AI Revolutionizes Chip Design: Over 50% of Advanced Silicon Designs Now AI-Assisted

AI is reshaping chip design tools, and the results are impossible to ignore

Why it matters: As powerful as AI may be, many industries are still struggling to find clear-cut applications that make a measurable, demonstrable difference. Thankfully, that is not the case when it comes to chip design software. In fact, since their introduction just a few years ago, AI-powered features have become a mainstay of EDA (Electronic Design Automation) tools from companies such as Cadence and Synopsys. Silicon designers quickly discovered that many of the complex yet often tedious tasks involved in their process - particularly the "grunt work" - could be automated or dramatically simplified by intelligent AI algorithms. From the automated layout of certain IP blocks to improved efficiencies in IP block interconnects, these AI features help accelerate the less creative (but still critical) parts of the workflow, allowing designers to focus more on the interesting and innovative aspects of chip development. In addition, AI-powered tools can drive impressive improvements in chip performance and energy efficiency. Case in point, vendors like Cadence have indicated up to 60% performance improvements on specific blocks within a chip because of AI enhancements. Silicon designers quickly discovered that many of the complex yet often tedious tasks involved in their process - particularly the "grunt work" - could be automated or dramatically simplified by intelligent AI algorithms. Power improvements of up to 38% have also been made possible thanks to these tools. Along the way, silicon engineers also discovered that AI-powered features could reduce the amount of time necessary to finish a chip design - in some cases, up to 10× faster. In short, these AI-powered EDA programs provide the kind of ideal AI-enhanced scenario of increased productivity and more engaging work that many organizations are looking for. Not surprisingly, this has also led to significant growth in the use of AI-powered capabilities in modern chip design tools. In fact, based on public data regarding the number of chip design tapeouts disclosed by major companies like Cadence and Synopsys, as well as their estimates of AI feature adoption, the industry is now crossing a critical threshold. Specifically, just over 50% of advanced silicon designs (those built with 28nm process technologies and smaller) are now believed to be AI-assisted. Looking ahead, it's easy to predict that this percentage will continue to grow significantly over the next few years. Given that there were zero AI-assisted tapeouts just four years ago, that's impressive progress. More importantly, it's a great example of how applied applications of AI technology can have a profound impact on a business's evolution. The fact that it happens to be in the chip industry (and, appropriately, likely involves a significant percentage of chips that are designed to accelerate AI computing!) makes the moment even more relevant and consequential. According to Cadence, these AI features can reduce chip design times by as much as a month, which is a significant positive impact. Plus, as mentioned earlier, it's a benefit that can be directly tied to the AI features - about as concrete an example of the technology's benefits as you could ever want. The power and performance improvements alone make the enhancements enabled by AI incredibly valuable. However, toss in the increased efficiency of the work that silicon engineers can achieve with these tools, and the story gets that much stronger. It's easy to see why so many people in the world of semiconductor design - including industry leaders like Nvidia, AMD, Qualcomm, MediaTek, Samsung Semiconductor, Marvell, and Broadcom - are so excited about the possibilities for AI in their product creation tools (as well as for the AI accelerators they're going to be designing with those tools!). The timing of the crossover point also ties in very nicely with a number of other semiconductor industry developments. Most notably, the past few years have seen a big increase in the kind and number of companies who are working on advanced chip designs. From cloud computing providers such as Google, Microsoft, and Amazon's AWS to device makers like Apple, Samsung, and more, there are many organizations pursuing the custom silicon route as a critical means of differentiation. However, the number of skilled chip designers in the world is still relatively limited, so having more advanced AI tools that can enable even junior designers or others with limited experience to take on more sophisticated chip layout tasks is critically important to keep the semiconductor industry advancing forward. Even for the long-time semiconductor players, these enhancements create new possibilities, including the ability to create more designs, build more customized options, and run more projects in parallel. Creating more customized designs, in particular, is something that many in the chip industry (and their chip-buying clients) have wanted for a very long time, however the practical realities of doing so with traditional design tools have kept that from becoming possible. But now all of these capabilities can translate into opportunities to build on the rapid growth the semiconductor industry has seen over the last few years. Another important point is that as semiconductor designs move into smaller and smaller process nodes and the number of transistors per chip continues to expand, AI chip design features are quickly evolving from a nicety to a necessity. The number of factors, permutations, and connections that chip designers face is quickly growing, and the work to create these sophisticated new chips demands the enhanced intelligence that a well-designed AI-powered tool can enable. While it's true that the speed of AI adoption and the extent of its influence haven't been as fast or as profound as many first expected in certain industries, it's also becoming very clear that in targeted applications, it's proving to be even more impactful than many hoped. With the transition to AI-enhanced chip designs crossing over this important 50% barrier, it's apparent that EDA tools are unquestioned beneficiaries of these advances. From a semiconductor industry perspective, it's also clear we're entering an exciting new AI era. Bob O'Donnell is the founder and chief analyst of TECHnalysis Research, LLC a technology consulting firm that provides strategic consulting and market research services to the technology industry and professional financial community. You can follow him on X @bobodtech

Chip Design Hits AI Crossover Point

This has also led to significant growth in the usage of these AI-powered capabilities in modern chip design tools. Specifically, just over 50% of advanced silicon designs (those made with 28nm process technologies and smaller) are now believed to be AI-assisted. As powerful as AI may be, many industries are still struggling to find clear-cut uses that would make a measurable, demonstrable difference. Thankfully, when it comes to chip design software, however, that is definitely not the case. In fact, since their early introduction Bob O'Donnell is the founder and chief analyst of TECHnalysis Research, LLC a technology consulting and market research firm that provides strategic consulting and market research services to the technology industry and professional financial community. You can follow him on Twitter @bobodtech.