpenAI Models Boost AMD Flagship GPU PPA by up to 14%

OpenAI Models Boost AMD Flagship GPU PPA by up to 14%

November 20, 2026 Technology Desk Technology

On Wednesday November 18, 2026, OpenAI revealed that its AI models helped AMD improve the power, performance, and area (PPA) of its upcoming flagship GPU design by up to 14% relative to heavily optimized baselines, as reported by The Next Web.

The Tech TL;DR:

  • OpenAI models rewrote SystemVerilog RTL circuit code for AMD’s flagship GPU, yielding up to a 14% PPA improvement.
  • AMD confirmed single-digit percentage PPA gains across key blocks of its upcoming next-generation flagship CPU.
  • OpenAI utilized Synopsys electronic design automation tools to verify that every model-generated code rewrite preserved block behavior.

The optimization work compressed what typically amounts to a year of conventional chip design effort into under five months, according to OpenAI. The models reasoned about Register Transfer Level code at the level of an expert microarchitect. AMD has already incorporated some of these automated changes into a recent design sent for manufacturing, while adding other suggestions to production design files for upcoming hardware iterations.

“We’ve observed the model reason about RTL at the level of an expert microarchitect,” OpenAI said in its published technical documentation.

Advancing chip design with OpenAI models. Colorful chip layout artwork beside the article title
Photo: openai.com

Alongside the AMD hardware collaboration, OpenAI detailed design metrics for its own custom inference processor, Jalapeño. Developed with early AI assistance starting in early 2025, Jalapeño silicon samples ran actual workloads as early as June 2026, as noted by Tom’s Hardware. Model-suggested optimizations reduced standard-cell area by about 13% in the vector compute unit and 10% in the matrix arithmetic unit, with these geometric and architectural savings proven during initial silicon bring-up.

To measure the progression of automated semiconductor design capabilities across successive generations, OpenAI tested its large language models on two open-source 32-bit RISC-V processor cores: PicoRV32, which is optimized for field-programmable gate arrays, and Ibex, utilized as an embedded processor within OpenTitan security chips. Each model received 112,000 tokens per execution run with instructions to maximize instruction throughput.

Testing yielded a clear performance hierarchy among model generations. GPT-6.2 Astra scored the highest optimization metrics and incurred the lowest overall compute costs when tasked with either improving existing chip architectures or designing a processor entirely from scratch based on written requirements. However, researchers said that designing processors from initial specifications proved more straightforward because the models could freely omit legacy features present in standard reference codebases.

Despite these benchmark gains in simulated and physical silicon, caution remains standard practice within hardware engineering divisions. Richard Ho, OpenAI’s chip chief, noted last month that he does not yet fully trust automated AI design changes without rigorous manual verification, ensuring every model-generated alteration undergoes strict EDA validation.

To embed intelligence directly into standard hardware engineering workflows, OpenAI is currently collaborating with Synopsys to train a specialized model designated as GPT-Synopsys. This initiative integrates frontier AI models natively with the electronic design automation tools engineers already rely on for semiconductor verification and synthesis.

While automated RTL rewriting accelerates the design cycle, maintaining structural integrity requires established verification pipelines.

OpenAI, founded in 2015 as a nonprofit with Elon Musk and Sam Altman as co-chairs, converted to a public benefit corporation in 2025 after a for-profit subsidiary created in 2019 was restructured. The release of ChatGPT in November 2022 accelerated the generative AI boom, making it the fifth-most-visited website globally by September 2026.

Always consult with certified IT and cybersecurity professionals before altering enterprise networks or handling sensitive data.

More on this story: TensorWave to Supply OpenAI 457MW AMD Compute Capacity

Share this:

  • Share on Facebook (Opens in new window) Facebook
  • Share on X (Opens in new window) X

Related reading

  • Norwegian Households, Firms See Inflation Far Above Target, Survey Shows
  • Moomoo: Wall Street Falls 2%; Nvidia Buy Bias Drops, Apple, TSMC Attract Buyers

Related

Search:

World Today News

World Today News is your trusted source for global journalism — breaking headlines, in-depth analysis, and reporting from around the world.

Quick Links

  • Privacy Policy
  • About Us
  • Accessibility statement
  • California Privacy Notice (CCPA/CPRA)
  • Contact
  • Cookie Policy
  • Disclaimer
  • DMCA Policy
  • Do not sell my info
  • EDITORIAL TEAM
  • Terms & Conditions

Browse by Location

  • GB
  • NZ
  • US

Connect With Us

© 2026 World Today News. All rights reserved. Your trusted global news source directory.
For contact, advertising, copyright, issues email: [email protected]

Privacy Policy Terms of Service