🔍 Read the full analysis: OpenAI’s AMD Chip Design Figure: What The Up To 14% Refers To on ThorstenMeyerAI.com
Get the latest gadgets 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 November 18 that its models rewrote parts of AMD chip designs, reporting up to 14% improvement in power, performance and area on a flagship GPU against heavily optimized baselines. The figure is a reported best case: OpenAI has not identified the blocks or average gains, and no independent audit has been published. AMD confirmed single-digit gains in key blocks of a next-generation CPU, according to OpenAI.
OpenAI said November 18 that its models rewrote parts of designs for AMD’s flagship graphics processor, reporting up to 14% improvement in power, performance and area (PPA) against heavily optimized baselines. The figure describes the best reported result, not a stated average across the chip; OpenAI has not named the design blocks or published an independent audit of the measurements.
OpenAI described work at the register-transfer level (RTL), where engineers write code specifying how a chip’s circuit blocks operate. Its models proposed alternative RTL implementations, while Synopsys design tools, rather than the models themselves, checked whether the rewrites preserved the original behavior and assessed their PPA trade-offs, according to OpenAI as reported by ANI and Inside AI.
OpenAI said AMD had already incorporated some model-generated changes in a design sent for manufacturing, while other changes were added to production design files for future versions. The company also said AMD verified single-digit percentage PPA gains in key blocks of its next flagship CPU. AMD has not published its own detailed figures, according to the supplied reporting.
The 14% claim is one of several distinct results, not a single measurement covering every project. For OpenAI’s own Jalapeño inference chip, the company reported standard-cell area reductions of about 13% in one unit and 10% in another, validated after the first chips returned. Reports differ on the name of the first unit: ANI calls it a vector unit, while Inside AI calls it a warp engine. OpenAI also described tests on two open-source processor cores, PicoRV32 and Ibex, involving 9,881 scored runs across five models.
Why Chip-Level Gains Matter
If the reported changes hold through later design and manufacturing stages, AI-assisted RTL work could help chipmakers improve efficiency or performance without relying only on larger or newer manufacturing processes. Power, performance and area are competing design goals: changing a circuit to improve one can affect the others. A useful rewrite must preserve correct behavior and deliver gains that survive the rest of the engineering process.
The report also points to a feedback loop: AI systems are being used to help design some of the hardware that runs AI workloads. That could matter to companies facing high demand for computing capacity. But the reported up-to figure is not evidence of a typical whole-chip gain, and the available information does not establish how much the changes affect AMD’s shipped products, their cost, or performance in real workloads.
There is also a commercial relationship relevant to how readers assess the claims. Since October 2025, OpenAI has held a warrant for up to 160 million AMD shares that vests as AMD chips are deployed, according to the source material. That arrangement does not by itself determine whether the technical results are accurate, but it makes independent, detailed reporting of methods and outcomes particularly relevant.
Products Worth Considering

Apple MacBook Air 15-inch (2026) with M5 chip, 16GB memory, 512GB SSD — Midnight

DJI Mini 4K Drone Combo with 4K Camera, 3-Axis Gimbal, and Two Batteries

Jackery Solar Generator 1000 V2 with 100W Solar Panel
As an affiliate, we earn on qualifying purchases.
From RTL Rewrites to Silicon
Chip design proceeds through several stages. At RTL, engineers describe circuit behavior in a hardware-design language such as SystemVerilog. Synthesis tools convert that description into logic gates; later processes place and connect those gates before a design is manufactured. The reported OpenAI work concerns the RTL stage, not a claim that its models independently designed or manufactured a complete processor.
OpenAI’s reported method uses models to propose alternative implementations of a function, then conventional tools to test equivalence and estimate trade-offs. Inside AI described examples from Jalapeño, including a multiplier redesign reported to use about 15% less cell area in synthesis while reducing latency from two clock cycles to one, and an adder redesign said to cut area by about 15% at the same latency. Those examples concern particular blocks and should not be conflated with the separate AMD GPU figure.
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. That is a characterization of the pace of the work, not a separately reported measurement of chip performance.
What the 14% Leaves Out
OpenAI has not disclosed which AMD GPU blocks produced the up-to-14% result, how much of the chip those blocks represent, or the average improvement across the design. The available reports also do not specify whether the figure was measured after synthesis or after placement and routing, later steps that can change estimated results. The comparison was against what OpenAI described as heavily optimized baselines, but full baseline data and measurement details have not been published in the supplied material.
The reports attribute the claims principally to OpenAI. They say AMD verified single-digit gains in CPU blocks, but do not provide a company-issued technical report or independent confirmation of the GPU result. It is also unclear which reported changes will appear in products available to customers, when those products might ship, or what their effects would be in real-world workloads. OpenAI’s own Jalapeño figures are separate results and do not independently verify the AMD measurements.
Evidence Needed From AMD
The next useful evidence would be a technical disclosure from AMD or OpenAI identifying the relevant blocks, the measurement stage, the baselines and whether results represent a best case or a broader average. Independent analysis could help establish how the changes perform after implementation and whether they retain their estimated gains in manufactured chips.
OpenAI said some changes had entered a design sent for manufacturing, but the supplied reporting gives no production date or public validation milestone. Readers will be able to judge the practical impact more clearly if AMD publishes specifications or test results for products using the changes. Until then, the 14% should be read as an OpenAI-reported maximum for parts of a design, not a confirmed improvement across an entire GPU.
Key Questions
What does the 14% refer to?
It is OpenAI’s reported best-case improvement in power, performance and area for parts of AMD’s flagship GPU design, compared with heavily optimized baselines. The company has not said that the whole chip improved by 14% or given an average.
Who checked whether the AI-generated circuits worked?
OpenAI said Synopsys design tools checked that proposed RTL changes preserved the original behavior and assessed their effects on power, performance and area. The account does not say that the model itself verified correctness.
Has AMD independently confirmed the GPU figure?
The supplied reports say AMD verified single-digit gains in key blocks of a next flagship CPU, according to OpenAI. They do not provide an AMD-issued technical breakdown or independent confirmation of the GPU’s up-to-14% result.
Are the changes already in manufactured AMD chips?
OpenAI said some changes were included in a design sent for manufacturing. That does not establish that finished products containing them are already shipping or show what gains customers will see.
Does the 14% cover OpenAI’s Jalapeño chip too?
No. OpenAI reported separate Jalapeño results: standard-cell area reductions of about 13% in one unit and 10% in another. Those figures concern OpenAI’s own chip and should not be treated as the AMD GPU result.
Source: ThorstenMeyerAI.com
