We'll get started now. Good morning, everyone. I'm Stacy Rasgon. I'm Bernstein Senior Research Analyst, covering the U.S. Semiconductor and Semic ap space, and it's truly an honor to have our guest here today, Dr. Anirudh Devgan, the President and CEO of Cadence. Before I start, I wanna mention if you have questions you'd like to ask during the presentation, you should have a QR code in your program that links to our Pigeonhole forum, where you can submit those questions. We'll have time for a Q&A at the end. Cadence has also asked me to read their safe harbor statement, so I will read that now. Today's discussion may contain forward-looking statements, including Cadence's 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 disclaims any obligation to update them. Great. With that, let's dive in. Anirudh, I'm so, I'm so happy you're able to join us today. Thank you so much for coming.
Of course. Thank you.
I wanna start at very high level. So, this is primarily a generalist conference-
Mm.
Not everybody here maybe is deep in the weeds on semis, and some of them may not know exactly what EDA is.
Yeah.
Can you just talk a little, like, just what is Cadence? What is EDA? What, what does it stand for? What does it do? Where does it fit into the semiconductor supply chain, and, and, and where does Cadence play within all of that?
Oh, yeah. Thank you for the question, and thank you for your interest. My kids don't know what EDA is. Yeah. Actually, they barely knew that there are chips in these things, you know, they were using, you know? But at least now everybody knows there are semiconductors in-
Mm-hmm
... phones and TVs and cars. But what we do, what Cadence do at the highest level is, you know, we basically make products to design semiconductor chips and electronic systems. So it's mostly, you know, software products.
Mm-hmm.
We have some kind of supercomputers to accelerate software development, but this is to design... and then all the chips that you see around you, and we are privileged to work with all the leading companies in the world, in all-
Mm-hmm
... kind of verticals, in all geographies. So that's it in a nutshell.
Okay.
Yeah.
So just to, I guess, simply, it's, it's sort of a software environment that's used to design the chips and supercomputers used at the. I guess this is what you call your emulation-
Yeah
... business, correct?
Yeah.
So simulating those designs before you actually put them in silicon-
Yeah
... and realize something is wrong.
Exactly. So the thing with semiconductors is this, this matured over the last 40 years and, of course, going through a resurgence now, okay? But they are- I mean, at the simplest level, they are very, very, you know, complicated, right? So they are cannot be designed with like hu-
Yeah.
It's not like your architect can draw, like, a five-bedroom floor plan-
Yeah
... of your house.
They used to do it by hand-
Yeah
... like, back in the day, right?
In the old days.
Yeah.
But now they, like, you know, some of these chips will have 100 billion, 200 billion transistors operating at 3 nanometer or 5 nanometers. So all of this is done by mathematical software-
Mm-hmm
... or engineering, what we call computational software. So we will have a high-level description, then we do synthesis, optimization-
Mm
... simulation. So it's a pretty involved area. And so all our customers will work our software to design-
Mm
... whether CPU, GPU, or phones or car chips.
Got it. Got it.
Mm-hmm.
Then who are your customers? Like, is it, is it just every semiconductor vendor on the planet buys your, buys your toolsets, or kind of?
Yeah, what we like to say is that... And also, Cadence is the most diversified-
Mm-hmm
... EDA or design software company.
Mm-hmm.
We have software to design digital things-
Mm-hmm
... you know, like GPUs and CPUs, software to design memory, you know, all the big memory makers, software to design analog, you know. So what we like to say is that almost any chip design in the world today uses some form of-
Mm-hmm
... Cadence software. So we are lucky to work with all the leading semi companies, and some of them are public, of course, and then also all the leading system companies that are designing chips. Because the other big thing that happened over the last few years is these big system OEMs, you know, the phone makers-
Mm-hmm
... or the car companies-
Yeah
... you know? And this is well known, right? Like, Apple has a whole line of, M-series and its chips. Of course, Tesla is designing their own chips.
Yeah.
All the data center companies are building their chips. So what happened in the last few years is not just working with the semi companies, which are great, and they themselves are becoming system companies.
Mm-hmm.
Like, NVIDIA is truly a system company.
Yeah. Yeah.
Qualcomm, truly a system company, Broadcom. And then the system companies are also doing silicon-
Mm
... so that adds to our customers.
Got it. How does working with those system companies differ from working with—like, do you have to do a little more handholding upfront or like... Again, I'm probably grossly simplifying, but-
No, these are remarkable companies.
Um-
I mean, and there, there's a range of capabilities. I mean, some of them are best in class, right?
Mm.
I mean, they don't require any handholding at all, right?
Mm-hmm.
But some of them, we do have some value-add services.
Mm.
It's not our main business, but sometimes we do provide some value-add services, especially to system companies. But yeah, I mean, but now they have a lot of them have really remarkable teams, and also they have an advantage of... See, when it's you're a system company, the advantage is, you know, this, anyway, they have a software stack, you know? Otherwise, you wouldn't be. So there is a lot of optimization that can happen with the software stack-
Mm
... and the silicon. And so things like emulation become more important in system companies.
Mm.
So what emulation does, which is our hardware business, you know, we are mostly software, but what happens is when you are designing at the highest level, so for people—I mean, for those who are familiar, this is like—basic stuff, if you're not familiar, it just— So what used to happen in the old days, and pardon my, you know, use of hands, is that you would do, like, silicon development, right? You would finish. It would take, like, four or five years.
Mm
... in the old days. Then you would do software development. That would take, like, one or two years, and then you would release the system, which is hardware plus software. Now, this is too long now, so they get overlapped.
Mm.
This perfect example is all the latest chips there. You do the hardware development and software development in parallel-
Mm-hmm.
So when the chip is finished, it just comes out, you know, within few months, and it has all the software running on it. So only reason that's possible, and this reduces the cycle time, because if you want new chips every year, you know, and the chips take, you know-
Mm
... time to design and software, you have to overlap it, okay? So the only reason this is possible is that when you're doing software development, you don't have a chip, but you can emulate it. So we sell these supercomputers-
They can run the software on the emulator?
Yes, exactly.
Oh.
So all these system companies and these big companies like, you know, NVIDIA is a development partner of Palladium.
Mm.
So they will run like-
We'll talk about NVIDIA.
Yeah. So because they're, you know, they have talked about it publicly at Jensen and NVIDIA-
Yeah
... so which we really appreciate, you know. But so NVIDIA will run software stack, whether it's Windows or, you know, CUDA or Android-
Mm
... or same thing with the system companies, well, before they have a chip.
Mm.
So, these emulators will run that software-
Mm
... I mean, slower than real life, but maybe 1,000 times faster than if you were to run it on a general purpose CPU.
Mm.
They become essential to the design of semiconductors and software.
Mm-hmm.
So the other thing that's different with the system companies, I mean, semi companies, a lot of them have software anyway, but the system companies definitely have software-
Mm-hmm
... you know, so.
Got it, got it.
Mm.
How do they used to do? They used to do FPGAs or something? Or, like, how did they-
I think it-
Make it slower to run it there?
It... Oh, yeah, it's slow. It was just more cumbersome. You know-
Yeah
... you can do it with running it on—in the old days, you would do it by running it on just regular x86 CPUs, but they are too slow to really do-
Yeah
... a software bring up and verify. So, the last, you know, we have done this for, like, now-
Mm-hmm
... you know, more like 20 years.
Mm-hmm.
The importance of these emulation systems became more and more critical.
Mm
... as the chips get bigger.
Huh. Do you think that's, like, one of the bigger development issues that the industry is facing? I mean, 'cause there, there's been a bunch of stuff. You have an increasing complexity of process-
Mm
... technology and, like, multiple patterning, shorter design cycles that you mentioned. Chip development is, in general, getting more-
Mm
... challenging and more expensive.
Mm.
What other sort of like, I guess, specific innovations is Cadence have they brought and are they bringing to the table to help reduce those cycle times, to reduce that design?
Yeah.
'Cause I've seen some of these charts which shows, like, design costs. I mean, I don't know that I believe the charts, 'cause I think if they were real, like, we wouldn't be designing any chips at all.
Yeah.
There's just no way.
Well, people like to project these-
Yeah
... you know? Just like, you know, right now, the power demand will go through.
Yeah, yeah.
Yeah. So I mean, the design costs have gone up to some extent, but I think what is important to remember-
But-
... which sometimes is missed in this-
Mm-hmm
... okay, is that those have to be normalized to the volume of silicon?
Mm-hmm.
Because, you know, this is upfront design cost, so if you're selling like 1,000 chips or systems or 1 million or 10 million-
Mm
... that's a very different economics. So the good thing is that because of silicon proliferation into everything, you know, the volume of things are going up a lot.
Mm.
So the cost normalized by volume actually be-
Mm
... going down. That's the reason that, you know, all the system companies want to do things-
Mm-hmm
... is because they have high volume also, whether it's data center AI-
Mm
... or phones or cars, right? Now, but we are always trying to make things more and more efficient, and this is the history of EDA, you know.
Yeah.
In the last 20 years, we have improved design productivity by 100x.
Mm.
Because, you know, remember in the early 2000, it would take like, you know, few hundred, 500 people to design a chip in, like-
Mm
... five years.
Mm.
Now, you can do that in, you know, six to 12 months-
Mm
... with probably 40, 50 people. So it's still more to be done.
Uh-huh.
The history of EDA, the A stands for automation, right?
Mm.
Of course, the big automation now going forward is AI-
Yeah
... you know.
We'll get there.
Yeah
... for sure, like you said-
So that's another level-
Yeah
... of automation. Yeah.
Got it. I guess just one more question, just on the broader market.
Yeah.
How do we think about EDA growth? Is it just-does it grow with long-term semiconductor growth, or is it like semiconductor R&D, a percentage of that? 'Cause it's-I always sort of looked at it as kind of the latter. It was-
Mm
... it was almost like, you know. And it's weird because it gets into the pricing discussion too, right? 'Cause I feel like historically, you were not really negotiating pricing with... It was mostly with the procurement folks, right? It was like, you know, we spend X on R&D, and we spend Y% of that on EDA, and like, that's it. Is that how it is? Is that dynamic changing? Or like, how do we think about the drivers of the industry driving growth of EDA industry?
Well, one thing I would like to say is that, just like the value of semiconductors have gone up, the value of EDA has gone up, especially the more advanced-
Mm
... node it is. You can see it in system companies, too. I mean, the value of silicon in system companies has gone up.
Mm.
Then, in our EDA interaction, we are having more and more interaction at the CEO level.
Mm
... rather than, you know, what you mentioned in the past. Because this becomes so essential to getting more-
Yeah
... kind of realization of value because they are already investing billions of dollars in these node transitions, you know, from five to three to two.
Mm-hmm.
You have to make sure EDA, the design part, is also as-
Mm.
So that has gone up for sure. But we are part of semiconductor R&D spend. Now, the other good trend in the last like, 10 years is EDA as a percentage of R&D has increased-
Mm-hmm.
Because there is more and more need for automation. And then we have new customers, like system companies.
Mm.
So I would say the three big things would be the—I think the importance of EDA is higher in our customers and also in the foundries.
Mm-hmm.
in all the major foundries. And then we are a higher portion of semiconductor R&D. And thirdly, I do think that, you know, this new system company is adding more R&D. And even semiconductor R&D, even though semiconductor had, like, a difficult 2023, and I think hopefully-
Mm
... 2024 and 2025 is better. There's realization because of all this competition, that they do need to invest in-
Mm-hmm
... design and EDA.
Do you think the system players value more what you do than the semiconductor ones because they're newer to it?
No, we love both our customer sets, you know. I think, I think they're all very capable. They're... And there's also movement between them, right?
Yeah.
Some of the people that, you know,
Yeah.
No, they are both fabulous, and we have the opportunity to work with them.
Okay.
Yeah, yeah.
Okay. Let's talk about AI.
Yeah.
Let's do this. So I mean, you certainly talked about improving efficiency-
Mm-hmm
... design efficiency, and so where does AI drive that? If you went from, you know, 500 engineers in five years, you know, to 50 engineers in two years-
Mm-hmm.
Like, does it go to five engineers in one year? Like, where, where, where are we driving efficiencies with this?
Yes.
And maybe even to start, like, a broader overview, like, what does AI and EDA, like, mean?
Yeah, yeah.
Like, how does that... What does it actually look like?
Exactly.
Yeah.
Yeah, yeah. So give me a few minutes...
Yeah,
... because this is a big topic. Because the one issue with AI is that everybody calls everything AI, so you don't know what is ... what is real AI and what is, you know. So, so, okay, let me back up again.
Yeah.
If you, like, you give me a couple of minutes to-
Yeah
... what I think AI is. Because one Webster dictionary definition of AI is, like, any automation that humans could do is AI. Well, for that metric, we're doing AI from-
Mm-hmm
... when I was in kindergarten.
Yeah.
Okay, so that's not my definition of AI. Okay, so now, to me, AI will have three kind of phases of deployment. So if you let me say that first, then I'll talk about it. So and, and that's true for any technology, and then people say, "Well, AI is like the Internet," okay. Well, I think it's more fundamental than the Internet, but we can take Internet as an example, okay? It's more like... If you really want to get into it, it's more like-
Mm
... you know, calculus or electricity. Okay. So but the first phase of AI is always or new technology, always the infrastructure, which is what we are in now-
Mm
... and probably next several years. And silicon is a key part of that infrastructure-
Mm
... so will Cadence participate in that?
Mm.
So all the AI companies making chips and systems-
Mm
... so this is a, this is a thing that's really different from six, seven years ago-
Mm
... where semiconductors was, like you said, maybe maturing industry.
Yeah, I was joking. We did a piece in 2015, a big one. It was called Playbook for Maturing Industry-
Yeah, mm
... and then 2017 or so.
Yeah. So I think one is this whole infrastructure build-out, which we are far away from, and that they use our products, regular products-
Mm
... not just AI products. Okay. The second phase of a new technology, second phase of AI, is applying AI to your products. You know, whether it's Search or PowerPoint-
Mm
... or EDA, you know, and we are, of course, doing that for at least last five, six years-
Mm
... including GenAI. And then the third phase of AI will be like new applications that will emerge that never existed before, just like what happened with Internet-
Mm
... with Facebook or social media. So I think AI will go through these three phases, first of all. So and we are participating in all three, and the first one is with the infrastructure build-out with our customers, whether they're system or semi. The second one is what you asked about-
Mm
... like, can we apply AI to ourselves?
Yeah.
And of course, we can, in a big way. And so and the reason that what AI can really do for EDA, which EDA never did, you know, the really new part of AI, so which is this? So if you look at EDA, you know, we write all this complicated software. You know, when you run it, it takes, like, one or two days to run some of the most... You know, this is, like, design software. So you give it some RTL description, then it will do the synthesis, place, and route, timing, performance, tell you how fast it runs, okay? And these chips will have, like, hundreds of blocks, so each designer is running, let's say, one block, just to make it-
What's a block? Sorry, what do you mean?
Block, like a-
Block
... number of transistors.
Yeah. Uh-huh.
This could be, like, a camera unit or a CPU unit.
Okay.
Typically, it's like, you know, a few million gates-
Mm
... or like, tens of millions of transistors. So one designer is designing, like, 20-30 million transistors using our tool.
Mm-hmm.
Okay. But what EDA... And these are, like, complicated things that run for a few days and give you a beautiful answer. But, you know, the design process doesn't happen in one or two days.
Mm-hmm.
You know, we'll be done, you know. So any design process is iterative, you know, so what happens is you run these tools, then you change something, you run it again, you change something-
Mm
... you run it again, and that's human intuition. So the driving of the tool is done by the h-
Is that really how they optimize it? Is it really just the intuition of the engineer, or there-
Of course, the tool is doing a lot of work-
Uh-huh
... but, like, some of the, of course, it's intuition.
I'm a dumb process engineer, so like the EDA-
No, no
... stuff is over my head.
It is, you know, it's the intuition of the engineer-
Uh-huh
... super critical. Also, the architecture, you know, this is what we want to do, you know, this-
Mm
... GPU should have this much, this CPU should have these features, and we want to go to TSMC 3 nanometer. We want to have this power level. So there's a lot of, like, high-level design-
Mm
... of course, that is the bread and butter of our customers. But then to implement that, it's an iterative process, just like in anything, right? So what EDA never did was that we were very good at what I call single run. Like, you give it an input, it'll give you a beautiful output.
Mm.
When you go to the next day or two days later and restart that, it has no knowledge of what happened in the previous run.
Oh, okay.
This is true in general in a lot of things, you know? Like when you open a new PowerPoint, it doesn't know what, you know, what you like in the pre-- So now, it's not that we didn't want to do that. We always wanted to do that, you know? But it was mathematically not possible to do it.
Mm.
Because you are transferring knowledge from one run to another. There are some ways to do it, but they were too expensive. You know, like-
Mm.
So what you're doing is you're searching the design space. I'll give you example, like, if you're doing CPU design, you're searching this design space, and you want to get a best power or-
How many, like, dimensions or variables are you, like, searching?
It's-- It could be a... I mean, like, even at the-- Now, when you design the things, it's like a huge number.
Mm.
Like, I said, $20 million-$30 million, sometimes $1 billion.
Uh-huh.
But even at the user level... Okay, this is at the tool level-
Mm
... it's doing millions of variables and billions of variables, but even at the user level, it could be something like 15-20 variables.
Okay.
Okay? Which is a lot, okay, in one case. So, you know, that's difficult for any user to do, so, so that's why the human intuition is important. Okay.
How many iterations are typically required?
A lot of it. So what happens is, you say, "Well, why don't you do it mathematically?" You know, the classical way to do it. Now, some people may call it AI, but the classical way. Now, now statistics is called AI, right? But there's a classical way in statistics to do. That's true. By the way, this is not true for companies. It's also true for universities. All my. You know, CMU, now they have School of AI, not School of Computer Science. You know, that's my alma mater. MIT, you know, it's all. Everything has to be AI.
Yeah.
What used to be computer science and math-
Yeah
... is now AI, right? So, in the classical statistics, the way to do that is design of experiments.
Yeah.
Sorry, taking a little longer just to... So that, the problem with that is that it would take millions of runs, even for 17 variables.
Mm-hmm.
So that's infeasible.
Yeah.
So then, the mathematical way was infeasible, and then the human way, of course, works, but it took, like, six to 12 months.
You may not find, like, the global optimum, right?
But still, it works great. I mean, it's still much easier than 20 years ago. And then, there are some other mathematical ways that don't work, but AI, you know, there's reinforcement learning and gen AI. The beautiful thing is that it can build a model of anything, so you can actually do that using AI. So for the first time, this is real, this is now real AI, you know, we can actually do this knowledge transfer from one run to the next run and do this workflow automation. So instead of the user just running one time, the machine runs, and so-
Mm
... in this case of the CPU example I was giving, with 15-20 variables. Now, it didn't take 4 million runs, but it take, like, 200 runs-
Mm-hmm
... or 100-200 runs, and it can do the design, okay? And then we can parallelize those runs because some of these things are parallel. The big thing is, what used to take one or two days in one run now can be done in one or two weeks. You know, instead of doing it in six months, you can do it in one or two weeks with, with AI.
200 runs.
Yeah, with AI, with 200 runs. So there's a lot of benefits to that. So one benefit is that, of course, they still iterate. Okay, they're not done in one AI run, okay? But still could be five, 10x better efficient-
Mm
... more efficient so far. That's a lot. Okay, so what that means is that one engineer, instead of doing one block, can do three to five blocks.
Mm-hmm.
So this is the real productivity benefit of, of AI. And the second benefit, which is, can be even more impactful, is that that optimal is better than what a human can do.
Yeah.
Because in a lot of cases, we are able to design chips which are lower power or faster-
Mm.
... and there are a lot of examples, sometimes 8%, sometimes-
Mm
... 10%, sometimes 15%, which is huge because you're going from one node to another node. The improvement is normally 5%-15%, and you're getting that by better search, you know, better optimization. So those are the two. Now, what is the impact to our customers and our business? So now, the typical thing with AI or automation, let's say this is another way of doing automation, and it's very. Now, we are all top 20 customers, all of them are engaged with our AI tools-
Mm
... because this is a real benefit. This is not like marketing AI to me.
Yeah.
Okay, this is a real AI. That's why I took a little time to explain what it is. The benefit is now this-
The first genuine use case, right?
Yeah. See, this is the other thing with AI: you need some level of automation already to apply AI on top.
Mm.
See, if there is no automation in the environment and you just apply AI models, it's more difficult to automate. Whereas EDA has done for years, you know, decades of automation, so we are actually pretty good use case. So then, what happens is then the other worry with AI is, okay, you have so much automation that you will need less people or less tools.
Mm.
Okay. So first of all, that's not the history of design, and the reason for that is that, in my opinion, I mean, there's several reasons, but one big reason is that's assuming that the workload is constant, and then you apply automation, then you need less work or less people. Okay, that may be true in some things.
Mm.
I don't know how many lawyers you need to process a document—... or whatever, okay, and not to pick on lawyers. But if you look at chip design, the workload is going up like this.
Mm.
So, you know, the biggest chips right now are, like, 100 million, 100 billion transistors.
Mm.
They're widely predicted to be 1 trillion by 2030. Okay. So that's 10 x bigger in size than the verification complexity, software complexity. So the design complexity-
Mm
... from 24 to 30 will go up by at least, okay, 30-40 years.
Does it scale linearly? I feel like it should scale more than linearly if the-
... No, but I'm coming to that, I'm coming to that.
Ah.
Let's say the design complexity goes up by 30,
Yeah
... 40x, okay? There is no way our customers or, you know, their system or semi company want to hire, hire 30, 40 times more designers or engineers. They're not even graduating-
Mm
... 30, 40 times. But I think they will still hire more because there'll be more silicon. But I would, you know, I would- if I were to guess, I would say they will hire 2, 3 times more.
Mm.
You know, that's reasonable for in the next five, seven years.
Yeah.
Still there's at least a 10x productivity gap that has to be filled, and which we have done in the past also.
Mm.
So that's what now AI can provide this gap for. Because, you know, even if something gets really non-linearly faster-
Mm
... or better, there is still some design iteration process. So actually, to, to really affect like a five times or 10x productivity in the real measurable-
Mm
... design process, requires component level to be even more efficient.
Mm.
So I think five to 10x is a, is a very good goal. You know, some people may claim higher, but I'm-
Mm
... talking about real end kind of output. And that can... And that's a very-
Mm
... the other good thing about semiconductors and application of AI, is because the workload is going up, so everybody wins.
Mm-hmm.
You know, the customer wins because now they have to use more automation and more compute, but they can get better designs and,
Mm
... and not invest as much in headcount. And we win as a provider, and then, of course, the customer wins because they have better-
Mm
... better solutions. Yeah.
Got it. No, that, that makes a lot of sense.
Right.
It does sound to me that if one believes that, I mean, the EDA should be taking a bigger percentage. I mean, it has, and it sounds like it should take materially more.
We should-
Does this-
Yeah.
Does this help with, like, I've always wondered about pricing in this industry, right?
Mm-hmm.
It goes back to my comment on, well, we spend X%, and like that's it.
Mm-hmm.
Does this give you... I mean, this, this sounds like a lot of value add. Like, how much of that can Cadence capture, like, as you're delivering this? Like, how do you think about that?
Well, the way I look at it is, you know, we just have to provide value to our customers.
Mm.
You know, everybody producing a product thinks they're not valued, and everybody buying the product thinks they are. So we want to make sure we provide enough value, and, you know, we save our customers-
Mm
... you know, enough things, and we capture part of that.
Mm-hmm.
You know? So I think I just want to make sure we provide value to the customers, which we will, and then, now they will keep some of the savings, and we should get some of that.
But everybody wins to you.
Exactly.
Right.
Yeah, because we cannot, you know, we are not. That's not our culture to just go on the pricing.
Mm.
You know, we want to make sure, and we have a lot of amazing customers helping us in a lot of ways. We have to deliver value to them-
Mm
... and they're always fair to us.
Mm.
We get our part of the-
Okay
... share of the gains. Yeah.
Let's talk about, like, what you're doing with NVIDIA.
Yeah.
So NVIDIA has—I mean, they have a whole CUDA library-
Mm
... set for that. I think, cuLitho, I think it's called.
Mm-hmm.
They have their new Palladium Emulator-
Mm
... Jensen.
Mm.
I'll let you tell the story, but Jensen's made some interesting comments around like
Mm, mm.
What, what exactly are you doing with NVIDIA? How are you guys helping each other around these lines?
Yeah, we have great partnership with NVIDIA, and NVIDIA, of course, is a remarkable company. Even more so in the last few years, our partnership has,
Mm
... has increased significantly. But we have a history of working with them more than 20 years.
Mm.
Okay. Now, one of the... And I think there, Jensen has said that they use almost all of our products-
Mm
... which we appreciate, you know, whether it is, you know, packaging tools or digital design tools-
Yeah
... analog tools, Palladium. Now, one—and sometimes we co-develop products with them. So, for example, Palladium, which is this emulator, you know, is NVIDIA is one of our main development partners.
Mm. Mm.
So they will kind of work with us years in advance-
Mm
... you know, based on what they need, you know-
Yeah
... the size of the chips and the requirements.
What did Jensen say about it? It was, it was,
He said, AI, I think he's—I mean, you can watch it. He says it's more important to him than any other appliance, you know, more important than a refrigerator. Jensen is also very funny, right? And then he said that, that basically, he said that you can't design Blackwell without-
Mm
... use of Cadence products and Palladium, which we really appreciate. And, you know, especially coming from-
Mm
... such a marquee company. So I think this is, but this is, this is good partner, great partnership. And then we also have all other R&D partnerships-
Mm
... okay, with on AI-
Mm
... because NVIDIA is doing a lot with NeMo-
Mm
... and NeMo Service, NIMs.
Mm-hmm.
So we have embodied them. We have partnerships on chip design using NeMo, and then we have partnership on the bio side.
Mm-hmm.
And we also use their GPUs. I mean, there's a longer discussion on Millennium and our other—you know, like Palladium is emulation for chip design. We have a new system, Millennium, which is emulation for system design.
Oh, okay.
Like cars and planes and phones, you know, like a system-level simulation.
Okay.
And it's a very remarkable-
Like thermals and mechanicals-
Yes, yes
... like, Really? Okay.
Yes.
Let's talk about that in a minute.
Yeah, I'd love to tell you more about that.
Yeah.
That's because one of the biggest problems going forward is power and thermal.
Mm
... at the system level, not just at the chip level.
Mm-hmm.
Right? Whether it's data center or it's a phone or a car or a plane. So we have a big partner. We call it Cadence Reality Digital Twin Platform-
Mm-hmm
... which also, because NVIDIA is using that to design their data centers.
Yeah.
And Millennium is the device. I can tell you more about that. So overall, I think, because I believe at the fundamental level, and this is a good understanding with NVIDIA, and you know, great partnership is, that there are three layers of the stack that has to be done. So the bottom layer is accelerated computing. Of course, GPUs play-
Mm
... a very important role. It used to-
Does Palladium use GPUs, by the way, or?
... No, Palladium is a custom chip.
Okay.
The reason for that is that, you know, when you do emulation of chips, you know, you're emulating Boolean things. You know, Boolean is 0, 1, 0, 1. Okay? So we have a Boolean supercomputer-
Mm.
It's a specially designed chip that we make ourselves, Cadence on Cadence, made by TSMC advanced node. So each Palladium rack, rack is, you know, this will have, like, hundreds of these chips.
Mm.
They're all liquid cooled. It's a pretty complicated system.
Mm.
Now, when we come to Millennium, which is simulation of floating point, you know, like, when you do simulation of thermal and all, it's a floating point-
Yeah
... matrix multiply kind of operation, which GPUs are great at.
Okay.
But for Boolean, you know, we had to do a custom chip.
Got it.
Okay. Uh-
I didn't mean to interrupt you, sorry.
No, no, no. That's, this, this is great stuff, and we have done this for some time.
Mm.
So, Cadence is the only company that builds... We are a software company by nature, but we have this special team that builds the entire systems.
Mm.
Of course, we know how to use our tools and, you know, great, great benefit.
Mm.
So we actually use our AI tools to design Palladium.
Mm.
And the
So you've got direct knowledge of, like, how, what, what's the efficiency gain, like, by doing this?
Yeah, so we got 15% power saving.
Mm-hmm.
This is amazing.
Okay.
15% power savings in the design of Palladium Z3-
Mm
... using our AI tools-
Mm
which is Cerebrus. And so, these are pretty, so this is like full rack system, this is the hardware part of our business, to design-
Mm
... other chips.
Right.
So, like, when NVIDIA and Jensen talks about you need a supercomputer to design and test-
Mm
... our supercomputer, he's talking about Palladium as the supercomputer.
Got it. Sorry, I, I interrupted you. You were talking about three levels, you said, to accelerate computing.
Oh, yeah, yeah, exactly.
Okay.
This is important. I think this is going to happen in all industries. Okay, so there is accelerated computing at the bottom, right? CPUs, GPUs, FPGAs-
Mm-hmm
... and also custom chips-
Mm
... you know, just like we do in Palladium.
Mm-hmm.
There's going to be a rich set of. This is what is very different from 10 years ago in semiconductors, okay? You can see even new things. I mean, we can talk a lot about that. Even, like, what Qualcomm announced-
Mm
... we worked with Qualcomm for a long time. It's CPU, GPU, plus NPU-
Mm
... and all the Apple M-series and all. So there is a lot of richness in the hardware, you know, the computing. Then the middle layer of the stack is what I would call, like, physical intelligence. This is, like, the basics. You still need the basics, you know. You know, what is coming from physics or chemistry or differential equations-
Mm-hmm
... the actual behavior of the transistor-
Mm-hmm
... or whatever system, the car you're trying to model. Okay? And this is more principle simulation and optimization. And then the top layer is data intelligence or AI, you know, which is more on-
Mm
... you know, fitting a model of behavior-
Mm
... based on input, output. So I think these three things will happen in all industries.
Mm-hmm.
Okay, and it's already happening. And of course, we have a lot of expertise in all three of them. And then you have to verticalize them for multiple end markets.
Mm-hmm.
I think the real value in the long run will happen in the verticalization of these things.
Mm.
Okay, because... So you know, one verticalization, of course, is chip design. That's a great vertical. I mean, there are other vertical, I mean, like, sell cars, you know-
Mm
... same thing will happen in cars. Okay, same thing happened with, with Search,
Mm
... you know, what, you name it. But from a Cadence standpoint, there are three verticals which are most interesting. I mean, there are a lot of interesting verticals, but which are more computational in nature.
Mm.
You know, we always want to do, like, computational or engineering software. So one vertical, of course, is our core business, which is chip design. The second, you know, is what I would call system design, which, you know, if you go to, like, aeronautics or something, or planes as example, or data centers, like, this kind of thermal things-
Mm
... I was talking about. You know, when chip design, we simulate like 100% of the stuff-
Mm
... is done virtually. Or 99, should never say hundred, but let's say 99%, you know? And, and then, you know, when the chip comes back, you know, it works, right? First time right-
Mm
... most of the time, you know. Especially if you use Palladium, it always is first time right. Well, that's the big reason-
Right
... to get first time right. Whereas if you go to planes and cars, you know, like planes, they simulate only 20% in the computer.
Mm.
Because not that they don't want to, because it's either too difficult or too slow.
Mm-hmm.
So now they still verify the rest by doing physical tests, like wind tunnel-
Mm
... or something like that. But it's very cumbersome, right? It's very cumbersome compared to chip design.
So-
Mm
... what do you guys, like, this one is, like, to give you a chance to talk about, like, what's different between the system design math? I mean, are you just doing computational fluid dynamics? Are you solving, like, Fourier's equations? Are you-
Yeah, I see it. Actually, mathematically, to me, okay-
Uh-huh
... it's simpler than chip design.
Yeah.
But we never applied it, so EDA never applied it to SDA.
Uh-huh.
Because chip design, the reason it's more complicated is because, one, we have a huge number of variables, right? These things have 200 billion transistors.
Yeah.
But transistor is the most nonlinear thing. It's a switch. I don't know how much you know about actual working of transistor. It's a zero-one switch. It's, like, the most nonlinear thing-
Mm
...by nature that you... So, so this is, like, very difficult to design.
Mm.
These are nonlinear-
Yeah
... large systems, okay? Whereas things like fluid dynamics or weather simulation is more linear-
Mm-hmm
... okay? You know, and also the math is a little bit simpler-
Mm
... than EDA. But EDA and their companies never did this, unless I started this in 2018-
Mm
... to do this. Because at that time also, we saw this thermal issue, and also when you go to 3D-IC, you know, that's the other big trend-
Yeah
... you know, we can talk about for a while, in chips, so we also need thermal and... So now to give exa- so that's the-
Looks like an offshoot of your chiplet efforts, like-
Exactly.
You do it anyway. Okay.
See, we had to move to the system level because we also have majority share in package design.
Mm.
Cadence is the most-
Yeah
... widely used. Allegro is the most widely used tool for package design. The other big thing, trend is, of course, System-in-P ackage and 3D-IC-
Mm
... last several years, okay? But we have been doing this for a long time now, last several years, very closely with TSMC-
Mm
... and now Samsung and Intel, okay, and Global and all the main foundries. So that also requires a system-level view and a thermal view, because 3D-IC, the biggest issue is thermal.
Mm.
One of the biggest issues.
Yeah.
Okay. So we had to go there anyway because of our 3D-IC, and then, okay, then why not apply it to-
Mm
... plane design or car design? You know, we have a partnership with, I don't know if you saw, with McLaren-
Mm-hmm
... also. That's a great partnership because in racing cars, I don't know if you follow F1, the biggest issue is aerodynamics.
Mm.
The biggest differentiator. And same thing is true for electric cars-
Mm
... you know. And aerodynamics is needed, or CFD is needed for thermal anyway, because airflow is a big issue.
Mm
... in your laptop or a phone or something like that. So that's why we are doing a CFD, and the results are phenomenal, especially on this three-layer stack, okay? So we have a new solution called Millennium that completely changes-
Mm
... the way CFD is done. Take, you know, we have this new tool from Stanford we acquired-
Mm
... called Cascade, which is a new way of doing CFD, which is much more accurate.
CFD is computational fluid dynamics.
Yeah
... by the way.
Which is simulating, like, planes and, you know, phones. And so, so we have this, a company we acquired two years ago, which is much more accurate CFD-
Mm
... out of 25 years of research at Stanford, okay? And they were already working with NVIDIA, and already had some AI efforts with-
Mm
... with Caltech and other universities. Okay. So always had this idea of using this tool, but accelerated heavily on GPUs, because this is numerical now.
Mm-hmm.
Okay? It's more like AI kind of calculation.
Yeah.
So GPUs is perfect for that. And also our partnership with NVIDIA.
Mm.
Then also put AI on top of it. This is what we call Millennium.
Okay.
This is, I think, one of the most fundamental products in that space in a long time.
Is there, like, a software stack that goes with that as well, or is it just, like, primarily the hardware sales?
No, no, it's a software stack-
Right
... of course. So the two top two layers are software, the bottom layer is-
Got-
So, yeah.
Like, how much of your revenue today is this kind of systems analysis?
Oh, overall system analysis is about 12% of revenue-
Okay
... growing at 20% a year, for the last several years. So this is a very good business for us, and it's also, you know, good profitability.
Mm.
Now, this new system in which we are combining hardware and software-
Mm
... that's new. Because we are taking, you know, what we did in Palladium-
Mm
... which is for chip design, and we want to have a hardware-assisted system for-
Mm
... system design, which never existed. You know, the system design never had these emulators, right?
Yeah.
This is a first, this is a new, new product, beginning of this year.
Okay.
Yeah.
Is this where the BETA CAE acquisition fits in, or... Because you've got a-
Mm
... a competitor of yours that is going larger into systems. They're buying Ansys, and we'll see. Like, is, like, it sounds like you've been working on this, at least on your own, be-
Yeah
... before that.
Yeah, yeah.
Mm
... I mean, I think we are doing this for a long time-
Yeah
... because we see all these trends. And other people may try to copy us now-
Mm
... but I think our culture is more of we would rather do it organically-
Mm
... in the right direction.
Mm
... because it's also better for investors in the long run. And, and also we are doing well organically anyway.
Mm.
Okay? Now, from time to time, we will do some tuck-in acquisitions we have.
What is this new one? Is it BETA?
Yeah.
Yeah.
I think after BETA, I think we are pretty complete in our systems portfolio.
Mm.
So, one thing that BETA brings is what is called structural analysis.
Okay.
Okay? So we did electromagnetics ourselves, which is things like, you know, on the, like, interference on the package.
Mm.
Then thermal, I talked about, that's a big thing. Now, one thing that was missing was structural. Now, structural is useful for a lot of things, like, you know, when you drop your phone, you know.
Mm-hmm.
It's a crash test. That's a stru-
Mm
... structural simulation, or if you crash your car-
Mm
... you know, that's a structural simulation. So BETA CAE is used by almost all the car companies and a lot of some of the other kind of electronics company. Now, what is happening... So that's a good business in itself, okay, and they're very well-regarded. Now, what is interesting is even in our core business, okay, there are structural effects coming in-
Mm
... because of 3D-IC.
Mm.
So if you look at the big thing is HBM, right? You know, if you follow the you know, this memory, and they have, like, eight layers of memory. These memory guys-
Yeah
... are very advanced, okay? Sometimes we don't realize-
Yeah
... how advanced...
Yeah.
And then eight-layer is going to 12 layers of memory with this new HBM, all this news, and this is one of the limiting things, also for AI adoption. So there are all kinds of structural effects that happen when you have 12 layers of memory.
Mm.
So not only it's a good business for designing planes and phones, it can be applied to the silicon structure.
Mm.
So that's the reason for BETA CAE, and it's a good company, and, you know, should hopefully close soon. It's not closed yet.
Mm-hmm.
It completes our whole systems portfolio, so we are in very good shape in that.
Got it. You have an IP business as well.
Yeah.
Right. Can you talk about that a little bit? Like, what- is that like Arm, like you're selling like cores or like, what, what, what is this?
Yeah, yeah, so first of all, Arm is a great partner of Cadence.
Mm.
You know, Rene and Arm, and we have worked together for more than 10 years.
Mm-hmm.
Okay? You know, majority of Arm flow is Cadence-based, okay? Majority of Arm's customers, when they design CPUs and GPUs, are Cadence customers. So we are very deep partnership with Arm... so this IP is slightly different than that.
Mm-hmm.
It's not like it's, you know—so either we have, like, some embedded processors, you know, like Tensilica, which is, like, doing, like, more DSP-
Mm
-kind of, or this is physical IP. So physical IP is the things like, you know, UCIe, like chiplet connections or HBM IP or DDR and these kind of things. So this is more like for a particular node. See, Arm is more higher level IP-
Mm.
This is more physical IP, and it's a good business. Now, normally, it's not as profitable as software business, but it's a good business, and we, we are, again, 12%-13%-
Mm
of our revenue is IP, and I do see a lot of opportunities for IP to grow-
Mm
because of AI and HBM.
Mm.
Also, a lot of the foundries all require more and more IP.
Mm-hmm.
So we have, and as you know, there are more foundries now, more. That's a very good thing. So, so apart from TSMC, I think earlier this year, we announced, like, a new partnership with Intel Foundry.
Mm-hmm.
With Pat and Intel, they are investing a lot in the foundry.
Mm.
Samsung is investing a lot.
How much do you want to say about Intel's foundry?
Oh, we work with Intel Foundry—you know, so we work with all the great foundries.
Ah.
And that's for our customers to decide, you know, what foundries-
Mm
to use, but we enable all our foundry partners.
Yeah.
Okay. Now, we have a much longer relationship with TSMC, you know, and-
Intel was pretty insular back in the past, I think-
Yes
... on their EDA efforts, right? So they're opening up now.
Yeah. So that's good for Cadence.
Yeah.
You know, it's like we were not in Intel in a big way, and now with the, your new design effort and foundry efforts. So we are glad to work with Intel-
Mm
... glad to work with Samsung, and then we have very long partnership with TSMC.
Mm-hmm.
They have been our development partners for, I don't know, 15 years.
Yeah.
Most of TSMC flow internally is Cadence-based.
Mm.
You know, like I said, most of TSMC customers are also big Cadence customers.
Mm.
we wanna make sure that, you know, we enable all the foundries in a good way.
Got it.
So, but IP is a part of that. So there could be good growth in IP. We expect good growth this year-
Mm
... with a combination of, like, AI-driven design-
Mm
more foundries coming online, and hopefully that continues in the future.
Got it.
Yeah.
Got it. You know, just around the competition, so it's a fairly-
Mm
... concentrated-
Mm.
There's not a lot of players. It's you and you, it's Synopsys and-
Mm-hmm
... I can't remember, I can't remember who got bought by, by Siemens. Was that a-
Mentor.
That was Mentor.
Yeah.
It was Magma back in the days.
Yeah.
I guess, like, how do you differentiate from your competition? And then also, like, I look at just the gross margins. Like, your gross margins seem to be higher. Like, why, why is that?
They are much higher, yeah.
Yeah.
Okay, one thing on gross margins, I mean, you guys are very smart anyway. So one thing we always look at, so there's gross margin or gross operating profit, but also operating profit, and SBC.
Mm.
Because, you know, stock-based compensation is-
Yeah
... is sometimes a lot of tech companies don't pay as much attention to, which is as important. Rather, you know, all our employees would rather get stock instead of cash.
Mm.
One key metric we take is operating margin minus SBC.
Okay.
Okay? And we print that in our CFO commentary-
Got it
... right? So if you really calculate that, we are really much more profitable-
Mm
... than anybody else, okay? Because our SBC, our profit margin is, this year, is 42.5%, is our guide, and then our SBC is 8%-
Oh
... as revenue, so-
50, over 50% then.
Yeah, so that's something to remember. But the difference between our competition-first of all, Cadence is more software centric, okay? Our competition-first of all, it's a good industry anyways, I don't want to-
Yeah
... and first of all, the importance of industry is very important because we want to have a good house in a great neighborhood, you know, not just, you know, like. So there's no point talking down our competition.
Mm
... too much, but just to
I'm not asking you to talk them down.
Yeah, yeah. But just to differentiate between the competition is that, that we are more software heavy, they are more IP heavy, so they have bigger portion of IP business, which is fine.
Mm.
I think the reason our margins are also higher is because we are more software centric.
Mm.
Okay. So we are more EDA, and they are more IP. And also, within EDA, we are more diversified.
Mm.
We have a big analog business. All the analog companies use us. We have big digital business and verification and packaging.
Mm.
So that's why, you know, a lot of the companies will... You know, we have the full set of portfolio, complete set of portfolio in EDA.
Mm.
And then, over the last seven years, we have done SD&A, you know, which is all the system stuff. So that's the kind of-
Mm
... difference. So we are very well-positioned, and also we are stronger in the TSMC ecosystem-
Mm
... because of our history. We are stronger in the Arm ecosystem because of our history. So, so that's kind of some differences.
Got it.
Yeah.
Got it. We've got time for, maybe one or two questions.
Yeah
... from the audience. Lightning round, I guess.
Mm.
Are China export controls a risk? And, like, what's the risk from China, like local EDA players?
Yeah, I mean, there are several companies there. I watch it carefully, you know, maybe like at least ten companies-
Mm
... in China doing EDA. Now, they are smaller and more specialized. So when I talk to our big China customers, I mean, they're still, you know, we are the primary flow.
Mm
... used in China. So to me, that's not a, you know, and again, I don't want to underestimate these things because we watch it very carefully, but that's not a immediate short, medium, or even what I can see in the long run, risk. Yeah.
Got it.
Yeah.
Got it.
Also, because all these things have to be certified by the big foundries-
Mm
... and also capabilities are pretty new.
Yeah.
We feel confident where we are. Now, China, I think the bigger risk in China, to me, is more the macro situation.
Mm-hmm.
You know, because the macro, I think, has been a little weaker, and, I mean, there are some signs it can improve, but we will see-
Yeah
... more than the local competition.
How much of your business is in China anyway?
... I think, like Q1 was like 12%.
Okay.
Yeah.
Okay. And the export controls, is there any? You are controlled, I guess, like, given what-
Yes
because what you can ship out
No, we follow all the U.S. regulations.
Mm-hmm.
Okay? So there are some controls on it, especially, you know, gate-all-around and things like that. But most of the U.S. controls are on manufacturing.
Mm-hmm. Got it.
We are on the design side.
Yeah.
Okay, so most of the controls are 14 nanometer and below manufacturing-
Mm-hmm
So it will affect, like, foundries or the equipment companies.
Mm-hmm.
Now, there are some controls on EDA, and we, of course, follow all of them. But a lot of the design activity in China, just for your reference, is for Samsung or TSMC-
Mm-hmm
-and other nodes. So, like, if Xiaomi is designing some phone chips and all that-
Mm-hmm
They're typically not manufactured in China.
Right. Right, right.
So, we work with all the companies that are doing the design effort in China.
Okay.
Mm-hmm.
Got it. So we have about 30 seconds left.
Yeah.
I'll give you your soapbox. You've got a whole room full here. Like, why should investors buy Cadence stock today?
Well, first of all, thanks for your interest, you know, like I said in the beginning. I mean, I think the good thing... Okay, I'll tell you one of the good things is that we are a software business, okay? So there are some advantages to that, because we have very high gross margin and, of course, high operating margin. You know, we have the highest margins, I think, software businesses. So that's. And because we are engineering software, we are much more differentiated, and we have a big moat around it.
Mm-hmm.
See, one issue with software business is that, like, is there too much competition or is how sustainable it is and all that. So we have all the advantages of being a software business, but, at the same time, having advantage of a big... I mean, it's very difficult to enter this market. You know, it's a big investment-
Mm
... a massive R&D investment, takes 10, 15 years to do it. So, so that's a good position.
Mm.
Now, we have, you know... And then underlying that is... So that's on the margin and differentiation side. But underlying all that, this software is applied to the semiconductors and electronic system, which is in a generational boom, at least for the next 10 years, in my mind.
Okay.
Maybe longer, but at least we can see the next 10 years, which is semiconductor, AI, you know, system companies doing more silicon.
Mm.
So you have a very well-run business. We have done this for a while financially, but in a very good growth market-
Mm
very difficult to enter.
Mm
-with a moat around it. So I think these are the three great characteristics of Cadence, and it took a hard time and a long time to get here-
Mm
But I think we are in a very good position. And also, the other thing I really enjoy about Cadence, and I used to be in like, you know, big . . . IBM long time ago, and then other companies, is that we do have the privilege of working with all the leading companies in the world, okay? Because they are all designing chips-
Mm
or electronic systems. So we do have a good front-row seats, and we can see the trends coming, and then we can invest in them and have great partners-
Mm-hmm
Like you have mentioned. So that's, like, Cadence in a nutshell. Yeah.
Got it.
Yeah.
That's wonderful. Thank you so much.
Thank you. Yes.