OpenAI report says its models rewrote AMD GPU RTL, moves into manufacturing
11/19/2026 — 11/21, 18:17·8 sources (all)·9 reports (all)·In the 2026-11-19 briefing
Key takeaways
- OpenAI has released a technical report stating that its models rewrote SystemVerilog RTL for AMD's flagship GPU, lifting power, performance…
Background & analysis
On 18 November 2026, OpenAI published the technical report Advancing chip design with OpenAI models, offering its first concrete account of its own models producing hardware design work. It said GPT-6 Astra had optimized circuits used in its in-house inference chip Jalapeño and improved RISC-V core designs. From 19 November onward, a wave of outlets covered the disclosure.
The headline figure concerned AMD. According to The Decoder and The Next Web, OpenAI said that in less than five months of working with AMD, its models delivered up to a 14% PPA improvement on AMD's flagship GPU IP, a chip design AMD engineers had already optimized for years, and that this amounted to roughly a year's worth of progress. Techmeme described the disclosure as OpenAI's second public account of models taking part in chip design, and OpenAI president Greg Brockman echoed the one-year framing, saying part of the gains had landed in a production-ready chip design.
The report also detailed work on OpenAI's own silicon. As reported by IT之家, engineers began using GPT-4o on Jalapeño in early 2025; redesigned multipliers and adders each cut synthesized standard-cell area by about 15%, with the multiplier also cutting latency from two clock cycles to one, and the chip's total area was reduced by up to 13%. On the AMD side, DEV Community and cnBeta reported that the models rewrote parts of the SystemVerilog RTL of AMD's flagship GPU, followed by functional verification with Synopsys EDA tools. Some AI-generated changes entered a design sent to manufacturing, while others were folded into production design files for future products; single-digit percentage PPA gains were also validated on key modules of AMD's next flagship CPU. The Next Web added that OpenAI is training a GPT-Synopsys model with Synopsys. The picture now stands at this point: the 14% figure refers to PPA gains, with the split across power, performance, and area not broken out, and AMD has not confirmed which products will use the optimizations.
AI-generated from 9 reports · updated 15 hours ago
Latest turnA technical report OpenAI published on November 18 reveals that its models rewrote parts of the SystemVerilog RTL for AMD's flagship GPU, achieving up to 14% PPA gains on highly optimized baselines. Some AI-generated optimizations have been integrated into designs recently sent to fabrication, compressing about a year of engineering work into under five months, according to OpenAI.
Related AI tools
- Heat index
- 62
- All sources
- 8
- All reports
- 9
- First seen
- 3 days ago
Reports on this story headlines open the original
A technical report OpenAI published on November 18 reveals that its models rewrote parts of the SystemVerilog RTL for AMD's flagship GPU, achieving up to 14% PPA gains on highly optimized baselines. Some AI-generated optimizations have been integrated into designs recently sent to fabrication, compressing about a year of engineering work into under five months, according to OpenAI.
cnBetaAI score 72
OpenAI published a technical report on November 18 saying its models rewrote SystemVerilog RTL for AMD's flagship GPU, delivering up to 14% better PPA, roughly a year of design progress. AMD has folded some model-written changes into a design sent for manufacturing, with the rest in production files for future products.
DEV Community · AIAI score 73
OpenAI president Greg Brockman said the company's models helped AMD fit roughly an extra year of chip R&D into under five months, with some optimizations already in production-ready designs, according to Sina Finance. An OpenAI research report published Wednesday put the gains at up to 14% on power, performance and area for AMD's flagship GPU.
36氪AI score 60OpenAI published a technical report, Advancing chip design with OpenAI models, saying its models helped optimize circuit design for its in-house Jalapeño AI inference chip. In work with AMD, the models improved power, performance and area metrics by up to 14% on key IP blocks of a flagship GPU, and some of those changes are already in AMD designs headed to manufacturing.
cnBetaAI score 55OpenAI published a technical report titled Advancing chip design with OpenAI models, detailing how its models helped optimize chip designs. The work delivered up to 14% PPA gains on AMD's next-gen GPUs and up to 13% area reduction on OpenAI's Jalapeño inference chip. Engineers began using GPT-4o on Jalapeño in early 2025, with some multiplier and adder redesigns cutting standard-cell area by about 15%.
IT之家AI score 66OpenAI says its models improved power, performance and area by up to 14% on AMD's flagship GPU IP. Some of those changes made it into a design AMD sent for manufacturing, marking a further step in OpenAI's push into chip design work.
TechmemeAI score 61OpenAI says its models improved power, performance and area by up to 14% on AMD's flagship GPU designs, with some changes already in a chip sent for manufacturing, The Next Web reports. AMD confirmed single-digit gains in key parts of its next flagship CPU, and OpenAI is training a GPT-Synopsys model with Synopsys.
The Next WebAI score 63OpenAI says a five-month collaboration with AMD produced up to 14% PPA improvement on AMD's flagship GPU IP, roughly a year's worth of progress, and AMD has folded some of the changes into a design sent to manufacturing, The Decoder reported. The method was also applied to Jalapeño, OpenAI's first custom inference chip.
The DecoderAI score 61
OpenAI published results showing its models optimized circuits used in Jalapeño silicon and improved RISC-V core designs, work it attributes to GPT-6 Astra. It is a concrete disclosure of the company applying its own models to chip design rather than a general claim.
OpenAI NewsFirst-partyAI score 48
Other stories people are talking about
- Heat index 290RisingGPT-6.2 Astra and Sol now generally available, with deployment comparisons5 sources
- Heat index 77Microsoft rolls out Copilot integrations and unveils MAI models at Ignite 20262 sources
- Heat index 65OpenAI launches GPT-Live-1.1 voice model with lower idle pricing1 sources
- Heat index 1143RisingAnthropic Reaches $1.6 Trillion Market Value in First Trading Week After IPO21 sources
- Heat index 1011Anthropic Mathematician Says Claude Formalized the Generalized Riemann Hypothesis21 sources
How is heat calculated?About the methodHide
Heat counts how many independent sources covered a story in the last 48 hours: one source counts once no matter how many posts it published, decaying with a 24-hour half-life. What ranks first is what many people are talking about.
This page aggregates public feeds. Headlines and summaries are machine-organized and remain the property of the original authors; verify important facts at the source. If you believe a headline or summary infringes your rights, email the contact address in the footer with the page URL and basis of your claim, and we will remove or replace it after review.
- Surge
- Discussion rising fast
- New
- First report within 6 hours
- Rising
- Still gathering discussion

