🔍 Read the full analysis: AI Models In Chip Design: What OpenAI Says Changed For AMD on ThorstenMeyerAI.com
Get smart everyday buys delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
OpenAI said on Nov. 18 that its models rewrote parts of AMD chip designs, reporting up to 14% improvements in power, performance and area for a flagship GPU against heavily optimized baselines. The company also reported single-digit gains in selected CPU blocks, but AMD has not published its own figures and independent verification is unavailable.
OpenAI said on Nov. 18 that its models rewrote parts of AMD chip designs, reporting up to 14% improvement in power, performance and area for a flagship graphics processor. OpenAI said some model-generated changes were included in an AMD design sent for manufacturing, but the figures have not been independently audited and AMD has not published its own detailed results.
The results appeared in an OpenAI post titled Advancing chip design with OpenAI models. The company described changes to the register-transfer-level (RTL) code that specifies how chip circuits behave. According to OpenAI, as reported by ANI and other outlets, its models suggested alternative implementations and conventional Synopsys design tools checked whether the rewrites preserved the original behavior and measured power, performance and area, or PPA.
For AMD’s flagship GPU design, OpenAI reported “up to” 14% better PPA compared with “heavily optimised baselines.” It did not identify the specific blocks behind that best result, how much of the chip they represent or the average gain across the design. OpenAI said some proposed changes had been incorporated into a design AMD sent for manufacturing, while others went into production design files for future versions.
OpenAI also said AMD had verified single-digit percentage gains in PPA for key blocks of its next flagship CPU. The company separately reported that model-proposed changes reduced standard-cell area in its own Jalapeño inference chip by about 13% in one unit and 10% in another. Reports differ on the first unit’s name: ANI called it a vector unit, while Inside AI described it as a warp engine. OpenAI said the area reductions were validated after the first chips returned from manufacturing.
AI Rewrites Reach AMD Chip Designs
The announcement matters because it describes AI being used to improve the hardware that runs AI systems. Chip design involves repeated trade-offs among performance, power use and physical area; even improvements in selected blocks could matter if they survive later design checks and manufacturing. OpenAI said its models produced about a year of design progress in less than five months, a characterization that refers to its work with AMD rather than an independently measured industry-wide acceleration.
The claimed results could affect future chips only if the changes work as intended in production. OpenAI said some GPU changes were in a design sent for manufacturing, but it did not disclose which changes, their share of the chip or the resulting product-level effect. A gain in a specific circuit block does not by itself establish a comparable improvement in a complete processor or in a customer’s computing costs.
The companies’ financial relationship is another reason to read the claims carefully. Since October 2025, OpenAI has held a warrant for up to 160 million AMD shares, which vests as AMD chips are deployed. That relationship does not establish that the reported engineering results are inaccurate, but it is relevant when evaluating results presented by OpenAI without a published independent audit.
Products Worth Considering

Samsung 32-Inch ViewFinity S7 4K UHD IPS Monitor

Logitech MX Creative Console with 9 Customizable Keys, Graphite

Canon SELPHY QX20 Compact Photo Printer – Dark Gray
As an affiliate, we earn on qualifying purchases.
From RTL Code to Silicon
The reported work focused on the RTL stage of chip design. Engineers use RTL code to describe a circuit’s behavior; synthesis tools then translate that description into logic gates, followed by steps that place and connect the components on silicon. A model can propose a different implementation of the same function, but the proposal still has to pass engineering checks and fit the later physical design process.
OpenAI said Synopsys tools checked that rewritten designs preserved the originals’ behavior and assessed their PPA trade-offs. This verification was performed by conventional design tools, not by the model alone. That distinction matters because a functionally incorrect change can make a chip unusable, and errors discovered late can require costly redesign or another manufacturing run.
OpenAI president Greg Brockman summarized the AMD work on X as “roughly an additional year of progress on @AMD’s chip achieved in less than 5 months using our models,” according to Stocktwits. OpenAI also tested five models on two open-source processor cores, PicoRV32 and Ibex, reporting 9,881 scored runs in total. That benchmark is separate from the AMD chip results; GPT-6.2 Astra, released the day before the announcement, scored highest in the reported test.
“Su praised the results on X.”
— AMD chief executive Lisa Su, as quoted by Stocktwits
The Missing Chip-Level Evidence
Key details needed to judge the headline figures remain undisclosed. OpenAI did not name the GPU blocks behind the up-to-14% result, state their share of the processor or provide an average across the design. The reports also do not specify whether the gains were measured after synthesis or after placement and routing, later steps that can change the final trade-offs.
AMD has not published its own detailed measurements, and no independent audit of the results has been reported. The source material does not identify which model-generated changes went into the manufacturing design or say whether the reported improvements have been confirmed in manufactured AMD chips. OpenAI’s financial stake tied to AMD chip deployments also forms part of the disclosure picture.
The 14% figure is a best-case claim, not evidence that the full GPU became 14% better. The CPU claim is described only as single-digit gains in selected blocks. OpenAI’s Jalapeño area results concern its own chip and should not be treated as confirmation of the AMD results.
Manufacturing Results and AMD Data
The next useful evidence would be detailed results from AMD or an independent reviewer, including the affected blocks, measurement stage, baseline and whether the gains held in manufactured chips. OpenAI said some changes were already in a design sent for manufacturing, but it did not give a date for the product or for public performance and power data.
For now, the announcement establishes that OpenAI says its models proposed and checked RTL changes used in AMD design work. It does not settle how broadly those changes improve AMD products, or whether the claimed gains persist through production. Further company disclosures and results from shipped hardware will determine what can be confirmed beyond the reported design-stage measurements.
Key Questions
What did OpenAI say its models changed for AMD?
OpenAI said its models rewrote parts of AMD chip designs at the RTL level. The company reported up to 14% better PPA in parts of a flagship GPU design and single-digit gains in selected blocks of a next-generation CPU.
Does the 14% figure apply to the entire GPU?
No. OpenAI described it as an “up to” result against heavily optimized baselines. It did not identify the blocks, their share of the chip or the average gain across the full GPU.
Have the AMD results been independently verified?
No independent audit has been reported. OpenAI said Synopsys tools checked that the rewrites preserved circuit behavior and assessed PPA, while AMD has not published detailed figures of its own.
Were any changes used in a manufacturing design?
OpenAI said some model-generated changes were included in an AMD design sent for manufacturing. It has not disclosed which changes or reported whether the gains were confirmed in manufactured chips.
Why does OpenAI’s relationship with AMD matter?
OpenAI holds a warrant for up to 160 million AMD shares that vests as AMD chips are deployed. That financial tie is relevant context for evaluating company-reported results, but it does not by itself show that the findings are wrong.
Source: ThorstenMeyerAI.com
