Will artificial intelligence disrupt EDA?

Generative AI has already disrupted search, is changing computing, and now it threatens to disrupt EDA. Despite all the buzz and claims of radical change ahead, it remains unclear exactly which fields will be impacted and how deeply.

EDA has two main roles - automation and optimization. Many optimization problems are NP-hard, meaning it is impossible to find an optimal answer in polynomial time, especially as designs scale. Over time, heuristics have been developed to get "good enough" results in a reasonable amount of time. While it is conceivable that AI could provide similar or even closer to optimal results, its impact on design is likely to be more evolutionary than disruptive.

Disruptive innovation often results in market changes. Consider the question: “How would the semiconductor industry be impacted if EDA could deliver optimal results in zero time?” Time to market would be faster and the PPA of the design would improve slightly. But whether this would be enough to cause a significant increase in the number of design starts or lead to the development of new markets is unclear.

Under these assumptions, design creation and verification will still be limiting factors. Generative AI may be able to improve this, and there are encouraging signs that it can do so. If design and verification time is significantly reduced, new markets will almost certainly be created.

EDA has faced disruption in the past, and the problem with disruptions is that they’re often not obvious until they happen. “In some cases, people knew disruption was coming, like Kodak knew about digital printing, but they just couldn’t get it to market,” says Prith Banerjee, CTO of Ansys. “There are three types of innovation horizons. The first is short-term. What features should the next version of the tool have? We know which ones because they exist in the market. You’re selling to the market, you’re watching the competition—70% to 80% of the investments of the big companies are focused on the first horizon.”

The second involves adjacencies. “For example, you’re selling a product that was designed to be on-premises but you want to move to the cloud,” Banerjee added. “It requires innovation, but we’ll find a way and be successful.”

Many of the disruptions based on computing fall into this category. “Computers used to have very little memory, and then they got bigger and bigger,” said James Scapa, founder and CEO of Altair. “We changed the way one of our tools worked, and that innovation was disruptive to the market. Essentially, we put all the models in memory. That change meant we were about 30 times faster than our competitors. A similar change happened in HPC. The business model associated with cloud computing is going to be one of the big changes in EDA. The business model that goes with it is going to be somewhat disruptive. It’s important to understand the evolution of computing, where it’s going, and how to leverage it.”

Another kind of transformation is still underway. “Think about parallelism,” says Jan Rabaey, professor emeritus at the EECS department of the University of California, Berkeley, and CTO of imec’s System Technology Co-Optimization unit. “People used to say that parallel computing was a very bad idea because we didn’t know how to compile. Instead, we should take a single processor and make it as fast as possible. Then, power problems came up and we couldn’t make them faster. So all of a sudden, parallelism became a good idea. It was a disruption.”

The remaining 10% of investment goes to the third type of innovation. “It’s not part of your current R&D and it’s not targeted at an existing market,” says Ansys’ Banerjee. “A classic example is Apple launching the iPhone. That was disruptive. Amazon launched AWS, their web service. That was disruptive. How does a large company do disruptive innovation because it doesn’t happen by accident? It requires a process, and you need to tap into where innovation is happening. It’s in academia, it’s in startups. You should constantly monitor what’s happening in startups, and then have a central R&D team try to invent something themselves. But this central team doesn’t have to invent everything. Part of it is something you do organically, and part of it is bringing the technology into your company.”

Looking back, we can see the disruption that happened within EDA. “If you go back to the 1980s, we saw a whole series of ideas that were born out of academia and startups that changed the way we do design,” said UC Berkeley’s Rabaey. “EDA started driving design using standard cells. When you first saw it, it seemed like a really bad idea. It was very limited. You put cells in rows and so on. But it made automation possible. That led to basically logic synthesis, where we could start thinking about logic functions, optimizing them, having a set of tools to help us turn that high-level description into something, and it was automated. We take that for granted today. There were other areas—simulation, verification, behavioral synthesis—that ended up causing some form of disruption.”

There has been little disruption in the EDA space over the past 20 years as the industry has largely been on a linear path, but that is changing rapidly as Moore’s Law moves from planar designs to multiple chips stacked in a package.

“Disruptive change is more likely to happen when the status quo is bad,” said Chuck Alpert, automotive R&D fellow at Cadence. “Take the design team. They may know something is wrong. Maybe the engineering budget is out of control, or they are trying to do a new design but don’t have the engineering skills. They have to do something disruptive. Today, we are seeing an explosion in design complexity. There is a lack of scalability. The design team may be facing something that forces innovation. These are all situations where the status quo is bad or is going downhill. For EDA companies, this can happen when you are not a market leader. You fall behind and have to do something disruptive to catch up. Or maybe you have been a market leader, but the code base is written in COBOL and no one knows it anymore. You have to change because the trend is going down and you are in a situation where you innovate or die.”

Opportunities for innovation are everywhere, especially in an innovative culture. “The advent of AI and large language models can be very transformative, while cloud computing enables rapid scaling,” said Altair’s Scapa. “The business model—not just the technology—is part of how you disrupt. Startups in the EDA space are really hard because there are only two companies that dominate. They have a long history of acquiring and eliminating startups and competitors. That stunts innovation.”

By looking ahead, some of the pressures can be identified and dealt with. "What are disruption cycles?" Rabaey asks. "There are a lot of them on the horizon. The good thing about roadmaps is that you can identify the problems that are likely to arise 10 years from now. That's what academics are good at - looking at these roadmaps and identifying the new paradigms that are likely to emerge from them. For example, will scaling last five years, or even 10 years. What do we do about this? You can't choose disruption. The only time you go down the path of disruption is when you hit a roadblock, when you suddenly realize, 'I can't go any further.' We have to rethink the way we design. One possibility is to start thinking in a third dimension, where you layer different technologies on top of each other. The easiest way to do this is to just map the old architecture onto this. But that doesn't get you a lot of benefit. You have to rethink how you use it."

Sometimes, change is externally forced. “Design is moving from chips to systems,” Banerjee said. “If the goal is to design an electric car, my requirements are not just RTL entry. My design requirements are an electric car that can go from 0 to 60 in one second, has a range of 500 miles, and it has to be Level 5. Those are my requirements. The EDA industry is focused on designing chips. You have to design power electronics, which is a power electronics simulation combined with battery, motor design, and then aerodynamic workloads. It’s a multi-physics world and it’s very complex. Then you have software, which has to be written and automatically compiled to system-level specifications and then verified.”

EDA companies have been quick to adopt some forms of AI in their tools. "Reinforcement learning is being used to solve optimization problems," said Stelios Diamantidis, senior director of AI solutions at Synopsys. "People are now using reinforcement learning to run experiments, collect data, build better metrics to drive optimizations, and automate those optimizations. The technology itself can be applied to other problems. We started with optimizing physical layout and floorplans, clocks in certain topologies, DTCO and other physics-type applications. From there, we applied the principle to problems like verification, where reordering tests or changing the seed to test vectors can help you achieve coverage faster or track down bugs in manufacturing tests."

But AI is unlikely to replace existing EDA tools. "I think the status quo is low because we have great EDA products and our customers are using them," Alpert said. "If we decide to use AI to build new products, we're going to pay a huge price. Maybe in the long run we'll get some benefit. If we take the entire product team and say, let's start over and build something new, it's going to be very painful. Eventually, you might succeed, but in the meantime, you're going to pay a huge price."

The key for the EDA industry is to maintain continuity and ensure that customers are provided with the tools they need to launch the next product. "We have to protect our $2 billion business," Banerjee said. "A startup starts from scratch. But it is still difficult for customers to accept new technologies to solve their problems. This is not just a challenge for EDA, but a challenge for the entire industry, which is why I see a third prospect - vision, partner with startups, and then acquire startups that have proven these technologies."

Alpert agrees. “Disruptive technology is hard for almost all industries to handle, not just EDA. They can invest some resources, but not too much. Or they can wait for someone else to innovate and buy it. That’s another strategy.”

But where have all the startups gone? “Over the last 10 or 20 years, the existing ecosystem has collapsed,” Rabaey says. “There was a time when there was a vibrant research space in EDA. Go to all the top universities and they’re all working on tools. You don’t find them anymore. They don’t exist. Maybe you have a good idea and academics can publish papers, but they’re not going to build that product. The role of startups was really important, and in the ’90s, it was a vibrant world. It was these small companies that came up with ideas and tried them out. That has collapsed, too. But the ecosystem may rise again.”

A lot of investment is pouring into GenAI, but much less into EDA. “GenAI is real and will deliver real results for us,” Scapa said. “But there’s too much hype and the amount of investment doesn’t match the returns we’re seeing today. GenAI will decline first and then have a typical slow rise because GenAI is really big business. We’re also doing some interesting things with traditional machine learning, which also has huge potential.”

But GenAI’s true potential in EDA seems somewhat beside the point. “EDA doesn’t create designs,” Rabaey said. “But it is driven by design considerations. AI will become a disruptive part of the design process. AI will become a design tool that helps us explore the huge space of options.”

The second wave of generative AI is addressing automation. “Specifically, some of the key industrial challenges,” said Synopsys’ Diamantidis. “It’s more about economics, geopolitical pressures, talent availability, and the ability to do more with less. In the second wave, we are able to take the data or the design environment. We are able to use that data to train models at a very large scale. Then, we are able to contextualize them for different tasks specific to the designer’s activities. We are certainly addressing the human-machine interface. We can now explore huge complexities.”

Perhaps the biggest return on investment for GenAI is productivity. “We’re committed to guiding people through the development process and helping them improve their problem-solving skills using generative AI,” said Erik Berg, senior principal engineer at Microsoft. “Where does this data come from? I believe the richest source of data we have is in the heads of our engineers. The tools I’m building not only provide solutions to our engineers, but also capture results, other data, and results from their heads at the same time.”

This is happening in many areas of the design world. “GenAI can definitely help non-expert users get better,” said Vidya Chhabria, an assistant professor at Arizona State University. “It can help non-expert users ask the right questions — more intellectual questions. It can help non-expert users quickly get up to speed on new designs and new EDA tools. Maybe it can help expert users be more productive or do their jobs faster.”

But will it be disruptive? “Despite all this technology, it still takes four years to get a chip into a socket,” Diamantidis said. “I’m talking about gathering requirements, architectural exploration, design entry, verification, inserting test, preparing instrumentation for silicon diagnostics and data mining—the whole process. It takes a lot of people, money, and time, which means it doesn’t really change the fundamentals or economics of the semiconductor space.”

Disruption is hard and often goes unnoticed until it becomes obvious. Many people have been watching the advancement of technology, changes in design practices, and the shifting landscape from chips to systems. In addition, everyone believes that all forms of artificial intelligence may help solve these problems. Looking at the landscape today, nothing seems disruptive.

Today is the 3838th content shared by "Semiconductor Industry Observer" for you, welcome to follow.

★ Important report on EUV lithography machine released by the United States

Silicon carbide "surge": catching up, involution, and substitution

★ Chip giants all want to “kill” engineers!

Apple , playing with advanced packaging

★ A historic moment for GPU!

★ Continental Group, developing 7nm chips

★ Zhang Zhongmou's latest interview: China will find a way to fight back

★ EUV lithography’s new “savior”