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

All right. Well, I think we're live. My name is Gary Mobley. I'm with Wells Fargo Securities. And the analyst who covers the EDA software sector, which includes Cadence Design and up on stage with me is the management team from -- or part of the management team from Cadence, including John Wall, the CFO; and Richard Gu, to my far left, who has Investor Relations. Guys, I really appreciate you joining us today. And I don't say this to all my management teams, but you're one of my favorite. So really glad to have you guys here. And so John, for the purpose of the audience here in the room and online, I was hoping that you could just give us an overview of Cadence. And give us a sense of the breakdown between your different products and exactly this, what it is that Cadence does.

Yes, sure. Absolutely. Just before we start, can I have Richard to do our disclaimer?

Yes, of course.

Yes. It keeps us clean.

That's reasonable.

So today's discussion will be -- will contain forward-looking statements, including Cadence's outlook in future business and operating results due to risks and uncertainties. Actual results may differ materially from those projected or implied in today's discussion. 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. All forward-looking statements during this meeting are based on estimates and information available as of today, and Cadence disclaims any obligation to update them. With that out of the way, back to you.

Great. And thanks, Richard. That keeps us clean with all our legal folks, our legal folks at Cadence are very conservative. They make sure we do this all the time. And thank you, Gary. I'm sure Gary said that about all his -- all the people he will be doing. So I know Gary knows this team. We've had a long relationship with Wells Fargo and with Gary. And he understands the space really, really well. The -- but for those of you that don't know Cadence, it was formed back in the late '80s, and basically a company that was created by engineers for engineers as when people realize that you're going to need other engineers to create tools that engineers need as they move down the process nodes and chase Moore's Law. The biggest -- maybe the best way to describe Cadence now is, I mean, we're always customer focused on solving the biggest customer problems and what customers have been looking for, for the last 5 years or so is they want ideally one unified platform. that takes you from chip design, chip verification, packaging and board through to electronic system analysis. But -- and ideally, they want to do all of that concurrently. Traditionally, it's been done in a chronological order that -- but the -- but they want to do this all at the same time so that if they tweak a design, they want to see what the electromagnetic impact of that is at the same time. But before that section, people generally when they're trying to design an electronic system, they'll choose some IP off the shelf and then focus their efforts on where they want to differentiate when they start their own chip design. That's a -- in Cadence's business, we have some IP. We focus on differentiated IP because the important thing there for us is just to have a seat at the table. It's the least profitable business at Cadence. It's about 10% of the business. But like I say that once customers have selected the IP, then the real work starts in terms of that chip design verification, packaging board and system analysis. On the chip design part, we have -- generally, there's analog, digital and mixed signal design that's where you have the mix of both. Cadence was born as like the analog EDA company, that, we started off in analog. Synopsys, our competitor is -- was a digital version of Cadence. They focused on digital. And over the early kind of decades, both of us converged in each other's space and kind of soaked up much of the other smaller EDA players, so that we're down to really 2 big EDA players, and Siemens plays with Mentor Graphics in some of the higher process nodes. The -- so there we have custom IC is our analog piece and digital is our digital business. We feel our destiny is always -- was always going to be the largest company in core EDA. So that's why we focus on this particular area. And the reason for that is that we started off maybe about 70%, 80% of the analog markets, we're probably higher now. But -- and then we always felt it was only a matter of time before we got to 50-50 on digital. And we think we're there with the exception of maybe 1 large customer in North America who's a particularly big spender with our competitor. But extracting that customer, we're kind of -- we're at 50-50 already, I think, in the market with everybody else. The -- so that's the core piece in terms of EDA. Then you go to packaging and board. In packaging, we have a product called Allegro, which is really, really popular. We think Allegro is probably 80% of the market in packaging. And then that's a useful bridge from chip design into system analysis. On the system analysis side, it's where our customers are basically running all their systems now as they're checking the electromagnetic dynamics and things like that, terminal analysis. There's 2 elements really, its finite element analysis, which are physical things. And then there's terminal analysis because when you start stacking chips, you create kind of heat around that and it causes kind of thermal issues, and you need to be able to model those. But -- so that's the kind of the overall business. I think about 50% of the business is the core EDA piece, the chip design piece. Like I said, 10% is IP, system analysis is probably 12% and growing. And then, of course, you've got verification in the middle of that too, our chip design verification. Verification has been really strong for us. When our customers use our chip design tools, what they're trying to do is optimize for power performance and area but then they want to verify all of that. The other thing is that we have this emulation -- we have emulation and prototyping hardware systems. We have software simulation as well for verification. But for hardware, we have some extremely popular products, particularly the emulation system. The emulation system at Cadence is called Palladium. And its Cadence on Cadence. So it's a Cadence custom chip in a hardware system that's designed to allow you to bring up your chip so that software developers can design the apps to put on the chip before there's a physical version of the chip. And it's been -- that's really disruptive, the order with which people have designed. In the past, you would do the chip, software guys would create the apps for that chip, and it's a long process, but we have some customers that are on like an -- if you forgive the pun, they're like on an annual cadence of product release. And to do that, you can't do that without bringing up a virtual version of a chip and letting your software guys design the apps to go on that. The reason I say it's disruptive is that we think some of the popularity for those hardware systems is coming from the software engineers. Because the software engineers have -- are able to collaborate more with the chip design teams and influence the chip design because the beauty of the emulation system is that when you bring up the chip, it doesn't need to be finished. It doesn't need to be final and locked in. The software guys can influence the chip design before it's hardened if you like. So that's been a really popular piece for us. And I think if you look at the overarching -- I suppose it's hard to simplify this, Cadence is really a computational software company. We do a little bit of hardware, but we do a little bit of IP, but it's a computational software company. And when you apply computational software to silicon, that's EDA. That's the front part of the design process. When you apply computational software to like a car or a building, that's system analysis -- or phone, that's the system analysis side of things. And then when you apply computational software to data that's overarching the entire process, that's AI. Well, I guess that's probably the best way to describe this.

Thank you for that comprehensive overview, John.

I wanted to talk about the growth that you've been able to put in, solid growth, year in and year out. Per year, most recent guidance, you're expected to grow at a mid-teens percent this year, which is consistent with your CAGR over the last few years. And I know you're not going to give us any sort of a preview into next year. But as we think about next year, how do you think about the different puts and takes? I know you've had a heavier mix of revenue coming from upfront sources in 2023, maybe that's not quite as repeatable for next year. But that repeatable recurring portion of your revenue has been growing at a 13% rate. And then also, we have to consider some recent acquisitions. So how do you put all those pieces together?

Yes. Fair enough. Gary always ask me questions to help you get smart. But, yes. Look, so we're very transparent. I'd encourage you all to go to our Investor Relations website. Richard does a great job there. We're very transparent, you'll see all kinds of interactive financials going back to 2012 because that it's -- we're farmers, not hunters. It's -- we're planting now. We're generating R&D now for future benefit for revenue in the future. The often the investment time line is you're investing a whole bunch of effort into R&D now, and it may not pay off for revenue for 2 to 3 years. The -- in terms of the model, we have a mix of upfront and ratable business. The software is mainly ratable. I mean, that core EDA side, that's all ratable, we get revenue on a daily basis from our customers, and that's probably 70%, 75% of the revenue. Then there's other things where there's kind of bonus elements in contracts and there's ones where there's consumption based with the drawdown over time. And it's recurring in nature of the royalty on the IP side. That's kind of recurring. All in all, our recurring has generally been kind of 85% to 90% of revenue over the last kind of 7 years since I became CFO in 2017. But over that period of time, it's been like 85% to 90%. That's a -- which means 10% to 15% is upfront. Now the 10% to 15% upfront can be cyclical because you get upfront revenue on the hardware piece, you get a upfront revenue on IP, and they're the main components of upfront. You get a little bit on software from time to time on what we call license compliance. If someone's pirating the software, we have call-home technology into software. So we know who is using the software. And then occasionally, the triggering event for a license compliance catch-up is that customer might want to purchase hardware to do the verification side, we'll know it's the same IP addresses and -- but -- I think, we would say, look, we can't sell you the hardware. It's our policy to not sell you the hardware unless you're compliant on the software side, and then we catch them up on the software, do a new 3-year deal with them or something like that. But most of our customers on the software side do 3-year baseline contracts. And then over the course of 12 quarters, they'll come back between 4 and 7x to purchase kind of add-on technology. When they purchase an add-on -- and they'll purchase add-ons either because new technology, we've released it, they didn't have access to in the underlying agreements and the baseline agreements or headcount growth that they've hired more engineers than they had planned in the original baseline contract. And most of the time, they have visibility into what they need for the first year, less visibility, second year, less visibility, third year. So the baseline contract is the first 3-year contract. The annual value of that is probably bigger than the total contract divided by 3, if that makes sense, because they'll have -- maybe it's 100% -- what they commit to is 100%, 80% and maybe 60% in the year 3. So it kind of averages 80% of their requirement. And then they come back and purchase add-ons, which adds the other 20% or so. And when we sell those add-ons, the -- we always try to co-terminate. I actually, I say it that way because I'm not aware of ever selling an add-on where we didn't co-terminate but we sell so many it is possible that won't exist. But generally, everybody doesn't add-on the co-terminates thing with the baseline contract. And by the time you get to the baseline contract renewal, it's -- you kind of start the dance all over again. Everything gets rolled into a new baseline contract. The -- we've drifted kind of higher on upfront revenue, partly because of the success of this hardware system. I mean, it's been so successful. Our previous hardware iterations, they come in -- it's been like, I think, in my time, the last -- between Palladium X2 -- or sorry, Z2 and Z1, I think, was 6 years. But the Palladium version before that was 5.5 years, I think. They're kind of long cycles. I always complain to the engineers, why so long, how come it takes like 6 years? And they say it's the paradox of long cycles by the time you get to year 3 or 4, and you're ready to launch something, someone will come out of the woodwork and say, if you can wait 4 more months, I'll have it ready for -- I'll have something for 2-nanometer. And then you're faced with the choice of, do I wait 4 months for that, knowing that if I don't wait, it could be another 4 or 5 years? But -- so it tends to take a bit longer. But they've been really successful. We launched the latest ones -- was it March 2021? I think it was March 2021. That -- so we're coming up on 3 years. And the -- just demand is like outstripped our ability to supply. We had to get more production lines back last January just to keep up. Our lead times are typically around about 8 weeks. And at the end of last year, we had about 28-week lead time with all the orders that we had. I think it's just really, really, really popular. And I think it's just such a big investment for our customers to -- when they're doing design work not to verify it because everybody wants to be right first time, you want your silicon right first time. If you don't invest in the -- in verification and the systems that come with that, you run a much higher risk of not having your silicon right when it comes back from the foundry. The -- but on your question in terms of modeling though, generally, the recurring revenue is the farming piece, right? That's where you have a much more stable kind of revenue recognition. It's averaged 13% growth the last 3 years. And then I think we're on track for 13% growth this year. So it's pretty consistent. And that's up from double digits. We generally aim to do double digits. I mean that's our model. And you know us. I mean, we always -- we like to underpromise, but overdeliver, right? So -- but our model is basically, every year, we aim to do double-digit revenue growth and 50% incremental margin because it's -- the EDA business is tremendously profitable. It's very sticky. Our customers can't operate without it. And when it comes to renewals, it's like -- it's practically 100% renewal rate. The only people that don't renew have either been acquired in which case the renewal has been done by someone else or they've expired. But there's very, very little -- I think we might have had 1 or 2 startups that were desperate for cash that went and designed us out because they wanted -- they just wanted cheaper tools or something like that, but they don't last very long. They -- if they don't expire -- if they haven't expired by the time they haven't renewed, they will expire shortly after. The -- so we've very, very sticky customer base on the EDA side. And then, like I said, because all these customers want this unified platform. They want to be able to do everything concurrently, I think that's put us in a position where our strategy has been to expand across that platform from chip design through to system analysis, and it makes everything more sticky. Because the more you adopt from Cadence and the more you have the full flow on everything with Cadence, the better your AI results or your analysis of data because you're using the same tools for everything. So that's been really, really helpful. On the upfront side, the IP, that's like services, it's like milestone because every -- most IP is customized to a certain extent. I mean, you'll buy IP off the shelf, but there's always some element of customization and depending on how much customization you either have -- revenue tends to trigger when you hand over the IP or when you hit certain milestones like -- so that can be -- that's upfront. We treat that as upfront. We take -- we include the royalty pieces recurring because it's quite predictable. On the hardware side, because you hand over a system, you're -- when you've delivered most of the value as soon as the system is installed and the customer is using it. The piece -- there's a piece of hardware that we've assigned to maintenance that gets taken over time like software. But that has grown from the kind of the low double digits to about 15%. It could be 16%. This would be the first time in my time where we're outside of the 85% -- we're under 85% for recurring revenue. And I think like your focus is really on so what does this mean for next year, I presume, right? Like when we think about that, recurring revenue is -- I mean, we don't guide next year until we complete Q4 because Q4 is always a huge bookings quarter for us. Now that's probably consistent over a number of years. Now this year was a bit odd in that 2/3 of renewals, like the software renewal base that was coming up for renewal in -- so like if you look at all of our software contracts that were expiring in 2023, I mean, we're spoiled for predictability because you know when those customers' licenses expire, they have to renew. They have to renew. When we get hedge fund people coming and tell us sometimes, "You could 5x the price on this and your customers have to pay." That is absolutely true. Because if your customers are in the middle of like multiple project designs, there's just no way by the time your contract comes around for a renewal that they can't renew with you. Now the trouble is still that we adopt the farming, the kind of long-term strategy that we partner with our customers, we would never do that to them, and that's why they're comfortable partnering with us. But I think the trade-off for that is we do want a predictable kind of rate of return. And we do want to get some return on the investment that we make because we spend about 30% of our revenue on R&D every year. And that 30% is sunk upfront. But about 2/3 of that doesn't contribute anything to revenue in the year that it's spent. But -- so although we have high margins, that if we had to match the expense timing with the revenue timing, we'd probably have higher margins again. The -- but generally, the way we start each year, is we aim for double-digit revenue growth. And then what we'd like to do is it scales really, really well. So we want to try and drive more than $0.50 of every dollar of revenue growth to operating income, which would mean the double-digit revenue growth turns into kind of low to mid-teen operating income growth. And then on top of that, we use about 50% of our free cash flow because we generate consistent free cash flow from our customers. But we use about 50% of the free cash flow to repurchase shares, which has the share count declining gradually over time, which means the kind of low to mid-teen operating income generally turns into high-teen EPS growth. We've averaged from 2017, periods to -- like if I just take the midpoint of the guidance, I think we've averaged about 22% EPS growth, but that was helped partly by tax benefits. It's probably high teens is what we've done, if you had a consistent tax rate, but taxes are lower now than they were back in 2017.

Let me stop you there. So I want to double-click on something you said in there, and that is it sounds like this year's fourth quarter has a disproportionately high amount of license renewals. How is that going?

Oh, it's so busy, so busy. I think since we started Q4, I don't think there's been a day that I haven't been talking to the sales force on something. It's tricky out there though. I mean in comparison to other Q4s, Q4 is always busy. I think people are a bit cautious. They always -- I think they see that their spend with us is an opportunity to save because the more you spend with us, the more you potentially save because you're -- if you can leverage the tools more, yes, and particularly with the AI capabilities that are there, that maybe you can throttle your headcount growth in your model, that when you're doing some of your long-term planning. That's been a challenge though from a pricing perspective, but I mean, Anirudh is really super proud and he should be, the R&D teams have created such tremendous AI tools. I mean he gave me an example of -- there was one customer who was getting 75% benefit. While he's rejoicing and the R&D guys are rejoicing with that, it gives me shivers because I'm thinking, we didn't price that right, if it's 75% benefit. And it's like, it's okay, we'll get them next time. But -- and we'll make sure we factor that into pricing for those tools with other customers going forward because we're all kind of learning through that. But Q4 is always a super busy quarter. Actually, this year was a bit unusual. I think it was the most back-end loaded half, first half versus second half than I've seen in a long time. It was about 1/3 of our renewals came up in the first half, 2/3 of the -- so like I said, if you look at your software space and which contracts are expiring, you look at the annual value of those, about 1/3 of the annual value that was expiring in the year was expiring in the first half, 2/3 in the second half, which meant you're going to have a much bigger second half bookings. Next year is about 40%, 60%, Gary, 40% first half, 60% second half, right now. It's hard to kind of tell what it's going to be because it's such -- there's still a lot to do between now and the end of December. It's -- yes, there's huge demand. The two things I want to point you to, though, is that I think on the earnings call, you all wanted us to talk about our AI tools and how much growth we're getting from AI tools. And we did, we give you a metric. We said, look, it's 3x, basically, the revenue we're generating. We have 5 AI apps that we're selling. And they're generating about 3x the bookings and revenue value this year than they did last year. And that's all ratable revenue because it's all on the software side. What I want to point you to though is that recurring revenue is still 13%, right? So that's part of the recurring revenue. And what I'm seeing on a bunch of contracts is that customers are adopting the AI tools. But if they had 100 licenses of this product X before, they're trying to make do with 85 or 90 in the baseline renewal to make room to buy the AI tools. And we saw that with -- when we proliferated the digital tools because there was a time when we were probably 20%, 25% of the digital market, we're up to 50% now, we think. But back then, when we launched new tools that were very popular with customers, they did the same way, and it proved to be a fairly fast economy in that they shave back their baseline configuration then discover within 6 months they can't manage without the extra licenses and they just do more add-ons later. But -- so I think the recurring revenue piece is pretty consistent. I mean that will go in line with the economy to a certain degree as well. I mean we always say that we're very, very resilient, but we're not immune to recession. And maybe there's a recession and a whole bunch of people are laid off, that will hurt us. But not so much as most other companies. On the upfront side though, what tends to happen, like I mean, I know when I'm doing budgeting and everything, I'll cut back on CapEx and things like that, that could hurt your hardware. But -- and we're up against tough comps for next year. The upfront piece, I mean, last year, I was very cautious because upfront revenue grew like 50% almost, I think, from '21 to '22 on the -- because of the popularity of those hardware systems, and I saw really, really tough comps. And I think I was trying to get you all to calm down this time last year because I was worried we had tough comps and then we've beaten that again this year. And it's -- upfront is probably up 20% again, isn't it? But again, I would just caution, it's tough comps for next year. You don't -- I mean, again, if there's any slowdown in the economy, people might slow down their purchasing. Can I talk a little bit about pricing just for a second?

Yes.

In terms of how people approach things. Because most of the stuff I'm going through right now in Q4 is I'm dealing with the sales force, because we do extensive pricing analysis. We look at what's the average selling price that we're getting from all our customers for each product at different volume levels. But -- and we compare what we're getting on an individual contract against that. and we call it a deal quality metric. So let's say, I'm looking at the average selling price for what's in your configuration and you're looking for $10 million worth of software annually, but you're coming off a contract where you were paying $8 million and you just have the budget, you're telling us you have the budget for $8.5 million. Nobody capitulates and gives us $10 million. But what happens is that we shave the configuration down to try and match their wallet. But -- and you might end up -- if they have the budget for $8.5 million, you might end up getting them to $8.7 million, $8.8 million. They might be stuck at $8.5 million, but they'll end up with a smaller configuration. And then that creates the add-on opportunities later. And I think that's playing in somewhat into -- you're seeing some AI revenue show up, but it's not really raising the overall top line number yet. I think that will come, though with add-ons. I just wanted to make that point.

Okay. I appreciate that. And it's probably a good pivot point to talk about what these 5 different AI tools from Cadence bring to customers, from an efficiency standpoint, right? And so the way I view the opportunity, there's about $90 billion a year spent on chip design R&D activity. And if you just simply divide that into the size of the EDA market and the semiconductor IP market, you can come to a conclusion that you're grabbing only a high-teens percent of the chip design R&D wallet share. And so if you can bring more efficiency to the chip design process, you can capture a greater percentage of that wallet share. Maybe you can just give us a sense of how that high level shift alters the core growth rate of the EDA market and your growth rate.

That's a great question, Gary. So let me start, but I want to bring Richard in a little bit, and I'll let him talk about -- would you talk about the AI tools maybe but in a minute. But overall, most of our customers, if you look on the EDA side, the tool spend is only about 10% of their total spend in R&D. One of the reasons we're resilient in a recession is because our revenue line is our customers' R&D line. And like in a recession environment, the last thing I'm going near is cutting R&D because R&D is for the future, but we'll cut back on G&A, we'll cut back on some sales and marketing efforts or -- but from an R&D perspective, we might even double down in R&D to design our way out of that recession. The -- so we're quite resilient that way. But when you look at our customers' R&D spend, they spend 90% on people and 10% on tools, particularly on the semi side. We've talked to some customers that are adopting AI tools and the whole pitch about the more you spend, the more you save is based around, let's say, for every $10 million a year you're spending in R&D, right now, you're spending $9 million on people, $1 million on tools. And without adopting any different approach or different methods, at some point, you'll double. And when you double, you're going to be spending $18 million on people, $2 million on tools. If -- on the assumption you can find the people because those in -- that engineering talent is really hard to come by. But what we explained to them on the AI side is that if you adopt more of our AI tools, you might need less people. And what you could achieve for $20 million, you might be able to do for $16 million if you're spending $4 million of that on tools. So the logic there is that we don't necessarily want them to cut back on people. We just -- we want to try and capture more of kind of an outsized portion of the increase in their R&D spend because I think that would be helpful for them. I think they will save more by adopting more tools. And what we should see over time is that the 90-10 split of their budget between people and tools, we'll start gradually work towards 75-25 and 50-50. I mean if you extrapolate long enough, there's just not enough growth in the human population to keep up with the growth and complexity of design. You're just going to have need more and more AI tools. The tools are going to have to do more of the heavy lifting, and I think we're in the right spot for that. But we have -- like we have 5 AI apps. So do you want to talk about some of those?

Sure, sure. I'll just touch on that quickly. So 5 AI products, cutting across the entire EDA and system design analysis kind of life cycle. And these AI tools bring dramatic tremendous benefit in threefold. First off, it helps the designers to design a better chip, okay, from a power performance and area standpoint. I'll give you an example. We have a chip design project with Renesas on one of the automotive chips. It has 17 variables. If -- for the longest history for EDA, it's all -- a single run process. You do one round, you split the different blocks amongst a lot of different engineers. And they'll use intuition to keep doing iteration and make it better. But with 17 variables, it will involve almost 4 million kind of runs and every run takes 1 or 2 days. So humanly, it's impossible to explore the entire design space. With the Cerebrus, one of our AI tools, that could be shrink down to weeks and 200 runs, we get them much better results. So the benefit of PPA is to the tune of anywhere from 5% to 20% depending on where your starting point is, or how good your design team is. So the second benefit for the AI tools is you have much better control for the destiny, right? And you're not dependent on the other merchandise companies to help you do that. And dramatically, importantly is, as John pointed out, the -- from a productivity standpoint, it helps companies and design companies and EDA company -- and chip companies to have much more design activities with less engineering resources. So these are really at a starting point of a big AI revolution for us.

Actually, can I add one more point?

Sure.

So the -- in that whole model of double-digit revenue growth and 50% incremental margin people often ask, 50% incremental margin, how long can you keep that up for because we've done that for like 7 years in a row now. And I think it's averaged 55%. But we start generally with 50%. We like to go with -- go low and then trying to overachieve with that. The -- we're really holding it down to 50% -- in that 50%, 55% range because we're -- that's our budget for R&D. We're trying to do more R&D because revenue growth tends to go in line with new product releases. We want to release more and more new products over time. But it's also giving us some funding for plugging some gaps on the M&A side. You might have noticed that we bought Intrinsix and we bought piece of the Rambus business. The -- and those are kind of long-term plays. We changed the IP leader at Cadence at the start of the year, and Anirudh spent a bit of time with them, wanted to know what did you want? And we went and bought those. I don't think they provide a huge amount of revenue in the short term because the way we did the -- I think we got those cheap because we allowed them to sell out into the future. Take it for a low price and then we'll eventually get the renewals on those when they come around. But there's the -- we've got that. We're also investing in the system analysis space. EDA is pretty -- I mean, it's down to 2.5 players, really. That's a -- but we did some IP because we have a new IP leader, and they've got to digest that. I would say, in terms of -- from an R&D standpoint, a lot of efforts go into system analysis because I think that's where there's system analysis and verification. Because I think that's where the -- we have kind of an above-average opportunity for growth.

Okay. Perfect. Well, I think we're actually up against our time line. I wish we had more time, but we don't. So John and Richard, I really appreciate you joining us up here on stage. I appreciate the people in the audience who joined, and as well, our people online. So again, thank you, guys.

Excellent. Thanks, Gary.

Thank you.