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

Hi, everyone. Welcome joining us at the 19th Annual Needham Technology, Media and Consumer Conference. Well, on this virtual stage with me today are Nimish Modi, Senior Vice President and General Manager, Strategy and New Ventures as well as Richard Gu, Vice President of Investor Relations of Cadence Design Systems. Before we begin, just some really quick housekeeping items. Well, since this is a virtual fireside chat, please feel free to submit your questions in the Q&A box. We will have some time on the back of this session to allow attendees -- to allow basically our guests, Nimish and Richard to address your pressing questions. As usual, we'll keep your questions anonymous. Next, I'm going to read the safe harbor statement from Cadence. Today's discussion may contain forward-looking statements, including Cadence outlook on future business and operating results. Due to risks and uncertainties, actual results may differ materially from those projected or implied in today's discussion. All forward-looking statements during this meeting are based on estimates and information available as of today and Cadence disclaim any obligation to update them. All right, cool. So Nimish, we've known each other for a while, but since this is actually the first time you join me for a fireside chat with investors, can you give us an introduction about yourself and your role at Cadence?

Sure, sure. Sure, Charles. Thanks for inviting me. Hello, everyone. So I'm on the strategy and new ventures group at Cadence, primarily responsible for corporate strategy, corporate marketing, M&A, and new ventures, which includes running our Molecular Sciences group, which has the OpenEye acquisition, the team that we acquired about 18 months ago or so. I've been at the company now for about 18 years. The first half of my tenure was in R&D, I ran portions of all our 4 business groups. And prior to Cadence, I was at Intel for 18 years in the CPA Group. Most recent role there was in charge of the server CPU R&D.

Great. So Nimish, look, so you have this hard core, right, I call it hard core semiconductor background, based on your 18 years at Intel. Well, your blood is probably still blue. But then you led the systems business and in your corporate strategy, M&A, more importantly, new ventures at Cadence. I think you might be one of the best people in the industry, EDA industry, to really talk about this convergence of silicon and system designs. So we all know this is not really a new trend, EDA. And I mean, Cadence has had this intelligent system design strategy for probably more than 7 years. Can you talk about how Cadence came up with this strategy quite a few years ago? And whether this strategy has evolved, how that has evolved over the past 7 years?

Yes, that's a great question, right? So I think it's a big topic. But what we've been seeing has been over the course of the last few years, there has been an accelerating hyper-convergence in the electronics industry. The hyper-convergence meaning of mechanical, electrical domains, of hardware and software, of semiconductor and system. And of course, all these different aspects are required in the realization of a system. But what's happening is that with the increasing complexity of silicon and systems, more software and the like, and the need for an optimized user experience and the like, the interdependency has grown much, much stronger. And so there's a lot more of co-development, co-verification, co-optimization, which has been needed. And then you can see this in terms of our customer base as well. You've got semiconductor companies, not just doing silicon, they are also moving up the stack, providing more software in this era of chiplets and the like doing more of this heterogeneous integration, system companies, building silicon, coming down, if you will, the stack in building purpose-built silicon. So we are seeing all of these different trends, which are -- have been coming along and accelerating. And so from a Cadence perspective, it's -- we kind of saw this. As you rightly pointed out, I mean, we've had a strategy in place for the last 6, 7 years, our intelligent system design strategy, and that was envisioned the view in mind to kind of help -- kind of to provide this full set of capabilities with straddle silicon and the system world, and we'll talk much more about those. But to enable our customers to kind of realize these different opportunities. The foundational aspect of our strategy at the core, the heart is our computational software expertise that we have, right? I mean, we've been in EDA now for over 30 years. We've developed a tremendous amount of capabilities and skills in computational software. And then when you think about our strategy, think about this as the best way to kind of describe it is as it was 3 concentric circles. The interval circle is silicon, the middle circle is about systems, the mechanical, electrical and the like. And then you got the data circle, which is the outer most circle. And so when you overlay our computational software to these 3 circles, you put computational software and silicon that's EDA, computational software in the system circle and that system analysis and simulations, what we call SD&A system design analysis. And then on the data circles is AI. And so that's basically our strategy. And we've been very steadfastly kind of executing to that. So what we realize is all of those -- the computational software heritage that we have on the core of ED algorithms, the techniques, the solvers and the like are very affordable and leverageable, if you will, to the computational part of the system and the data space. So we are absolutely confident when we embarked on the strategy 7 years ago, we're absolutely confident that there was not just a need for this continuum, if you will, to deliver to the challenges of the hyper-convergence, which are happening, but that we could make a disruptive kind of influence, have a disruptive influence on that. And we're really happy that it's played out. I mean, to the second part of our question, how has it evolved. The strategy hasn't really changed. It's just more methodical progress on the delivery of that strategy. At the core is EDA. We got to make sure you've got best-in-class tools there, but got to make sure that we invest accordingly to make sure that they continue being best-in-class. And they evolve to meet the increasing needs of the market and of the customers. And then we invested -- we did organic investments initially in the system analysis space. And then we have methodically built up a portfolio over there. And the latest of that was with the acquisition that we announced of BETA CAE and we can talk more detail about the systems portfolio. And then we also looked at this in the context of what's the next big thing likely to be, and we feel life sciences is that vertical. And so we also made an investment in a molecular modeling, molecular simulation with this acquisition of OpenEye. So that's basically how we view the strategy, how it's evolved. I mean -- and we feel very, very good about the strategy in terms of that's the -- what the vision is and very pleased with the execution through both organic and inorganic means. And very thankful, obviously, to our customers for their faith and confidence in these solutions. Charles, you are on mute.

Yes. So Nimish, I attended Cadence Live, Silicon Valley last month. By the way, thanks for the invitation. I recall the fireside chat between Anirudh and Jensen Huang of NVIDIA. Jensen is basically saying NVIDIA uses several Cadence products. He actually gave some really specific shout outs to a few products, Virtuoso being one, that's Cadence Custom Design tools and Palladium, you spent a lot of time talking about Palladium, that's Cadence emulation hardware, right? And I think you also touched upon something called the Reality 3D, that's the data center simulation tool. So I'm sure NVIDIA probably buys more than just these 3 products from Cadence. But I found it quite interesting that 2 out of the 3 products Jensen highlighted are not quite the traditional EDA software tools, but more like system design tools, both hardware and software, right? So can you kind of give us an overview Cadence system product portfolio and why they are relevant to the future of the industry, not just the semiconductor industry, but probably the more broader electronics industry or even broader than that, especially in the context of the AI?

Yes. Yes, Charles, I mean, it was great. Cadence Live was great. It was great to have Jensen over there and listen to his views about the massive kind of global transformation which is happening, driven by AI and how NVIDIA is helping accelerate that transformation and talk about the partnership with Cadence as well. I mean, we are really privileged to have NVIDIA as a customer, not just a customer but also a valuable development partner, a teaching partner. And so when you talk about the macro level kind of systems kind of perspective, we talked about the hyper-convergence happening hardware, software, mechanical, electrical and the like. And then you tie that to look at Palladium, I mean, that's a classic manifestation of all of those different things together. Very complex supercomputer that we've built. And this is something which has got hundreds of reticle-limited kind of dies in there, multiple racks, liquid cool, very complex. And it's solving some of the most complex challenges our customers have. I mean, you heard Jensen talk about saying that it was very -- it was essential for designing a [ blackbox ] and so I think this is -- the hardware piece of it is just one aspect of it. Now taking a step kind of extracting it and talking about your question on systems. So when you look at our system portfolio, it's got 2 main components. There's the system design piece, there's the system simulation analysis piece of it. And system design is where we got our packaging, our PCB solutions. We've been in it for over a couple of decades now. And now with this -- all these trends happening on 3D-IC, on chiplets, heterogeneous integration, like it's -- we are in a really, really good position to kind of deliver to the needs of the moment, if you will. And it's a very complex set of capabilities which are needed to realize this promise of chipsets and the like. And we've got our platform called Integrity, 3D-IC platform, which is the only platform out there, which is -- uniquely and natively integrates a lot of -- all of these capabilities under one umbrella. So you -- of course, you need the implementation, Think about this as 3 layers over here, you absolutely need the implement analog custom digital design implementation. Then, of course, you need packaging. I mean, you can't have a packaging solution without our packaging technology there. So the advanced packaging, which is needed. And then on top of all of that, you need system analysis because the challenges of analysis speed, electromagnetic speed, thermal are very acute in the manifestations of these chiplets. And so you need in design analysis capabilities there. So we provide all of these through our integrity platform. And so that's that on the system design piece of it. Now I've said this before as well. But unlike when you look at core EDA, there is -- the Moore's Law is you're forcing function. You have the need to retool and reinvent yourself, keep up with the latest technology changes every 2.5, 3 years, else you kind of fall behind, you become irrelevant. There's not been that forcing function and system simulation. There are tools out there, which have been kind of there for a while, legacy tools. They are good tools, but there was a great opportunity for kind of coming at it and disrupting that market, as I mentioned, with our core EDA heritage computational software leverage, we felt we could make a disruptive difference. And that was our decision to go into the system analysis kind of space. And we started off with organic development, both clarity cells, electromagnetic and electrothermal solvers were organically developed. And they were disruptive. I mean, clarity came in with 10x more performance, no loss in accurate, you had massive capacity. So you don't have to piecemeal your design and then simulate different pieces, stitch them together. You could kind of -- be it the car, be it the airplane and the like. And so this was a very -- a big step forward in terms of kind of starting to have a disruptive influence in that market. So we think that when you look at the space, and you heard Anirudh talk about this, that when you look at the EDA space, a lot of this, 99% of the design is all similarly done digitally and the like versus when you go in the systems space, for various reasons, only about 20% is. So this whole shift of that paradigm, where you want to get away from doing a lot more of the physical testing and move it up in the chain and not just flush out issues earlier, but also come up with a better design because you're exploring more at an earlier stage, I think those are -- that's the opportunity and that's the promise and the potential of the future. And so now we've kind of built up our generative AI portfolio as well. These tools kind of sit on top of our simulation tools, what you call our principal simulation tools. And a couple of them in the systems space, Allegro, for example, our PCB automation, there's never really been much automation in the PCB design space. And now with Allegro X AI, we are really showing some tremendous results for our customers 4x to 10x productivity benefits. One of the customers talked about that at Cadence Live. And then for systems, we have the Optimality Explorer. So this is not just simulation. I mean, simulation, of course, is needed. But it's also about optimization, right? So we obviously want to do productivity and do things faster, but you also want to come up with a better design. And that's the really big opportunity which we've seen a lot of that in EDA, but now to extrapolate that and make that happen in the system space. So that's, at a high level, how we see our systems tools be it design simulation and how simulation tools overlaid with the AI tools and how they are being really helping our customers across multiple verticals.

Great. Thanks. I think I definitely have more questions to follow up on some of the things you mentioned, like you guys want to make a disruptive difference, right, in the systems space, when the simulation kind of -- simulation coverage in the space is only like 20%-ish based on what you just said. But before we go there, right, I think I have to throw in more of a near-term question on hardware. I mean, hardware, I don't know, it's probably still considered as a part of the more traditional semiconductor business, but they had something that feels like a more software, hardware codesign, more of the system. There is a system angle to that. So we know Cadence has been executing well overall, but relative to the high expectation, right, we can debate whether they are too high or not. But the last quarter, well, that was not perceived as a spectacular results or print. So one of the reasons for the guidance, I think that was probably seen as the issue, right, coming in slightly below consensus. It seems like it's a hardware revenue being a little bit less than expected. And part of the reason is the transition -- product transition from Palladium Z2, Protium X2 to Palladium Z3 and Protium X3. So can you provide us some color on why this transition has had an impact on the financials and why investors should or should not worry about it?

Sure, sure. So let me start off with a macro level kind of view of emulation and verification. Verification is a is an NP-complete problem, right? You're never really fully done with that. And it's probably the most challenging problem that our customers face, probably accounting for the largest portion of R&D headcount as well. And then -- so when you're looking at where the market is going, the trends going and the chips are becoming much more complex, Blackwell, right, 200 billion transistors, running many more sophisticated applications and software. The verification challenges grow exponentially. And so when you add and transistors, you're just raising a number of states, right? There's -- and the verification complexity goes up by 2 rates to end. And so the traditional ways of doing verification, which is by simulation, just doesn't -- absolutely doesn't cut it anymore. You're just not going to get too far with. And that's where hardware assisted kind of verification like emulation, FPGA-based prototyping kind of come in. And so when we look at that and you say, well, okay, you've got the simulator running in kilohertz, you got the actual chip running in gigahertz and you got -- in between, you've got the emulation platform that you've got a prototyping platform. And depending upon the use cases and the -- where you are in your design cycle or early-stage RTL late-stage RTL software bring up, you use these 2 platforms. And several of our customers are using them in tandem. Now Z2 and X2 are already the best products in the market. And our R&D team, actually hardware R&D team, performed a near miracle. I mean, they came up with Z3, X3 in just about 3 years, virtually half the time. Now just using Blackwell as an example, right? 200 billion transistor chip, was designed with 8 racks of Z2. Now when you look at Z3,16 racks out there, and we actually can handle about 48 billion kind of gates, which is about -- designs 5x the size of Blackwell, right? So it's really future-proofed in that regard. Hundreds of chip in a rack, hundreds of billions of transistors overall across the rack, Reticle Limited, liquid cooled, InfiniBand, Optical Connect, right? And also, by the way, aside from capacity and performance, there are also some new capabilities that we provide, which is for state emulation, right? So that's really great for low power kind of verification and also analog mixed signal modeling and the like. And so this is a very complex supercomputer, but it's also solving some very, very -- addressing very complex challenges for our customers. So when we -- of course, we have had beta customers. We talk about NVIDIA, ARM and AMD as our beta customers. And so when we looked at the interest which we are seeing from customers when we announced these products, there's a lot of interest from customers on this, and we fully expect a very strong kind of ramp-up of that happening in the second half. Production of these systems happens in Q3. We're taking orders now. And so -- and we are building -- we are planning to scale, right? We're building up for scale right now to meet the demand. And so Q3 is going to be much more compared to Q2 in terms of that. And then Q4, obviously, more than Q3, and it's going to go into 2025 as well. And so given the strong interest in these platforms, we've just been kind of prudent on what Q2 kind of meant and looked like for Z2, X2 and being a little bit more conservative with our assessment over there. But we're absolutely confident about this is the right step to take long term. It's addressing these problems. And when you look at AI and with workloads of AI and the AI chips and the like, I mean, these are just all very, very big chips. More you can put on, the better they're addressing the needs, right? And so you can fully expect this to continue. So this is the right thing for the market, the right thing for our customers. So we kind of came out with these systems, [ RC ] systems, and we're going to be ramping up very aggressively there. And then, of course, Charles, as you know, we always remind investors, right, that, hey, we manage our business for the long term and not to look at any 1 quarter or 2 quarters, first half or whatever. We had a record year on hardware last year, and we fully expect this year to be another record year. So that's how we look at that.

Thanks. Fantastic. Well, especially when you said this new Z3,X3 can -- is capable of design 5x a chip -- 5x larger than Blackwell. Well, the future is definitely very -- going to be very interesting and exciting. So let's abstract or zoom out a little bit more. We know emulation prototyping hardware. Well, it's very successful for Cadence. But in January, you guys also launched another hardware product called Millennium M1. And it's also a supercomputer, but it's not for semiconductors. It's for CFD simulation. I'll leave it to you to explain what is CFD to investors. Even Jensen, I think, he said at GTC that Cadence as a software company is now making hardware, right? I think he's not only referring to Palladium, but also Millennium M1. So can you tell us why this product? And why does Cadence believe the company can become successful in this hardware supercomputer CFD business relative to the incumbent solutions? And please remind us, what are the incumbent competition you have.

So it's a good question, right? So again, let's start off from a strategic perspective. So when we talk about the realization of our intelligent system design strategy, CEO, Anirudh, right, he talks about this in the context of a 3-layer cake and which is very apt over here. Where you look at -- because all these 3 layers are optimally -- the optimal benefits come when you are taking of all the 3 layers. The middle layer is what we call -- is our simulation optimization layer. All our software tools are over there in that layer. We call it principal simulation. It's first principle based, right, physics-based, chemistry-based and the like. And it's providing the physical intelligence. Now on top of that is our AI layer. Our generative AI tools, LLM copilots and the like, call it the data intelligence layer. And then all of this is running on specialized compute. Now could be CPUs, DSPs, GPUs, our own Palladium, for example. And so this is -- you can call it -- it's accelerated compute. You can call it the compute intelligence later, right? So you've got the principal intelligence, you've got the data intelligence, and you've got the compute intelligence layer, right? So when -- Palladium is a classic example of that. Now when we talk about this in the context of system analysis and computational fluid dynamics, we mentioned, right, earlier that in the chip design, 99% of their design is all done digitally, virtually. And you get 99% coverage, if you will, before you tape out. Now the system world, for multiple reasons, largely because of accuracy reasons and the time it takes to do all the simulation. And it's just about as we talked about, 20% -- for example, an airplane, only 20% of the flight envelope is really simulated. And when you're talking about that -- and that's really much more in the context of steady state. But when you're talking about takeoff and landing and the like, there's a lot more complexity, turbulent flows and a lot more effects to be considered even there, only certain portions of that are done with simulation. Of course, the whole thing is still validated, but done by the traditional physical testing and like. And the goal is to kind of move this up as much as possible. Now the traditional CFD tools out there in the market are just -- they're just not capable of handling this level and there's a reason why the state of simulation is the way it is. So we -- when we -- a couple of years ago, we acquired a small company called Cascade from Stanford, very good heritage, excellent team, excellent technology, and they've come up with some remarkable algorithms on CFD, which got much, much more higher accuracy. And of course, you've got to do much more compute for that stuff and the like. And so CPUs were just not good enough. But there also was very, very easily scalable to GPU and you're getting a really remarkable kind of speed up by running that on a GPU. And so that's basically the promise of the potential of this. So when we announced the Millennium M1 supercomputer, it really is that. So we are creating -- it's similar to the Palladium. As you mentioned, it's our own supercomputer, but targeted for a different kind of set of verticals over here and solving a different problem, if you will. And the response from customers has been awesome because one rack, one Millennium rack is replacing like 32,000 CPUs. It's just -- so when you look at that and say, wow, you've got this tremendous performance enhancement and accuracy, this is really going to be helping this whole ship left paradigm to be realized much more, and we expect that to be more and more broadly proliferated. And since the announcement, and of course, we had beta customers at the time as well. But we -- since that early days as well, we've actually had customers who come back and gotten more of this and proliferating it across more of their designs. So it's still very early stages, Charles, but this is, again -- this is a long-term kind of endeavor and we see the potential to come up with something which is disruptive to solve a real and growing problem for our customers. And so that's basically how we view it.

Got it. I kind of feel like you guys -- your software is helping your customers design GPUs. But at the same time, you guys are also trying to use the best technology, like GPUs, into like Millennium type of the supercomputer solutions for customers, who may still be using some really legacy stuff like actual physical testing or CPUs, for example, right, like you said?

Yes. So just on that one, Charles. I mean, just to add on to that. I mean, that's a very good observation, right? So when we think about it from the context of AI, right, and by the way, this -- AI is another, like I said, the data intelligence layer, right? So it's not just the accelerated compute piece of it and the simulation. There's the AI layer on top, and that's the overall value prop, which just enhances it tremendously. So when we look at it from a different lens of how is AI really helping our business, right? In that context, what you just said is very true. So not only are customers using our tools for designing their own AI chips themselves. That's one aspect of it. We ourselves are using AI internally for our tools. We witness all the generative AI tools that we have to make them better and more full feature and more capable for our customers to build better solutions. And then you've got the opportunity with this AI to go into these newer markets or newer areas which were previously not possible to kind of do. And so be it with the reality, digital twin solution that we have, Millennium is an example over here, going into bio and the like. And so I think there's multiple layers of goodness which come together. And so we look at all of this in a very holistic manner and making sure that each of these layers that we are very well positioned on each one of these. And the real value comes in when you look at it as a slice of that 3 layers, like Millennium as a slice, which is providing all pieces of that and going through. Yes, very exciting stuff.