The startup Moores Lab AI seeks to use agentic artificial intelligence to shorten the chip-design cycle, initially focusing on design verification, debugging, and test coverage. The company, which was founded last year by specialists in the chip industry, says its tools aim to automate the chip-design cycle from start to finish and accelerate the development of systems-on-chip (SoCs).
This approach comes at a time when growing demand for AI chips is pressuring design teams to deliver products within shorter periods. Moores Lab AI believes that building AI tools from within chip-design expertise differs from the approach of companies that try to develop AI models first and then apply them to semiconductor problems.
From Months to 48 Hours
According to Shelly Henry, the company’s CEO and co-founder, Moores Lab AI has deployed its system at about 10 sites, and customers use it with different types of intellectual property units. She says the greatest productivity gains have appeared during the initial debugging setup stage.
Teams typically need three or four months to prepare the test plan and test environment, run the first simulations, and find the first bug. Users of the company’s platform, however, Henry says, consistently reach their first bug within 48 hours and then continue using the platform to discover additional bugs more quickly.
Why Does Human Expertise Remain Essential?
The article does not present AI as a self-sufficient replacement for chip engineers. Henry says that providing a specification to a large language model or AI agent can result in the loss of 20% to 30% of its fine details. In an industry where a chip cannot be corrected after it has been manufactured, a single error could become a $10 million problem, because addressing it may require rebuilding the entire chip set.
For this reason, the company focuses on verification and is working to divide the complex design cycle among multiple disciplines, including design engineering, architecture, verification, design-for-manufacturability testing, physical design, and synthesis. According to the article, around 200 engineers with different specializations participate in developing a single chip, making coordination among agents across these fields an engineering challenge in its own right.
What Is Changing in Practice?
The most important result so far is not proof that agents can produce a complete chip without intervention, but rather the reduction of the time needed to reach the first useful testing and debugging cycle. This could help design teams identify defects early, before project costs accumulate, but it does not eliminate the need for comprehensive verification or experts capable of interpreting specifications and reviewing the system’s outputs.
Moores Lab AI also faces the challenge of separating measurable results from the marketing noise surrounding AI. The company says it presents a single project implementation to the customer and compares the results with what its engineers accomplished on that same project before moving on to other work.
This strategy shows that the practical value of agentic AI in chip design will be measured by the time required to reach reliable results and by the system’s ability to handle specification details, not merely by its generation of code or design suggestions. As for the claim of building chips in less than six months, it was made by the company, and the article did not provide independent details about the projects or verification metrics supporting it.