How OpenAI Evaluated AI Gains In AMD Chip Designs
AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: How OpenAI Evaluated AI Gains In AMD Chip Designs on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get hardware and tech essentials delivered free with Prime

  • Fast, free delivery on millions of items
  • Prime Video, Amazon Music and more included
  • Member-only deals all year
Start your free Prime trial Free trial for eligible customers · Cancel anytime
As an affiliate, we earn on qualifying purchases.

TL;DR

OpenAI says its models rewrote portions of AMD chip designs and found up to 14% improvements in power, performance and area against already optimized baselines. The company says some changes were included in a design sent for manufacturing, but AMD has not published its own figures and no independent audit is available.

OpenAI said November 18 that its models rewrote parts of AMD chip designs, reporting up to 14% gains in power, performance and area on AMD’s flagship GPU design. The company said some model-generated changes were included in a design sent for manufacturing, but the reported figures have not been independently audited and AMD has not published its own detailed results.

The work targeted register-transfer-level (RTL) code, which describes how a chip’s circuits perform their functions. OpenAI said its models proposed alternative implementations, while Synopsys design tools checked whether the changes preserved the original behavior and assessed their effects on power, performance and area, commonly shortened to PPA. The models generated the rewrites; conventional verification and design tools evaluated them.

OpenAI’s reported 14% figure is a best case, measured against what it called heavily optimized AMD baselines. The company did not identify which GPU blocks produced that result, how much of the chip they represent, or whether the figure came before or after physical placement and wiring. OpenAI also said AMD had verified single-digit PPA gains in key blocks of a next-generation flagship CPU, but AMD has not released supporting measurements.

The announcement included results beyond AMD designs. OpenAI said changes to its own Jalapeño inference chip reduced standard-cell area by about 13% in one unit and 10% in another, with the first chips returned from manufacturing used to validate the changes. Reports differ on the name of the first unit, calling it either a warp engine or a vector unit. OpenAI also tested five models on the public PicoRV32 and Ibex processor cores, totaling 9,881 scored runs; according to Inside AI, GPT-6.2 Astra scored highest.

At a glance
reportWhen: OpenAI published the results November 1…
The developmentOpenAI reported that its models helped improve parts of AMD chip designs, including a best-case PPA gain of up to 14% on a flagship GPU design.
Crypto market snapshot
Fear & Greed Index
10/100 — Extreme Fear
Bitcoin BTC$62,868▼ 0.4%
Ethereum ETH$1,778▲ 0.5%
Tether USDT$0.9992▼ 0.0%
BNB BNB$586.59▲ 0.5%
USDC USDC$0.9996▼ 0.0%
XRP XRP$0.9735▼ 0.7%
Solana SOL$74.14▲ 0.4%
TRON TRX$0.3106▲ 0.6%
Live data · CoinGecko · alternative.me (24h change)

AI Tools Reach Chip Design

The results suggest that AI models may help engineers search a design space that is costly and labor-intensive to explore manually. A verified improvement in a chip’s area, speed or power consumption could affect manufacturing cost, computing capacity and the electricity needed to run data centers. OpenAI’s claim that some revisions entered a design sent for manufacturing makes the announcement more consequential than a software-only benchmark, although it does not establish the size of any real-world benefit.

The financial relationship between the companies also matters when weighing the claims. OpenAI holds a warrant for up to 160 million AMD shares, issued in October 2025, that vests as AMD chips are deployed. That connection does not show that the results are inaccurate, but it gives readers reason to distinguish OpenAI’s account from independently verified performance data. For chip customers, investors and researchers, the missing block-level measurements and outside audit limit how confidently the reported gains can be generalized.

Products Worth Considering

As an affiliate, we earn on qualifying purchases.

From RTL Rewrites to Silicon

Chip design proceeds through multiple stages. At the RTL stage, engineers describe circuit behavior in code; synthesis software converts that description into logic, followed by steps that place and connect components on a physical chip. OpenAI’s account concerns rewriting RTL, not having a model independently design and manufacture a finished processor. The distinction matters because an apparent improvement at one stage may change or disappear as the design moves toward physical implementation.

OpenAI said Synopsys tools checked that proposed changes preserved behavior and measured PPA effects. Inside AI reported examples from Jalapeño in which a redesigned multiplier used about 15% less cell area in synthesis while reducing latency from two clock cycles to one, and a redesigned adder cut area by about 15% at the same latency. Those are examples for OpenAI’s own chip, not evidence that every AMD block achieved similar results.

OpenAI president Greg Brockman described the AMD work as “roughly an additional year of progress” achieved in less than five months, according to Stocktwits. OpenAI’s broader announcement was published under the title “Advancing chip design with OpenAI models.” The reported results therefore combine company-level claims about AMD products, measurements on OpenAI’s own chip, and benchmark tests on open-source processor designs; these are related but distinct forms of evidence.

“Su praised the results on X.”

— AMD chief executive Lisa Su, as quoted by Stocktwits

Missing Independent Performance Data

No independent audit of the AMD chip-design results has been published in the supplied reporting. OpenAI’s “up to” figure does not reveal a typical result, and the company has not named the GPU blocks, their share of the overall design, or the measurement stage. Those omissions make it difficult to determine how much of the reported improvement would remain in a finished chip or apply to other designs.

AMD’s reported CPU verification is also described through OpenAI, rather than through a separate AMD technical release with figures and methodology. The source material does not provide details such as the exact test conditions, the number of blocks assessed, or how power, performance and area were balanced. It is also unclear how many of the GPU changes sent for manufacturing will be retained in production silicon, or what results will emerge once that design is tested.

The relationship between OpenAI and AMD is an additional point for readers to keep in view: OpenAI’s warrant can vest as AMD chips are deployed. That financial arrangement is confirmed in the supplied material, but it does not by itself resolve whether the reported engineering gains are representative. The underlying methodology and outcomes remain subject to further disclosure.

Production Results and Further Disclosure

The next meaningful evidence would be AMD’s own technical data or independent measurements showing which changes reached production silicon and how the finished designs perform. As the design sent for manufacturing proceeds, physical-chip results could clarify whether the reported RTL-stage improvements translate into measurable gains in power, speed or area. No specific date for such results is provided in the source material.

Readers should also watch for fuller methodology from OpenAI or AMD, including the blocks tested, baseline definitions, measurement stages and variation across results rather than only the best case. Until those details are available, the 14% figure should be read as OpenAI’s reported maximum, not as a typical improvement across an AMD processor or a verified outcome for chips already in use.

Key Questions

What did OpenAI say its models changed?

OpenAI said its models rewrote portions of RTL chip-design code for AMD products and its own Jalapeño inference chip. Synopsys tools, rather than the models alone, checked behavior and evaluated reported PPA effects.

Does AMD’s flagship GPU have a confirmed 14% improvement?

OpenAI reported up to 14% improvement in power, performance and area against heavily optimized baselines. The figure is a reported best case; the available reporting does not include an independent audit or enough detail to establish a typical chip-wide gain.

Have the changes reached manufacturing?

OpenAI said some model-generated changes were included in an AMD design sent for manufacturing. The source material does not identify which changes will remain in production silicon or give finished-chip performance results.

What remains unknown about the results?

The reported accounts do not identify the specific GPU blocks, their share of the chip, the full test methodology or the design stage at which each gain was measured. Independent verification has not been published.

Source: ThorstenMeyerAI.com

Nothing in this article is financial or investment advice. Cryptocurrency and precious-metal investments carry significant risk — do your own research and consider a licensed advisor.
You May Also Like

According to Yi He, Binance Undertook Over 120 Internal Investigations With Assistance From US Law Coordinators.

Get insights into Binance’s extensive internal investigations and discover how these efforts could reshape the future of the cryptocurrency exchange. What revelations await?

The Continual Learning Research Map: Where the Memento Constraint Stands in May 2026

An update on the research landscape of the Memento Constraint in continual learning, highlighting current approaches, timelines, and remaining challenges.

Inside Frontier Lab’s Innovation: AI And The New Head Of Leasing And Energy

Anthropic reveals major staffing shifts with new leaders in leasing, land, energy, and compute infrastructure, highlighting a focus on capacity over research.