Cadence Design Systems, Inc. (CDNS) Earnings Call Transcript & Summary

All right, everybody. Everybody is filtering in. Thanks for joining us. Our next presentation is with Cadence. And we've got Anirudh Devgan, President, CEO and Member of the Board of Directors. First, a brief statement by Richard from the Investor Relations team.

Hey, good morning, everyone. Thank you for having us. I'll be real quick, just get us out of the way. So today's discussion will contain forward-looking statements. Due to risks and uncertainty, actual results may differ materially. For information on factors that could cause actual results to differ, please refer to our SEC filings, including our most recent Forms 10-K and 10-Q. With that out of the way, back to you, Jeff.

All right. Thanks, Richard. All right. Anirudh, so you became CEO in 2021. Before that, you were President. And before that, you were Executive Vice President and General Manager of Digital & Signoff and System Verification groups, so quite a career at Cadence. It's been remarkable to watch Cadence's growth under your leadership. It's one of the best performing stocks on Nasdaq since then. So what have you done to drive some of this remarkable growth in your time as CEO?

Well, first of all, thank you for all the interest. Thank you for the question. Yes, we have done well in the last 5, 10 years. And what I would say is that we are probably better positioned now than we were 5 years ago. And there are multiple reasons for that. I mean, the -- it's a combination of like our own products and culture and our strategy and then what's happening in the marketplace, the importance of silicon now, the importance of system companies doing silicon. So I think it's a combination of good team, good technology and what the customer trends are there. And we are well positioned to capitalize that going forward.

Fantastic. So let's do just a quick overview and a refresher of Cadence. If you could, for the audience, just kind of talk about EDA, where does it sit in the broader kind of semi's ecosystem? And then as you think about the EDA landscape, what are some of your competitive strengths?

Yes. And it's like difficult to give an overview because some of you might be like super experts already and then some of you may not know. So it's always tricky to give overview to it. Like some of them may be more experts than even me. So I think the main thing about what we do is we make software. And we call it computational software because this is not regular software, this is very mathematical, numerical, technical software to design chips and electronic systems. Because these chips are, as you may know, are very, very complicated. Like the chips these days, 1x1 inch may have like 100 billion transistors. So there is no way that they can be designed by hand. And they haven't been designed by hand for like 20 years. So we make the software that makes that possible. So we are used by all the companies that are designing chips and in all geographies, in all vertical segments. So it has -- the business has become more and more essential to digital transformation and AI. And then -- so that's the EDA part. That's our core business. And we are the most diversified, the biggest portfolio. We are the leader in core EDA. And then the question has been in EDA, okay, you're good at designing chips, but can you do other things with it, right? So we do some IP, which is like existing building blocks. But the real expansion in the last 5 years is into system design and analysis, so what I call SDA, along with EDA, and of course, in AI. So the big drivers are SDA and AI. And the reason for the system move is because right now, about 45% of our customers are system companies. And they are designing more silicon, so we sell them EDA. But also, there are big issues of thermal, of power management, of aerodynamics. So we have this whole area of SDA, which can double our TAM with SDA. And then AI can further increase it. So in terms of our growth drivers for the future, apart from core EDA, the drivers are SDA, AI and then this emergence of chiplets, or system in a package.

So let's kind of take those in order. So as you kind of think about the past couple of years, a lot of companies have faced macro headwinds from the economy. You guys have continued to grow. What are some of the technology drivers behind that in terms of driving that growth?

I think the main reason in my opinion is that we are part of the R&D process. And R&D is anyway, and especially for these big companies, is essential for growth. So even though there is some fluctuation in revenue, even in some of our semiconductor customers in the last couple of years, these things take years to design. So the customer still invests in R&D for the future. And so I think the -- this is essential part is that we are part of R&D of our customers. And the second reason of our resilience is that most of our revenue is recurring in nature. So we have 85% to 90% recurring revenue. So we normally have a 3-year contract cycle. So that gives a certain predictability. And for that reason, we invest heavily in R&D because we have recurring revenue source. And then third reason is that we are very, very diversified. So almost all geographies where chips and electronic systems are designed, all verticals. So what happens is sometimes some verticals are doing well, some are not doing well. But there's always some new thing. Of course, AI is a big thing or data centers or automotive. So if some verticals are weaker or some geographies are weaker, some others are stronger. And in general, there is going to be more and more silicon. So in general, the verticals that are stronger are more stronger than the verticals that are weaker. So I see that continuing for some time. There will be more and more silicon design. And as long as we are best-in-class, we will be critical in that process there.

And so one of those end markets that's obviously top of mind right now for everybody is artificial intelligence. So can you talk a little bit about what you've been doing to drive product innovation around AI and what's in your product portfolio today? And where does the future take it?

Yes, AI is a big topic. And of course, everything -- everybody calls everything AI. That's a little bit of an issue right now. But that's a good in some ways. But we have done AI for some time. Because even EDA, the core algorithms, a lot of kind of computational methods are used for a very long time. And the way we look at the world is that there is silicon, then system and then data, right? A perfect example is like an electric car, you have all the navigation data, then you have the actual car, which is electrical plus mechanical, hardware plus software, and then the silicon that drives the car. So then if you overlay our core strength, which is computational software, this kind of numerical software, into these three circles, so the innermost circle, that's EDA, right? Then the next circle, if you put computational software, that's SDA, that's simulation and analysis and system design. And if you put computation software on data, that's AI. Because a lot of the AI algorithms are very similar to what like inference is like matrix multiply, training is like some kind of a conjugate gradient. So we have done these kind of algorithms for a very, very long time. So then as these emerge, we can apply this into our own products and also the build-out of AI in general. So to me, AI has like three phases to it. And some of you may know all this anyway, but -- so the first phase is like the build-out of the infrastructure, whether it's GPUs and then -- because NVIDIA is a big partner of Cadence, or like car companies doing AI like Tesla. Tesla is a big partner of Cadence. All these hyperscaler companies like Google and all these others have talked about their own silicon, so -- and then other companies, GPU companies, so -- and then there's AI in these phone chips and all that. So the first part is the infrastructure build-out. So we are very, very well positioned. And a lot of the -- not just from a software standpoint but also from what we call hardware-assisted verification, we have special systems that design these AI chips. So that's one. And that is still -- we don't break it out of the $4 billion that's our revenue. But that's a big portion and driving a lot of growth. So that's the first way we participate in AI. Because anyway, you can't design any of these systems without using products like ours. Now the second part of AI is applying it to our own products, so -- and we have now five major AI platforms or gen AI platforms, which are based on all these reinforcement learning, generative AI technologies. And they are giving -- I can give some examples, giving like massive improvement in productivity. But also, what's even more important, I think, is improvement in performance. So what would take like 3 to 6 months can be done in about 2 weeks. But what is more interesting is that it can give better results than what a human can do in lots of cases. And I'll tell you why that is. But we have five major AI platforms right now. And we have done that for -- released it probably 2 years ago before all the buzz about AI. And that can drive a lot of growth for the future. Because these chips are getting more and more complicated. So you need AI to make the problem more tractable. So that's the second big part of AI. And I think that was your main question. But I come back to that. But the third -- I just want to comment on the three phases in my mind. And then the third phase of AI is like new verticals that will emerge because of AI. Now this -- the second part is applying AI to our existing products and make them much better. But I think AI will also drive new solutions, of course, just like Internet did or other things did. So then the question is like what is that new thing? And there could be a lot of them. But in my mind, the biggest one has to be life sciences. So we invested last year in biosimulation. And you will see a lot more activity for us. Now this is a little bit further out because the third phase takes the longest but could have the most transformative effect. So just like we can do chip design software and then we can do system design software, like planes and cars, similar methods can be applied for molecular design and drug discovery. So we acquired a company last year more for the long run. So those are the three phases. But the second phase, it can drive a lot of productivity benefit, performance benefit. Like I'll give you an example. In lots of cases, it's not that the design is faster, but the design is better. Because automation has not been applied to certain parts of the design process that we can do with reinforcement learning now in AI. So in some cases, we're getting 8% to 10% power performance in what we call PPA, power performance area benefit. Just to put that in context, if you go from one node to another node, the improvement these days are like maybe 10% to 20%. Because there's more maturation of Moore's Law. So you're getting half of that or 2/3 of that from better software. You don't have to go to the -- and of course, people still go to the next node because there are benefits to that. But that's a huge benefit that you can get by a better software.

Well, then -- and that's a good thing because we have seen Moore's Law kind of slowing down, which leads to a lot of the interest in kind of 3D integrated circuits. You mentioned chiplets as one aspect of that. But maybe you could talk a little bit about how you're positioned in that kind of core chip design area as it gets tougher and tougher to get to the next node. How does that...

Yes. And I think Moore's Law has been slowing down for a very long time. Of course, some people say Moore's Law is alive and well. Some people say it's dead already. I think it depends on what you -- this problem with everything. Like what do you define as Moore's Law to be? Same thing with AI. So the Moore's Law, the classical Moore's Law, I think, is done for a long time already. I used to work with Bob Dennard in IBM, who actually wrote actually the Dennard Scaling. So by that definition, it's done like 15, 20 years ago. What has happened in the last 5, 10 years is area scaling. That means when you go from 7-nanometer to 10-nanometer, let's say, just pick any 2 nodes, when you go -- I mean, it's easiest to look at 10 versus 7 because 10 by 10 is 100, 7 by 7 is 49. So when you go from 10-nanometer to 7-nanometer, what's happening is effectively number of transistors are doubling on the same area. So you have double the transistor. So you can do more things with them. So if you look at even last 10 years, it's not that the chips are getting faster. I mean, you're still -- mobile CPU is still 3 gigahertz or so. But instead of 1, you have like 8 and then you have like GPUs in it, you have neural engines, all these kind of -- so the main thing that has driven Moore's Law is area scaling. And that's why GPUs have also done well because they have a lot of cores and you can have a lot of them, right? So now what is the extension of that area scaling? And I think this can continue, like we are at 3-nanometer right now, we go to 2, 1.4 and 1, so it can continue for like 10 years or so easily. But in parallel, the other way to do area scaling is instead of just putting more things in a chip and reaching the reticle limit, you can put multiple chips in a package. And of course, we talked about it in the '90s, too, right? It's nothing new. But I think the economics now make it more feasible to augment Moore's Law, whether you want to call it Moore's law or some other dimension of area scaling, which is in a package. And it has a lot of other advantages, as you know. Because like if you look at these chips now, especially it started in HPC, high-performance computing. And I think it will percolate through the entire value chain is that like even like these are all public, like Amazon has Graviton, it has like 7 chiplets in a package, right? And then when you go to the next version, you don't have to redesign all 7. You can redesign some critical pieces of it. So it's much more cheaper and cost-effective to do next generation and IP reuse and design reuse. So for area scaling benefit and also for cost benefits, I think chiplet is going to be a huge trend. So then the question is how are we positioned in that? So the good thing with Cadence is that we -- to do chiplets well -- now I mean, from a design software standpoint, I'm not talking about manufacturing. Because in general, we focus on the software part of it. There are unique three basic components. So you need the IC design, of course, tools. So we have the leadership in analog and digital. Because these are mixed-signal chips. And because some of them are IP, some of them are interface chiplets, some of them are compute chiplets. So that's first layer. Then the second layer, you need package design tools. Of course, it's a system in a package. And then the third layer on top is you need analysis tools, which are specific for that. And the biggest thing is like thermal simulation or electromagnetics. Because they have a lot more power being generated. So if you talk to TSMC, they will say thermal is a big issue in 3D IC. So the three -- and 3D IC is a generalized term for 2.5 and 3D. Not everything is stacked, I mean, some of them are stacked. But some of them are next to each other and interposed. But to really do it well, you need the IC tools, the packaging tools and the analysis tools. So Cadence is the only company that has all of them. Because we are the leader in packaging tools. Allegro is a flagship. Now some of this is good luck because Allegro has been there for 30 years and now packaging became super critical. It was not that critical 5, 10 years ago. And some of it is good planning, so Allegro. And then 5 years ago, we started a lot of system analysis tools. So we have a lot of thermal tools and electromagnetic tools. So we have one -- so when TSMC announced like 3Dblox last year in all the big foundries, so we are working closely with them to enable this kind of 3D IC. So I feel we are very, very well positioned in this chiplet world.

And so then as we kind of move up from the silicon design to the SDA segment, as you mentioned, the kind of system design and analysis, what all goes into that? And how do you kind of define that opportunity for Cadence?

Yes. So I mean, the chiplet is like a bond between the semi and the system side. So it's a little bit of system but not really all of it because -- but the fact is that 45% of our customers are system companies, these big phone companies or car companies or automotive companies. So of course, we can work with them on the silicon side. But silicon is just one part of the system. Like there is actual mechanical/electrical system, the hardware/software system and all the data. So the key thing is that a lot of those problems, there is a lot of R&D synergy with EDA. To give you an example, like I mean, without getting too -- into the details like computational fluid dynamics, right? Like if you have an electric car, the range of the electric car depends on -- that's why they're all rounded now. It depends on the shape of the car. That's probably the biggest -- or if you look at an F1 car, we have all this collaboration with McLaren and others, the aerodynamics is the fastest determinant of which car is the best, right? And -- or thermal is a big issue in cars or phones and data centers. So like thermal simulation and CFD simulation in terms of R&D algorithms are very similar to EDA. Actually, they are simpler than EDA. Because EDA you're doing transistors at very large scale, like 100 billion transistors. And transistors are nonlinear and this is the most difficult simulation problem. So about 1/3 of EDA is simulation. And we are the world's best at all these simulation algorithms. So we can apply those same algorithms to simulate thermal and aerodynamics. And we have done that organically and with some acquisitions. So that's the first part, there's a lot of R&D synergy. The second part is that there's a lot of customer synergy. Because the big customers, we have a much deeper relationship with them because of silicon. And third part is that there is a lot of growth of simulation and also a very profitable business. Because simulation is always profitable. And then this emergence of digital twin and whether it's for designing planes or cars or molecules is going to be a big area. So that's an entire big space. And that's about a $10 billion TAM expansion opportunity for us. And it's very synergistic in R&D, synergistic with customers. And it's a good area to grow. So that's why we have -- now that's about roughly 12% of our revenue. This year, it's about -- so we started this about 5 years ago. So it's about $500 million revenues, growing like about 20% to 25% a year for the last several years.

Okay. Now with ARM's IPO this year, intellectual property, or IP, a lot of buzz. You mentioned before, Cadence has an IP portfolio. So how do you think about that part of your business?

Well, IP is a good business because we have software to design these things. And then in some cases, we also sell like predesigned components. Now our IP business is very diversified. So there are parts of IP which are very differentiated, like ARM. ARM is very differentiated. And also, they -- and they're a great partner with Cadence, by the way, for the last 10 years. And typically, you buy that and you harden that with software like Cadence. And then part of the IP business is more commoditized, like vanilla, like PLLs or interface IP, things like that. So we want -- we have a good IPs. And about $500 million of our $4 billion is IP. That's also roughly 12%, 13%. But we want to make sure that we focus it on the more profitable parts of it, right, so -- but it can grow a lot. So that's the good thing. But we want to make sure the growth with profitability. That's always our motto, even in our regular business. Because in the end, the investors want not just -- they want take-home pay, so what's the profit, what's the EPS, right, not just what's the growth. So we just want to be -- so we have a -- like Tensilica is a big part of our IP business. That's very profitable. So Tensilica model is very similar to the ARM model. So these are specialized DSP and audio, vision processors, AI processors for the edge. And we have royalty in that there. It's very similar to the ARM model. And then in the interface IP, we do some critical IPs, like UCIe, DDR, PCIe, which are critical for our chiplets and AI infrastructure. So it's a good business. And we will continue to invest in it with a focus on growth plus profitability.

Fantastic. Well, I've got plenty more to go through, but I do want to take a pause if there's any questions from the audience.

Can you comment on the China exposure of the business and how that differs from the non-China part of the business?

Yes, China, I think, is roughly 15%, 16% of revenue for Cadence. It moves around a little bit. It has grown over the years, last 5 years, even though there are some -- sometimes there are some regulatory changes in the past, right? So China, I think, it's true in any other region, I mean, the semiconductors are so essential that there are a lot of companies in China designing chips for a variety of applications, right? Whether it's washing machines or phones or TVs or cars or -- and that's true in a lot of other parts of the world as well. So we are pretty diversified. But we are working with all the major Chinese companies that are designing chips, just like we are in other parts of the world. Now there are some regulatory environment. But that's roughly stable for Cadence. So most of the regulatory is -- we have some effect on it, but most of it is focused on certain kind of chips, like certain -- in the news even yesterday, some kind of AI chips, and then certain kind of manufacturing below 14-nanometer. But we are on the design side and we are in design of all kinds of things. So even though some small like AI chips may have an effect, but given that we are so diversified, it's not that pronounced.

All right. Yes, the one in the back there.

I think we've heard -- we've seen from semi caps, especially this past couple of quarters, strong manufacturing activity in China. And so does also assume that you'd see also strong design. But I think one of your closest competitor on the last call mentioned weakness in China. So I was wondering how -- I mean, how you can reconcile that? Are you seeing any strong design activity in China?

Yes. I mean, I think there are two parts to the China, I mean, just like any other -- there is one extra part. But there's the regulatory environment. I think it's roughly stable at the moment. There's some small changes. And the other is the macro environment. So I think the macro will dominate the regulatory, in my opinion, in China, just like it would do in the U.S. So we'll see. We still see sort of design activity. I mean, the same thing in China, like here, some parts are very strong like, for example, auto is phenomenal, right? And all those auto companies are -- it's all public designing their own chips and moving. It's remarkable what's happening there. And some parts may be weaker if the recovery is not as strong. But overall, I still see a lot of activity. So we just have to -- in terms of what we see, we just had -- this year is good. We just have to see where we end up. Then January, February, we'll have a better idea for next year. But I think it will be more macro issues than regulatory issues is my guess.

One right down here.

With your expertise, which you've described in computational software, can you bring that to areas other than semiconductors? I think you've mentioned biosciences in the past. And can you update us on where you are and pushing into sort of other areas and how much resource you'll put into that?

Absolutely. I mean, that's the whole -- I mean, that's a huge growth opportunity for Cadence is -- so the first area -- I mean, bio is there. Bio, I am very optimistic long term. But in the middle term or short to middle term is the system simulation. So if you look at system simulation, there are like 80,000 -- there could be 70,000, 80,000 customers. So the number of customers is much higher than for silicon. And there are three or four big areas in there. So one big area is what they would normally call like high-tech. So that's like similar to our customers, like electronics companies and silicon companies, but a lot of electronics companies, like they have to do all the simulation of thermal or electromagnetics or aerodynamics. But the other two parts of the system business is, of course, automotive and then aerospace. So we are now working with a lot of the big automotive companies. Now some of them are also designing chips, so that helps. But some of them are not. I think a lot of them are not. So then -- but simulating, like I was saying, the range of the car or simulating the heat of the car. And then same thing, we are working with -- and we did several acquisitions also with a lot of aerospace companies, the design of planes and -- so in the system simulation business, now bio is a different area. But it's the high-tech electronics or semiconductors plus electronics plus cars plus planes. Now of course, there's a lot of commonality in that too. So that's about a $10 billion market. And the EDA is about $13 billion, $14 billion. So that's a very good opportunity. And Cadence is leading that merger of these semi end systems.

Awesome. Well, we are out of time. Thank you, everybody, for your questions. Anirudh, thank you for sharing so much about Cadence.

Thank you.