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

Hi, everyone. Welcome to join us at the 20th Annual Needham Technology, Media & Consumer Conference. On the virtual stage with me today are Nimish Modi and Senior Vice President and General Manager for Strategy and New Ventures as well as Richard Gu, Vice President of Investor Relations from Cadence Design Systems. Before we begin, really a couple of housekeeping items. I'll first read the following safe harbor statement on behalf of the Cadence team. Today's discussion will contain forward-looking statements, including Cadence outlook on future business and operating results. Due to risks and uncertainties, actual results may differ materially from those projected or implied in today's discussion. Next, since this is really a virtual fireside chat, please feel free to submit your questions in the Q&A box. We'll have some time near the end of this session to allow our guests to address your pressing questions to directly to Nimish or Richard. As usual keep your questions anonymous.

All right, Nimish really welcome back. Last year, you are here with me at the same conference. That was the first time, I think we had -- we've known each other for a little bit longer than that for sure. But welcome come back. A lot has changed over the past 12 months, right? But it feels like Cadence hasn't really changed much. I'm making it in a positive way. The business is still growing. The top line in low double digits, your operating margin still in the 40s and still expanding, right? While there's increased worry about macroeconomic conditions as it relates to maybe AI infrastructure buildup cycle and maybe more specifically, the semiconductor cycle as of today. Partially due to some really meaningful change and as we have seen today, the volatility in the policy landscape. So let me ask the first question this way, Nimish. You guys definitely talk about resiliency of the business through any market uncertainty, tell us why Cadence is uniquely positioned in such a way in the semiconductor industry and what leads to this resiliency, which looks like a structural inherent to the Cadence business?

Sure, Charles. First of all, thank you for inviting me back. It's good to be back. And at the macro level, as you said, in many ways since the last time we had this chat, a lot has changed, a lot of movement at the macro level. But at the same time, this just underscores some of the cornerstones about our business that makes us that much more buffered and resilient. So if we kind of break it down, first of all, we are essential to our customers. The semi and system companies are primarily our customers. We are tied to the customers' R&D budgets and their design cycle. So you remember that the customer's design activity of today is translating to the product, which is coming out in the market 2 to 3 years down the road, right? So they want to be ready to come out with their competitive products on the other side of any potential downturn, which may be the often. And so core R&D is usually one of the very last things which gets impacted. Secondly, from a Cadence perspective, we are very diversified in terms of products. We got all kinds of products, right, end-to-end core EDA products. We got IP, growing IP portfolio, which we'll talk about later, system products, hardware. We have a very diversified customer base across our semi and systems customers. We don't have any 10% customer. And we're also very broadly geographically diversified as well. And lastly and very importantly, we have a predominantly recurring business model, right? It drives a lot of visibility, predictability. So very high gross margins, strong exit Q1 backlog. So when you -- [indiscernible] model, which is predominantly recurring. So when you put it all together, we feel pretty good about how we are set up to withstand this. Now again, to be clear, we are not completely immune to -- if there's a significant prolonged downturn, we'll see. By the way, we are situated very well. And you can see the multiple elements which are driving the structure resiliency in the business and which is why in this environment notwithstanding, we beat unraised and when we reported Q1.

Thank you, Nimish. So maybe moving on to some of the or more recent events. I want to talk about the product announcements you guys made at a Cadence Live, which happened a couple of weeks ago. One -- yes, one is Millennium and the other is about the --around the Tensilica. I want to ask you more about the Millennium here. . I think Jensen Huang, NVIDIA CEO, he showed up, but he had a fireside chat with Anirudh, but he talked about buying 10 units of Millennium, you guys. I thought -- because thanks to you guys, right? Millennium I think you guys launched not this year, right, but probably in the prior year, I thought that it was a hardware product for something called the CFD, right, computation of fluid dynamics. And I recall the target customers more like car companies, aerospace, defense companies, but now NVIDIA is a customers. So can you tell us what Millennium is about? And why Millennium today feels more like a new kind of EDA hardware and much, much more relevant for the semiconductor customers? And how should we think about like product positioning between, let's say, Millennium and other existing Cadence hardware products like famously, the Palladium and the Protium?

Yes, great questions. So we are super excited about the Millennium M2000 supercomputer, which we launched last week. I mean it's got tremendous potential to really bring some significant change to our customers across the engineering workload, science workloads. We have a very strong and growing partnership with NVIDIA. And as you mentioned, Jensen on stage announced the PO for the 10 units, but there's a lot which has been leading up to it. We've integrated our massively scalable solvers with NVIDIA Blackwell systems. And the numbers are in terms of performance, energy efficiency, ADX that we talk about in terms of performance, 20x more energy efficient. So as you pointed out, last year, we launched the first application of Millennium, which was CFD. But your question about how is it extensible beyond CFD? So without going into too much detail, but I take a step back and let's look at the architecture, the software elements of traditional GPU-based systems, right? So historically, GPUs have been very, very good. They have been exceptional for a highly parallel, very dense computation, right, especially matrix multiply operations. So it makes them very ideal for AI workloads, training deep neural networks that involves billions of dense multiplications. And CFD simulations are very similar in terms of structure over there. So it lends itself -- these kind of simulations lend itself to themselves to more regular computation. EDA is different. EDA workloads are fundamentally different. They involve what we call a sparse matrix operations require double precision floating point or FP64, memory accesses are very irregular. So all of these characteristics have made it very hard for EDA to be ported over and do well optimized in GPUs. But what has happened over time is GPUs themselves have become more and more general purpose, if you will. So when you have architectures like Hopper, like Blackwell, these GPUs they support very efficient sparse matrix. They have very large on-chip memory, very high bandwidth and also support these irregular flows, which are data dependent, which is very common in EDA. So anyway, the point being that the GPU architectures have evolved to make themselves more in line with what the requirements are for EDA workers, number one. Number two, Cadence, instead of just boarding over our EDA workloads over to the GPUs, we rearchitected the solvers, right, to exploit the GPU parallelism, to minimize the data movement bottlenecks and the like. And then thirdly, I think NVIDIA has just done a phenomenal job with their GPU optimized CUDA library that they help tremendously in terms of just out-of-the-box kind of performance improvements are like. So you put it all together, and then that's where you see all those kind of very strong numbers. And on EDA workloads, it's very impressive -- particularly impressive, not just the speed up, but you can do things which were just not able to -- immediately accept that they can now run voltage drop simulation that was just not possible before. So that's what's changed. And that's why it's so exciting. And then now you can apply Millennium across EDA, SDA and bio as well. I mean we gave some numbers on Bio -- treat on biosciences as well as a reference. Now to your question about Palladium and Protium and Millennium, how are they all different? So they are used for very different purposes. They're very fundamentally different. So Palladium and Protium are for what you would call accelerating kind of bullion operations, bullion computations, which are very prevalent in functional simulation. So the architecture is much more structured, much more simpler and Millennium is more for numerical simulation. So these 2 are complementary to each other. Specific different purposes and both are very much needed for different elements of the workflow. And lastly, on your question on Tencent -- we also announced Tensilica NeuroEdge, I mean it's really a coprocessor designed to complement any neuro processing unit. And handle stats, which are best offloaded by the main NPU. So again, it does very well. We talked about seeing 30% smaller area, 20% dynamic power savings. So exciting as well. So yes, I mean this was a good batch of excitement -- of exciting announcements and -- Millennium especially, we feel really, really -- given the breadth of the applications that it can help accelerate. We and our customers are very excited about that.

Thanks, Nimish. If I may, I do want to ask a follow-up on Millennium as it relates to EDA. So it sounds like you're basically talking about GPU is moving closer to EDA workloads, and you guys are pushing EDA workloads closer to GPU. So it looks like increasingly, it sounds like EDA can be run on GPU platform, but Millennium is a hardware product. So I just want to ask you this follow-up question. Do you more imagine Millennium to be a hardware business or maybe it's more like a cloud business? Or what do you think about the future here?

Yes, yes. So Charles, the way to think about this is Millennium is actually not just a hardware product, it's a hardware plus software co-optimized offering that we are providing. So this is where we talked about our solvers, EDS solvers that we have rearchitected to make them much more "GPU friendly" to take advantage of some of the GPU enhancements in the Hopper and Blackwell platforms. And so -- but now when you provide that in that context, we are still offering our solvers stand-alone for traditional deals that you have the 2-year, 3-year term deals that you have with the software and customers can use it that way. In this context, we also apply -- or have our -- the solvers available in the public cloud. So in the public cloud or Amazon or other hyperscalers, they have CPUs available, they have GPUs available and the like. So that can also be -- and then we also have our own Cadence on cloud. This is our own data center, and we offer this and offering through the cloud as well. So you can either buy the Millennium box as an prem appliance or you can also access it through the cloud. So a lot of flexibility in terms of use models. And of course, the business models will evolve with that as well. But the key thing is the Millennium is not just -- think about the Millennium itself, it's the sum of the parts, and that's where you get exponentially good value when you're co-optimizing all these things, the solvers, the GPU, the libraries, all together, creating that box and you're providing and that's where you're getting that special extra in terms of performance and energy efficiency.

Thank you. This is such a new thing. And many, many optionality possibilities and definitely excited about that. We'll see as we go and see how you guys really lead in this category. So maybe the next question. We talked about hardware, right? I want to switch gears a little bit to the more traditional hardware. Palladium, Protium, these are, I mean to say, upside drivers for Cadence for a while, right? So you guys launched Z3, X3 last year, want to ask you this, which inning are we in over that Z3, X3 product cycle? I mean, the reason why I ask this is very simple, right? When we look at -- when we think about the product life cycle, how much revenue opportunity, it can be for Cadence, we look at Z2, X2. Z2, X2 Launched in, I think, April 2021, right, revenue, you guys are not given the disclosure, but we can look at upfront revenue, right? It's ramping up throughout 2022 very strongly into at least the early part of '23, right? But it looks like now we are -- now then you guys transitioned to Z3, X3. I'm sure you guys have done some market sizing. How much bigger do you think the Z3, X3, the product life cycle can be versus Z2, X2 plus my earlier question, which inning are we?

Yes. So I think the way to think about this, I mean, again, the emulation, the prototyping, which is really for verification, verification is a very, very significant, probably the most complex kind of problem which the -- which our customers deal with because there's a lot of push on first-time pass silicon and the cost of respends, the market opportunity and it's just something which is driving more and more of this verification to be done as early upstream as possible. And while there are multiple ways to do verification, I mean, simulation and virtual prototyping and formal analysis, it's really the hardware accelerators, which really are the ones which can run at tremendous speed. And for the complexity of the design. So verification is an NP-complete problem, right? You're never done with verification. You're never done with verification. You just get -- got to get enough confidence before you tape out that you've really flushed out all the significant issues and the like. And so with that, and then you look at the workloads, which are out there and you look at the systems and AI super cycle, a lot of these chips, which are -- the AI chips are all most advanced nodes, reticle limited and then all the software workloads on top. So you have verify, not just at the chip, but at the system level with the software. So all these drivers are what's driving the need for more and more verification and emulation and prototyping are very key cornerstones of the verification kind of portfolio. So as you said, I mean, the last cycle -- or the last Z2 and X2 2021 is when we launched them and then we came out with Z3 and X3 in 3 years after that. And I mean, the demand has just been incredible. I mean we can't build them fast enough. And this -- the -- when you talk about the innings of this, the inning statement is actually kind of relevant if the game is constant. The game itself is changing. And then you are -- the innings is changing as well, if you will. But I would still feel like we are still in the early stages of the cycle. And we are seeing customers, again, very broadly diversified, but it doesn't matter whether you're a big digital customer, AI customer or you're doing mixed signal designers, like verification something you really, really, really got to continue kind of really flexing and making sure that you're pushing through the system level piece of it. So I think stay tuned. I mean, I think we crossed over from Z3 -- from Z2 to X2 sometime late last year and we're continuing to kind of ramp up the first half of this year. So we honestly believe based on customers' feedback that the best system out there in the market from emulation is Z3 and the next best one is Z2. I mean and then just talking about the previous question that you asked, same thing, we offered the emulation as on-prem devices, but we also offer them in the cloud as well for our customers to avail themselves. So yes. I think this is -- we feel very good about how we're situated and our customers' feedback just underscores the importance of these systems.

Yes. So just to really to check some of the statement management previously made, right? I think '24 was the record hardware revenue year and '25 million will be another record year. Is that still the case?

We've had multiple record years in a row. And yes, that is correct. We expect 25% to be another record year.

Got it. So now let's ask -- let's discuss something more around the core EDA software. One of the key debates among investors, I would say since March, is Lip-Bu becoming Intel CEO means for Cadence. I'm sure you have stories to tell as you worked under Lip-Bu's leadership at Cadence for many years, right? So from the investment community, I'm hearing 2 sides of the arguments, I would say. On one hand, some folks, I think Intel is going to do a lot of cost cutting. Well, they are doing that not going to and headcount reduction, right? This sounds like possibly a negative for overall EDA spending by Intel. But on the other hand, Intel still runs legacy in-house EDA has historically favored one EDA vendor, one competitor. And Anirudh has multiple times has said Intel is a low water mark in terms of your market -- EDA market share, right? I mean, amongst all the global customers, which could be upside for Cadence going forward? That was the bull case, right? So between the bull and the bear cases, which one do you think will prevail in your opinion and why?

Yes. I mean, there's obviously -- we get asked this question a lot these days. And let's just say, we are obviously very pleased with the announcement of Lip-Bu becoming the Intel CEO. I mean, in a bigger picture view, we all know about his incredible track record in the industry. He drove this remarkable turnaround Cadence all these years ago. And we think there's a great deal of excitement about what he'll bring to Intel. Now last week at Cadence Live, we wanted to have him for a fireside chat with Anirudh. And during the chat, I mean, Lip-Bu stressed his priority to focus on innovation to reformulate that strategy on the AI strategy particularly and the importance of first-time pass silicon, increasing the productivity and the effectiveness of the monthly engineering teams, right? So he also stated and expressed his desire to move away from custom EDA, from custom IP and embrace standard workloads, embrace standard IP. So I think this approach, this mindset really, I mean, should enable Intel to not just rationalize the internal investments they are making in an informed manner, but more importantly, it increasingly allows them to move to broad industry tested best-in-class technology to help with their development, with their designs. Now from a Cadence perspective, as you said, Anirudh has mentioned this before, we have made progress, of course, at Intel on the foundry side as well as on the product side. A lot more to do over there. And as you know, Charles, I mean, our solutions have been used by the marquee companies at the most advanced nodes and other foundries. So we think there's much more opportunity for us to help Intel, not just on the chip design side, but even beyond that on packaging. I mean, Intel has got a great packaging solution. We are the best in 3D-IC and [indiscernible] solutions out there on system analysis. And so we look forward to the opportunity to help and work with Intel. And as Lip-Bu said, there's a lot more opportunity for us to do stuff together and to learn from each other. So that's basically stay tuned.

Yes. Stay tuned. We're definitely looking forward to more updates. So let's pivot to IP. This has been an area, Cadence seems a little bit behind your closest competitor, synopsis. In recent years, we've been hearing from Anirudh on earnings calls about this StarIP strategy. I think maybe it's a little bit not fully elaborated to investment community. Maybe tell us what the StarIP means and why Cadence think this is the right strategy.

Yes, yes. So I'd be, again, a huge opportunity with us. And you look at what the trends are, again, kind of like the previous question, more and more customers want to outsource IP, outsource their standard-space interface IP, for example, There's really no good reason for them to continue doing it in-house. And so more of our customers want to focus on their own unique innovation and then these kind of things, okay, more outsourcing is happening. So that's number one. The foundry ecosystem build-out is happening. So that is well. I mean you've got now 4 big foundries, right? We're all focused on most advanced nodes they all need IP, which is optimized for their technology. That's happening. From a Cadence perspective, we have kind of built out our portfolio over time. And about a few years ago, our focus changed, if you will, or was refined to move not just from growth by any means to scalable and profitable growth. And so what that meant was that we made the decision to say, okay, this is what we're going to really focus on. We think the biggest opportunities were with, let's say, AI, HPC customers, they were designing chips at the most advanced nodes, systems companies like hyperscalers building their own silicon who needed outsourced IP. I mean it's one thing for the other companies and the semiconductor guys who have been building IP over the years and want to outsource the hypescalers, there's nothing to outsource. They didn't have it in the first place. So all this is very synergistic. This whole thing on AI, HPC, most advanced nodes with our digital business, where we also focused on providing tools for the most advanced node designs for customers, the most bleeding edge. So that's what we mean when we talk about StarIP from high-value, differentiated IP, targeted for specific verticals and delivered at the most advanced nodes. So think about this as titles like PCIe, all the different versions of the DDR, high-bandwidth memory and also our Tensilica computing IP. So what Cadence has done then over the past few years has methodically built out this portfolio. Its customers have been giving us feedback and our IP is also getting better. It's getting -- giving better PPA, better quality. And so based on customer needs and guidance through organic investments, and we have been investing more and more over the last 3, 4 years and also inorganic, meaning acquisitions. We augmented our portfolio with high bandwidth memory IP from Rambus. We added internally chiplet connectivity IP like UCIe and security is becoming a key care about. So we signed this agreement with Secure-IC earlier this year. And then I mentioned the foundry ecosystem, Samsung and TSMC, of course, but then Intel foundry, Rapidus, we had this big deal we talked about in Q4 of last year. So -- and then most recently, you saw the announcement on the Artisan foundational IP, right, as well, standard sales, IO, memory compilers, so these are all going to be optimized and validated at different process nodes at different foundries. And so we are in good shape. IP grew 40% in Q1. And yes, we feel like we're in a really good place. We continue investing, continuing delivering and growing the portfolio. And yes, we're excited about where we are in IP.

Thanks, Nimish. Maybe let's move forward to the next topic. I want to ask you about China. We know that last year, probably not a good year for Cadence overall China business. You guys began the year thinking maybe China could be flat I mean 2024. And I think ended 2024 at minus 16% growth for the overall China revenue. But still, kudos to the team for really maintaining and actually slightly raising the growth outlook through the year despite the China headwind. So what's the outlook for this year? Minding, if you reiterate some of the things that you guys recently talked with investors about. And how should we think about what could lead to the upside or the downside for your overall China business compared with the baseline outlook you just laid out in the last earnings call.

Yes, Charles. I mean I think the key thing over here again is that we are driven by design activity, right? And design activity continues to be strong. I mean, that's what we said in the earnings call, we continue seeing that across not just China, I mean, broadly as well, but also in China. It's -- we continue to see strong design activity. I mean, again, keep in mind that customers are designing products of today that are going to hit the market tomorrow, 2 years, 3 years down the road. Dynamics are very similar. There's a lot more silicon being built, domain-specific workloads, a lot of accelerators and many systems companies wanting to build their own silicon, hyperscaters and others. I mean Alibaba, ByteDance, Tencent, right? And so all those dynamics are very similar, as we see in other parts of the world as well. And that's driving much more design activity. And then the other thing to point out is that we sometimes kind of given all the hype and the excitement AI, kind of just focus on data center and AI and the like. So of course, that's there. That's a key part of it. But it's not just that, right? It's also the physical AI, right? The emergence of physical AI. And in there, you're seeing autonomous vehicles, robots and drones, there's a lot of design activity, which is happening over there as well, right? And autonomous especially on EV, tremendous activity happening in China on that. So from a Cadence perspective, we are pleased with the 19% year-over-year Q1 growth in China. And as we have said, I mean, we are -- we think it's prudent to be prudent, and we'll see how it goes. But we are assuming at this point the China 2025 revenue stays flat year-over-year. We'll see, obviously, as we get deeper in the year, we'll provide more updates. But at this point, happy in Q1 when maintaining our resumption of feed flat year-over-year for the year.

Thank you, Nimish. I want to spend the last question a little bit of a more kind of maybe you can call it a blue sky kind of discussion, a bit longer, for sure. So Anirudh in the past has laid out how AI benefits Cadence, right, roughly 3 main aspects; more companies designing more AI chips, that's number one; more customers adopting Cadence AI products that's number two; and Cadence expanding into adjacent areas such as drug discovery. Well, the area that can be disrupted by AI, right, that's number 3. I think investors are largely thinking that the first one, more companies, especially system company, hyperscalers designing more AI chips have not fully played out in a way that is meaningfully reaccelerate Cadence growth. So imagine this, right? I mean this is the thinking behind that. If Mag 7 companies -- magnificant 7 companies become more like semiconductor companies, each chips and from the beginning to the end, and each one of them and, let's say, put a number there contributes 5% end of Cadence revenue, I know it's probably not there, probably not for most of them. But 7x that's 35%. That could be 35% of Cadence revenue. I'm sure you guys really not quite there, but when will that happen if that happens at all? And what is Cadence doing to make it happen? I think this is a top of long-term questions I think everybody have in their mind.

Yes, that's a great question. A big pick -- like is said, blue sky, big picture question, right? I mean, first of all, just reemphasizing, we got a very diversified customer base, no 10% customers. And we've said this in the past as well about that the top 40 customers bring in roughly 55% to 60% of our total revenue. Now when you talk specifically about AI, I mean we have said this, right? It's a generational mega trend. I mean we are still, I think, in the very early stages of this secular cycle. It's got to be a multiyear, some say multi-decade super cycle. And of course, as I said, HPC data centers are what folks typically think about when speaking about AI. And the demand for hyperscalers is ginormous, right? I mean massive training requirements, the scale at which they are operating. Everyone is designing their own custom chips, AI chips, very complex chips, reticle-limited, most advanced nodes and they're requiring very sophisticated design tools, very sophisticated verification tools. And that's what Cadence has been providing. And some of those things that we talked about, the AI-driven tools that we are providing be it on or PPA optimization, be it for verification like Verisium or system analysis and the like. And so all -- so that's in the context of the chip. Now we have talked about the needs are being beyond just silicon. Chiplet architectures are becoming more prevalent, means you need advanced packaging, and it's not just packaging by itself, you're running into more and more thermal issues, electromagnetic issues, warpage issues, right? It requires a lot of in-design analysis. So it's the -- the workflow itself, the way that the engineers, customers that designing these things are different as well. It's not just designing the chip and then, okay, ship it over to the package or the systems guy. It's in design that you've got to kind of go and do all this stuff. So it's the type of designs are changing, the way you do the design development is changing. And so even when we talk about scalers and the way they are doing the design, I mean, obviously, they're one of our fastest-growing verticals and -- but their flows are evolving. We see hyperscalers going from doing ASIC flows, right, where you do the front end of the microarchitecture, the logic verification, then you ship it over to the ASIC vendor for the implementation. But some of the hyperscalers and more and more of them are looking at doing it in-house, bringing it all in-house, doing a COT, a customer-owned tooling flow. So even that workflow is changing. And then as I mentioned, that the need is not just in data center and hyperscalers, there is also physical AI, autos, robots, drones, now there are several applications across all these multiple verticals there. So anyway, the point being, there's a lot of demand, much more AI chips being built. It's a virtuous cycle. The more the demand for AI silicon, the more the demand for our tools and then it pulls in our tools and our AI-driven tools to deliver to those complex designs, which allows them to do even more complex designs. And so that's basically is virtuous cycle that we see. And then beyond electronics, as you mentioned, we also view life sciences as an area which is pretty ripe for disruption. I mean we've talked about this in the past, Charles, that 99% of design in chips is done, right, virtually, digitally, about 20%, 25% in systems. But in life sciences, only 1% is being done with in silico methods. And so I think this is, we think, a huge opportunity, 5 years, 7 years down the road for us to kind of look at that. So anyway, that's the way we look at this. So we think we've got several exciting growth vectors for different phases of the AI cycle as it plays out. And I think we'll keep reporting on how the progress we're making over there.

Thank you, Nimish. And I think folks we have probably 4 more minutes, and we can take some questions from the audience. [Operator Instructions] So far, there's no question in the queue, but maybe Nimish, let me start a question. As a quick follow-up you mentioned about you guys acquired the Artisan business from Arm. And it's the -- I think they call it the physical IP. Can you tell us where does it fit in the Cadence portfolio? And quite frankly, Arm is also an IP company, right? They are #I in IP. You guys are #3 in IP. And why Artisan -- why Cadence is a better owner of Artisan than Arm, why do you guys make that transaction? What's the rationale? Why it's good for both of you?

Yes. So from -- Charles, from our perspective, on Cadence, like I said, we have been building out our portfolio on IP, and we're going to layer on these different capabilities to make it a more full-featured portfolio. So we have the compute IP with Tensilica. Then this interface-based IP, the StarIP that we talked about earlier. Then we added on high-bandwidth memory, which was being driven by AI. And then we got the security IP, Secure-IC, although the acquisition hasn't closed yet. So when we looked at it from that perspective, Foundation IP is something we have had our eye on the past, but never felt that there was a compelling need to add it to the portfolio, because we are focused more on layering on these other things, which are much more mainstream, number one. And number two, we felt the market was adequately serviced by what was available out there at the time for the opportunity. Now the opportunity has changed. Now you've got the foundry ecosystem building. So when you have these new foundries coming in, they are looking for their own the standard cells, the memory compilers, the IO buffers and all these, what we call foundational elements to be optimized for the foundry. They need help with that. So that when the opportunity came along from Arm's perspective, Artisan a very well-known brand in the market, very prevalent out there. And from an Arm's strategic perspective, they were willing to divest that. And so we had these discussions. So the opportunity came along, the need was there. And we had this -- the great opportunity to pick up some really well-established IP. And so we signed a definitive agreement on that. It's not closed yet, though, yes.