Tình hình của ngành EDA năm 2024
In what has become a yearly custom, I recently spoke to Jay Vleeschhouwer, Managing Director of Griffin Securities, for an update on his view of the state of the electronic design automation (EDA) industry. My inquiries were based on his presentation at the 2024 Design Automation Conference (DAC). With his long background as an informed EDA industry follower, I knew it would be an enlightening and interesting discussion, and it was.
Smith: During your presentation, you pointed out that the industry’s consolidation has come down to four big companies. Is there room for startups?
Vleeschhouwer: I think so. I would not want to rule out an opportunity for a startup, given the always dynamic new problems and technical challenges that face semiconductor and systems design and engineering. Through the long history of EDA, we’ve seen a broad range of technical requirements, dozens of steps and dozens of types of tools and functions needed to have a design infrastructure for a complete design flow.
As complexity continues to become more and more challenging, there are always going to be certain headaches that need to be cured. In principle, assuming sufficient funding and technical abilities, a startup can have a role. This has been the case over the last four decades.
However, the number of startups is less than it used to be, a function of consolidation and the fact that the largest companies have been focused on their own internal development. There’s been an occasional small acquisition. Synopsys, for example, has focused on areas like silicon lifecycle management with several acquisitions. Cadence is growing its broader Engineering Software strategy that it calls Intelligent System Design (ISD), where it has moved into adjacent areas that are not strictly EDA, but complements their core EDA products.
The four largest EDA companies that account for 90% market share have been focused on building out their internal development capabilities. That leaves a 10% share for everyone else. Ten percent of a more than $13 billion industry is a considerable amount of potential revenue and even this segment has seen consolidation or, in some cases, expansion of non-EDA engineering companies into the EDA space. Examples include Altium, now part of Renesas, Keysight with its own Engineering Software diversification strategy that includes EDA, and PDF Solutions.
For smaller companies, the opportunity is to solve a hard, specific problem that is inherently or newly important to current and/or future nodes or device types or complexities. However, it is more challenging now to gain a foothold as the largest companies seem to have less inclination than two decades ago to go out and write a check and buy emerging companies and technologies.
Smith: Are EDA products generally becoming aligned to support the “arms” race in software and silicon development?
Vleeschhouwer: The answer would be inherent in the bookings and backlog trends. Customers are investing their engineering dollars wisely and appropriately. We wouldn’t see large multi-year contracts, including both renewals and expansions, if the products were not sufficiently aligned or relevant.
If EDA products were not sufficiently aligned or suitable for addressing the silicon and software development needs, we would see a decrease in the book-to-bill ratio, indicating an issue with growth and adoption. But this does not seem to be the case as exemplified by Cadence and Synopsys, which have both had a positive book to bill for each of the last 14 years, since 2010.
EDA companies continuously spend over 30% of their revenues on R&D to innovate and create or update products that customers find important and essential. It is a very competitive marketplace and customers often engage in benchmarks to determine which solutions they should invest in. Product mix also matters. An EDA company needs to align its product portfolio to meet the current or expected growth or technical requirements of its customers. Customers ultimately decide wisely with their dollars to invest in the technology that best suits their needs.
Smith: The big EDA companies engage with the foundries as they flesh out new processes. It’s in the foundry’s best interest to make sure that when it’s ready to go to manufacturing, tools are in place to support designs.
Vleeschhouwer: Of course. Once upon a time as DAC began, a barrage of press releases from the EDA vendors announced reference flow certification. We don’t see that quite as much, though it remains an important part of the process. EDA requirements are predictable because there’s a well-understood semiconductor process roadmap and parameters need to be addressed in software, capacity and functionality.
This is why five or six years ago we concluded that even though the synthesis category had been flat it would be entering a new period of growth and rejuvenation. It has in fact experienced a much better period of new and more sustained growth, tied to new requirements for sub-10 nanometer design implementation. One signal was that internal investments were ramping up at Synopsys and Cadence. Interestingly, those investments are continuing with further development of synthesis, one of the oldest and most critical categories in EDA
In certain instances, the precise technical needs for new or updated EDA tools are going to be obvious. The other interesting aspect is that some EDA categories are correlated with each other for process and technical flow reasons. If one category may be about to see renewed growth or renewed new requirements, there are often associated categories that will grow along with it or need reinvestment.
Smith: Is EDA missing out on the huge artificial intelligence trend?
Vleeschhouwer: The answer must be no. While difficult to quantify, the contribution to the EDA companies is emblematic of this phenomenon for both machine learning and AI. Perhaps ML is the more relevant, having more to do with pattern recognition. EDA tools deal with massively complex patterns that lend themselves to massive computation. Clearly semiconductor design lends itself to these kinds of techniques.
More broadly, Engineering Software and design software, not just EDA, are well suited to enable and benefit from this phenomenon. Cadence and Synopsys are both purveyors of multiple AI- and ML-enabled products.
Even if semiconductor companies don’t employ specific AI/ML enabled tools from Cadence and Synopsys, they are using EDA technology to design their own AI products. For example, Microsoft, Google and Nvidia rely on EDA products to design their products, which are then used in computational use cases for AI and ML
With respect to this AI arms race, EDA companies and other companies like Adobe and Autodesk are positioned to deliver this technology given their existing intellectual property. They have technologies and capabilities they can build upon, coupled with the massive scale of their R&D resources. In the long run, every company that claims to have AI/ML will require a sustainable differentiated capability to be competitive. Certainly, Cadence, Synopsys and Ansys are well suited and positioned because the needs and use cases in their customer base are readily available to them.
Smith: The EDA product mix, specifically hardware-assisted verification, has changed in the past decade or so. What’s causing this growth?
Vleeschhouwer: What’s interesting about hardware-based verification is that once upon a time, it was an EDA backwater and not a large revenue category. Certainly it was not a source of predictable revenue, as is often the case with hardware. It was deemed to be useful but not essential to what customers needed for engineering their products.
Within the last decade, the breadth of use cases, capacity and capability of the systems from Synopsys, Cadence and Siemens have expanded, particularly for systems customers that do hardware-software co-design and co-verification and prototyping and software bring-up. The category was bumping along at a couple of hundred million dollars and now surpasses a billion dollars, which is a considerable part of their businesses. For Synopsys, hardware along with IP has been a significant contributor to its revenue growth.
Smith: What key areas do you currently see that EDA needs to invest more for the future as the semiconductor industry heads toward a trillion dollars?
Vleeschhouwer: Operationally, the EDA companies will have to continue devoting substantial resources to R&D as a percentage of revenue, which will remain a larger percentage than is invested in sales. The major customer base is finite and well-defined, as compared with the industrial side of Engineering Software where there are one to two orders of magnitude more customers that have engineering software requirements that need to be addressed.
The whole scope and structure of the go-to-market strategy for the large number of Engineering Software consumers is different than in EDA. Proportions spent on R&D and sales are different between EDA and non-EDA Engineering Software. Cadence, Synopsys, and Siemens are going to have to continue to invest over 30% of revenues in R&D, an ongoing structural need to address their major customers. And it’s interesting to consider how these companies are thinking of complementing or diversifying their portfolios given the significant differences between their approaches.
Cadence is investing in its Intelligent System Design strategy through several acquisitions in the computational fluid dynamics area to move into complementary Engineering Software spaces. That led to a relationship with Dassault Systèmes, one of the largest Engineering Software companies.
Should it occur, the pending acquisition of Ansys by Synopsys will change the complexion of the industry in many ways. Synopsys would become the largest Engineering Software company, where Siemens has been to date. Roughly 80% of Ansys’ revenue is not connected to EDA, moving Synopsys into that new world, a large part of the intent behind the acquisition.
On the operational side, particularly for Cadence, it is connected to this strategy by having to make investments in its go-to-market. Cadence is focusing on expanding its sales channels with resellers and partners. It will be interesting to see how that plays out.
Similarly, if the Ansys-Synopsys combination does occur, Ansys has a large reseller channel business of roughly half a billion dollars a year covering all its end markets, significantly additive to the sales reach that Synopsys has. Therefore, an interesting consideration in this combination is the sales channel that Synopsys will be picking up as well.
In terms of ongoing investments, a persistent need is for applications engineers (AEs). For Cadence and Synopsys, it is typically the second largest number of openings after R&D and product engineering.
We look at AE positions that each of the Big Four is looking to fill. This is an important and indicative function as both a coincidental and leading indicator. Oftentimes the need for AEs can be indicative of customers’ new product and new capacity adoption.
Smith: You mentioned that Cadence’s strategy is broader and moving into analytical software such as computational fluid dynamics. Is this a key reason behind the proposed Synopsys acquisition of Ansys?
Vleeschhouwer: By revenue, Ansys is the largest provider of computational fluid dynamics (CFD) across all end markets. We now have a new subset for CFD as applied to EDA, particularly as it relates to new forms of packaging, new kinds of mechanical effects or 3D ICs and electro-thermal effects. Down at that level, new forms of multi-variable optimization problems need to be solved.
Jay Vleeschhouwer, managing director of software research at Griffin Securities, has more than 40 years of research analyst experience in the technology sector, including software, semiconductors and computer hardware.
