Chipmakers turn to AI to speed design, verification and defect detection

Industry executives say AI is helping shorten chip-verification cycles, improve hardware-software co-design, detect wafer defects and reduce learning time for engineers

Artificial intelligence (AI) may be straining the global semiconductor supply chain, but chipmakers are increasingly turning to the same technology to make chips faster and more efficiently.

 

From improving hardware-software co-design and shortening verification cycles, to detecting manufacturing defects and cutting learning time for engineers, AI involvement in semiconductor design and production is deepening, industry executives said at Semicon India 2026.

    

“One of the things which generally semiconductor companies lack is they never talk to software guys. They don't care how software is developed. How the operating system has to use it. Now, with AI, we have seen hardware-software co-design,” said Prakash Raghavendra, senior fellow, software systems, AMD. 

  

AI not only optimises the software. “It also tells you can't optimise further. That feedback goes to the hardware… And if we practice software and actually redesign the code, hardware and software can optimise it,” he added.

 

“The opportunity cost is where the AI creates the best value with first-time drive and verification. Otherwise, it's going to cost hundreds of millions of dollars of delay,” Navin Bishnoi, vice president and India country manager at Marvell Technology, said. The verification cycle for chips involves hundreds of engineers and several methods. 

 

Micron has implemented machine learning techniques along with data centralisation to develop a live solution at semiconductor sites. “(This) is detecting the patterns or detecting new defect patterns, or existing defects, and how we're going to catch it. We're seeing a huge difference in how much machines used to say versus the AI used to say,” Shisheer Kotha, director of smart manufacturing and AI, India, Micron, said. 

 

Machine, here, refers to Micron’s advanced defect classification system that inspects wafers where detected defects mean significant losses. 

 

Ritesh Tyagi, VP global head silicon at Wipro, highlighted that the company uses AI to shorten the learning cycle. “We created many of these agents in the last couple of years… We call them learning agents,” he said. 

 

There is a repository of agents, which are not only being used in the actual design. This AI initiative actually democratised the learning class, he added. 

 

“Most of the chips and the circuits and the architecture that we talk about is native to our company. It's not public knowledge. So the models that are built in are only 20 per cent public data with 80 per cent (as) institutional enterprise data. A lot of these model algorithms have to evolve over the data that's inside the enterprises,” Kotha added.