🔍 Read the full analysis: What OpenAI’s 14% Figure Means For AI-Assisted AMD Chip Design 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 said on November 18 that its models rewrote parts of AMD chip designs, reporting up to 14% improvement in power, performance and area against heavily optimised baselines. The figure is a best case, not a disclosed average, and neither the affected blocks nor the measurement stage have been made public. AMD reportedly verified single-digit gains in key CPU blocks, but no independent audit or AMD-published figures are available.
OpenAI said November 18 that its models rewrote parts of AMD’s flagship graphics-processor design and achieved up to 14% improvement in power, performance and area, or PPA. The company said some generated changes were included in a design AMD sent for manufacturing, but the figure is a reported best case: OpenAI has not disclosed which chip blocks produced it, the average result, or the precise measurement stage.
The work focused on register-transfer level (RTL) code, which describes how a chip’s circuit blocks behave. OpenAI’s models proposed alternative RTL implementations, while Synopsys design tools checked that the rewritten versions preserved the original behavior and evaluated their PPA trade-offs, according to OpenAI as reported by ANI and other outlets. The models proposed the changes; conventional verification tools, rather than the models themselves, checked their functional equivalence.
OpenAI described the GPU comparison as a gain against heavily optimised baselines—existing designs that had already received substantial engineering work. It said some model-generated changes were in a recent design sent for manufacturing, while others went into AMD’s production design files for future versions. The company did not identify the affected blocks or specify how much of the overall chip they represent.
OpenAI also reported separate results. It said AMD had verified single-digit PPA gains in key blocks of its next flagship CPU. On OpenAI’s own Jalapeño inference chip, the company said proposed changes reduced standard-cell area by about 13% in one unit and 10% in another, with the results validated after the first chips came back. Reports differ on the label for the first unit, calling it a warp engine or vector unit. These figures describe different designs and measures; they are not the 14% AMD GPU result.
AI Reaches Chip Logic Design
If the reported results hold up, AI-assisted design could help chip engineers test more ways of implementing a circuit without changing what that circuit is meant to do. Even limited improvements in power, performance or area can matter: they affect how quickly a processor runs, how much energy it uses, and how much silicon it occupies. Those trade-offs shape chip capability and manufacturing cost.
The manufacturing claim gives the announcement more weight than a result confined to a software simulation: OpenAI says some changes reached a design sent to a fabrication partner. But that does not establish that the manufactured chip has delivered the reported PPA gain in final silicon. The public information lacks block-level results, an independent assessment and a clear account of how the improvements were measured.
The report also matters because it describes AI being used to improve hardware that can run future AI systems. OpenAI characterized the work as “roughly a year of progress achieved in less than 5 months.” That is the company’s comparison, not an independently measured development timeline. The result could point to a useful engineering workflow, but one company’s best-case figure cannot establish how broadly or reliably the approach will work.
Products Worth Considering

Samsung 27-Inch M5 (M50F) FHD Smart Monitor

Dell XPS Tower 8960 Desktop, Intel Core i7-13700, NVIDIA GeForce RTX 4060, 16GB DDR5, 512GB SSD

FLEXISPOT EN1 Electric Standing Desk, 48 x 24 in, Black
As an affiliate, we earn on qualifying purchases.
What the 14% Measures
OpenAI published a post titled Advancing chip design with OpenAI models on November 18. The source account says its automated request to the post was refused, so the reported figures and methods here draw on coverage by ANI, The Decoder and Inside AI, which agree on the main claims. OpenAI’s president, Greg Brockman, summarized the work on X as “roughly an additional year of progress on @AMD’s chip achieved in less than 5 months using our models,” as quoted by Stocktwits.
At the RTL stage, engineers write code describing a circuit’s operation cycle by cycle. A model can propose different logic, divide a calculation differently across pipeline stages, or reduce shared hardware or intermediate storage. Synthesis tools then convert RTL into logic gates; further design steps place and connect those gates on silicon. A gain measured after synthesis may not carry through unchanged to the completed layout or manufactured chip.
Reporting on Jalapeño gives examples of the kind of change involved. Inside AI described a multiplier redesign that used about 15% less cell area in synthesis and reduced latency from two clock cycles to one, and an adder redesign that cut cell area by about 15% while keeping the same two-cycle latency. These are examples from OpenAI’s own chip, not evidence that AMD’s GPU achieved the same results.
OpenAI also tested five models on two open-source RISC-V processor cores, PicoRV32 and Ibex, in 9,881 scored runs, according to Inside AI. GPT-6.2 Astra scored highest in that test, which the source says was released the day before the announcement. Those benchmark results are separate from the claimed AMD design gains.
OpenAI and AMD also have a financial relationship: 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 connection does not disprove the engineering claims, but it is relevant when weighing figures that have not been independently audited. The source also disclosed that its article was made with AI assistance, including Anthropic’s Claude, and said the companies involved neither reviewed nor sponsored it.
“Praised the results on X.”
— AMD chief executive Lisa Su, as quoted by Stocktwits
The Missing Measurement Details
The key unknown is what the “up to 14%” figure represents in practice. OpenAI has not named the GPU blocks involved, said what share of the full design they occupy, disclosed an average across changes, or specified whether the result was measured after synthesis or after placement and wiring. “Up to” signals a maximum reported outcome, not a typical gain across the chip.
The available account also does not include an independent audit of the AMD results or AMD’s own published measurements. It is not clear whether the reported improvements persist in the final manufactured parts, how they affect power, performance and area individually, or whether a gain in one measure came at a cost in another. The report says a design containing some changes was sent for manufacturing; it does not provide final silicon test data.
OpenAI’s commercial connection to AMD is a further reason to distinguish the company’s claims from independently confirmed results. The disclosed warrant does not establish that the reported results are wrong. It does mean readers should look for AMD’s own technical disclosures and third-party validation before treating the headline figure as a general result.
Awaiting AMD and Silicon Results
The next useful evidence would be AMD’s own technical account: the blocks changed, the baselines and measurement stage, and the separate effects on power, performance and area. Results from completed silicon testing could show whether the reported design-stage gains carried through manufacturing. Neither an AMD technical publication nor independent validation is identified in the source material.
OpenAI says some changes are already in a design sent for manufacturing and others are in production design files for future versions. The public information does not give a release or testing timetable. Until those details emerge, the 14% figure should be read as an OpenAI-reported best case for parts of a design—not as a confirmed improvement across AMD’s entire GPU or a guarantee of results in future chips.
Key Questions
What does OpenAI’s 14% figure refer to?
OpenAI reported up to 14% improvement in power, performance and area for parts of AMD’s flagship GPU design, compared with what it called heavily optimised baselines. It has not published an average or said which blocks produced the best result.
Does the figure mean AMD’s whole GPU is 14% better?
No. The report describes a best-case result for parts of the design, not a 14% improvement across the complete GPU. The affected blocks and their share of the chip have not been disclosed.
Did OpenAI’s models verify the chip changes?
According to OpenAI, the models proposed RTL rewrites and Synopsys tools checked functional behavior and evaluated PPA. The source does not say the models themselves verified the changes.
Has AMD confirmed the results independently?
OpenAI says AMD verified single-digit gains in key blocks of a next flagship CPU and that some GPU changes reached a design sent for manufacturing. The source material says AMD has not published its own figures, and no independent audit is available.
What evidence would clarify the claim?
AMD’s block-level measurements, details on the baseline and design stage, and tests of manufactured silicon would help establish how broad and durable the reported gains are. Independent review would also help assess the results.
Source: ThorstenMeyerAI.com
