Technology

OpenAI Shows How AI Models Are Helping Engineers Improve Custom and RISC-V Chip Designs

0

San Francisco, United States, November 19, 2026

OpenAI has detailed new work showing how its AI models can assist engineers in improving processor designs, including circuits used in its custom Jalapeño inference chip and open-source RISC-V cores. The research, published on November 18, describes how the company is applying AI to hardware development, extending the use of large models beyond software coding and into the design of the chips that power artificial intelligence systems.

The work builds on OpenAI’s earlier disclosures about using its models to accelerate development of Jalapeño, its custom inference silicon. The latest publication provides more detail on the engineering tasks involved and describes how the company is applying similar methods to third-party hardware, including work with AMD.

AI Moves Into Processor Design

Chip design requires engineers to balance several competing factors, including performance, power consumption and physical area. Even relatively small changes to a processor’s architecture can affect how quickly it executes instructions, how much energy it consumes and how difficult it is to manufacture.

OpenAI said it tested successive model generations on two open-source 32-bit RISC-V processor designs: PicoRV32 and Ibex, the embedded processor used in OpenTitan. The models were given existing hardware descriptions and asked to improve throughput, which measures how many instructions a processor can execute over a period of time.

The company evaluated the resulting designs through simulation and applied constraints to prevent models from producing changes that significantly increased circuit area. OpenAI also acknowledged that some evaluation conditions allowed models to omit certain features from the original designs, which could influence the results.

Work Extends to AMD and Industry Design Tools

The research also describes collaboration with AMD on improvements to production hardware and with Synopsys on connecting AI models to electronic design automation tools used by chip engineers. These tools are central to designing, simulating and verifying complex semiconductor systems.

The collaboration with Synopsys is intended to help models work within established engineering processes rather than operate as standalone code generators. This distinction is important because a hardware design that appears correct in a model-generated description still needs extensive verification before it can be used in a production chip.

OpenAI said it is working to train a model that can use Synopsys tools in ways similar to experienced chip designers. The goal is to broaden the range of design tasks that AI systems can support while retaining the checks required in semiconductor engineering.

Potential Impact on Chip Development Cycles

AI-assisted design could help engineers explore more alternatives, identify optimisation opportunities and reduce time spent on repetitive development tasks. The potential value is significant as demand for specialised processors grows across data centres, consumer devices and industrial systems.

However, the technology is not a replacement for hardware verification, physical synthesis or manufacturing validation. Chip designers must still confirm that proposed changes meet timing, power and reliability requirements under real-world conditions.

A New Role for AI in the Semiconductor Industry

The publication illustrates how AI development is increasingly connected to the infrastructure needed to run AI itself. As companies invest in custom accelerators and specialised processors, tools that can help improve chip design may become an important part of the semiconductor development process.

The next stage will depend on whether model-assisted improvements can consistently translate into verified, manufacturable designs across more complex commercial workloads.

Bajaj Finserv and Pennant Technologies Invest $4 Million in Indian AI Startup Soket AI

Previous article

Microsoft and Intel Expand Long-Term Partnership to Build Azure’s Next-Generation AI Infrastructure

Next article

You may also like

Comments

Comments are closed.

More in Technology