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

So good morning. Thanks, everyone. I'm Vivek Arya. I cover semiconductors and semi cap equipment at BofA. And I'm really delighted to have the team from Cadence join us: John Wall, the CFO; and Richard Gu, VP of Investor Relations. And what we'll do is we'll start with a small statement from Richard, and then we will jump into Q&A. But please feel free to just raise your hand if you have any burning questions in between, and we'll get to you. With that, welcome, John. Welcome, Richard.

Thanks, Vivek.

Thank you. Thank you for having us. So this is my 1 minute of glory. Today's discussion will contain forward-looking statements and 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. With that, I'll turn it back to you.

Excellent. If you sing it next time, [indiscernible].

I think we say anything we want.

That is right. So maybe, John, for a number of people who might be new to the EDA industry, if you could just give us a sense for what is the right way that you forecast growth for the industry, right, for the next 3 to 5 years? Is it rising capital intensity? Is it a broadening customer base? I imagine there are 3 or 4 big inputs. So how do you decide whether this is a 7% industry, 10% industry, 15% growth industry?

Okay. Great question, Vivek. Yes, I mean, we're spoiled a little bit with Cadence because so much of our revenue is ratable or recurring. It's very easy to predict our revenue in the future. But just taking your overall question, we generally look at design starts and the design activity that's out there and the demand for basically complex technology. I mean there's just an insatiable demand there. And where we play, I mean, our space is really chip design, chip verification, packaging and board and system analysis. There's a little bit of IP. I mean, Cadence is a computational software company with a little bit of hardware and a little bit of IP, I don't know how we're an IP tends to catch the headlines a lot because that's the small percentage of our revenue that's upfront, so it tends to grab headlines when things happen there. But if you look over a longer time horizon, it's computational software and so much of its recurring. We see our customers' R&D budgets really increase over time. We see that there's a race there where, I mean, you could simplify our customers' R&D budgets and to spend on engineers and spend on the tools that go into the tool bag for those engineers there. And if I go back like 20 years, maybe 92% of the spend was on people and 8% was on the technology that was going in the tool bag. Maybe now it's closer to 88% and 12%. I think over time, over multiple decades, that probably gets to 50-50 because there's no change, there's no letup in the demand of the complexity for technology, but there's natural limitations in the number of engineers. And if there's natural limitations in the number of engineers, you're going to lean more heavily on the tools and the tool bag. And with AI and machine learning technology, I think there's more value will be assigned to the tool bag over time.

Got it. Yes. Thank you for closing the door. I was just about to suggest that. It's a little noisy with the lunch crowd. So this broadening into the new customer vertical, right, of systems companies wanting to do their own internal silicon. Do you think we are at -- is that still a growth industry or is that plateauing? Because I ask that as we look at the next generation of 5 nanometers or 3 nanometers, I imagine that the cost of that is so enormous that is that likely to deter that part of the customer base and force them to rely more on traditional semiconductor companies as opposed to do -- like could we see the pie reverse in the other direction?

Yes. So I think if you look at the -- broadly speaking, if you look at what Apple, right, as one of the examples, they -- I mean really is one of the trailblazers see there, right? They launched M1, M2 yesterday. You see the sort of the productivity gain in terms of CPU and GPU just mind blowing. I think that's -- we're still at a very early phase of that kind of transformation and tailwind. Broadly speaking, if you also look at the autonomous driving, IoT, hyperscalers, everybody is trying to make sure they are developing the best verticalized kind of ships in order to have a tighter integration between software and hardware. So I think we're still at early innings. And cadence as a company, design activities, obviously, is one of the drivers broad business. We stand to gain and benefit from that megatrend.

Got it. Right. And one other thing, John, that I find very interesting and useful in the CFO letter that I told you with the every earnings call is the way that you showcased either 3-year growth CAGR.

Yes.

Why has that accelerated over time? I can understand it growing or saying, look, semiconductor growing companies are growing at X, we have kind of grown a little bit faster. But it's accelerated from like 8% to 10%. And now I think this year, it like close to 14%. What has driven that acceleration over time?

Well, I think it's all -- Cadence has been a great compounder for investors, but it's all these incremental improvements that we make in each year, and they all compound on each other over time. But we always tell investors that, I mean, if you think you can time the market, you're better off somewhere else. With us, if you want to store or stash money away for a decade, it's a good place to be. But if you look at over the last several years, I do think that the industry itself has become much more rational from a pricing perspective. Our competitors should take credit for that, too. I think it has got down to a smaller number of players at the lower process nodes, like it's more rational. And we partner with those customers. I mean a lot of our revenue comes from our top 100 customers, despite the fact that we have thousands of customers, a lot of focus on those top 100. We disclosed the top 40 is 55% to 60% of our revenue. We still feel we're quite diversified across that group of companies. But we pay particular attention to those top 100 customers and partnered very closely with them over the last few years. And in return for that partnership, we're gaining a greater share of wallet there. I think we've also fulfilled our destiny to a certain extent. We always thought that it was our destiny to be the largest core EDA company. I think we are that. But we felt that, that was our destiny because we always felt our strength in analog from our roots, but we had a significant market share in analog and that it was only a matter of time before we go to 50-50 on digital. And you saw we announced that we had a record contract with our U.S. marquee semiconductor company back in Q1. It's quite interesting if you look at the RPOs. RPOs have improved the contracts that customers have committed to us and our main competitors have been significant over the last few years. If you look at the next 12-month revenue on those RPOs as well, I think you'll see this -- I think we probably have the largest next 12-month revenue in RPOs because your backlog is a mixture of annual value and time, right?

Right. That's right.

And but if you look at your next 12 months revenue, that typically tells you the annual value and how much is, you could have a bigger number if you do more time. But we tend to stick to -- you mean -- you mentioned a 3-year CAGR. The reason we focus on 3-year revenue CAGR is most of our contracts with customers are 3-year baseline contracts. And then the average duration for our contracts in any bookings year is typically 2.4 to 2.6 years. It's around about 2.5 years. That's because during the course of a 3-year baseline contract, we'll have a customer come back and purchase anywhere from 4x to 7x during the contract to purchase add-on technology. And when we do add-on deals, they co-terminate with the underlying baseline contract. So when you net it all out to all those add-ons are, by definition, less than 3 years. And when you net them all together, it tends to average out about 2.5 years. So we're pretty consistent with our duration. And the model is fairly boring. What do you get to is at the end of any year, because of the consistency and duration, you probably have about 55% of what's in your backlog burning off in revenue in the next 12 months. That tends to be about 75% of next year's revenue. And then the other 25% of revenue comes from a mixture of new recurring revenue, which is about half. So about 12.5% of the revenues coming from new recurring revenue in that year and the other 12.5% is coming from new upfront revenue from like hardware and IP.

Got it. So let me push a little bit, John, on this growth number. So let's say, we had discussed this 3 years ago, right, when you were kind of growing in that 10% to 11%. And I ask you, predict to me the growth rate 3 years from now? Would you have given me 14%? Or would you have said 10% or 11%? Because I asked the question that just as you went from 10%, 11% to 14%, is this 14% now the new baseline? Or is it just kind of an above-trend number that you're just going to go back to 10% or 11%?

So our focus has been not just on revenue growth solely but on Rule of 40, right? So the Rule of 40 being revenue growth plus operating margin. And what we focused on is trying to grow that overall Rule of 40 metric. And I think we're up into the 50s now. But essentially, our target has been to -- we want to continue accelerating that 3-year revenue CAGR. But also, we're trying to achieve over 50% incremental margins. In 2016, every dollar of Cadence revenue cost us about $0.74, but we knew that we scaled really well that every dollar of revenue growth since 2016 has cost us about $0.46. That -- so if we are continuing to drive operating margin or incremental operating margin over 50% that will -- that naturally creates -- there's no near-term ceiling on operating margins. And then on the -- our focus on R&D. What we did there was -- it was funny really because I got the [indiscernible], who is the President of R&D. When I got me CFO, [indiscernible] was made President of R&D, and we had our Head of Sales. And the 3 of us go into a room, and we had the discussion that, look, we've been around with Cadence lifers. So we said, you know what happens when we fail or when something goes wrong in a contract and we don't get what we forecast is that sales guys will point to R&D and blame them. They'll say the technology is not good enough. The R&D guys will point to the sales guys and say the sales team doesn't know how to sell it. That -- so what I agreed with them upfront was we have to identify the metrics so that we know the reality ourselves. What is -- when is it an R&D problem? When is it a sales problem? So we came up with deal quality metrics. We had a measurement for every product. We measure basically the price we were getting on any deal against the average selling price. And we came up with a deal quality metric for that. But -- so for example, if I take, you want a specific configuration and it costs $10 billion, if I look at the average selling price for that and you have a budget of $8 million. then your deal quality metric right there is 80%, right? So we're trying to improve that. But you never capitulate and give us $10 million. So what happens is that our guys will [indiscernible] the configuration down to try and match your wallet. And then we leave room for those add-on opportunities through the contract cycle and that you get revenue growth from that. On the deal quality side or on the product quality side, what we agreed was we've measured things like customer call center volume and customer change requests because when those fight, the odds for two bugging when we were initially -- originally introduced them. And that was a drain in R&D time. So what we needed to do to improve the output for R&D was to make sure that they weren't releasing because customers -- our customers are so technically brilliant. But as soon as you've got a solution for something, they want it. They don't care if it's [indiscernible]. So it takes a bit of resolve to try and wait and make sure it's right before it goes out. So we put product quality metrics into our R&D team's bonus plans and gotten to focus on not releasing the technology until it was ready because we didn't want R&D time spent on maintenance. We want R&D time spent on true product innovation, creating new products. And 5 years ago, close to 2/3 of our R&D spend 5 years ago was on -- our R&D time was being spent on products that were generating revenue in that year. Now with less than 50%, less than 50%. And where you're seeing that turn up in metrics is that the pace of innovation, the pace of new product launches has accelerated.

So is that then, John, to a point that it is new product introduction that has accelerated?

That's exactly right. It's new products, products being launched more frequently at higher volumes, that execution against our Intelligent System Design strategy, and there's a long runway on that. But we're achieving that 50% plus incremental margin, while we're building out a multi-physics platform and new tools in the system analysis.

So why can't I assume 14% as a new baseline growth rate for MSA?

You can assume what you like.

What do you assume? That's my question.

That's it. So what I aim to do is improve the Rule of 40 metric, that if that was the slowdown, I would want higher margins.

I see. So the point you're making is that, let's say, if growth were to slow down to whatever 10%, 11% from that, it would...

And I see no sign of that slowing down. Just -- I mean when you look -- I know everyone's talking, we're asking people here at the conference, what are you hearing from different people. But now we have the benefit. When I'm at the CFO roundtable forums and things like that, I tend to get teased a little bit by my peers because we're selling to customers where they're purchasing -- our revenue is derived from our customers' R&D budgets. And often, even when I'm sitting with customers and their CFO, is they're telling me that a lot of their revenues derived from IT spend. And they know that if there's a recession environment, IT might get cut before R&D gets closed. But -- so we haven't seen any slowdown. I mean the design activity is still ramping up across the world. The -- and it's just amazing what some of these companies are doing with the technology. The -- so again, I don't see any near-term slowdown in operating margin. And it's the same with the 3-year revenue CAGR. It's continued to accelerate since we started this. And again, I don't see any near-term setting on that either.

So it can be higher than 14% in the future.

But the aim is to continue to accelerate this. There will be a natural level. But I focus on the 3-year revenue CAGR because that's what you're trying to do. I mean as we continue to -- we're landing and expanding because we're adding new system analysis tools all the time. So that lends to pretty easy revenue growth.

Got it. Understood. Pricing power, as -- how much of the sales growth has a pricing component to it? Because we have seen semiconductor companies right being able to pass along a lot of cost inflation. I know you don't have quite the same kinds of cost inflation. What are the kind of inflation you have from a cost perspective? And how successful have you been in terms of passing those along? And then how sticky is that pricing?

Gu, do you want to start at that?

Sure. I think the pricing power is a function of how mission-critical and how important your technology is to customers, right? I mean if you look at our technology, it's essentially engineering design, engineers designing tools for engineers to go change the world. And it's essential for the customers to come up with the best inbred kind of technologies to go disrupt the world. So I think from that perspective, would bring a lot of value to the customers. So that validates model too, right? So I think as a company, our philosophy has always been, hey, as the customers become more successful, they're going to share the success with us. And we historically, we haven't seen any issues in terms of passing along inflation from a P&L standpoint to our customers. And we feel it's a very collaborative cliche of kind of relationship, which we intend to continue to drive.

Yes. I think that's great. I mean, generally, it's people, right? I mean our cost is a very asset-light model, Cadence. It's all engineers, it's all people. And like I say, you could simplify our customer spend. Their R&D budgets are spent on people and tools in the tool bag. They buy the tools from us. They -- and if when you were spending $100,000 for an engineer, if the tool bag was costing $10,000, I think it's natural and people expect no less than if it's costing me $200,000 as an engineer, it's going to cost me at least $20,000 for the tool bag. And I do think over time, you'll spend more and more on tools because you'll get more economy from -- and productivity from using the tools. But...

Got it. So is there any cyclical aspect to the business? I ask that because today, investors are more concerned, right, rightly so about what is the downside risk, right, as opposed to the upside, right? So if you look, John, at prior cycles, what happened during years when say semiconductor sales were flat or negative to the ability of those customers to invest in EDA tools. Were they able to keep up those investments? Or do you think that there is a cyclical aspect to your business also?

Well, generally, what we found in the past is that our customers, when they go through downturns, try to double down on R&D and design the way out of it. It's -- I mean there have been extraordinary times where it's normally the last way you start cutting in R&D. You typically take it out of G&A and other areas first. But if you have less engineers in R&D, then you have less demand for our tools. But we have 3-year baseline contracts. So you could cut now, but -- and you might still have 18 months to run, and it depends on what you need in 18 months time. And in the meantime, what we found in the past is those engineers get picked up by someone else who then has to purchase tools for those -- another tool bag for those engineers. It's kind of like the way your hyperscalers getting involved in. So we have traditional semiconductor companies using our tools. They don't stop doing what they're doing just because a hyperscaler decides to do their own project, do their own chip design. So we're -- so generally, we kind of plot along that through this, like I say, if you can time the market, you're better off somewhere else, we're kind of boringly predictable, and we'll go with that. There's a secular trend. I think in the past, we've had some cyclicality to hardware design cycles, but that was when like 10 years ago when we had maybe 1 hardware system. Now we have multiple hardware systems, and it's -- I think it's less cyclical.

And hardware is also recognized ratably? Or is it recognized upfront and...

So hardware used in the cloud is ratable, but maintenance on the hardware systems is ratable but the product portion of hardware is recognized upfront on delivery when it's an outright sale.

And how much of it is cloud-driven versus on-prem?

A very small percentage is cloud-driven, but that's growing.

That's growing.

Yes.

Okay. Got it. Okay. Is there a simple formulation, John, of how many engineers use your tools times average price, just so that we get a sense for over time, where has been that growth? Is it being value or let's like [indiscernible] content per customer as opposed to a number of customers?

I think it's very hard to simplify it like that. I mean when you look at Cadence, it's a bit like -- your Cadence itself is -- like there's a business -- there's like 5 different businesses under the umbrella Cadence. And they all have slightly different dynamics to them. I mean if you look at our Analog franchise, our custom IC franchise, Virtue also is a great franchise product for us, but it's an interactive tool. If we're selling to a 100-person Analog design house, you're not -- they're probably buying hundreds licenses of Virtue also. We're not going to sell 110 until they hire 10 more engineers. So in -- on that account, your growth is going to come from a combination of engineers growth in the number of years and growth in pricing. But now thanks to you, that is a very robust and sticky product, and there's good pricing power there. What we found on the simulation side, and there's extensive simulation expertise internally at Cadence, but what we found on the simulation side is that if you had 1 engineer, it can use like 5 licenses, then if that engineer has the compute capacity to be able to use 10 or 20 licenses, volume can go nonlinear on you. So your revenue on those simulations is that's our most profitable business a cadence. The area that we're growing into in this Intelligent System Design strategy, in stimulation is actually our most profitable area. So that's also helping margins for us. But it's because revenue can go non-linear when an engineer can use multiple licenses. So it's -- so to answer your question, I guess, it really depends on each individual business.

Got it. So I guess, a different way of asking the same question is, if you look at that acceleration of growth from what used to be a 7%, 8% CAGR to double-digit plus where we are today, has it come more from selling more number of seats or more on a per seat basis?

So we don't typically sell [indiscernible], its licenses.

Licenses, sorry.

But I think it's a combination of both. I think pricing has become more rational, and it's probably added a couple of points of growth over the last few years. But there is a -- we're naturally getting a bigger share of wallet because our customers are depending more on the tools. And when you take the likes of a Cerebrus product, that Cerebrus is a really interesting tool in that it allows -- it's like self-driving for digital design. So a digital engineer will use a product of ours called Innovus, but that's an interactive tool, 1 license for 1 engineer. That Cerebrus, it creates a cockpit that allows that engineer to use 10 licenses of Innovus. So not only do you get up from selling Cerebrus, but there's also an upsell from the fact that if you're using Cerebrus, now you have to use 9 extra copies of Innovus. But it changes the dynamic in that now you've taken a revenue stream and a profitability profile for that business from an interactive style business where you're more dependent on pricing to drive revenue growth to one where you have much more opportunity and volume.

Got it. On the competitive landscape, we'll get to the marquee customer. I reserved some marquee questions for that customer. But how is the competitive landscape? Are you selling products that are complementary to the other peers? Is it more competitive where you have to go? And because the perspective that I've had is you guys are extremely good in Analog, right, competitors are very good in digital? Or is that not the case that there is a lot bigger overlap in your product portfolio outside of this customer, just from an industry perspective.

So I think with all of the competitors that we compete against, we're very respectful of all of them. I mean they're all great people, and so they have great tools. I think we all drive each other on to perform better. I think we can all peacefully coexist.

Does a customers buy from all of you or only from 1 of you or 2 of you? Like, what customers tend to...

Well, generally -- so on chip design and lower process nodes, you probably have ourselves and [indiscernible]. But -- and it's really down to 2 players, most of the time. And the higher process comes, you might see [indiscernible] or Siemens, I guess. But on the system side, you'll probably see DENSO and ANSYS. I think we can all peacefully coexist. I mean we can operate side-by-side. Ultimately, what our customers are trying to solve is they want to be able to do chip design, chip verification, packaging and board and system analysis, all in parallel on 1 platform. And no company has that yet. I think we're the closest to being able to do that because our center of gravity is chip design, chip verification, is core EDA, that we've always had a packaging and board business and system analysis is the next natural adjacency for us there. And with the expertise we have in simulation so much of core EDA is based on simulation expertise and applying that expertise to the system analysis market has made us very, very competitive there. But ultimately, I think we're swimming downstream into that space when I think others are trying to swim upstream against this.

Got it. So would you say this revenue acceleration, is that just because the industry is getting bigger or because you're taking share?

Well, we're also expanding TAM, right? I mean we did say as part of our Intelligent System Design strategy. We would take the TAM from $10 billion to $30 billion by 2025. But I think we're on track. But in terms of being able to expand our time there with the road map that we have. So that's helping a lot, too. So we've got a number of tailwinds.

Got it. And then the marquee customer win that you mentioned, how has your relationship expanded with them? Was this like a competitive thing that you won? Or is it just that as they are expanding into other markets as well that just the TAM is growing, and that's why you're getting pulled with that? Or was it really just a head-to-head competitive thing that you won?

I think that's just a natural evolution at that customer. I mean the biggest win there was probably back in 2018 in terms of meaningful. Because in 2018, we had a meaningful win there where they started using our technology more broadly in a number of designs. And that meant that they were going to see the real advantages of using our tools. I don't know how to say this, maybe the best way to say this. I always feel that whatever your greatest strength is, is probably the source of your greatest weakness. But -- and if you have -- I mean, this particular customer has incredible engineering talent. An incredible engineering talent can use tools in the tool bag that are maybe not the best and still have an incredible performance. But the trouble is that you can only go so far and if your competitors are beating you because they're using better tools. But at some point, you've got to, of course, correct, right? And I think 2018 was an opportunity for us to gain some share in that account. And we did -- and what we did -- what we announced in Q1 was a record contract with this customer because they plan to expand and proliferate more significantly with us. Now I don't think that's going to be terribly damaging to other people they expand with. I think their intention is to spend more over time. We'll probably gain a greater share of the growth in that account for the next few years...

Relative to history.

Yes. But it's really positive for us, and it's been a long time coming. We used to joke in the past, people would say, "Oh, explain your overnight success." And I don't think there is any really overnight success in EDA. It's years and years of working with and partnering with those customers. So it's a good proof point, I think. Anything to add there, Richard?

Yes. The only thing I want to add in there is this is just a relationship that blossomed over the past 2, 3 years, right? It's certainly we're sowing the seeds step by step. And it's an interesting industry in the sense that people look at each other, "Hey, if you're successful, what kind of tools are you using, right?" You look at the TSMC, which we have a very close partnership with, and they're winning the market and then we're part of the reason. And then I think it's just very natural. I think it's a strong validation to the portfolio and the strength of our technology, which is great.

Yes. Absolutely. Your IP business, right, is an important part of the business, but it is smaller than what we have seen, right, with some of your peers. John, how do you decide that this is the right level of IP business, right, that I don't want to make it much bigger? Or how did you decide that? Because I just find it fascinating that 2 players in the industry have such different views about how large the IP business should be. First of all, what is the IP business in your portfolio? Does that have a per unit kind of royalty aspect to it? And then over time, do you think it becomes a bigger part of your business? Or it just kind of grows with the overall business?

So the IP business is tremendous at Cadence. Now it's not as big as some other players. And I think from a strategic perspective, if you look at our Intelligent System Design strategy, it's all based around our center of gravity being the largest core EDA company. What we figured it for the largest core EDA company, the natural place for us to get a greater return on investment was to compare adjacencies to that. On the IP side, when I first analyzed that, I thought there was too much commoditized IP there. I think It's almost like, I guess, if you're looking at homebuilders, right, there's a reason CEMEX isn't your biggest homebuilder, right? I mean there's -- it's foundational, right, in terms of it goes into the foundation, and it's an important piece but there's a lot of folks that sell CEMEX, right? But the -- and if you were designing a house or designing a property or something like that, you want tools, you want to architect something beautiful, but it's not that big a deal. And there's not a huge amount of pricing power in there. But I mean, some elements, there are, but it was an opportunity for easy revenue growth, but it wasn't profitable. It was the least profitable business for Cadence back in 2016 when I first analyzed this. And I felt that if you chase revenue growth -- that's why we moved to this Rule of 40 metric, as I felt you could chase the revenue growth. Anyone can get revenue growth if you're willing to sell $20 bills per $15, right? So we wanted to focus on profitability. It was a mix of revenue growth and profitability, and we keep driving the culture of treated like it's our family business. And the team that we have there has done an absolutely fantastic job. But I generally set them a target to achieve low-teen growth for us in IP. Over the last 3 years, they've averaged mid-teens growth. I think it's been 16%, 17% the last 3 years, on average. They did 17%, again in Q1. They started off this year tremendously well. I set the target, and the guide actually only has 13% in it.

That's nice.

So -- but I said to guide that way to give them the license and the autonomy to be able to walk away from less profitable business and pick their spots. We prefer to focus on differentiated IP. So we have Tensilica IP, where there's some royalty business that we get from that royalty revenue that we get from that based on unit volume. There's also a license revenue we get, so some people who don't want to pay a royalty, They'd rather pay the license. We've adopted that approach with some of our customers this if you can't afford to pay the price for our technology, in many cases, we're happy to invest alongside you, right? So we'll prove high tools -- but if we invest alongside you, we have to share in the outcome. You might end up paying more later. So if you're on a royalty type model, you're not paying upfront, but you probably pay more over the long term. If you're royalty some of our customers, they might be more confident in the design that they want to sign up to a particular fixed fee for using that technology, and we don't get paid on the units at the end. So there's a mixture there. But our focus is mainly on differentiated IP. And there's -- I mean IP is important, but overall in our strategy -- again, if you're the largest core EDA player, I think Intelligent System Design strategy that we're operating and executing against makes perfect sense for us. If you're not, it makes perfect sense to go and build franchises in other areas like software integrity and IP. So I think that strategy fits our competitor better than if it's us.

Got it. How do you segment customers? Like, is it by -- or these are kind of traditional semiconductor companies, these are system companies? How do you segment -- so I understand how you might segment people by the kind of tools, right, that they buy, et cetera? How do you segment your customer base? Like do you literally see different buying behavior among your different customers.

So our marketing group does a great analysis of our semi and system companies base that our system customers. And all I ask of them is that they're consistent in how they apply that because I focus really on the trend information. I remember 5 years ago, they told me that 40% of our revenue was coming from systems companies. And datafied systems company does not [indiscernible]. So they basically...

The largest smartphone company would be considered as semiconductor company or a systems company?

Well, this is what I run it to difficulty because I don't know how you would delegate Samsung. I mean Samsung is such a big player...

Samsung or Apple for that matter. What they be...

So all of asset marketing group is to make sure that what they provide me is consistency, that when we do pricing, I cover the name. I don't even want to know the name of the -- it's the same because when I review my team, I review metrics. I don't even look at people's nits. But it's an opportunity to remove any kind of bias that you might have. So you're not looking at the names you're looking at the actual outcome, the metrics, how people are performing and you judge based on that. So when we do bonus allocation often, I'll do that without looking at the name. When we do pricing, we don't price to semi companies differently than we do for systems companies. But our strategy, Intelligent System Design is naturally taking us more into the systems space. So I think systems is growing a little bit faster for us than semis. And the marketing team tells me that we're at 45% -- we're 45% now?

45%, yes.

For systems and 55% for semis. All I know is it's consistent is the way they're...

And that is including people like Samsung or Apple as far systems rather than semis?

Yes. That's right. And to use a consistent methodology for allocating systems and what semi for those accounts.

Got it. Okay. So if we set them aside, the reason I'm kind of going in this line of question is that as we go to these more advanced geometries, do you think it becomes more expensive for customers to use this. And they may depend more on traditional semiconductor companies, right? And I'm including Apple and Samsung as part of traditional semiconductor companies that it kind of reverses the trend in some ways.

Do you want to start?

I think if you look at the -- what all these verticalized, the shift in that trend, I think it's a big paradigm shift which you have seen over the past couple of years, there's no signs of stopping, right? Because the benefit from the customer's standpoint is huge, right, in terms of the productivity gain, the power savings, the performances. And I think in the grand scheme of things, I think that they are reaping huge benefit. And so that's why I think it's the more complex, the more sophisticated those vertical as chips at the main specific chips becomes, I think the better off would be. So we see that still as very early innings. And then I think that's one of the 3 mega trends that's driving our business to your earlier point. Okay.

And then just the last one, John, in a few minutes. So China and the possibility of some export restrictions. So last quarter that you reported China, I believe, was almost 16% of sales, right, much stronger than what we had. I saw that it actually went down as a percentage of sales last year but was higher the year before. So what has caused this kind of -- I know looking at percentage of sales can be a little bit, I guess, misleading. It's grown. But what is created this cyclicality in demand from China customers?

Yes, that's the issue. Now of course, you had a customer that was added to the entity list that played into part of it. And then this also -- if I look at 3-year revenue CAGR. As individual years, you can get a skew of like if there's a lot of hardware purchasing in 1-year compared to the next it can drive up revenue in 1-year and then next year's growth looks very muted by comparison. But when you look at the underlying recurring revenue growth is fairly consistent. What we've seen in China is generally we've been growing at about 20% a year over the last few years. It's probably getting up into the mid-20s now. So I think we went from 10% a few years ago of our revenue coming from China to 13% like you say, Q1 was big. I don't think it will stay that high for the rest of the year. But we could be heading to like 14%, 15% of our revenue for the year coming from China. There's huge design activity going on out there. But to your point, that -- so we had customer added to the entity list back in 2018. I mean the engineers that worked at that customer have ended up with other customers right, and those customers have purchased tools from us for those engineers to use there. 1 learning I had from back in the early days, back in like 5, 6 years ago, we were focused on like customer consolidation. We saw a number of partners were acquiring each other. And we were tracking what was going to happen to our revenue and our revenue forecast. But -- and we have 1 customer that acquired another and completely removed the entire team that we're using our tools for Analog design in Sweden. And we still had about 18 months to run in the contract. But from a forecasting perspective, we thought, okay, write that off. We're not going to get that in 18 months' time, that we've got to find a replacement for that. And what we discovered when you look back at it was that those engineers were picked up by another customer. We got revenue from them, and we had kind of a double up on revenue for about that 18-month period. And when that -- when the customer that had led that engineering group go when their renewal came around, they still renewed all of those licenses. They were just doing that Analog design work in a different country in a different location. So it's remarkable just how fluid our customer base is. Our geographic mix of revenue is based on consumption. It's where the work is being performed. And the engineers are quite...

Does that mean multinational in China would be classified as China revenue? Or they won't have as many design [indiscernible]...

So a multinational in China, that seek purchases $10 million a year or spends $10 million a year with us and does 10% of the design work outside China, we will reflect that as $9 million China revenue because only $9 million is being done in China. In the U.S., if we sell to a customer here in the Bay Area, and similarly, 10% of their design work is being done in China, we'll represent 10%...

Ho do you know how much is being done?

So we have call home technology in the stock. We know where all of our software is, but we also work with our customers. They tell us where they want the key ship to. So that -- so we know exactly where the licenses are being used. And we set up revenue like that. So China revenue in our like 10-K in our 10-Q is slightly -- so it reflects what's being used in China, but it would be inflated or if you view it in terms of the business that we do in China. It's -- there's not that much difference, probably only a couple of percentage points. Throughout, its like 14% this year, maybe 12% is coming from sales into China and 2% is being transferred in net from outside. So what we find is that our -- the designers and the engineers that use our tools are quite fluid. They move across borders and at different companies very, very quickly. And you can see the usage profile change. It's very hard to predict what geographic revenue is going to be, but it's very easy to predict what total revenue is going to be.

That's true. Yes. [indiscernible]. Good. John, Richard, thank you so much. Really appreciate your time. Thanks, everyone.

Thanks for having us.