AIThis post was created with the assistance of artificial intelligence (AI).

TL;DR

OpenAI said on November 18 that its models rewrote parts of AMD chip designs and produced up to 14% improvements in power, performance and area against heavily optimized baselines. The figure is a reported best case; OpenAI has not named the affected blocks or published an average, and AMD has not released its own figures.

OpenAI said November 18 that its AI models rewrote parts of circuit designs for AMD processors, reporting up to 14% improvement in power, performance and area on a flagship GPU design. The companies have not published block-level results or an independent audit, so the headline figure remains OpenAI’s account of a best case, rather than a result outside researchers can assess.

Before you orderOffer from Amazon

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
Start your free Prime trial Free trial for eligible customers · Cancel anytime
As an affiliate, we earn on qualifying purchases.

OpenAI described using models to rewrite SystemVerilog register-transfer-level code, which specifies how chip circuits behave. According to the company’s account, engineers set a target of improving power, performance or area while preserving behavior; Synopsys design tools then checked whether a proposed rewrite behaved like the original and assessed its design trade-offs. The reports do not provide details that would allow outsiders to reproduce the AMD results.

For AMD’s flagship GPU design, OpenAI reported gains of up to 14% against “heavily optimised baselines.” It did not name the blocks involved, give an average across the design or specify how each of the three measures changed. OpenAI also said AMD had included some model-generated changes in a design sent for manufacturing, with other changes added to production design files for future versions.

OpenAI separately said AMD had verified single-digit percentage gains in power, performance and area in key blocks of its next flagship CPU. That claim, too, comes through OpenAI’s description: AMD has not published figures of its own. AMD chief executive Lisa Su praised the results on X, but the source material does not include a detailed technical statement from her.

At a glance
reportWhen: Announced November 18, 2026; independen…
The developmentOpenAI reported that its models made power, performance and area improvements to parts of AMD chip designs, with a best-case gain of up to 14%.
Chart titled Four claims, one source. AMD flagship GPU design, up to 14% better power, performance and area: stated by OpenAI, no AMD figure, no independent check. AMD next flagship CPU, single-digit percent gains in key blocks: stated by OpenAI, verified by AMD according to OpenAI, no independent check. OpenAI's Jalapeno chip, about 13% and 10% less cell area: stated by OpenAI, validated at silicon bring-up according to OpenAI, no independent check. RISC-V test on two public cores, 9,881 scored runs: stated by OpenAI, public cores, no independent check. Highlighted: no figure has an independent check yet
Who stated each claim and who checked it. Source: OpenAI, November 18, 2026, as reported by ANI and Inside AI; Lisa Su on X, as quoted by Stocktwits. Chart: Thorsten Meyer AI.

AI Reaches Chip Design

If the reported gains hold up, model-assisted chip design could help engineers improve hardware that powers later AI systems. A chip can trade among power use, speed and physical area; a change that improves one measure may affect the others. OpenAI’s announcement suggests models can propose useful circuit changes within that engineering process, while verification tools check whether the rewritten code preserves the intended behavior.

The practical scope is still uncertain. A best-case gain in selected blocks does not establish a similar improvement across an entire processor, or show how the changes affect a finished product’s cost, energy use or real-world performance. The claim that the work achieved “roughly a year of progress” in less than five months is OpenAI’s characterization, not a separately reported industry measure.

There is also a financial connection relevant to how readers assess the announcement. Since October 2025, OpenAI has held a warrant for up to 160 million AMD shares, which vests as AMD chips are deployed, according to the source material. That relationship does not establish that the reported gains are inaccurate, but it makes independent data especially useful.

Products Worth Considering

As an affiliate, we earn on qualifying purchases.

From Code to Silicon

The reported work takes place at the RTL stage of chip design. Engineers describe a chip block’s behavior in code; synthesis tools convert that description into logic gates, followed by later steps that place and connect components on silicon. OpenAI says its models proposed alternative implementations of the same functions, and Synopsys tools checked the proposed changes.

The announcement also included results from OpenAI’s Jalapeño inference chip. OpenAI said model-proposed changes reduced standard-cell area by about 13% in one unit and 10% in another, with the results validated after the first chips returned. Reports differ on the name of the first unit, calling it either a warp engine or a vector unit. The available material does not resolve that discrepancy.

OpenAI also reported testing five of its models on two open-source RISC-V processor cores, PicoRV32 and Ibex, across 9,881 scored runs. It said GPT-6.2 Astra, released the previous day, scored highest. These benchmark results are separate from the AMD GPU claim and do not independently verify it.

The Missing Block-Level Data

The public account does not identify which AMD GPU blocks produced the up to 14% result, show the distribution of gains or state the average across the design. It also does not give enough detail about the baselines, test conditions or measurement methodology for an outside party to repeat the comparison.

OpenAI says some changes entered a design sent for manufacturing, but the source material does not identify that design, say when its manufacturing results will be available or establish what performance the finished chip will deliver. AMD’s reported CPU verification is likewise described only as single-digit gains in key blocks. No independent audit of either company’s reported results has been published.

Evidence From Manufactured Chips

The next useful evidence would include AMD’s own technical figures, identification of the affected design blocks, and measurements showing how the changes performed in manufactured chips. Further disclosure could clarify whether the reported gains apply to selected circuits or translate into measurable improvements at the processor level.

For now, readers can distinguish the separate claims: OpenAI reports a GPU best case of up to 14%, says AMD verified single-digit gains in key CPU blocks, and reports area reductions on its own Jalapeño chip. The size and broader effect of the AMD improvements remain unverified in the material available for this report.

Key Questions

What does the reported 14% gain refer to?

It is OpenAI’s reported best-case improvement in power, performance and area on parts of AMD’s flagship GPU design, compared with what OpenAI called heavily optimized baselines. The company has not published an average or named the blocks.

Did AMD independently publish the results?

The supplied reporting says AMD has not published its own figures. OpenAI says AMD verified single-digit gains in key blocks of a next flagship CPU, and Lisa Su praised the results on X.

Were the AI-generated chip changes checked?

OpenAI says Synopsys design tools checked whether rewrites behaved like the original and measured design trade-offs. No independent audit of the reported AMD results has been published.

Have any changes reached manufacturing?

OpenAI says some model-generated changes were included in an AMD design sent for manufacturing. The supplied information does not identify the design or provide measurements from manufactured chips.

Source: Thorsten Meyer AI

You May Also Like

178 Major Sites Were Hit By Malware—What AI Users Should Review

Report URI says malicious JavaScript reached visitors on 178 organizations’ sites through Optimizely. The company says it fixed the vulnerability.

Data Center Power For AI: Why DC And Batteries Are In The Mix

Vantage reserved transformer capacity and NiSource signed a Microsoft power deal, highlighting how AI data centers are changing power plans.

Pasqal Reports Third-Quarter Revenue Decline As QPU Services Revenue Grows

Pasqal reported €3.3 million in Q3 2026 revenue, down from €5.7 million, while QPU-related services revenue rose 53% year over year.

How Far Has Anthropic’s AI Coding Team Come? A Top Engineer Explains

A Business Insider headline says an Anthropic engineer discussed her team’s AI coding use. The article’s details and evidence are unavailable.