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

So we're off to our sixth session of the day, which means we've tipped in the second half of today. And thank you, everyone, for continuing to listening in to our sessions. I'm extremely happy to have Nimish Modi with me today. He is Senior Vice President of Marketing and Business Development at Cadence. And I'm sure we'll have a very interesting 45 minutes or so. Hi, Nimish.

Hi, Gal. How are you?

Very good, thank you. I appreciate. Unfortunately, not in person, but it's nice to be able to connect anyway.

Absolutely. Thanks for inviting me.

Anytime. And on that note, yes, I want to -- I got quite a few topics prepared for us today. And I think the first one is, which is the one that I'm very kind of passionate about. I'm being spoiled for choice sometimes when I look at my companies in coverage and really the amount of innovation and enablement of the next-generation development of the systems and electronics and everything that they do. And when I think about Cadence, I really put it at the forefront of that innovation. So for me, it doesn't get more complex. It doesn't get more cutting-edge than that, right? We cannot imagine anything. We cannot imagine a Zoom session today without Cadence. We cannot imagine our smartphones or the laptops without it, right? So this conference would not exist if you guys weren't part of the ecosystem in enabling it. So first thank you. And then secondly, there's a lot of trends that are helping you right now. It was interesting, I had a chat this morning with a couple of industry analysts, and one of those was from SimData as well. His observation was, he said, we just went through a pandemic, and it was interesting from his perspective that every company was affected in a negative way throughout that period of time, except for EDA companies, which just shows you how important it is and how much of a trend -- of a structural trend it is, I guess, as we go into this. So I'm going to stop rambling. I'm going to ask you some of the recent trends that you have mentioned that have supported your growth, have been the applications like artificial intelligence, the cloud computing, continued move to the cloud, 5G as well, we're starting to roll it out. Which of those drivers, maybe the ones I mentioned or the ones that I've missed out, are the ones that have kind of been the most incremental to your demand just in terms of mind share and maybe financially as well. And then how do you see those things play out over the next 3 to 5 years? Is it the same things? Or are we looking into some other areas as well?

Yes. No, that's exactly a very topical question. But before I answer that, I need to take care of some logistics or safe harbor statement. So let me just read that first. Today's discussion will contain forward-looking statements that will make use of certain non-GAAP financial measures. Please see our most recent 10-K, 10-Q and website for a discussion of risk factors and our use of non-GAAP financial measures. Okay. So before I just answer that, Gal, maybe just a high-level kind of background of Cadence, and a quick intro, if you will. So we are really a technical software, more specifically, a computational software company, providing software tools, IP, hardware verification platforms to semiconductor customers and increasingly system companies to help them design their packages, their chips, PCB boards and the like. And we've got a customer base across multiple verticals, several of them driven by the trends that you just talked about, and we'll talk more about that in a second, but more also upcoming in industrial and medical, aerospace, defense and the like. And then we partner very, very closely with our customers, what you called them, the market-shaping customers helping us drive our roadmap, defining the roadmap effectively. Very strong innovative culture. Truly believe that you lead with innovation, lead with technology, lead with the best product and help their customers deliver their best differentiated product, be it in terms of power performance area or other system attributes there as well. 40% -- about 40% of our revenue does go into R&D. And over the last 3 years, we've developed and introduced more than 20 significant products. And so our strategy in what we call the Intelligent System Design strategy, the base layer is core EDI design excellence. Just talked about that. We provide end-to-end EDA flow as well as IP. And then we're expanding beyond EDA into the system space, and we'll touch upon that as well. And then there's a third layer, Pervasive Intelligence, which we are applying our tech know-how and AI to our products. And so bottom line, there's a lot of excitement around all the design activity, which is driven by these generational drivers and as we execute to our strategy, we also believe that we can triple -- we are tripling our current TAM from a $10 billion to about $30 billion over the next 5 years. Now, what we are seeing play out right now is that the core EDA industry is benefiting from a number of these drivers, we call them generational drivers because we think they are going to play out over several years. This is not short-term facts. These are each one, so be it 5G, ML, AI, hyperscale computing, autonomous driving, the whole industrial revolution 4.0, these are all at different stages of their maturity cycle, and they're just going to continue playing out over several years. It's all our data. You heard Lip-Bu talk about that. There's tons of new data being generated every day, I mean, 90% of today's data didn't exist 2 years ago, right? And all this data needs to be processed, transmitted, stored, analyzed and requires a lot of innovation for high-performance computing, high-connectivity bandwidth, high-density storage and everything has got semiconductors at the foundation. Then you click -- double click on the next layer down, and then you see what the trends are, which are happening in the space of let's say domain-specific computing, it's all about workload optimized computing. General purpose compute fabric is not always the right answer. You should be able to take a CP or a GP or DSP or FPGA, ASIC, whatever is needed to provide that optimized experience. So that's one trend, very strong trend. The other thing is we are seeing increasingly vertically integrated system companies, in that they are building their own custom silicon for their differentiated IP, a whole host of new silicon startups coming up. And then, as I said, the digital transformation occurring in other industries as well. So all of this is actually driving a ton of design activity, it's got silicon at the foundation. It's a renaissance of semiconductors. And we see every one of them contributing towards providing a pretty strong tailwind over the midterm and the longer term. And so ours -- from our strategy perspective, we got a very broad EDA end-to-end flow, right? We got the IP portfolio. And increasingly, we're providing more and more of the system stack. And we're in uniquely ideal position to kind of capitalize on these opportunities. So that's -- it's really exciting times to be in this industry.

Absolutely. And it seems like that. It sounds like that. And that also -- it's an industry that's fairly concentrated as well in terms of the mind share and being able to deliver those solutions, right? So it's kind of a sweet spot for investors, who like competitive advantage driven by R&D effectively, right, because that is unlikely to change anytime soon, I guess. So one thing that you touched on, and you said a few really important things that we're going to expand on a little bit. But one thing that maybe I'd like to start with is this constant -- I don't think it's a tug of war because it's a good thing. But the drive between the leading edge, the market-shaping customers that you referred to, who are more -- who are all about Moore's Law, right? Who are all about shrinking the chip, making it more powerful and more efficient. On the other side, you have the -- you mentioned systems companies, who are asking for different things. They are asking you for parts of IP that they can use in order to build the building blocks, but, again, with a special purpose in order to put it somewhere within their technology stack, let's say. Those 2 forces are interesting because on one side, it diversifies away a little bit of your exposure that you've had previously on the largest chip, leading-edge chip customers. But then at the same time, it also provides this long-lasting tail of growth. What have you been seeing over the last couple of years in terms of that balance between the core, let's call, core, the leading-edge customers, and then let's call the systems companies and maybe start-ups and smaller companies as the driver of that incremental demand? Which ones have been more powerful to you in terms of your performance? And where is your incremental maybe R&D spend going, which ones?

Yes, good question. The commonality over here is, as I said earlier, very strong design activity. So what we have seen over the course of the last few years is that, that tide has lifted kind of both boats, right? It's the semiconductor side as well as the system side. And that's really, really been a strong driver, and all these trends that we talked about are going to continue in that mode. Now having said, just double-clicking a little bit on some aspects of the Moore's Law, for example, that you mentioned. While Moore's Law has definitely slowed down, there still absolutely continues to be a relentless kind of march to get to the lower and lower process nodes. There's still gains to be had, right, by going to these lower geometries. And while they may not be as much as what has been seen historically in terms of benefits, they're still meaningful. And especially when you talk about certain vertical segments, be it the leading-edge mobile guys, the hyperscalers, AI companies for that matter, Moore's law drives the -- the first sort of the digital segment, right in EDA. That's the largest TAM market in EDA. And we have talked about what we've done in that space with respect to revamping our entire portfolio, focusing on the leading quality of results in power performance area, and we've got some really, really good momentum there with the market-shaping guys, proliferating not just individual products but this trend that we are seeing, where it's now more the full flow. And we give commentary on that in our calls as well. We had 45 full flow wins last year. Now, the Moore's Law trend over here in this regard is not just turning to semiconductor companies per se, number of system companies in these verticals I talked about are also developing silicon at the most advanced node. They're also pushing Moore's Law. The difference though is that the needs of the system companies are much more than just at the chip level, right? They need -- or the eSoftware, I should say, they need IP, you referenced that. I mean, we talk about high-speed SerDes IP. They're pushing the boundaries on advanced packaging, hardware-software conversions, increasing the system analysis. I mean -- so there's a whole bunch of other requirements that they need. And if you look at hyperscalers, they are a great example, in that they present a very good opportunity for both our semi portfolio, meaning chip level portfolio as well as the system solutions. And we have got very strong momentum on this. We are closely collaborating with them on their designs. We call out aerospace and defense as another example of a vertical that we are doing well in and where we've got strong momentum in, in our most recent earnings call. So I think what you're seeing is that over this increased design activity plus the portfolio that we have, which is the IP and then you talk about the next layer, it's a more staircase strategy as we build up our portfolio. You've got the IP at the bottom, then you're serving the chip with the EDA portfolio, you've got packaging, you got the PCB, system analysis on top of that, our partnership with Green Hills for embedded software and -- embedded software, right, sitting right close to silicon. And so I think we're servicing more and more of these requirements from the system companies. Our engagements with system customers are steadily deepened over the past few years. And a few years ago, we had said that revenue mix was about 40% coming from system companies, and now it's closer to 45%. And so while the tide has lifted both boats, both are doing well. The ramp, if you will, that has accelerated some in the system space over the past couple of years or so.

That's really helpful. And that's probably a trend that kind of continues because there's -- it's just the power of these companies are still very, very large. But then at the same time, there's many of them, right?

There's many of them. And to that point, Gal, I mean as we proliferate our portfolio, right, as we go into systems analysis, we can talk about that in a minute, but it just opens up a whole vista of other system companies that we are not participating in today, right? So I think that's just the overall trends over there are absolutely in line with what we just talked about.

Absolutely. And you mentioned -- so we also talked about some of the trends, and I want to make sure that I touch on that because you said autonomous vehicle is a good example of it. Vehicles, effectively are smartphones on wheels, on steroids as well, right? So smartphone cycle was a good cycle for you as a company. Your growth definitely corresponded to that, and it was a nice thing. But if we say that automotive becomes the next smartphone move over the next decade or so, that alone is one of the lasting drivers that kind of sustain, right? It's not just about the hyperscalers, even though like you said...

No, no, no. I think -- exactly. So this goes back to the drivers that we were talking about, right? We're seeing the 5G, AI, ML, hyperscale autonomous as 1 of those 5 in this whole digital transformation.

No, everyone has developed in their efforts, that's what I was saying.

Exactly. There are all the different stages out there of their maturity curve. And this is why we call them generational. They're going to play out over several years. I mean -- and they're all in different stages. And the commonality there is you've got -- each 1 of those is huge by itself, right? And the commonality is that they are all to trying design activity, right? And so that's good news. And the broader the portfolio you have, the better off you're in a position to service the needs of them.

One question I had is, when we think about that split between the systems and the leading-edge companies. When -- leading edge, when you're developing a new node, the whole set of tools effectively adjusted because they are not 100% out of the box. So you continue to innovate alongside with your customers, right? Because of that, historically, when most of the exposure was to the leading-edge companies, the margins kind of reflected that as well at the same time. On the other side, I wonder whether now that you see increased lasting demand to the systems. The interesting thing there would be that perhaps the R&D is more upfront, and then it's a lot more scalable in terms of the technology. Because like you said, there's many, many more customers that can utilize the same technology that you have developed. Is that the right way to think about it?

Well, I think, again, R&D, we put a big premium on R&D. As I said, 40% of R&D -- of our revenue goes into R&D, and we are very methodical and focused on ensuring that we -- there's a beat rate to this thing is where you invest, how much you invest and when you invest, right? I mean -- and you got to keep on kind of continuing looking at that. And all of this is driven by the direction we get from our customers. We are looking at the top-level trends. We are looking at where the customers are going. We're looking at what the customers are asking us to provide them and then you kind of invest in there. And obviously, you're investing way ahead of the curve, for -- when you see -- we talked about on our call, for example, that on digital that we already got -- we are in collaboration with 2 nanometers right now. So that investment is a way ahead. And then the needs of the customers are different, to your point, whether the semiconductor folks or system folks. And we are looking at -- as we go through these engagements and you figure out, right, what is in the end, what is -- you're leading with innovation, what is the right product to build? And then we are not just talking about point to invest best-in-class products. That's a basic first order of business. You've got to have tools, solutions, which are differentiated. But more importantly and I alluded to that earlier in different context is there are platforms. You're building platforms. You look at a digital full flow, that's a platform. Verification suite, it's a platform. RF platform, we are building up a multi-physics platform. And so it's very important to look at the strategy from a holistic perspective. And look at it -- again, be guided by our customers and then make those investments in a very cognizant manner to meet the needs of the customers, while still making sure that from a margin perspective as well that we are paying heed to that very closely.

Yes. It definitely seems like a natural development, and it's almost like -- it's -- if the initial work was the hardest work, the incremental R&D dollar returns on much better scale, better yield than what historically has done.

Right. And different 2 parts of the portfolio has different beat rates in that regard, right? I mean, digital is closest with Moore's Law where you've got to retool stuff every 2 or 3 years, right, versus a simulator, which is different, right? But in the end, you still want to continue innovating, be it driven by process technologies or be it driven by the complexity of designs and the needs of our customers, right? So always got to stay in this innovation treadmill, and we're really happy about the pipeline that we have on that front.

That makes sense. The other topic I wanted to focus on is kind of a continuation of what we just talked about is just specialty systems companies are way more open to using some IP building blocks, the technology that you have developed in sometimes in cooperation with them and then resell to many other customers in the space. You entered the market, you've been extremely successful. It continues to outgrow the overall EDA segment as well. Even though as we talked, EDA is a healthy growth business today. How penetrated is the IP market today? And what I mean by that is people always develop their own tools, they always develop their own solutions. But at the same time, they will have a trade-off between developing and reusing something that perhaps is not their competitive advantage. So do you think there's a lot more room for IP to continue to outgrow that EDA part by a few percentage points per year?

Yes. I think the IP has been a very good business for us, as you pointed out. It's been growing faster than EDA for us. And then about 13% or so of our revenue last year did come from the IP business. And when you look at -- and there are the different parts of the IP business, right, there's the standard-based IP, interface IP business, memory IP, and then we got our processor in Tensilica IP. But just talking about the design IP, if you will, I think customers are going to continue outsourcing that, the standards-based IP, right? USB, PCI, whatever protocols. But this type of IP is really not a really differentiator for the customers' product, right? As long as there's high-quality interface IP, or memory IP for that matter, as long as you can obtain that from commercial vendors like Cadence in the process node in the right time that you need it in, the customer is much better off focusing their R&D investments on what they can uniquely differentiate themselves in. And there, you can see that the system company, some of these hyperscalers, for example, that we talked about, there's no legacy to outsource to begin with, right? I mean it's like they are starting from scratch, and so they're just outsourcing from the beginning. And so yes, I think the outsourcing trend is going to continue. There's still a fair amount of IP which is still being done in-house in different companies. And I think, increasingly, that trend will continue. So macro, I'm just talking on macro level, right? We don't guide business by certain segments or certain quarters, a year or so. But overall, I think the IP trajectory is in that regard. And by the way, these IP building blocks that we are talking about, be it high-speed SerDes, PCI, they are very complex, right? I mean a high level of domain knowledge is required, not just the domain knowledge of the protocols, but digital mixed-signal design, verification. I mean just -- so the point is, why do it if you can get it from somewhere else when it's not really a true differentiator, but it's a checkmark. You do need to have it. And I think that trend is going to continue to be favorable.

I think that's an important point because some people might say, oh, if it's an IP, if it's a building block, why doesn't someone else come and just do it, commoditize it, do it cheaper or something. But the point is, it's still very, very -- the moat is still there. It's still very differentiated and complex. It's just -- it's not a driver of differentiation for that customer itself because they want to focus their limited resources on differentiating in other areas. Is that fair?

Yes.

Okay. Awesome. Why -- so I guess answers that -- well, in part answers my last question on the IP side, which is why do customers decide to outsource part of R&D? At what stage are they at the risk of basically losing that competitive edge, competitive advantage versus doing it in-house? How much can they outsource in order to still have that competitive advantage?

Right. And that kind of goes back to our previous conversation on this, right, which is if that piece of IP is really -- it's a needed piece of IP, but you can't really differentiate yourself per se from that, such as these protocol -- interface-based protocol IP, DDR, PCI, HBM, whatever, then it's -- you're not really losing a competitive -- there's not that differentiation to begin with. Obviously, you need to make sure it's high-quality IP, you get it when you need it. And that's basically where commercial vendors, who focus on this, stand out. And I think that's where we've seen a good tailwind to our business in that regard. I think these companies are just doubling down on those pieces of IP, which they really uniquely can differentiate themselves in. And so I think it's a good symbiotic relationship over here. And we continue seeing a lot of pull for IP, especially at these advanced nodes when we talk about our STAR IP business and our design IP portfolio business like DDR, PCI, high-speed SerDes. We talked about this on the call that we have similar design wins on some of these at 5 nanometers and lower process nodes, so they're really pushing the envelope on the process technology on the complexity, and we are partnering with them to provide them with that. High-speed SerDes, for example, when you talk about hyperscalers, right, that's a key component of the infrastructure and the 112-gig SerDes, I think we talk about Xsight Labs who's got working silicon over there. Again, very complicated IP, but -- and great to be partnering with many of these customers where we can help them in that regard. Tensilica, by the way, is different, right? That's the DSP IP, and it's something which we're really excited about. The legacy of that stronghold has been in audio applications per se. And that's obviously proliferating to wireless stereo and as we've talked about smart speaker applications, for example. But increasingly, we're seeing that in automotive as well, right, wins in functional safety and radar, digital audio, radio and ADAS with a vision processor aspect of it. And we believe that given the characteristics of the IP in a high performance, low power, we think it's highly configurable. It's ideal for edge inferencing applications, for example, right? So we've got a SKU on that, our DNA, deep neural network accelerator SKU. And so again, early stages, but this -- we're all talking about where the trends are and how we are positioning ourselves with the portfolio we have to best help our customers take advantage of it.

Awesome. And what you alluded to is that IP does not come just on the digital side, it comes from both the digital and the analog side as well, right? So -- and your business is today very well mixed in terms of both in terms of the exposure, right? Okay. So shifting a little bit into the other area, which is something that we and some of our investors would be very familiar with. Cadence specifically has been investing over the last couple of years a lot more into that systems area, especially from the physics perspective. You've been effectively doing simulation, but let's call it full flow, right, digital flow from design to verification on the chip level for a very long period of time. And you can only do that virtually. Now you're looking into the area of the system engineering, where a lot of the physical properties are still being tested in real life with mechanics, right? They're still labs that don't exist in electric design. So my question there is, do you see the convergence of that EDA approach more into the -- through simulation into the mechanical space as well? And is that the reason why you've taken steps recently for -- first for verification of the EDA tools, specifically for your solvers that you introduced, I think it was 1.5 years ago when you first introduced the portfolio there? And now even with the recent acquisition of NUMECA, which puts you more into other types of physics, which is computation of fluid dynamics. Is that because you see that conversion happening?

Yes. It's a great question there, right? So absolutely, the trend. When you look at the bigger picture on the electronics ecosystem, the worlds of mechanical, system analysis, electrical, embedded software, I mean, you can look at them as kind of overlapping Venn diagrams, so that overlap increasingly kind of happening and a lot more requirements by the end system companies, for example, to participate and they need all aspects of these portfolio, right? I mean -- so I think that trend is driving -- is one aspect of it. The other piece of it, as you're saying, the shift left paradigm, if you will, simulate prototype before you do the physical prototype or the physical chip is available, clearly there are multiple manifestations of that, both pre silicon as well as post silicon in the context of digital twins and the like, yes, those trends are absolutely happening. And Cadence's kind of view on this from a strategic perspective was very simple. I mean we've got a very strong base on computational software competency in the EDA, built over the last 30 years. And as you said, we are already doing simulation at the chip level, be it functional simulation, circuit simulation and the like. And it's a natural extension. We are very near adjacency and natural expansion into system-level simulation. Obviously, in this case, in the physics domain and the goal over here is to apply all the goodness that we have had into this new domain and bring the power of that innovation that we have developed and delivered on the solvers and apply and extend it to this new area and build up a multi-physics platform. The space is very exciting to us, obviously, a $6 billion TAM opportunity growing at high single double digits. And it's meaningful, not just to several of our existing customers. Because remember, we've got a base of customers, who are not just our EDA customers and IP customers, who are also PCB customers, right? And so it's not just applicable to several of our existing customers, but helps us expand into newer end markets with a whole vista of new systems customers. So what do you bring to the table, right? This is where the goodness of the computational software and strong innovation that we've done in the EDA space helped us deliver, as you alluded to that, clarity on Celsius 18 months, 2 years ago. And we started off with electromagnetics electrothermal and brought the differentiation of significant performance improvements, right, 5x 10x performance improvement with the reduced memory footprint requirements and while making sure that the accuracy is not compromised. And we give progress reports on our calls about how they are going. Bottom line, I mean, we're very excited by the reception we have seen and it's ramping well. We're seeing repeat orders. And that's always a good sign. When you have repeat orders, it just tells you that what you are delivering is registering with the customers. And obviously, still in the early stages, but off to a really good start. Now we've decided that, okay, with Clarity and Celsius, we're addressing about $800 million TAM market of that overall $6 billion TAM. And then NUMECA, we said, let's move into CFD, computational fluid dynamics and NUMECA was a great opportunity for us to partner with them. And now we have tripled our system analysis TAM to $2.4 billion and again, go back to what are we trying to do? We're trying to build up a multi-physics platform, methodical manner, a thoughtful manner. And NUMECA is great. I mean we closed it, I think, 2 or 3 weeks ago, and they provide all the geometry processing, the measures, the solvers, post-processing optimization capabilities and as I said, it opens up to a whole bunch of new markets because we've got more than 450 customers. We have referenced Ford and Honda and NASA and it's across aerospace, automotive, industrial, marine, right, some of the existing markets we're in, but to a different class of users in those customer base as well as some of these newer markets -- end markets there. So bottom line, very excited about our prospects in this space. And the one thing which I will say is also, in addition to the technology, the innovation, the new customer opportunity is also the margin profile of this business, we are talking about in simulation, I mean in a steady state, obviously, we're investing ahead of the curve right now, but in a steady state, we absolutely believe that the margins are going to be equal, if not better, than EDA.

That makes sense. I mean we do have data points that kind of prove that as well. And I guess my observation to that would be that because there's still a lot of replacement of physical testing going on, the market can gradually grow into itself much more than if it were just a completely penetrated market today to start with. So I guess you also shared that view that we're just in the first stage of that simulation.

Yes, simulation is going to be -- it is big, it's going to be bigger. I mean all that shift left also points to the democratization...

You didn't buy it because you want to be in $6 billion market. Right. You didn't want to be in $6 billion market. You obviously believe that the market goes from $6 billion to something else. Is that the point?

Yes. We'll take $6 billion for starters, right? We're in a $10 billion market right now for EDA and IP roughly. And that's a significant, you know, but yes, that market, as I said, has got a lot of the right characteristics for us to be very interested in.

Nimish, your focus on go-to-market as well, obviously, based on your title. And my question there is how does NUMECA and having Clarity and Celsius and how does that augment your go-to-market approach from what you were doing previously, where I guess it already changed when you were going more into the systems companies to start with. You didn't have to -- I don't think Intel needs an introduction into who Cadence is and what they do and the same with AMD and all that stuff. But when you go into the Teslas of the world, the Volkswagens of the world, it's a little bit different. It's a different discussion, right? So my question there again is how are you adjusting to that life in making sure that you're addressing those opportunities?

Yes, that's a great question, right? So while we talk about -- you've got to lead with innovation, you've got to have the right product, but then you got to need to have a delivery mechanism to reach the right customers there. And the first part of that, obviously, is to raise awareness of what you -- what Cadence is or who Cadence is and what we bring to the table. And so yes, we've got a lot of initiatives, both on our marketing front, digital marketing front as well as on the absolute go-to-market augmentation that we need. One thing you referenced there, actually, it's also in terms of the volume of customers, right? When you talk about core EDA and IP, it's relatively concentrated. You're talking in a number in the hundreds per se. With PCB now, you're going into the 1,000s, but with system analysis and the like, the totality of the opportunity there you're looking at tens of thousands. So obviously, you need to augment your existing go-to-market to increase our reach, right, to get to these new customers. Now our core business, traditional business largely use the direct sales model. And of course, with PCBs, we do have the indirect channel. We have partnerships with our value-added resellers and distributors. Now with -- what are we doing with respect to the systems side of the house, so for the direct channel, obviously, we already added a systems overlay, right? And we are working on expanding our footprint with [ AVaaS ]. But given the nature of the business, we also believe, for example, the cloud can be a great delivery mechanism over here as well, right, and over time, can expand that because especially for the small and medium companies there, maybe even the touch -- it's a lower touch model than your full-fledged EDA solutions and the like. So I think there are multiple aspects of the go-to-market that we are working on augmenting, enhancing, but it's a very important aspect of this. I think it's something -- yes, you've got the product, you focus on the R&D aspect of it. But in this case, there's going to be a big change as well, which we need to make for enhancement, I should say, to make sure that we maximize the opportunity with that -- the full customer potential base out there.

Yes. And so you said you're addressing about almost half of the market. CFD is a huge area, obviously, of the whole CAE market. Are you -- what's the plan? Are you stopping there? Do you think those are the key applications, you think about other types of physics over time as well? What's the end game for Cadence?

Well, I think the endgame over here is, as I said, to build up a multi-physics platform. And the same formula, if you will, or philosophy or approach that we use in EDA would bring over here, listen to your customer, right? What do those customers want? When you get a customer, what are they wanting? And how can you provide more and more of that in a differentiated manner to add value to our customers? So yes, we are absolutely going to be continuing to listen to our customers over here. Now, we are still in the very early stages. So now we've got Clarity, Celsius and NUMECA portfolio, and we want to make sure that we are absolutely maximizing that, proliferating that, growing that, integrating it, coming up. And so this is a long journey. We are in the early stages of the journey, and we want to make sure that we -- as we march up in terms of realizing the buildup of the platform that I'm talking about, we got to do it in a very methodical manner and thoughtful manner. And so yes, stay tuned. I mean we will continue providing progress reports as we go along on how we are doing. But clearly, I mean, as we said, there's a ton of opportunity there, and we're going to do our best to maximize that both for our customers as well as providing the technology to them.

Got you. Nimish, one of the -- when we talk about the market, obviously, one of the markets that's been very, very hard for you, even though we've had trade restrictions in place, has been China. I can't even imagine what it would have been if you didn't have restrictions of selling to particular customers. But this historically has been high single-digit revenue contributor to your business. Now it's high-teens, kind of becoming mid- to high-teens, let's call it, mid-teens for the full year last year. I guess my question there is as China continues to develop their own semi strategy and the know-how, is this a new normal? Is this a new normal where China should be spending a fraction of U.S., where the U.S. is spending? However, it shouldn't be 1/10 of what U.S. is spending, it should be probably a little bit more and as such, it's just a normal development of the market maturity? Or is there anything specific to that?

Yes. I mean I think macro level -- I mean I'll touch upon China, but I think it's important to understand, it's not just China, I think overall APAC, right. We've done very well over there. Taiwan has been really strong, Korea as well. System companies in Japan doing well. So I think overall, the APAC region has been a strong growth region for us, and we believe presents the growing opportunity. Now China, specifically over the past several years, I mean, they've talked about the need to get to autonomy, if you will, and it's committed to growing out -- it's building out its semiconductor ecosystem. A lot of investments going in it from the government. Many companies going to IP or the STAR market, very high valuations there. And you look at the amount of design activity which is going on there, because clearly still the consumption is far outstripping what their indigenous production is. You're talking about more than 1,500 tablets companies out there. Long tail, for sure, but the thing is they're doing a wide variety of designs. I mean some of them are very sophisticated designs for most advanced nodes. So it's a very big opportunity, I think, for the breadth of our portfolio across end-to-end EDA, IP hardware, system solutions as well. Now you mentioned some of the restrictions and, of course, we're going to do all we can to support our global customers, but we are a U.S. company, and we're going to comply with the government export regulations. The good part about this is that Cadence has a very diversified global customer base, right? We got now 10% thus far. And so I think China, last year, you alluded about 2020. Now 2020 strength obviously was there. It was unusual trend. There was an unusual year for many reasons. And we had -- during the second half, we had some specific macro-level tailwinds, primarily in Q4, the 53rd week, for example, right, is one aspect of it. John also talked about we managed to kind of recover $26 million of collections that we previously thought were uncollectible. So that was also largely a Q4 thing. And then in China, we saw a higher-than-expected mix of our upfront revenue products. And we have assumed that going forward, this year, we would see it settle down to a more traditional mix of recurring revenue versus upfront. So when you put all of this together, we are kind of lapping some tough comps, and we decided it has to be prudent until we get more visibility into it. So yes, it's a fluid situation of ongoing uncertainty, but the macro level, stepping aside, taking a step back from one specific quarter, one specific year, I think it's just overall, there's a strong demand activity, and we're going to do the best to try and capture that and do the best to support those customers there. The one thing I would say is that for our guidance as well, back to the export control thing, we are obviously going to assume that the export control regulations that exist for these certain customers [indiscernible]. Right. There's no change. Anything changes that's meaningful enough, okay, we'll talk about that.

That makes sense. Yes, it's interesting what you mentioned then. Is there an argument when you see trend? Because that's been happening as well in the other areas, not just in China, specifically. But what we've seen is when you see strength in upfront, especially the hardware purchases, software tends to follow on the back of it, right? It's almost like it's a platform to do more design starts in the future as well. So whilst I completely understand it because it's upfront revenue, it's a little bit more volatile, it's a bit more lumpy. But it does reiterate that someone is investing CapEx in order to do more ongoing development going forward, right? So from that perspective, it should be a proof point of the design activity is there to stay.

Exactly. I mean -- it's exactly. It's a proof point of design activity. It's just that when you look at things like emulation, like these are more product-specific decisions on timing of where you are in the product cycle. I mean you're coming close to tape-out or -- and the like, you want to run a ton of more verification runs and the like. And so this is why we say that there's lumpiness in 2 regards. There's lumpiness in terms of when the need is coming through and the secondly, because it's upfront revenue, well, it adds to the lumpiness, you see it right away, right? But the underlying theme over here, as you point out -- correctly pointed out is, yes, strong design activity, another proof point of very strong design activity.

Got you. Awesome. We're pretty much coming to the end of our time. So I'm going to squeeze one last question in, and I'm going to try to be very specific. And it's one thing that we've talked a lot about throughout the day today, even within the Cadence simulation and mechanical simulation about the adoption of cloud. And the general feedback has been that customers today when you almost are segmented to start-ups, they don't want to have the upfront purchases of hardware that are necessary in order to do some of that. So they utilize more of a cloud technologies, cloud-based workflows as well earlier on. But the other side, we're hearing that to high end, you would call them leading edge, but some other companies have exposure to, let's say, large OEMs or whatever, they tend to not utilize as much cloud unless the -- it's a one-off demand for kind of burst capacity. Is that the same driver in the EDA market as well? Because you have had a very, very clear, I'd call it, a cloud strategy, if your products are cloud ready, but it almost feels like that the market will be cloud ready as they grow up into the market because a lot of them still have their own infrastructure that they run stuff on.

Yes. So I think from our perspective, all our EDA products as well as our hardware, Palladium platforms are available on the cloud, in our cloud solutions, our offering customers. I mean what is the bottom line, it offers the flexibility to the customers for a different range of -- full range of different use models, customer manage as well as Cadence manage depending on what customers prefer. And then they can get compelling productivity and scalability benefits, right? And especially when you look at those kind of workloads or applications, such as simulation, characterization, right, workloads, we re-architect it to be cloud optimized, cloud ready, and they can really benefit from the elastic compute, if you will, which is offered by the cloud. And we gave some commentary on the call about how that's gone. It was a great year for us in the cloud, more than [ 160, 175 ] customers and there are different types and sizes using cloud offerings, to your point, and either all cloud and/or hybrid cloud kind of environment, right? And customers who would fit the model that best aligns with their current needs, so it often means that the largest companies, in general, might be more interested in addressing pre-cloud needs, while the small and medium-sized companies would probably be considering more full flows. So I think we are still, again, in the very early stages of this adoption from the EDA industry perspective and design in the cloud per se. The adoption is going to accelerate. And from our viewpoint, we are exploring new business models to go along with different kinds of configurations and packaging possibilities to provide customers even more options, right, and more flexibility in productivity. And I referenced earlier in the context of system analysis go-to-market, where cloud could be, that's another kind of avenue for us. It's not just a delivery mechanism, but could be means for customers to actually accelerate in many ways, some of the -- some of their business needs per se. And so yes, it's -- we are off to a good start. We're here. This is going to play out again over several years. But I think given the portfolio and most importantly, the relationships and partnerships we have built up with our infrastructure providers as well as the foundries, I think we're in a good position to help customers scale their offerings in the cloud as well.

Yes, nice to see another driver. Perfect. Thank you so much, Nimish. We're out of time, unfortunately. But again, I appreciate your time as always. I'm a looking forward then to the next debate. Cheers.

Absolutely. Thanks for having me. Thank you.