🔍 Read the full analysis: AI For AMD Chip Design: What OpenAI Tested And Left Unmeasured on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
OpenAI says its models rewrote parts of AMD chip designs and produced up to 14% gains in power, performance and area against heavily optimized baselines. The company has not disclosed which GPU blocks produced that best result or an average, and no independent audit has been published.
OpenAI said November 18 that its AI models rewrote parts of AMD chip designs, reporting up to 14% improvement in power, performance and area on an AMD flagship GPU design. The company also said some model-generated changes were included in a design AMD had sent for manufacturing, though the public figures and account have not been independently audited.
The reported GPU result compares changes with heavily optimized design baselines, according to OpenAI. It is a best-case figure: the company did not identify the blocks behind the 14% result, say what share of the processor they represent, or provide an average across the design. OpenAI described the work as roughly a year of progress achieved in less than five months.
OpenAI also said AMD verified single-digit percentage gains in power, performance and area in key blocks of its next flagship CPU. AMD chief executive Lisa Su praised the results on X, according to the source material, but AMD has not published its own performance figures. OpenAI said some GPU changes had entered a design sent for manufacturing, while other changes went into production design files for future versions.
The models worked at the register-transfer level (RTL), rewriting SystemVerilog code that describes chip behavior. OpenAI said Synopsys design tools checked that the proposed versions preserved the original behavior and assessed their power, performance and area trade-offs. Those checks are distinct from the AI model’s proposals: conventional verification tools, rather than the model alone, tested whether the rewrites behaved as intended.
OpenAI reported a separate result on Jalapeño, its custom inference chip: model-proposed changes reduced standard-cell area by about 13% in one unit and 10% in another. The company said the changes were validated after the first chips came back. Reports differ on the name of one unit, calling it a warp engine or a vector unit. OpenAI also tested five models on two open-source RISC-V processor cores, PicoRV32 and Ibex, across 9,881 scored runs; its newly released GPT-6.2 Astra scored highest, according to the cited reporting.
AI Moves Into Chip Engineering
The announcement matters because chip design is part of the infrastructure behind AI systems: improvements to processors could affect the cost, power use and performance of future computing. If model-generated RTL changes prove reliable across complete designs and manufacturing stages, AI could assist engineers with work that normally involves extensive iteration and specialist review.
But the reported numbers do not yet show a broad, independently verified improvement across AMD’s chips. The GPU figure is explicitly “up to” 14%, with no average or named blocks, and the CPU result is described only as single-digit gains in key blocks. The claims support interest in AI-assisted chip design, not a conclusion that AMD processors as a whole will be 14% better or that design cycles have been reduced by a year.
There is also a commercial relationship relevant to evaluating the claims. Since October 2025, OpenAI has held a warrant for up to 160 million AMD shares, which vests as AMD chips are deployed. That tie does not establish that the reported engineering results are inaccurate, but it is a reason to distinguish OpenAI’s account from independent verification.
Popular right now

De'Longhi Magnifica Start Automatic Espresso Machine with Milk Frother, Silver

Breville Barista Express Manual Espresso Machine with Built-In Grinder, Brushed Stainless Steel

Samsung 55-Inch QN70H Neo QLED 4K Smart TV (2026)
As an affiliate, we earn on qualifying purchases.
From RTL Rewrites to Silicon
RTL is an early stage of chip design. Engineers use it to describe how circuit blocks behave; synthesis tools then translate the code into logic gates, followed by physical steps that place and connect those components on silicon. A rewrite can preserve the same function while using a different logic structure or pipeline arrangement, potentially changing area, speed and power.
OpenAI’s report describes a workflow in which models proposed alternative RTL and Synopsys tools checked behavior and measured trade-offs. Inside AI reported examples from Jalapeño, including a multiplier redesign said 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. These are reported examples for OpenAI’s chip, not evidence of the same gains across AMD products.
The announcement also comes amid a direct commercial link: AMD chips are part of OpenAI’s plans, and the share warrant connects its value to AMD deployments. OpenAI published its account on November 18; the available reporting attributes the technical claims to that post, rather than to a separate public AMD results document.
The Missing Measurement Details
OpenAI has not publicly identified the AMD GPU blocks behind the 14% best-case result, their share of the chip, or the average gains across the design. The reports also do not specify whether the figures were measured after synthesis or after physical placement and wiring, later stages that can change the final trade-offs.
No independent audit of the chip-design results has been published, and the source material does not include a detailed AMD technical paper or a dataset that would let outside researchers reproduce the findings. AMD has not issued its own figures for the CPU claims. It is also unclear which proposed changes have completed manufacturing validation and what gains, if any, they deliver in finished products.
The Jalapeño reports use different names for one design unit, and the open-source core benchmark is a separate test from the AMD results. The figures should not be treated as directly comparable. Until more methodology and results are available, the scope and repeatability of the reported gains remain unsettled.
Awaiting AMD and Silicon Results
The next useful evidence would be a technical breakdown identifying the affected blocks, the measurement stage, the baseline and the spread of results—not only the best case. Independent verification, or detailed figures published by AMD itself, would help establish how much of the reported improvement survives later design steps and manufacturing.
Some changes were reportedly included in a design sent for manufacturing, but the announcement does not provide a timetable for public results from that design. OpenAI and AMD may release further information as those chips and future versions progress. For now, the confirmed public development is OpenAI’s report of AI-generated RTL changes and its account of validation—not an independently established performance improvement for AMD products in general.
Key Questions
What did OpenAI say its models changed?
OpenAI said its models rewrote SystemVerilog RTL for parts of AMD chip designs and for its own Jalapeño inference chip. Synopsys tools were reported to check that the rewrites preserved the original behavior and to assess power, performance and area.
Does AMD’s GPU improve by 14% overall?
That has not been established. OpenAI reported up to 14% improvement against heavily optimized baselines, but did not disclose the blocks involved or an average across the GPU.
Did AMD confirm the results independently?
OpenAI said AMD verified single-digit gains in key blocks of a next flagship CPU, and Lisa Su praised the results on X. The source material says AMD has not published its own figures, and no independent audit is available.
Have the changes reached manufactured chips?
OpenAI said some model-generated changes were included in an AMD design sent for manufacturing. It has not provided public results showing the performance of finished AMD products or clarified which changes have completed manufacturing validation.
Why does OpenAI’s relationship with AMD matter?
OpenAI has held a warrant for up to 160 million AMD shares since October 2025, with vesting tied to AMD chip deployments. That financial connection is relevant context when evaluating OpenAI’s account, but it does not by itself show that the technical claims are wrong.
Source: ThorstenMeyerAI.com
