Cadence Design Systems, Inc. (CDNS) Earnings Call Transcript & Summary
October 27, 2025
Earnings Call Speaker Segments
Richard Gu
executiveThanks, operator. I want to extend a welcome to all of you for our conference call about earnings for the third quarter of 2025. With me on the line are Anirudh Devgan, our President and Chief Executive Officer, and John Wall, our Senior Vice President and Chief Financial Officer. You'll be able to find a webcast of this call along with a copy of the remarks we prepared for today on our site, cadence.com. During today's discussion, we'll be making forward-looking statements, which include our outlook regarding future business and operating results. Because of risks and uncertainties, actual results could end up differing materially from what is projected or implied in what we discuss today. If you'd like information about factors that might cause actual results to differ, please see our SEC filings, which include our most recent Forms 10-K and 10-Q, the CFO commentary and today's earnings release. Every forward-looking statement made during this call rests on estimates and information available to us as of today, and we disclaim any obligation to update them; furthermore, unless otherwise specified, all financial measures we discuss on this call are non-GAAP. The non-GAAP measures shouldn't be looked at in isolation from or as a substitute for GAAP results. Reconciliations of GAAP to non-GAAP measures can be found in today's earnings release. <<operator_script>> Now I'll hand the call over to Anirudh.
Anirudh Devgan
executiveThanks, Richard. Good afternoon, everybody, and I appreciate you being with us today. For Q3 of 2025, Cadence posted outstanding results, thanks to robust operational and financial performance throughout every product category and region, as we kept up the disciplined implementation of our strategy. Bookings came in above what we had anticipated, with backlog rising past $7 billion, which highlights our ongoing technology leadership and confirms Cadence as the trusted partner that helps customers succeed. Because our business remains so strong, we are lifting our outlook for the entire year to roughly 14% revenue growth and 18% EPS growth. John will share more specifics about our financials in a moment. The AI megatrend is still speeding up, and it's producing a design boom unlike anything we've seen, touching everything from hyperscaler infrastructure, to physical AI — autonomous driving, drones, robotics, which is growing fast — and now sciences AI, which is just starting out. AI is introducing exponential design complexity along with new system architectures, and that puts Cadence in a one-of-a-kind spot to grab this generational chance, because our portfolio is differentiated and broad, covering EDA, IP, 3D-IC, PCB and system analysis. With Cadence.Ai, you can see how we think about design for AI and AI for design: customers get what they need to build global AI infrastructure, and at the same time we're embedding AI in our own offerings so we can deliver major gains in automation and productivity. Thanks to tight collaboration with AI trailblazers, chip manufacturers, and system giants—alongside a broad offering spanning chips through complete systems—Cadence keeps delivering breakthrough improvements in power, performance, and area as well as major efficiency boosts, leaving us well positioned to keep growing throughout the AI period. In Q3, we expanded our alliance with Samsung considerably, spreading our foundational EDA tools and our systems software extensively throughout PCB, sophisticated packaging, and systems analysis; likewise during Q3, we deepened a relationship we've held for years with a top semiconductor firm, distributing our primary EDA, IP, and systems offerings on a wide scale, while jointly developing agentic AI EDA solutions for the coming generation; additionally, we extended our years-long collaboration with TSMC to enable next-generation AI workflows backing TSMC's N2 and A16 process nodes. The Integrity 3D-IC offering from us delivers full support for TSMC's latest 3DFabric die-stacking setups. Plus, our design-in-ready IP — HBM4 and LPDDR6 on N3P among it — makes next-generation AI infrastructure possible. Broadcom highlighted successful deployment of the complete Integrity 3D-IC flow for hyperscaler high-capacity ASICs during TSMC's OIP conference. Momentum stayed solid for our IP business in Q3, driven by IP demand picking up speed across the globe and more customers adopting our growing IP portfolio. With our profitable and scalable IP strategy focused on AI, HPC and automotive verticals, we're well positioned to keep growing. Interconnect protocols are growing more complex because of AI and chiplet architectures, and new foundry opportunities are emerging too — all of which creates powerful tailwinds behind our IP business. Demand came in above what we had projected, making for a robust bookings period. Several competitive victories landed within our design IP portfolio, coming from leading AI and memory clients — one example being a hard-fought engagement with a prominent memory firm, which chose our HBM4 and DDR5 IP to power its latest AI design. Our design IP portfolio also received a lift recently when we finalized the Arm Artisan Foundation IP transaction, adding standard cell libraries, memory compilers and IOs optimized for cutting-edge nodes at leading foundries. We also captured numerous design wins for our Tensilica audio and vision DSPs, alongside Neo AI accelerator NPUs, with major customers in the U.S. and Asia across the mobile, automotive and data center segments placing those wins. Strong results came out of our core EDA business, fueled by wider uptake of the AI-driven design and verification solutions we offer. On the digital side, Cadence Cerebrus AI Studio — the industry's first agentic AI, multi-block, multiuser design platform — keeps delivering unmatched PPA and productivity gains. A SF2 design was taped out by Samsung U.S. using Cadence Cerebrus AI Studio, hitting a 4x productivity improvement. Separately, Samsung relied on Cadence Certus, Tempus and Innovus to quickly close and sign off a multibillion instance AI design on SF4, achieving 22% power reduction along with first-pass silicon success. We saw a lot of momentum behind Virtuoso Studio and Spectre. Customers are embracing the automated design migration and optimization features, and the AI-driven capabilities and workflows in those platforms caught on quickly. When it comes to AI designs, our hardware verification platforms have become the default pick, with performance, capacity and scalability that lead the industry. Hardware had a record third quarter, driven by a number of major expansions, especially with AI and HPC customers. Our work with OpenAI grew more extensive, since they increased their commitment to our Palladium emulation platform during Q3. Verisium SimAI also saw wider adoption, thanks to the big gains it brought in debug productivity, test bench efficiency and faster coverage closure. During the CadenceLIVE India event, success stories involving SimAI were shared by NVIDIA, Samsung and Qualcomm, which underscored verification throughput gains ranging from five to ten times. Our system design and analysis segment delivered yet another robust quarter, powered by a widening portfolio of inventive offerings and rising uptake among an ever-broader set of customers. In Q3, we made substantial additions to the Cadence Reality Digital Twin Platform library, incorporating the NVIDIA DGX SuperPOD model along with DGXGB200 systems to speed up AI data center deployment and operations. As three leading memory suppliers move toward a complete Cadence flow for advanced IC packaging, their use of clarity and security rose considerably, pushing out rival offerings. BETA CAE continued building on its strong trajectory, securing several wins where customers switched from competing solutions — a reflection of its superior precision and speed. One notable victory came at a major Tier 1 automotive supplier based in China. During the third quarter, Infineon Technologies made the Cadence AI-powered Allegro X platform its standard for PCB design work on upcoming products. Just last month, we reached a definitive agreement to purchase Hexagon's D&E business, which includes its MSC software unit, so that Cadence can incorporate best-in-class structural analysis and multi-body dynamics capabilities. This deal strengthens our multiphysics offerings, speeds up our growth in SDA, and positions us to lead in capturing fresh opportunities spanning automotive, aerospace, industrial, and the fast-growing realm of physical AI. In closing, I'm happy with how Q3 turned out and with the considerable momentum we're seeing throughout our business lines. The marketplace offers tremendous opportunities thanks to AI, and because we optimize all of our offerings together with an emphasis on AI and accelerated computing, Cadence occupies a unique position as the trusted partner that delivers transformative, AI-powered solutions across a wide range of industries. With that, I'll turn it over to John, who will walk through more details on the third-quarter results and our updated 2025 guidance.
John Wall
executiveThanks, Anirudh, and good afternoon to all of you. I'm happy to share that Cadence posted solid results for Q3 of 2025, with momentum spread broadly across every one of our businesses; we came in above guidance on Q3 revenue, operating margin and EPS, and we're lifting the full year outlook for those same key metrics. Based on the updated outlook and using the midpoint, we now anticipate 2025 revenue growth of roughly 14% compared with a year earlier, keeping us on pace for double-digit growth in every product category this year. Bookings during the third quarter were robust, leaving us with a backlog of $7 billion. Let me walk through a few financial highlights from Q3, beginning with the P&L. Total revenue came to $1.339 billion. GAAP operating margin was 31.8% and non-GAAP operating margin was 47.6%; GAAP EPS was $1.05, while non-GAAP EPS was $1.93. Now moving to the balance sheet and cash flow. Our cash balance at the close of the quarter was $2.753 billion, and the principal value of debt outstanding stood at $2.5 billion. Operating cash flow reached $311 million. DSOs came in at 55 days, and we spent $200 million buying back Cadence shares. Before I walk through our updated outlook, I want to point out that it rests on the customary assumption that export control regulations in place today stay substantially similar through the rest of the year. With that in mind, for Q4, we now expect revenue between $1.405 billion and $1.435 billion, GAAP operating margin between 32.5% and 33.5%, non-GAAP operating margin between 44.5% and 45.5%, GAAP EPS between $1.17 and $1.23 and non-GAAP EPS between $1.88 and $1.94. Consequently, our updated outlook for 2025 is revenue between $5.262 billion and $5.292 billion, GAAP operating margin between 27.9% and 28.9%, non-GAAP operating margin between 43.9% and 44.9%, GAAP EPS between $3.80 and $3.86, non-GAAP EPS between $7.02 and $7.08, operating cash flow between $1.65 billion and $1.75 billion, and we anticipate using no less than 50% of our annual free cash flow to repurchase Cadence shares. As we typically do, we've posted a CFO commentary document on our Investor Relations website, which covers our outlook for additional items along with further analysis and GAAP to non-GAAP reconciliations. To wrap up, I'm pleased with our Q3 results, a strong 2025 as we keep deepening strategic partnerships throughout the ecosystem. As always, I want to finish by thanking our customers, partners and our employees for their continued support. And with that, operator, we will now take questions.
Vivek Arya
analystAs I see it, your IP business is currently on pace for growth above 20% in year two. Anirudh, I was hoping you could help us understand what is fueling that expansion, since your rival voiced significant worries regarding their IP operation — be it in China, at Intel, or simply concerning IP visibility overall. They also seemed to reference a fresh business model, I believe. So how do we reconcile that with the growth you're experiencing? How long can this growth last? And what kind of visibility do you have into your IP business?
Anirudh Devgan
executiveAbsolutely, and thank you, Vivek, for raising that. I'm genuinely happy with how our IP segment has been doing. We never judge things on just one quarter. That said, if you examine our results from the prior year — sure, this quarter stood out — but taking the full year together, along with the backlog and the activity we're seeing as we head into the next year, the IP business on the whole is doing quite well. There are several factors behind that. To begin with, our IP business is not the same as others. My view is that it's considerably more profitable — granted, it doesn't reach the profitability of our EDA business — yet I'd say it beats the typical IP business, since we also own Tensilica, which delivers profitability that's nearly comparable to software; meanwhile, a great deal of the expansion is coming from design IP. Why is that? Because our IP business concentrates on AI and HPC at the most advanced nodes. We entered the IP business later, so we aimed it at where things are headed — AI, HPC, and chiplet-based architecture. That means a lot of SerDes, PCIe, and HBM4 IPs. And that portion of the market is performing strongly worldwide, in fact. The second factor is that, as you're aware, more foundries keep joining the fray, particularly at advanced nodes. We've had a partnership with TSMC for a long time, and also work with Samsung, Intel, and now Rapidus; so at leading nodes there are now no fewer than 4 major foundries. That, in my view, is the second factor positioning our IP business well. And since our IP business has been performing better — the PPA — our PPA happens to be relatively stronger in design IP, and many customers are eager to move to Cadence; so customer demand, I'd argue, is the third factor, as our strengthened IP business means we're witnessing momentum there. So for those 3 primary reasons, I feel fairly optimistic about the IP business; and looking toward the next year — I won't make predictions about the next year — but simply as an indication, it would catch me off guard if our IP business failed to outgrow the Cadence average, which it ought to, considering the profitability profile. We'd like that to be the case. When profitability runs a bit below EDA, then growth ought to exceed the Cadence average. So all in all, I believe that would amount to something like a 3-year pattern. And on the whole, our IP performance makes me happy.
Jason Celino
analystOkay, so during the prior quarter, I believe you brought up that the back half of the year would bring a solid chance to renew deals with a few of your biggest accounts. Given how backlog has risen, my guess is a portion of that momentum came from certain of those renewals. Still, when we look ahead to Q4, are there more renewals still lined up?
Anirudh Devgan
executiveYeah. Appreciate the question. On the timing around those renewals, I'll hand it to John to address. That said, taking a step back, I genuinely believe our Q3 results came in much — came in above what we had anticipated; and the main driver — and this holds across every region. Still, I'd say the biggest factor is the build-out of AI infrastructure, which, as you're aware, keeps speeding up, right? And when it comes to designing and building that AI infrastructure, we're indispensable. Naturally, we — I've stated openly that in my view AI unfolds in 3 major phases, with AI infrastructure as phase one, physical AI as phase two, and sciences AI as phase three; though the bulk of our attention and spending is, naturally, directed at phase one. And as you've witnessed over the past 6 months, things are speeding up. Beyond that — it's a privilege that we get to partner with all the Mag 7s, and spending on in-house chip design is also picking up pace, alongside, naturally, the major merchant silicon players such as NVIDIA and Broadcom and AMD. So I'd say that's reflected in our booking activity for Q3. And to this point, we're seeing that robust demand carry forward into the future.
John Wall
executiveAbsolutely, Jason — one more point from me: the blend looks solid too, spanning EDA, IP, hardware, and SDA. On top of that, our core EDA and IP backlog leans toward multiyear recurring deals, which underpins lasting double-digit growth.
Joe Vruwink
analystThank you. What really stands out to me is how many times people have said "acceleration" during this call already. Third quarter bookings came in far above what we'd anticipated, and that points toward faster growth ahead. I realize it isn't typical to hear thoughts about 2026, though Anirudh has already offered some for the IP segment. So I'm curious whether you could begin setting up how we should think about the coming year, given what's currently in front of you — it definitely looks like the pieces are falling into place. At this stage, do you have enough visibility to say something about that?
Anirudh Devgan
executiveYes. What I'd like to express is this: our habit is to evaluate the company by how strongly our offerings are performing, right? As you're aware, we break things out into roughly 5 reporting lines. And from where I sit right now, every one of those 5 lines is doing really well. You can observe it this year — my expectation is double-digit growth across all 5 of those lines; on top of that, we're doing well across every geography. So when it comes to offerings and geographies, which is where we concentrate, are we in step with the top firms? Are we a trusted partner to the companies shaping the market? So if you consider offerings, geographies and customer alignment together, we're in a strong spot. Naturally, as you're aware, whenever a new year approaches, our outlook always stays cautious, and we'll provide an update on the coming year around January, February time frame. That said, I believe Cadence is positioned very well — better than it has been, in my view, for — relative to the past several years, and we're eager to keep working with our customers going forward.
John Wall
executiveCorrect, Joe — no guidance for FY '26 is coming from us today. That said, as we close out FY '25, we're carrying what's likely a record backlog, plus momentum spread widely, driven by strategic and trusted partnerships that keep getting deeper throughout the ecosystem, and that leaves us well positioned going into next year. What you should expect is that our framework stays disciplined. Our usual ambition is double-digit growth on the top line, ongoing operating leverage, and capital allocation that's balanced. Underneath all of it sits secular AI demand spanning chip to systems.
Lee Simpson
analystFantastic, and well done on yet another strong quarter. What I really wanted to ask about was China. It seems like you're up roughly 53% compared to the same period a year ago, performing nicely within the mix, reaching 18%; that appears to be more than simply business coming back after the BIS letter restrictions hit last quarter; it does seem like there's real momentum building there; so I'm curious whether you could walk me through what's behind this. Is it IP? Is it hardware? Is it core EDA? What are the vectors here?
John Wall
executiveAppreciate that question, Lee. Yeah, we witnessed strength across the board, and design activity in China continues to be very robust. For us, the region got back to normal operations during the second half, following the removal of export regulations that shifted for EDA in early July; however, Q3 actually came in just a little ahead of what we had anticipated, and our current expectation is that China will grow year-over-year for fiscal '25. Anirudh, would you like to chime in regarding what's going on in China?
Anirudh Devgan
executiveYeah, Lee, good question regarding China. So on the whole, based on what I can observe, things there have returned to how they normally are. Naturally, Q2 saw some disruption, and the reasons for that are obvious, because of the policy during Q2. But what we're observing now in Q3 is that activity has gone back to normal. A significant portion of that came from us, for instance, making hardware deliveries a priority — ones we couldn't complete during Q2 — and moving them into Q3. Still, on the whole, design activity across China is robust — I mean, every nation needs semiconductors as essentials, and China keeps putting money into semis; but on the whole, I'd describe our strength as broad-based, not especially linked to any single geography. Also, there was a bit of catch-up moving from Q2 into Q3; predicting what comes next is hard right now, but based on what I'm seeing, nothing out of the ordinary is happening in China. A question might arise, for instance, about whether activity is being pulled in from later quarters; from what we can observe, that isn't happening, and we're also seeing broad-based strength overall in other geographies.
Sitikantha Panigrahi
analystThanks. Nice work on delivering another solid quarter. Anirudh, I'd like to ask about your system design side, specifically the simulation analysis market. Could you walk us through what your approach is? There was an acquisition you completed last year, BETA CAE, and now this year you've come out with the MSC software announcement as well. Give us some insight into how you plan to set yourself apart from the rival in that space; it's clearly a market that's expanding. Any detail along those lines would be welcome.
Anirudh Devgan
executiveAbsolutely, Siti — I appreciate the question. Look, I'm quite happy with how SD&A has performed overall. And just as a reminder for everyone, Cadence was the company that kicked all of this off back in 2017 and 2018; these days it's simply taken for granted that silicon and systems will converge. We've been discussing that idea for ages. What I'd say now is that the acquisition we closed this quarter leans more toward what's coming next, because, as I said, there are 3 horizon technologies: horizon 1 is infrastructure AI, horizon 2 is physical AI, and horizon 3 is sciences AI. That's where our focus sits. The bulk of our spending goes to horizon 1 — though of course, roughly 70%, 80% is horizon 1, around 20% is horizon 2, and a few percent is horizon 3. But horizon 2 — cars, drones, and robots — could become a really enormous market down the road. And here's the thing: AI will transform horizon 2 as well. You're seeing plenty of reports indicating the world will shift away from LLM-based AI toward a world model-based AI, where for robots, it's no longer text data that does the training — it's the physical movement and all of that. And a major obstacle when training robots or cars is that sufficient data simply isn't there. With an LLM model, the data is essentially sitting on the Internet, and language data is there too, but that's not the case when you're training a robot, right? So the data has to either be produced by hand — say, sensors are attached to a person and that individual lifts the object, and that counts as data. But gathering data that way is extremely slow. The optimal approach for producing data for a world model is simulation. That's something we've also discussed for ages — the 3-layer cake. So then fundamental simulation of multi-body dynamics turns out to be critical for horizon 2 physical AI. And Hexagon possessed a top-tier simulator for multibody dynamics, together with structured simulation, which assists with every kind of electronics and automotive work. So I'd say I'm fairly confident this can put us in a good position for the second horizon, physical AI; and regarding what that does for our SD&A business, the way I see it, once this acquisition is done, our SDA business will have 2 solid pillars; and actually, the run rate ought to go past $1 billion in 2026 assuming the acquisition closes; and 3D-IC and chiplets will drive one of those pillars. Allegro belongs to our SD&A business. Allegro is the de facto standard worldwide for package design. So if you take Allegro and pair it with Sigrity, Clarity, and Celsius — our electromagnetics and electrothermal tools, more or less — that forms one essential piece of this silicon-and-system merger. We're going to be extremely, extremely solid there. Our work with TSMC, plus what we're doing alongside every major AI company such as NVIDIA, puts us in a great spot with Allegro and 3D-IC. That should account for about half of our SD&A business, since this chiplet-based architecture is set to expand a great deal. Then the other portion involves physical AI, structural analysis, and bringing together BETA — the pre-post processing leader — with Hexagon, which holds plenty of solvers, such as multibody dynamics structural. We also picked up an excellent new CFD solver from Stanford 2 years back. So once every solver is combined with BETA, that's close to 50% of our SD&A business, and we're positioned really strongly for physical AI. Taking the whole picture, what Hexagon brings is 2 robust pillars within SD&A, in the segments poised for the greatest future growth: first, 3D-IC and HPC; second, physical AI and connected technologies.
Jim Schneider
analystI'd like to ask whether you could give us a sense of the tailwinds that you anticipate emerging across the coming 2 years because AI capabilities are being folded into your offerings on the core EDA side. Perhaps you could share whatever productivity measurements are available regarding time to market or developer productivity, and explain how those might convert into either revenue or adoption rates for that technology and those features.
Anirudh Devgan
executiveSure. That's an excellent question. Our AI approach, as we've mentioned previously, consists of 2 components, and we refer to them as design for AI and then AI for design, all right? The build-out of the AI ecosystem, be it infrastructure or physical AI, is what I'd consider the first component. And with every leading player, all the Mag 7 companies, we are extremely well positioned; and your question now, I believe, concerns the second component, which is, naturally, applying AI to design; so we highlighted a number of examples even this time. So we have no fewer than 5 major platforms. And among the big examples is, for instance, SimAI, which leverages AI to speed up verification. In chip design, verification is practically an exponential task. And with SimAI, we are observing 5 to 10x gains in logic simulation efficiency and coverage, which ranks among the most heavily utilized tools in verification. And Samsung and Qualcomm and NVIDIA even highlighted this at CadenceLIVE. So these are demonstrated benefits at customer sites being highlighted by the customers themselves, all right? Physical design, the back-end physical design with Cerebrus AI Studio, is the other area. Once again, Samsung recorded 4x improvement in productivity and also 22% improvement in PPA. Incidentally, these are enormous figures because when you move from, say, 5- to 3-nanometer, 3-nanometer to 2-nanometer, a node migration, for which the industry is spending billions and billions of dollars, typically delivers something like 10% to 20% PPA improvement. And should we achieve that through better optimization, through better AI, that represents tremendous value for our customers. So the encouraging part is that adoption of AI tools is practically regarded as a de facto matter. Every one of the big customers is adopting our AI tools. And I've stated previously that monetization of that requires some time. It always requires 2 contract cycles. And I believe we ought to manage that or slightly better. So — but the productivity gains from applying AI to EDA are enormous. And the reason I believe it differs in EDA compared with other things is, first of all, there are multiple reasons. One is that automation has been our focus for 30 years. The chip design process is highly automated. Roughly 80%, 90% of it is already automated; so we possess a great deal of history with automation, and then AI represents the next 10x of that automation that can occur. What I'd say is that over the past 20 years, we've likely made chip design about 100x better, and AI can deliver the following 10x. And there's another distinction, I think, between chip design and other sectors: the workload grows exponentially. Five years from now, chips will be roughly 5 to 10x larger. With software and chiplets, complexity will be 20 to 30x greater; so AI productivity is required simply to keep pace. That means our workload is exponential. It's quite unlike a workload that isn't exponential. So customers expect us to provide more productivity, and they're willing to deploy that in their designs.
Harlan Sur
analystGreat job on the quarter, as always. On the third-gen upgrade cycle for the emulation and prototyping platforms — you're about 5 quarters in, and it drove record revenues in Q3. Looking back at the second-gen launch, the team strung together 3 years of record revenues after it came out. The same drivers are still there, right? Design, software complexity going up exponentially, the Cadence of new chip program introductions, more customers coming on board faster — like OpenAI, which you mentioned on today's call — and all of that spreading into new markets such as automotive and software-defined vehicle. Given the lead times on your Protium and Palladium systems, I assume you're already booking into next year. What's the shape of demand at this point? And for the hardware platform, are you anticipating that the momentum and growth will carry on through 2026?
Anirudh Devgan
executiveYes, Harlan, you've consistently got a sharp read on where the market as a whole is heading. Hardware is indeed doing extraordinarily well, and my expectation is that this will keep going; so would '26 come in above 25%? That's our thinking. As for how much above? We tend to stay conservative there, because unlike software, where we enjoy full-year visibility, hardware doesn't give us that. The result is that going into any year, our visibility extends only 6 months. That's why our hardware guide always reflects prudence. If our performance unfolds according to plan, similar to what happened during this year, we'll have room to raise our outlook for the remaining quarters. That said, this really ties back to how we approach guidance—our aim is to lower the risk embedded in what we project for those who invest in us. Turning to the underlying tech and market dynamics, the conditions are highly favorable for hardware, largely because no other firm designs its own complete systems the way we do. Our proprietary silicon, manufactured by TSMC, consists of full reticle chips. You really ought to take a look at them. Each of these racks contains 144 chips that are liquid cooled, linked through InfiniBand plus optical connections, and a typical customer setup will tie around 16 racks together. Designs of roughly 1 trillion transistors can be emulated on that. Honestly, nothing else out there measures up to it. On top of that, hardware demand keeps climbing, and it isn't only due to growth in AI designs — as the industry moves from 3-nanometer to 2-nanometer, then 1.4 and eventually 1, a transition spanning the coming 7 to 10 years, chips just keep getting bigger, which drives ever greater need for hardware. Taking everything into account—our competitive standing along with where the market's heading—I'd say our hardware position is strong. That said, when it comes to any particular year, we keep the guidance conservative. John, feel free to jump in if you'd like to add something.
John Wall
executiveRight, right. Harlan, the thing I would put on top of that is that demand is still really robust, especially in AI, HPC, and auto. We've witnessed expansion — we've been expanding our production capacity and working to shorten lead times. On top of that, hardware gross margins have gotten healthier; our attention stays on throughput so we can satisfy the heightened need coming from AI designs. And when you look at our numbers for this quarter, you'll notice we've been accumulating inventory in an effort to satisfy the demand shown in the pipeline covering the next 6 months.
Jay Vleeschhouwer
analystAnirudh, you walked through a number of instances involving customer activity, customer engagements, and similar items. Now I'd like to ask you something regarding the announcement made recently about the joint effort that NVIDIA and Intel are going to undertake. Would it be reasonable to assume that work combining GPU and CPU would inevitably raise demand along with capacity requirements across multiple kinds of EDA tools, plus IP, and in all likelihood hardware too? So a general uplift would come about because of that combined effort. Yet at the same time, would it likewise mean you'd have to boost your investments, say in AEs, the way you did back when you achieved that breakthrough with Intel several years ago?
Anirudh Devgan
executiveAbsolutely, Jay — you've raised a solid point when it comes to pairing CPUs and GPUs. Incidentally, for roughly 15 or 20 years now, I've been saying that these two need to operate in tandem, since EDA represents an extremely well-tuned workload — it's computational and mathematical software, quite close in nature to AI. What transpired across EDA's history is that, naturally, plenty of SIMD-style jobs exist that a GPU-type machine can handle, yet equally there are many conditional jobs that must run on a CPU-type machine. That's why we've consistently sought both, and moreover wanted them positioned near one another. Honestly, credit belongs to NVIDIA and to Jensen for Grace Hopper and afterward Grace Blackwell — they were among the earliest to spot and follow this direction. Today, examining every major design coming from other firms as well, you find CPU and GPU paired together. That explains why, across the past several years, we've already been porting our workload onto CPU plus GPU. A prime illustration came when Millennium was unveiled earlier in the year — we're shifting not only system analysis workloads, which suit GPUs more readily, but additionally EDA workloads, essential for speeding up EDA and 3D-IC, onto the CPU, GPU pairing. So what I'd say is this: it genuinely pleases me that the entire industry is now heading toward CPU plus GPU combinations — check Apple's chips, AMD's chips, and naturally NVIDIA's remarkable platform. Our partnership with NVIDIA and Intel benefits us by delivering a fresh sort of x86 plus GPU. On top of that, we've enjoyed a long-standing relationship with NVIDIA. As Intel deepens its collaboration with NVIDIA, that likewise helps our broader conversations with Intel, which I believe are moving along nicely. Intel, in my view, must put money into both its foundry ecosystem and its own offerings. Lip-Bu understands this, I think, and seeing investment on both fronts is encouraging.
Jay Vleeschhouwer
analystTo put it plainly, apart from the internal porting work required for your own tools, your assumption is that when it comes to demand, this customer activity would inevitably drive up EDA consumption.
Anirudh Devgan
executiveWhen it comes to what customers are doing, I'd say the main thing is this: if our EDA tools improve as a result of CPU and GPU systems being tuned, customers will generally jump on board. We're constantly searching for methods to make our tools better, and this provides one more avenue for boosting tool performance. That benefits every customer. Then, with this specific partnership, there's particular design work that has to be completed — I won't go into too much detail — involving NVLink-based IP. So yes, we're collaborating with those specific companies on design to bring this design to life, the same way we'd collaborate on any of the top designs. So yes, there's a specific customer activity tied to NVIDIA and Intel. And broadly speaking, customers gain when our tools are better optimized on this platform.
Gianmarco Conti
analystNice work on yet another strong quarter. Perhaps returning to the topic of China, particularly considering the fantastic quarter you've delivered — naturally, a portion of that came from Q2 catch-up. Still, how should we view a growth rate there that can be sustained, separate from what got recouped in the prior quarter? And if possible, could you shed some light on whether a fresh ban in that region poses any genuine danger? Clearly, there's been some news out there. I believe investors would like to stay somewhat cautious regarding what actually constituted a real threat to EDA versus what might simply be a wider macro-level effect? Any thoughts on that would be appreciated.
Anirudh Devgan
executiveRight. Regarding China, as I stated, design activity appears to have returned to its usual state in my view. And as we noted — naturally, at the start of the year, we exercised a lot of caution, since as I've said previously, during my trip to China a year ago, they anticipated a quite harsh macro backdrop — geopolitical circumstances, and that indeed materialized in '25; therefore we stayed very cautious with our China guidance early in the year, and that proved right; at this stage, as John brought up both previously and now, we anticipate growth in China. The amount of that growth remains uncertain. We'll get a clearer picture. Predicting it is quite hard. By the end of the year, we'll have a clearer picture. Still, I do anticipate growth in China this year. And it's encouraging to observe — predicting the geopolitical situation is extremely hard, and I certainly have no desire to attempt that; yet it's encouraging to observe plenty of dialogue happening between the 2 presidents and among major economies. So whatever stability and predictability exists there benefits our business. Thus we're looking forward to that. Still, I do anticipate strong design activity. And provided nothing unexpected arises and conditions remain stable, that ought to benefit our business. Also, let me point out that our Q3 strength received a boost from China's performance, yet it's quite broad-based, for all the factors we cited regarding the build-out of the AI infrastructure, the emerging design of physical AI, the overall AI megatrend; so we're happy. So we aren't tied to any single country, yet it's encouraging to observe that conditions in China are getting better.
John Wall
executiveYes, and Gian, one thing I want to flag is that when we look at Q4 and the full year, our guidance is built on the assumption that the export rules in place right now stay more or less the same. And we consistently build in a cushion for shifts on the regulatory side. And we will keep complying strictly with — while standing behind customers around the world. And as Anirudh mentioned, we're witnessing momentum in every business and in every region.
Joe Quatrochi
analystCould you maybe give us some insight into how the OpEx is behaving? My sense is that 3Q came in somewhat above what we'd anticipated, while 4Q looks somewhat below. Does that tie back to when the Artisan deal is expected to close? Any clarity on that would be useful.
John Wall
executiveAbsolutely. Yes. That said, what's driving it is simply the fact that certain hardware deliveries are landing in different periods, moving from Q3 into Q4. Even so, for the full year we're running a touch better than our prior outlook, and we feel good about execution across the board, with robust demand in every product category. Core EDA software is doing quite well. Hardware remains solid. We keep gaining ground in SDA, and our IP has ongoing momentum, plus renewals are healthy going into Q4.
Joe Quatrochi
analystPerhaps I could ask something regarding operating expenses...
John Wall
executivePardon me, could you say that again?
Joe Quatrochi
analystRegarding operating expenses, was your question about when those costs land?
John Wall
executiveRight. On the OpEx front, a modest restructuring gave us a lift in the third quarter. Our gross margins on hardware came in quite strong during Q3. Then in Q4, that gets partly counterbalanced by added costs tied to acquisitions we've recently brought on.
Yu Shi
analystAnirudh, congratulations on the strong results, and John, same to you. My question: when I examine your overall company growth rate across the past 3 years, it's been holding in that roughly 40%-ish plus or minus band, which is genuinely impressive. It seems like you barely slowed down at all. Yet once I dig deeper, there are plenty of moving pieces, correct? For instance, let's put last year side by side with this year. During last year, China performed poorly. Hardware was decelerating somewhat. I believe that mostly stemmed from your hardware transition into the Z3, X3. I mean I'm using the upfront revenue as a way to gauge your hardware growth. However, this year, both of those things flipped much more net positive, such as your upfront revenue probably expanding somewhere nearer to 50%. China appears to be at minimum expanding faster than the corporate average; so I'm curious, when we look at — consider next year, do you believe hardware and China can both sustain the present momentum? Perhaps particularly for hardware, judging by what we observed in the Z2, X2 cycle, I think that fell somewhere between '21 and '24. Once you reach roughly a third year, the growth rate during the Z2, X2 cycle, it slowed down somewhat. So my question is, might this occasion be somewhat different regarding the hardware growth rate moving forward? And might any worry among your customers about a hardware transition to, say, Z4, X4 within maybe the coming 1 to 2 years trigger some slowing of hardware revenue? I realize this is a lengthy question, but I consider it the most crucial one when we think about the Cadence outperformance heading into next year.
John Wall
executiveThanks for the question, Charles; we're attempting to unpack it. So I think I wouldn't focus too much on any single quarter or even any single half when it comes to results. If you remember last year, the revenue curve's shape was somewhat back-end loaded. And Q3 over Q3 comps can be somewhat skewed, especially also with China, since we had that temporary restriction in China from May to early July. But generally, when you're discussing hardware, demand is extremely, extremely strong, but — and we've been seeing a secular trend in hardware demand for many years now because growth in complexity keeps going unabated. But we're seeing a very strong pipeline for the next 6 months. And we're building up inventory for some large orders that we have to fill over the next couple of quarters. But — so we're seeing lots of momentum. And we expect to — I mean if I go back, I think the last 5, 6 years, and it's typical of Cadence, Q4 bookings would exceed Q4 revenue. Alright, so at the close of Q3, we wrapped up with a backlog totaling $7 billion, and that's the highest level we've ever recorded. Based on how renewals are scheduled to fall in Q4 and what we can currently see, our expectation is that we'll finish '25 at a new peak. And since that composition looks really strong across every one of the businesses, I believe it points to good things ahead for next year.
Yu Shi
analystLet me ask one more thing really fast. Anirudh, the way you see it, our existing Z3 and X3 hardware can handle 1 trillion transistors, yet given how quickly AI is advancing, when do you think we'd have to go through another hardware refresh? Can you give us any insight there?
Anirudh Devgan
executiveCharles, yes, I feel quite strongly about where we stand on hardware. We've discussed Palladium — nobody else both creates their own chips and offers Protium FPGA systems like we do, and that business is performing nicely too with the Dynamic Duo; and as John mentioned, demand looks solid. Let me also point out that our guidance always leans conservative, since hardware can't be forecast the way software can, yet even though we booked what was essentially upfront revenue, what's actually going on is that these major customers purchase nearly annually; it isn't that they're purchasing — their purchasing pattern has shifted from what it was 4 or 5 years back, since they're designing so much that for all the truly large customers, it's nearly turned into a yearly subscription, though on the books it's naturally recorded as upfront; so will this hardware momentum keep going? Honestly, at this moment I can't find any reason it wouldn't. That's why I believe '26 will outperform '25. By how much, we'll know better later; as for what comes after this generation, R&D investment never stops for us; our R&D spending is enormous, as you're aware, with 35% of revenue going into R&D. And — looking at expenses, roughly 65% goes to R&D while about 25% goes to application engineering. That means over 90% of what we spend and of our headcount sits in engineering, customer support and R&D; the same holds for hardware; so we're — I won't go into every detail, but you can take it that we're already well along in designing our next generation of hardware systems, and they'll arrive when the time is right. One upside: our present systems already handle 1 trillion transistor designs, and that's expected to arrive by 2030. Yet ahead of 2030, a next hardware generation will be ready, supporting it through the following 5 years. So overall I feel quite good about our hardware road map. As for demand itself, since Harlan, you understand this space deeply, AI means chips just keep growing larger. And there's another thing happening — even with Blackwell, you no longer have a single chip, you've got several chips plus Grace working together. What's happening is that buyers aren't only mimicking a single chip, which doubles at each node — they're mimicking whole clusters of chips, such as Grace paired with Blackwell, or in cases where you have chiplet designs. Because of this 3D-IC, then, the appetite for hardware could advance more quickly than Moore's Law alone or simple technology scaling would suggest. Still, we'll watch how that plays out; our positioning is strong. Time will tell how it develops, yet from a systemic standpoint, there's no problem when it comes to hardware demand or where we stand competitively.
John Wall
executiveCharles, your question covered quite a bit of ground. I believe you also brought up upfront recurring revenue. What I mean is, we're still describing '25 as roughly 80 to 20, recurring versus upfront, measured on a rolling 4-quarter basis. And as I think you noted in what you asked, the quarter-to-quarter swings come mostly from powerful upfront businesses such as hardware and IP, plus the timing of China ratable revenue earlier in the year. Yet since core EDA is expanding so nicely, we feel good that 80 to 20 is likely the appropriate business mix going forward for the foreseeable future.
Gary Mobley
analystCongratulations, everyone. I just wanted to clarify something, or rather ask a question so I could get that clarification. If my memory serves, because of when the export control repeal happened — July 2, as I understand it — the backlog for China wasn't included in the backlog for the June quarter that just ended, though I assume it is now. With that $600 million in revenue, or $600 million difference in backlog, what portion came from adding China's backlog rather than from the previous quarter?
Anirudh Devgan
executiveSure. I'll give it a try, and then — I believe you have that right: backlog went from $6.4 billion up to $7 billion. That means an increase of $600 million. So roughly — I'd put it at around 25%, so about $150 million, representing catch-up moving from Q2 into Q3, while the remaining increase reflects strength growing throughout our business. John?
John Wall
executiveYes, that's right. No, that's exactly right.
Clarke Jeffries
analystAnirudh, I appreciate what you said about how the IP business is strengthening and, in particular, the design IP demand you're seeing tied to AI projects. I'd like to dig into how the wallet opportunity shifts as those AI projects evolve. More precisely, as the industry builds more customer-specific technologies — incorporating chiplet or custom memory designs into those AI and HPC designs — do you foresee any growing pains or reduced profitability in serving them? Has Cadence altered its investment plan or selling motion to address that more bespoke nature the industry requires? Or is that not even necessary?
Anirudh Devgan
executiveYes. Great question. Look, custom chip design is a major movement. We've discussed this topic for quite a while—firms that build systems creating their own processors. And as you're aware, roughly 45% of what we do comes from system firms, while semi companies account for 55%. OK, so because of this — particularly when it comes to AI — custom silicon is speeding up. And one thing that's different now compared with half a year ago, or a year ago, is that when I look at these large system companies, their commitment to custom silicon keeps growing. Naturally, our partnership with NVIDIA is strong, and NVIDIA will do incredibly well — but custom silicon will too. You can see that in Broadcom's results, and we also collaborate very closely with Broadcom and with the customers themselves. So there are opportunities here, since demand is so high — if you look at all these major customers, they expect AI compute demand to roughly double annually over the coming years. I'd say everybody participating in that stands to gain. And when it comes to custom silicon, particularly for inference, the payoff can be large enough that these companies are ready to put money into EDA and designing chips in-house. So for our customers — and naturally, these are the largest corporations on the planet — the financial upside plus the tailoring advantage of going the custom silicon route is substantial. Just consider the major players: Google, Meta, and others such as Microsoft, Amazon, Tesla. My view is this trend will speed up. And as internal design efforts expand, naturally they'll have to spend on EDA, IP, and hardware. I'd say things are moving in a good direction on that front. When it comes to profitability, the situation is much the same, and we aim to stay disciplined about how we set prices. That means our margins look comparable, yet the upside for the companies in our system is substantial, since they're building their own silicon.
Ruben Roy
analystAnirudh, I have a quick question — hopefully it's about something you said in your prepared remarks regarding working alongside a customer on next-gen agentic AI solutions. I'm curious: are you observing that kind of thing among a broad group of your end customers? If yes, could you possibly take us through what that might imply — for instance, in what ways could those joint efforts around that sort of solution be turned into revenue over a longer horizon? And how do you view agentic AI in general as opposed to specific — it sounds almost like bespoke offerings tailored per customer, versus a wider agentic AI portfolio that Cadence might bring to the larger ecosystem.
Anirudh Devgan
executiveThat's an excellent question; we could spend quite some time on this one. And we feel fortunate to have partnerships with multiple companies in AI. I don't mean only AI design, but AI used for design within our offerings, and particularly regarding agentic AI, since that's a newly emerging space; we have something like five major AI platforms; yet what makes agentic AI distinctive is, naturally, everything related to gen AI. And if you consider even among the largest uses of AI is things like vibe coding, or software development. Well, when you examine it, a portion of chip design is coding too; we've automated, as I said before, 90% of the workflow for chip design; however, there's one segment of the workflow that isn't automated — customers still must write RTL. RTL functions as a language, register transfer language, that specifies the chip. And that takes place at the very earliest stage of the chip design process. So that step remains manual. But the algorithm that assists with vibe coding or C++ coding for ordinary software development — these agentic approaches, roughly speaking — can likewise assist with RTL development, right? And it can deliver substantial value for that 10% of the workflow that isn't automated. So because of that, we've made a huge investment in agentic AI, which you'll see as we roll out additional products down the line. And we currently have multiple partnerships in this area, and we're shining a light on one of them. And the manner in which we're taking that to market is longer — it's via JedAI. I've discussed JedAI previously. So JedAI is the joint enterprise data and AI platform. So it does include certain standardized elements. The database is standard. Every model is accessible. AI models have interfaces to all our AI tools. So a portion of JedAI is standard for every customer, and we collaborate with foundries and others to kind of train our models. Now a portion of it might be specific to a customer, all right? And when that happens, the data stays on the customer's premises. This is exactly why, right from the start, we built JedAI to work both on-prem and in the cloud — since there are times when a customer prefers the cloud, yet other times, if they need their data kept local, they'll want it on-prem. That's the reason we've spent years putting investment into this sort of distinctive platform, JedAI, which lets us not only create unique solutions such as RTL development and verification plan development, but additionally roll it out either broadly or in a more tailored fashion for one specific large customer. Still, I feel quite optimistic about the way agentic AI could take over the rest of that portion which was manual, and once more, let our customers concentrate on higher-level work while getting rid of a few of the tedious jobs like RTL coding, verification plan generation, and similar things.
Josh Tilton
analystNice work on such a solid quarter. Since we’re short on time, I’ll get right to a fairly straightforward follow-up question. John, I believe this one’s mainly directed at you. If you do end up seeing certain effects in the China region, because of the tariff talks still going on during the upcoming quarter, do you think — or could you help us grasp how you account for the revised guidance in light of any possible negative developments in that region?
John Wall
executiveJosh, honestly, that's a fantastic question. I wish I could predict what lies ahead. The — what I mean is, as we always do, we build in caution regarding every sort of regulatory uncertainty. And our guidance rests on the assumption that the current export framework stays largely the same as we move forward through the close of 2025; yet forecasting what will actually occur is extremely, extremely difficult. Still, based on all the accounts we've received, we hold the view that geopolitical tensions are not as high as many anticipate.
Anirudh Devgan
executiveThanks, everyone, for being with us today. This is a dynamic moment for Cadence, given solid commercial traction and expanding possibilities among semiconductor and system clients; backed by an exceptional team, we keep executing against our innovation road map while striving to exceed the expectations of our customers and partners. Our Board of Directors and I appreciate the ongoing trust and confidence that our customers, partners and investors place in Cadence.
Operator
operatorAnd that wraps things up for everyone — thanks so much for joining us on today's Cadence Third Quarter 2025 Earnings Conference Call; that brings today's call to a close, and you're free to hang up now.
Read the full transcript via the API
You're viewing the first half of this call. Get the complete Cadence Design Systems, Inc. transcript — plus 132,000+ transcripts from 7,000+ companies, speaker segments, AI summaries and full-text search — through the EarningsCalls.dev API.
Get the API View API docs →This call discussed
For developers and AI pipelines
Programmatic access to Cadence Design Systems, Inc. earnings transcripts and 132,000+ others is available through the
EarningsCalls.dev REST API. Plans from $39.99/month — full transcripts, speaker segments,
full-text search, and the recently-added /api/v1/transcripts/recent polling endpoint for ETL pipelines.
More tools for stock research
OptionsBell: an unusual options activity scanner that tracks options flow across all US stocks and emails you when large trades appear in Cadence Design Systems, Inc. or any ticker on your watchlist.
StockMarketScan: a stock screener with alerts: 24 pre-built screens refreshed every 5 minutes, watchlists and email notifications. Both offer API and MCP access and fair monthly pricing.