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

🔍 Read the full analysis: What’s Behind OpenAI’s Claim Of AI Gains In AMD Chip Designs? on ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on monitors, keyboards and dev gear

  • Business-only prices and quantity discounts
  • Tax-exempt purchasing
  • Multiple users, one account, clear invoices
As an affiliate, we earn on qualifying purchases.

TL;DR

OpenAI said on November 18 that its models rewrote parts of AMD chip designs, reporting up to 14% improvements in power, performance and area on a flagship GPU design. The figures come from OpenAI, which has a financial relationship with AMD, and the companies have not published an independent audit or enough detail to determine how broadly the best result applies.

OpenAI said November 18 that its models rewrote parts of AMD chip designs and produced up to 14% improvements in power, performance and area, or PPA, on a flagship GPU design. The claim points to AI being used to optimize hardware for future AI systems, but the headline figure is company-reported, not independently audited, and OpenAI has a financial relationship with AMD.

OpenAI described the work in a post titled “Advancing chip design with OpenAI models.” The company said its models proposed changes to register-transfer-level code, known as RTL, which specifies how chip circuits behave. According to accounts of the post by ANI and other outlets, conventional Synopsys design tools checked that the rewritten code preserved the original behavior and assessed its power, performance and area. The models proposed the changes; the verification and measurement were performed with engineering tools.

OpenAI’s GPU result is an “up to” figure compared with what it described as heavily optimized baselines. The company said some model-generated changes were included in a recent AMD design sent for manufacturing, while others went into production design files for future versions. It did not identify the blocks involved, say how much of the chip they represent, or report an average result across the design.

OpenAI separately said AMD had verified single-digit PPA gains in key blocks of its next flagship CPU. The company also reported area reductions of about 13% in one unit and 10% in another on its own Jalapeño inference chip; accounts differ on the name of one unit. OpenAI said those changes were validated after the first chips returned from manufacturing. These are distinct results and should not be treated as one overall improvement across AMD products.

At a glance
reportWhen: Announced November 18, 2026; reported N…
The developmentOpenAI reported that its models made power, performance and area gains in parts of AMD chip designs by proposing alternative RTL code.

AI Tools Reach Chip Design

The report matters because chip design is one of the expensive, time-consuming stages in building the hardware used to train and run AI systems. If model-generated RTL changes can reliably improve a design, they could help chip engineers explore more alternatives before manufacturing. OpenAI characterized the AMD work as roughly a year of progress in less than five months; that is the company’s description of the work, not an independently verified measure of the time saved.

The practical effect is not yet clear. A local improvement in a particular block does not necessarily produce the same gain across an entire processor. Chip teams must balance power, performance and area, and a result reported at one design stage may change after later steps such as placing and wiring the circuits. Whether these methods deliver repeatable gains in finished products remains an open question.

The work also shows a potential feedback loop: AI tools may help design the chips used to run future AI systems. That could be commercially valuable to both companies. OpenAI has a warrant for up to 160 million AMD shares, issued under an arrangement announced in October 2025, that vests as AMD chips are deployed. This relationship does not disprove the technical results, but it makes independent evidence and clear methods especially relevant.

Products Worth Considering

Kindle Paperwhite Signature Edition 32GB (2024 Model) - Metallic Black

Kindle Paperwhite Signature Edition 32GB (2024 Model) – Metallic Black

Logitech MX Creative Console with 9 Customizable Keys, Graphite

Logitech MX Creative Console with 9 Customizable Keys, Graphite

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 what a circuit does, cycle by cycle. Synthesis software translates that description into logic gates; later design steps place and connect those gates to produce a layout for manufacturing. Changing the RTL can offer a different way to implement the same function, but a rewrite must preserve behavior to be useful.

OpenAI said Synopsys tools checked the proposed changes against the original designs and evaluated the PPA trade-offs. The source accounts describe this as a tool-assisted workflow, not a case of a model independently proving that its own design was correct. Inside AI reported examples on Jalapeño in which a redesigned multiplier used about 15% less cell area in synthesis and reduced latency from two clock cycles to one, while an adder redesign cut cell area by about 15% at the same latency. Those examples concern OpenAI’s chip, not the unnamed AMD GPU blocks behind the 14% figure.

OpenAI also reported testing five models on two open-source processor cores, PicoRV32 and Ibex, with 9,881 scored runs. The source says GPT-6.2 Astra scored highest, but this benchmark is separate from the AMD results and does not establish gains on a commercial GPU. OpenAI’s post was published November 18; the reported figures were relayed by outlets including ANI, The Decoder and Inside AI.

The Missing Measurement Details

The public accounts do not identify which AMD GPU blocks produced the up-to-14% result, what portion of the chip they represent, or whether the measurement came after synthesis or after later physical-design steps. They also do not provide the average result across the tested changes or a breakdown of the individual power, performance and area trade-offs. Without those details, readers cannot tell how representative the top figure is.

OpenAI is the source for the AMD figures in the material available so far. AMD’s reported verification of single-digit gains in CPU blocks is relayed through OpenAI; the source says AMD has not published its own figures. No independent audit of the results is described. The source material also says OpenAI’s automated request to its announcement post was refused, so the detailed accounts rely on reporting by other outlets.

It remains unclear how much of the design work was accepted into production, whether manufacturing will confirm the reported gains, and whether similar results can be reproduced across other chips and design teams. OpenAI says some changes are in a design sent for manufacturing, but the available information does not establish what final products will contain them or when those products might be available.

Manufacturing and Independent Evidence

The next meaningful evidence will come from fuller technical disclosures and results from designs that proceed through manufacturing. OpenAI says some model-generated changes were included in an AMD design sent for manufacturing; the source material does not give a release date or say when measurements from finished chips will be available. AMD could also publish its own account of the work, including the blocks tested and the measurement conditions.

For now, the reported findings are best read as an early account of a specific engineering workflow, not proof that AI has broadly shortened chip development. More detail about the baseline, design stage and tested blocks, plus independent confirmation, would help establish whether the reported gains hold beyond selected cases.

Key Questions

What did OpenAI say its models improved?

OpenAI reported up to 14% PPA improvement on parts of an AMD flagship GPU design. It also reported single-digit gains in key blocks of an AMD next-generation CPU, relayed as AMD-verified by OpenAI.

Does the 14% figure apply to the whole GPU?

The available accounts do not say that. OpenAI described an “up to” result against heavily optimized baselines, without identifying the blocks, their share of the chip, or an average across the design.

How were the proposed chip changes checked?

OpenAI said Synopsys tools checked that rewritten RTL preserved the original behavior and assessed power, performance and area. The source material does not describe an independent audit of the results.

Has AMD confirmed the headline result independently?

The source material says AMD’s chief executive praised the results, and OpenAI reported that AMD verified single-digit gains in CPU blocks. It also says AMD has not published its own figures for the results, and no independent audit is reported.

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, according to the source material. That commercial link does not settle whether the technical claims are correct, but it makes transparent methods and independent verification useful for evaluating them.

Source: ThorstenMeyerAI.com

You May Also Like

The AI Defense Takeaways From Binance’s Brief, Costly Crisis

A reported $6.36 billion Binance theft unfolded in minutes, raising questions about AI defense speed, coordination and human control.

Why DuckDB 2.0 Is Faster

DuckDB 2.0 alpha separates S3 downloads from query workers and rewrites recursive CTE execution. Early speed tests are limited to one machine.

Outlook Zero-Day Under Attack: The Patches To Prioritize Now

Microsoft and CISA say Outlook flaw CVE-2026-124194 is under active attack. See affected Office builds, the federal deadline and what remains unknown.

Telegram’s Android Call Flaw: What We Know, What To Consider

NebuSec says Telegram Android calls could be chained with two other bugs to compromise a phone. Telegram fixed the reported flaw in version 12.10.4.