Cadence Design Systems, Inc. (CDNS) Q2 2023 Earnings Call Transcript

Cadence Design Systems, Inc. (NASDAQ:CDNS) Q2 2023 Earnings Conference Call July 24, 2023 5:00 PM ET

Company Participants

Richard Gu - Vice President of Investor Relations

Anirudh Devgan - President & Chief Executive Officer

John Wall - Senior Vice President and Chief Financial Officer

Conference Call Participants

Jason Celino - KeyBanc Capital Markets

Charles Shi - Needham & Company

Jay Vleeschhouwer - Griffin Securities

Ruben Roy - Stifel

Blair Abernethy - Rosenblatt Securities

Joe Vruwink - Robert W. Baird

Gary Mobley - Wells Fargo

Andrew DeGasperi - Berenberg Capital Markets

Harlan Sur - JPMorgan

Marco Conti - Deutsche Bank

Joshua Tilton - Wolfe Research

Operator

Good afternoon. My name is Lisa, and I will be your conference operator today. At this time, I would like to welcome everyone to the Cadence Second Quarter 2023 Earnings Conference Call. All lines have been placed on mute to prevent any background noise. After the speakers’ remarks, there will be a question-and-answer session. [Operator Instructions] Thank you.

I will now turn the call over to Richard Gu, Vice President of Investor Relations for Cadence. Please go ahead.

Richard Gu

Thank you, operator. I would like to welcome everyone to our second quarter of 2023 earnings conference call. I'm joined today by Anirudh Devgan, President and Chief Executive Officer; and John Wall, Senior Vice President and Chief Financial Officer. A webcast of this call and a copy of today's prepared remarks will be available on our website at cadence.com.

Today's discussion will contain forward-looking statements, including our 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. For information on factors that could cause actual results to differ, please refer to our SEC filings, including our most recent forms 10-K and 10-Q, CFO commentary and today's earnings release. All forward-looking statements during this call are based on estimates and information available to us as of today, and we disclaim any obligation to update them.

In addition, we will present certain non-GAAP measures, which should not be considered in isolation from, or as a substitute for, GAAP results. Reconciliations of GAAP to non-GAAP measures are included in today's earnings release.

For the Q&A session today, we'd ask that you observe a limit of one question and one follow-up.

Now, I will turn the call over to Anirudh.

Anirudh Devgan

Thank you, Richard. Good afternoon, everyone, and thank you for joining us today.

Cadence delivered excellent financial results for the second quarter of 2023 with strong ongoing customer demand for our innovative technology. We exceeded our guidance on all key metrics and are raising our financial outlook for the year yet again, resulting in 14% year-over-year revenue growth and 19% non-GAAP EPS growth. John will provide more details shortly on both our Q2 results and the updated outlook for the year.

With its unparalleled promise, Generative AI is beginning to make a significant impact globally. Our dedicated focus on AI over the past several years combined with our computational software expertise and invaluable data that lies at the core of AI it uniquely positions us to deliver to this tremendous potential of this transformational knowledge.

With escalating design complexity, increasing design starts and a growing talent shortage, AI-driven design automation is crucial in empowering customers to fully realize their innovation potential. Our customers are increasingly adopting our chip, package, board, system, Generative AI portfolio, as they're achieving exceptional quality of results and productivity benefits with these solutions. Customers are ramping up their R&D spend in AI-driven automation, opening up significant opportunities for Cadence.

Our solutions are enabling marquee AI infrastructure platform companies to deliver their next generation-compute, networking and memory products. For instance, in his Computex keynote earlier this quarter, Jensen Huang of Nvidia noted that Nvidia is a big Cadence customer and commented on the expanding strategic partnership between Nvidia and Cadence to accelerate EDA, system analysis, AI and digital biology.

Cadence has successfully collaborated with Tesla on the development of their game changing DOJO AI supercomputer. Tesla utilized a broad array of Cadence solutions across digital, custom analog, verification, 2.5D and 3D IC and system analysis for developing DOJO chips and solutions. We are very excited to extend our partnership to Tesla's next generation DOJO and FSD platforms. Our customers like Tesla are able to leverage the power of revolutionary Cadence Cerebrus Generative AI technology to optimize the quality of results of their groundbreaking AI chips and related solutions.

And from a products perspective, we began by providing AI-driven solutions for core ED applications, then expanded our AI portfolio to include system design and analysis and plan to extend it to life sciences in the future. With AI as they're underpinning, other generational trends such as hyperscale computing, 5G, and autonomous driving, continue to spur robust design activity across semi and system companies, creating ample market opportunities for our differentiated technology portfolio.

Now let's talk about our key highlights for Q2. In Q2, we deepened our long standing partnership with a marquee electronic system company through a broad ranging expansion of our core EDA, hardware, IP and systems portfolio. Ambulation and prototyping have become a must have part of the chip tape out and software bring-up flows and secular demand for our hardware platforms drove our verification business in Q2 to a 27% year-over-year revenue growth.

Following a record 2022 and Q1 ‘23 are Palladium Z2 and Protium X2 hardware platforms delivered a record Q2 as market demand accelerated for these best-in-class solutions. With 14 new customers and 45 repeat customers, more than half the orders during the quarter included both platforms. Demand for our hardware solutions was broad based with particular strength seen in AI, hyperscale computing, and automotive segments.

A global communication services leader successfully deployed the Z2 and X2 systems to significantly accelerate the development of their data center chips for internal use, including those designed for AI applications. Our hardware platform enable them to accelerate their verification workload and software bring-up, enabling first-pass silicon and software success. Verisium, our AI-driven verification platform that's built upon the ZI database and natively integrated with our verification engines is gaining traction at market shaping customers.

For instance, a large mobile chip company is working with us on developing a new Verisium regression optimization app and a Japanese semi customer is actively deploying Verisium apps across his automotive SoC and already saw a 4 times improvement in the identification of erroneous source code check-ins. Our digital IC business had another solid quarter with 15% year-over-year revenue growth largely driven by proliferation of our digital full flow, especially at the most advanced node at market shaping customers.

Our Cadence Cerebrus solution leverages breakthrough AI-driven technology to explore the entire design space and automatically optimize the digital full flow to deliver transformational results. Adoption and proliferation of Cadence Cerebrus accelerated. And with multiple new marquee wins, is now deployed at eight of our top 10 customers. A leading hyperscaler doubled their run rate through a major expansion that included our digital full flow and Cadence Cerebrus.

Another market shaping hyperscaler successfully used our digital full flow and Cadence Cerebrus to achieve a 16% leakage power reduction on their latest custom silicon. And [Indiscernible] achieved 2 times to 3 times better productivity with a new Joules RTL design studio. Through analysis, efficiency and significantly reduce iterations between RTL and implementation.

On IP, our scalable and profitable growth strategy continues benefiting from ongoing outsourcing trend, as well as the opportunities offered by expanding foundry ecosystem. Demand for our design IP was strong, led by our multiyear agreement with Samsung Foundry to expand the availability of our design IP portfolio for Samsung's advanced process technology. We signed a definitive agreement with Rambus to acquire their five IT asset along with the talented team. The addition of their leading HBM GDDR5 solutions, and SerDes IP enhances our established IP portfolio, providing complete subsystem solutions for demanding networking, hyperscaler, and AI application.

Our system design and analysis business continued its strong momentum in Q2, delivering 23% year-over-year revenue growth. With Moore's Law slowing down, and chip complexity and cost increasing, companies are looking to multi-die, 3D IC, and chiplet-based architectures to achieve better performance and greater cost savings.

In Q2, we expanded our collaboration with Samsung Foundry through the delivery of reference flows and package design kits based on our uniquely differentiated integrity 3D IC platform. The industry is only unified platform that includes system planning, packaging, and system level analysis in a single cockpit. Our multi-physics solutions boosted by our optimality Generative AI technology, leverage differentiated system simulation and optimization techniques to deliver superior results.

Our CFD portfolio that includes recently acquired high fidelity simulation technology, one large renewal and add-on business with top Aerospace and Defense company. We were pleased with the new win and growing repeat orders for our multi-physics portfolio from customers across multiple end markets.

In summary, I'm pleased with our Q2 results and the continuing momentum of our business. AI-driven automation in chip and system design offers massive opportunities for Cadence over the long-term. That seamlessly align with our computational software, core competence, and intelligent system design strategy. We continue to invest in our world-class EDA system design and analysis and AI capabilities to deliver cutting edge innovation to our customers and partners.

Now I will turn it over to John to provide more details on the Q2 results and our updated 2023 outlook.

John Wall

Thanks, Anirudh, and good afternoon, everyone. I'm pleased to report that Cadence achieved strong results for the second quarter of 2023, driven by the broad-based strength of our business. Our customers are increasingly adopting our Generative AI portfolio as they are achieving exceptional quality results and productivity benefits with these solutions.

Here are some of the financial highlights from the second quarter, starting with the P &L. Total revenue was $977 million. GAAP operating margin was 30.7% and non-GAAP operating margin was 41.8%. GAAP EPS was $0.81 and non-GAAP EPS was $1.22.

Next, turning to the balance sheet and cash flow. Cash balance at quarter end was $874 million, while the principal value of debt outstanding was $650 million. Operating cash flow was $414 million and we used $325 million to repurchase Cadence shares in Q2. We are increasing our outlook for the remainder of the year, due to continued broad-based strength across our technology portfolio. Demand for our functional verification hardware solutions remain particularly strong and as a result, our updated outlook for the second-half of the year reflects approximately 15% revenue growth, compared to the second-half of last year.

Before I provide our updated outlook for the year and our expectations for Q3 I'd like to highlight that our outlook contains our usual assumption that the export control regulations that exist today remain substantially similar for the remainder of the year. Our updated outlook for fiscal 2023 is revenue in the range of $4.05 billion to $4.09 billion.

GAAP operating margin in the range of 30.2% to 31.2%. Non-GAAP operating margin in the range of 41.2% to 42.2%. GAAP EPS in the range of $3.35 to $3.41, non-GAAP EPS in the range of $5.5 to $5.11.

Operating cash flow in the range of $1.3 billion to $1.4 billion and we expect to use at least 50% of our annual free cash flow to repurchase Cadence shares. A large number of hardware systems are slated for delivery in late September and early October. For the purposes of providing our outlook for Q3, which ends on September 30th, we thought it was prudent to assume that the vast majority of those hardware deliveries fall into October and Q4.

With that in mind, for Q3, we expect revenue in the range of $990 million to $1.01 billion. GAAP operating margin of approximately 29%, non-GAAP operating margin of approximately 40%. GAAP EPS in the range of $0.76 to $0.80. Non-GAAP EPS in the range of $1.18 to $1.22. And we expect to use approximately $125 million of cash to repurchase Cadence shares. As usual, we published a CFO commentary document on our Investor Relations website, which includes our outlook for additional items, as well as further analysis in GAAP to non-GAAP reconciliations.

In conclusion, we delivered a strong Q2 and first-half of the year. With the increase in our outlook, at the midpoint, we now expect revenue growth for the year to exceed 14%, which would take our three-year revenue CAGR to approximately 15%. And non-GAAP EPS growth of approximately 19%, which would result in an average annual growth rate of 22% over the past three years.

As always, I'd like to close by thanking our customers, partners and our employees for their continued support.

And with that, operator, we will now take questions.

Question-and-Answer Session

Operator

Thank you. [Operator Instructions] Our first question comes from Jason Celino with KeyBanc Capital Markets.

Jason Celino

Great. Thanks for taking my question. John, Anirudh. Good quarter, I think if we take a step back, you know, up 14% for the year, still pretty impressive. Sounds like hardware at another record quarter and was up 27%, but I think some investors will still wonder, you know, could it have been even better? Maybe could you speak to any areas that maybe weren't as strong that you were hoping, or how do we kind of level set the performance?

Anirudh Devgan

Yes. Hi, Jason. Good question. In general, we manage our business for the long-term, and we are pretty positive with the results. For the full-year, like you mentioned, it's more than 14% revenue growth and earnings are growing at 19%. And then our strategy of doing EDA plus STA plus AI with the computational software expertise. So I think overall, we're pretty pleased with how all the businesses are performing. And, you know, continue to drive in all fronts, working with customers and partners to deliver the right solutions.

John Wall

Jason, as Anirudh says, we're very happy with -- very pleased with the first-half performance. If there's anything unusual, essentially in the numbers for the first-half, I think one unusual thing for me would probably be the operating cash is flat in our guide to last quarter, and that's because as Anirudh says, you know, we manage the business for the long-term. There's an opportunity, we've had some opportunities to build inventory. We've had very strong hardware demand, so we need to buy inventory and we were offered some larger discounts for payment upfront. So we've embedded that into the guidance. It'll help us with gross margins in future years, but it doesn't help with operating cash for this year.

Jason Celino

Got it. And then my quick follow-up, the IP acquisition from Rambus. Is it included in guidance? And then maybe can you speak to strategic rationale? I guess what goes into the decision, build versus high as early as IP? Thanks.

John Wall

Yes, Jason. So everything we know is included in our guidance, but it's immaterial to us for the year.

Anirudh Devgan

Yes. And it's a great opportunity for us to acquire a very good set of IPs and a talented team. And, you know, what is particularly good in this case is that it is a existing, you know, commercial operation. You know, sometimes we have opportunity to acquire teams from, you know, other semiconductor companies, but in this case, it was existing commercial operation with key IPs like, you know, HBM and GDDR. So we took that opportunity, and we glad to, you know, partner with Rambus here and it really helps, you know, IP for AI and 3D IC with HBM and GDDR IPs. So overall it’s good acquisition, and we continue to invest in all our businesses, you know, including IP. So…

Jason Celino

Excellent. Excellent. Thank you both.

Operator

We'll take our next question from Charles Shi with Needham & Company.

Charles Shi

Hi, good afternoon. Thank you for taking my questions. First, I want to ask about bookings, then I have a follow-up on hardware. So, John or Anirudh are you guys still seeing strong bookings potential in the second-half? I definitely saw your bookings implied bookings are actually already up in Q2, but I think that people are probably accustomed to, like, a $1 billion plus per quarter, kind of, bookings. So when can we get to that -- back to that level? Do you still expect the second-half bookings to get strong as let's say maybe first-half ’22? Thank you.

John Wall

Yes, Charles, great question. Yes, we'd expect a stronger second-half for bookings. As we said last quarter, the -- we had very few contracts expiring in the first-half of this year, and we have more contracts expiring in the second-half of 2023. So we'd expect bookings to be higher in the second-half, compared to the first. When we look at backlog, we filed our 10-Q by the way. So you'll see our backlog at the end of Q2 dropped from $5.4 billion to $5.3 billion and RPO at the end of Q2 dropped from $5 billion at the end of Q1 to $4.9 billion.

But the current value of that RPO, the annual value, as you know, backlog and RPO is a combination of the annual value multiplied by the time essentially that you've contracted with customers. And our current RPO is up from last quarter, it was $2.7 billion last quarter. It's up to $2.8 billion this quarter. So we've improved the annual value, and that's typically what we focus on, making sure we're improving the annual value. But yes, we're very, very pleased with the bookings performance for Q2. It was slightly stronger than we had anticipated or expected. And second half, we think it will be better again.

Charles Shi

Yes. So is it safe for us to -- for you guys to call maybe Q1 was probably the fall in terms of booking, and we are way past that going forward?

John Wall

Yes, yes, Q1 was definitely a low bookings quarter for us. Q2 was a lot stronger.

Charles Shi

Got it. Maybe a follow-up question around hardware. You kind of said for Q4, you're expecting a good amount of hardware revenue. Maybe it's parked in Q4, but when folks look at your full-year guidance, your Q4 isn't really up by a lot. They imply the guidance for Q4. But given you just said that the bookings are expecting relatively strong. I would think that ratable part of the business should see a very strong sequential growth into Q4. But -- and layer on top of that hardware you're expecting in Q4, can you kind of reconcile what you guided for Q4 versus the commentary you made about hardware and your bookings? Thank you.

John Wall

Yes, Charles, I think you're referring to the step-up from Q3 to Q4 in our guide. So we're guiding about $1 billion of revenue in Q3 with a balance of just over $1.70 billion in Q4. Now that's slightly exaggerated the Q3 to Q4 step-up. When we originally looked at the top-down forecast for the second-half, there was a lower step-up in revenue mainly due to software increases in annual value and in software bookings that we expect.

But when we looked at the hardware and the timing of when hardware deliveries happen and there's so many that fall late in the quarter, we thought it was prudent to assume that those bookings, we should include them in our implied Q4 guidance, so to speak, rather than including it in Q3 because I’m not really sure about the timing of those.

Charles Shi

Okay, thank you.

John Wall

Okay, thanks.

Operator

We'll take our next question from Vivek Arya with Bank of America.

Unidentified Analyst

Hi, this is [Indiscernible] on for Vivek. Just wanted to start out, if you could provide some insight on how you quantify benefits from AI? Just looking at the semiconductor landscape, there have been several AI-related design programs that started a few years ago. So would you say you've already seen the benefits in your financials? Or should we expect even more growth acceleration in the future? And if you could comment on the attach rate of your AI-enabled products to baseline contracts, that would be helpful? Thanks.

Anirudh Devgan

Yes. So the AI has multiple implications for us. The first part is we enable the development of AI system side. This is more like a horizontal application. So like we talked about our partnership with Nvidia and there are other companies building AI systems like we talked about today, our partnership with Tesla, which is building both server side AI and on the car AI and then other kind of data center companies. So that's -- and we are in a unique position to supply to all the -- whether it's GPU or data center or automotive applications of AI to do the AI build-out.

And that I think is still continuing. There'll be more and more systems designed, and our tools and IP can play a key role in that. So that's kind of the first application. And I believe we are still in the beginning stages of that kind of build-out and monetization.

And then the second part would be more vertical, applying it to our own products. So we have JedAI platform, which is with a unique position as a kind of a data platform for AI and then five unique AI applications, all the way from analog, digital, verification, package and board and systems. So I believe we have the most comprehensive product portfolio for chip package system design.

And that also I think we are in the early stages, but we are seeing a lot of adoption of those kind of vertical applications. And we are also using them internally. We design our own chips for Palladium systems, right? And then, of course, we have a big application engineering team that works with customers. So not only we are providing solutions to our customers and partners on EDA and SDA, but we are also using them internally and seeing good benefits.

And then the third part, which is more out in the future, I think AI will also drive new kind of applications, and one of the biggest ones is definitely going to be life sciences. So we are laying the groundwork for that with our OpenEye and there's a lot of opportunity to apply their -- AI there in digital biology, which will pay out in the next five to 10 years. So for all these kind of three big -- there's a horizontal play for building out the AI systems, GPUs and data center other things, vertical plate of EDA and system, both to our customers and ourselves and emerging areas that will emerge in life sciences. So overall, I feel we are very well positioned to really capitalize on this megatrend.

John Wall

Yes. And to answer your revenue question, essentially, I mean, I think it's clear that it's showing up quickly in the hardware space. We feel on the software side, it will probably take a couple of contract cycles for that to flow through. We're seeing strength in pockets on the software side. But the software side, because it's recurring in revenue will take a while to play through. But you're definitely seeing the impact of AI in hardware -- strong hardware demand.

Unidentified Analyst

Got it. Helpful. And then just quickly as a follow-up. I hope we can discuss your capital allocation policy as well. Just given the pretty solid cash position and significant cash generation of the company, is there any interest in allocating more cash towards share repurchases? And also, if you could comment on any interest in issuing the dividend in the future as well?

John Wall

Yes, [Blake] (ph), yes, we look at that all the time in terms of our cash allocation policy. In Q2, we used $325 million on stock repurchases. $200 million of that was due to an accelerated share repurchase. Now the timing of that was to coincide with -- was times deliberately to coincide with our merit and promotion cycle. You might have noticed as well in our guide that expenses take a step-up and margins slightly down second-half, compared to first-half. That's because our pay rise and promotion cycle kicks in on July 1 at Cadence. And that's typically when we do our stock refresh for employees in the company.

And because we granted a lot of stock at that time of the year, we -- in our stock repurchase approach, we want to make sure that we're offsetting the dilution first. And then ideally, we want to reduce that share count over time. We've talked about dividends with the Board from time-to-time, but we think it makes most sense to use our excess cash for stock repurchases. Our policy is essentially to use at least 50% of free cash flow to repurchase our shares. And that should more than offset dilution, see the share count continues to decline over time and give us enough to grow the business.

Unidentified Analyst

Great. Thank you.

John Wall

Always.

Operator

We'll take our next question from Jay Vleeschhouwer with Griffin Securities.

Jay Vleeschhouwer

Thank you. Good evening. Business question first and a technology question as a follow-up. So Anirudh, as you know, for the last couple of years, we've often spoken about how semiconductor companies and systems customers can become increasingly alike. The question though is, in what important ways do they remain dissimilar in terms of sales cycles, conversion of RPO to revenue, services requirements, account profitability? Anything you care to mention that continues to distinguish those two classes of customers?

Anirudh Devgan

Yes, Jay, I mean, like you've also written, I mean, the semi companies are becoming system companies and system companies are becoming semi companies. And some traditional semi companies are much more system companies now with software, and a lot of traditional system companies are doing a lot of semi, one of the big -- several big semis have done in system company. So there are a lot of similarities.

You're asking in terms of differences, I mean, I guess one key difference has to be that the system companies, by definition, have hardware and software together. You are a system company because you have software and also have electrical and mechanical together. I mean a good example is a car, right? It's emerging electronics and mechanical and it's emerging of hardware and software.

So for that reason, I think like the strength we talked about in the hardware business, a lot of it is not just for chip design, it is for software bring-up. And we are seeing that in AI applications, a lot of the software has to be built along with the hardware. And even for Nvidia, which is traditionally semi company, has a lot of software, of course, And then a lot of the data companies that are doing their own chips have a lot of software.

So I think the enablement of software becomes key. And the other thing is this merger of electrical and mechanical. That's why our whole investment in system design and analysis in thermal simulation, which is a big thing for data centers and cars, electromagnetics, CFD. So the key thing I think is unique is different is hardware plus software and electrical plus mechanical.

Now in some cases, we do strategic services for some system companies to get them started. And I think that's also a unique difference. So sometimes, we will provide help them to get going with chip development. So I would say, Jay, these are probably three big differences compared to. But it's good to see the growth in the semis, too. We are happy to have a whole new market with system companies, but we still love all our semi customers, and they're doing phenomenal, and we are partnering with them in all kinds of ways.

Jay Vleeschhouwer

Okay. As the technology follow-up at DAC two weeks ago, there was a quite interesting presentation by Cadence at the conference related to AI and there was the use of a very interesting with regard to productivity or where productivity is going to come from in future for EDA. And that term used by Cadence was task abstraction. So a new kind of abstraction from the decades of extraction we've seen in EDA for the last many years.

So how are you manifesting that new concept of abstraction in terms of product development, packaging, all those sorts of things going forward since that term seems to be very much a critical part of your thinking going forward for EDA productivity?

Anirudh Devgan

Yes. Jay, if you step back, right, you look at fundamental innovations, of course, we talk about a lot of things. But if you look at mathematics or computer science, I mean, there are two big innovation. One is abstraction, the other is parallelism. And in the history of EDA is moving up the abstraction layer, right, from -- all the way from polygons to transistors, to gates, to RTL.

And also parallelism, I mean, in the last 10 years -- I think if you look at EDA -- history of EDA, it started with abstraction first, as you know. And then in the last 10 years, it went back to more parallelism with use of cloud computing and CPUs. But with this AI, there is opportunity to do more abstraction, move even higher level. Basically, the real opportunity is to automate mundane tasks, right?

So what was done by the user can be done by the AI engine. So abstraction will provide in the end a lot of automation and productivity improvement. And we are seeing that we can apply to even technology -- design technology codesign with leading foundries. We can apply AI to design of floor planning and higher-level functions, which should previously manual apply to 3D-IC partitioning with integrity. So overall, this theme of applying AI, doing more abstraction, doing more of the mundane work and let the user focus on higher-level tasks and be more productive is only the beginning of that and is enabled in all these aspects, digital verification, system design and so on.

Jay Vleeschhouwer

Okay. Thank you.

Operator

Our next question comes from Ruben Roy with Stifel.

Ruben Roy

Thank you. Anirudh, if I could start from where you left off there, I had a question similar to Jay's. But I was wondering, I think you mentioned that SerDes with AI was deployed at eight of your top 10 customers. I'm just wondering today, can you talk about how AI is being used? Is it mostly placed in route? Or are you seeing some of these other areas being implemented in leading-edge designs, whether it's packaging or even some of the floor planning, et cetera? How far along are we?

Anirudh Devgan

It's getting pretty pervasive. The first thing I'm ascertain of now compared to like two years ago, we started on this two, three years ago in terms of working with SerDes and our customers. One thing I'm positive now is that a few years from now, almost all our customers will be using these tools because they are -- it's almost like an irreversible trend.

And then or placement route was the first one in some ways. So it has a lot of adoption. But it is extending to, like I mentioned, working with foundries, working with floor planning, higher-level functions, verification. See, in the end, I do believe verification will have a most profound impact within EDA on AI will be on verification because it's an unbounded NP-complete problem, right? So you can do more and more verification you're never done. So that's why you see us investing heavily in Verisium. And we are the first to apply AI to verification. We talked about couple of examples with customers in the prepared remarks.

And then the other thing is areas that traditionally haven't had a lot of automation, like package and PCB. So with Allegro X AI, that area has not seen automation for like 20, 30 years for the first time is possible with AI. And then in system simulation, see the system -- EDA by its nature has had a 30-year history of optimization and automation. Now it was not with -- it was more with classical techniques and now with AI. But if you look at the system area, like CFD or electromagnetics, they could barely simulate stuff, forget optimizing it. So not only now we can simulate things, because our algorithms are faster and we are also accelerating them on GPUs and all to give massive speed up on system simulation.

So now for the first time, you can stimulate a very, very high Fidelity simulation of the whole car. But then on top of it, optimality and AI, you can do the design of the car or the shape of the wing, which is what, in the end, the user wants, right? They just don't want to simulate. And all this applies exactly to this big market of SD&A. So I think that's a huge opportunity. And we're already seeing that, like even with some of the car companies we work with, we can optimize the shape of the wing that affects performance like for a racing car, you'll be using AI. So we already have pretty promising results, and we will talk more about that in the future. So I would say, first of all, I think it will be deployed across all these platforms.

We started with PNR, but already we are seeing a lot of potential. And in the end, in EDA, the huge potential will be verification. In the other area, I'm very optimistic is system simulation and AI.

Ruben Roy

Thanks very much, Anirudh for that. A quick follow-up for John. Just on operating margin. Down a little bit in Q3 with what sounds to me like hardware, kind of, strengthening towards the end of the quarter and obviously into Q4. How are you thinking about that? Or how should we think about operating margin and sort of the moving parts between Q3? You bumped it up a little bit for the full year. So if you can help us out on how you're thinking about operating margin in the context of hardware attach rates going up as you exit the year, that would be helpful?

John Wall

Yes. No worries, great question. But as you saw, I think the first half was roughly just under 42%, I think, for non-GAAP operating margin. For the second half, the guide essentially implies 41% to 42% operating margin. Now it's slightly down as it has been in the last few years in the second half compared to the first half. But that's because July 1 is our merit cycle. That's when we do pay rises and promotions and generally pay increases. The vast majority of our expense base is people and engineers, about 90% of the employees are engineers.

So pay increases impact us from the first day of Q3 and that step-up in cost is really the headwind you're seeing from Q3 when you compare Q3 and Q4 combined or second half compared to the first half. Now I've exaggerated that a little bit by being prudent with the Q -- our Q3 guide on the hardware side because there's a number of hardware systems that are scheduled to be delivered in late September.

And late September is the holiday period -- sorry, late September, early October is the holiday period in China and other parts of Asia. And I'm just worried that some of those will slip to Q4. So I'd assume the vast majority of them slipped to Q4 in the guide just to be prudent. And that, of course, is impacting the operating margin that we're showing in the guide for Q3. But I think if you look at the second half as a whole, generally, it's 41% to 42% for the second half when you take out the impact of that prudence.

Ruben Roy

Got it. Makes sense. Thank you, John.

John Wall

Thanks.

Operator

We'll take our next question from Blair Abernethy with Rosenblatt Securities.

Blair Abernethy

Thanks gentlemen and great quarter. Just wanted to dive in a little bit on China. The growth this quarter on a year-over-year basis was quite strong. And anything you'd -- any color you could give there on what's driving that and how this continues would be helpful?

And secondly, just on -- back on the hardware side, John, just you commented in the past about your production capacity ramping up. Just wondering where you're at midyear here and how lead times are doing?

Anirudh Devgan

Yes, I'll start first and then John can comment on the inventory. But overall, China, we are pleased to see continuing strong design activity in all kinds of end markets and also good strength for hardware business. So we've also had a good hardware business in China. And particularly in Q2, I would say there was a lot of strength from mobile vertical. A lot of companies, some big system companies designing their own chips, and also automotive, especially EV automotive in China was particularly strong. So I mean, China is a -- just like rest of the world is pretty diverse market, and we play in all aspects of it. But those two verticals were pretty strong in China. John, do you want to…

John Wall

Yes. And in terms of your question in relation to lead times I think, at the start of the year, we ramped up production capacity because lead times have exceeded 26 weeks. And we thought that's going to hurt us and we're competing for hardware business. We put a big dent in that in Q1. We ramped up production capacity. We kept at those levels for Q2 and intend to keep it at the higher production levels throughout the rest of the year and going forward.

We're bringing down those lead times. Our goal is to try and get it down into the 10- to 13-week kind of time frame, but we're not there yet though. I mean in terms of the guide that I've given you for Q3, even with prudence and all in -- let's say, if I took the prudence out, that we don't need $1 of Q3 hardware bookings to meet those numbers. So we're slightly over the 13 weeks, probably around the 15-week mark in terms of what we have in backlog right now to deliver.

So we won't actually get through all of the hardware orders that we have in backlog in Q3. The -- but demand is still strong. We are reducing -- it's probably around the 15-week mark now, but I'd like to get that down to 10 to 13.

Blair Abernethy

Okay. Great. And actually, just one more question on the hardware, if I can. So the Palladium Z2 and the Protium X2 have been out in the market now for a bit of time. When can we expect to see next-generation versions?

Anirudh Devgan

Yes. It's been only out for two years, which is -- our last system it was like, I don't know, five, six years between launches, and I think this one has been out two years. And even when we launch Z2, X2, even Z1 and X1, we're doing great. So right now, we are pretty pleased with these systems. And we also can add -- we keep improving the software. I mean, that's the hardware part of it, but we're always improving the software and compile times that go on top of these systems.

So there is ongoing software rhythm to them anyway, independent of the hardware. But overall, I mean, this phenomenal performance by the team, and we really appreciate all the support from our customers and partners.

Blair Abernethy

Great. Thanks very much, guys.

Operator

We'll take our next question from Joe Vruwink with Baird.

Joe Vruwink

Hi, there. Let me dig in a bit of a leading question. So you're now talking about second half revenues growing 15%. Some multiyear CAGRs are closer to 15% to 16%. There's -- things on the horizon in that next couple of quarters not trailing, but longer as just in terms of how your customers are acting and also what Cadence is offering that would seem to be better than what you've experienced, both in terms of R&D intensity and your potential wallet share?

So the question is, instead of this 15% rate of growth being the high end of what historical growth rates have been, do you think we're maybe seeing a resetting of the overall base case for growth rates and what might be available to the company going forward?

Anirudh Devgan

Well, first of all, I would say 15% growth rate and 19% EPS growth is an amazing performance of any company. So we are very proud of the team to deliver this. And in terms of design activity and R&D intensity, I mean that is strong, if not getting stronger. So of course, the overall macro environment as tough as it has been last year and this year, but we do see a lot of semi and system activity. So that's not any different than before, but we have to carefully monitor how things go in the future, of course. But overall, we are pretty pleased to see the design activity.

John Wall

Yes, Joe and you'll see it in the -- we published our CFO commentary in the second page of the CFO commentary, you'll see we track three-year CAGRs. But we're really trying to grow the accounts through a complete contract cycle, and most of our customers do three-year baseline contracts to do add-ons throughout the year that coterminate with those baseline contracts. So we think the three-year CAGR is a good metric.

And if you look at the revenue growth rate on a three-year CAGR basis, it's continuing to decline. Now it's certainly the pace of acceleration seems to have slowed down this last two years in the kind of mid-teen mark that -- but our focus is on driving revenue growth -- profitable revenue growth. But we're using -- we generate huge amounts of cash flow. And we aim to use more than 50% of that cash flow to buy back shares and reduce share count.

So we're trying to drive that non-GAAP EPS higher and be disciplined with our stock-based compensation. You'll see our stock-based compensation has consistently been around 8% of revenue, and our non-GAAP EPS CAGR, it's been 20% or higher for the last five years. And with this 19% rate that we're on now at the midpoint of the guide, that has us on track for 22% EPS growth on a three-year CAGR basis for 2023.

So our focus is really on bottom line. We care about EPS and growing EPS for investors and growing profitability, and revenue growth is a means to an end, I guess, in that respect. But we're very, very pleased that is continuing to increase.

Joe Vruwink

Okay, thank you for that. And then one AI question. I think it's a follow-up on Ruben's question. But just a point about -- it seems like maturity, certainly customer experience, referenceability, it's greatest the back-end area of its greatest around Cerebrus. When you think about the pace of adoption on the things you've introduced into markets since last fall, do you think it's any quicker or slower than the experience with Cerebrus so far? So like is there any reason that debone verification would not be a prime candidate for your underlying technology? I'm just wondering if you can maybe compare and contrast the adoption curve?

Anirudh Devgan

Yes, that's a good question. I would say that the adoption curves are, at this point, I would say they are similar. Especially Verisium, as I said, verification is a big area. Now they're easier in some sense that there is already a proof point that the customers have seen with Cerebrus. So they are confident in our ability to deliver overall AI solutions. And same thing with like Allegro X, Virtuoso Studio. So -- but also how it is in this business, right, the customers have to try the things and they tried on some design, then they do broad adoption. So that's always there in new areas as well, so like with Verisium and Allegro. So I would say, overall, it is similar and there's a lot of interest.

I mean the amount of interest in other AI solutions apart from the original like Cerebrus is very high. But I would say the rate at this point is similar of adoption. And this is going to take some time, which is okay. It is a natural process of trying and deploying. Now some areas, it's faster like hardware because of the nature of the revenue recognition. But overall, the adoption is great and moving along

Joe Vruwink

Great. Thank you very much.

Operator

We'll take our next question from Gary Mobley with Wells Fargo.

Gary Mobley

Hey, guys. Thanks for taking my question. I wanted to ask about the commercialization of the five different tools under the JedAI umbrella. And I wanted to confirm something, first of all, is that a baseline -- are those products in baseline license renewals with customers? Or are you licensing those on a per chip design basis and have the tools major way from price discovery phase to general availability?

Anirudh Devgan

Gary, so these all are new to products. So they're not in previous contracts. So all of these products have to be purchased, and they can be purchased with the renewal or outside of renewal, as you know, in add-on business, but they are not part of existing contracts with customers. And we have different business models in which we introduced these products just like our base product.

And then as we talk to customers, some of them are in very wide deployment. I think I gave you example in the prepared remarks on one hyperscaler. We had a very, very wide deployment of some of these AI tools. Some are in the early stages. But these are all new products versus what was there in previous contracts.

Gary Mobley

Got it. Got it. Thanks you for that, Anirudh. And I wanted to ask about what you're signaling with the acquisition of the five business from Rambus. I know you went through a period of time where there was a decision to maybe refocus some of the energy, some of the investments from the company away from IP into areas like SD&A. And so now you're making an acquisition to grow some aspect of your IP business. And so I'm wondering if you can give us an update on how you view the IP business as it fits with the overall business in terms of investment in effort?

Anirudh Devgan

Yes. IP is an important segment of our business. It's like having five kids, right? We love all the five kids, all five lines of business. Now one of them is very young and growing like SD&A, growing 20% plus. So that's great, but that doesn't mean we don't take care of all the lines of business. So IP, this was a great opportunity to get a right team and IP, which is synergistic, useful for AI and 3D-IC. So I think, at the right time, we will make investments in all lines of our business. And we have done so like with Virtuoso, with digital. And so we want to make sure all lines of business are competitive here.

Gary Mobley

Got it. Than you, Anirudh.

Operator

We'll take our next question from Andrew DeGasperi with Berenberg Capital Markets.

Andrew DeGasperi

[Indiscernible] thanks for taking my question. I guess the first one I have is on the individual segment. I noticed the IP business in Q2 came down a bit and also system analysis, a slight deceleration. Just wondering if there's anything you would call out for those two? And I have a follow-up.

John Wall

Yes, sure. Thanks, Andrew. But in relation to IP, we've always been expecting more IP to be delivered in the second half compared to the first half. It's actually the opposite of 2022. I think we called that out on last quarter's call as well. So we're expecting a stronger second half for IP, and IP can be a bit lumpy at times because it's based on the timing of deliveries.

On system design and analysis that we're very, very pleased with the strength of the growth there. Again, I mean all of our businesses are growing strongly. I think when I look at the second half, and the 15% that we expect our total revenue to grow compared to the second half of last year, I think all of our businesses are growing double digits in that outlook.

Andrew DeGasperi

Got it. And then in terms of your guidance, if I were to look at the high end of the range for the top line, specifically, what exactly would -- assumptions would we have to make for you to get there or even exceed it? In other words, is there a potential for surprises -- positive surprises on that front?

John Wall

Well, I'm certainly -- I'm very confident in the second half guide. But I'm now confident also in the Q3 guide -- I was less confident when I saw all those hardware deliveries time for kind of the final week and the final few days of Q3, and I didn't want to take the risk of including those in the Q3 guide. But for the second half, I'm very confident in the second half.

But -- and as we reduce the lead times on our hardware, we found that we're certainly seeing the demand for our hardware strengthen and we have strong demand across all lines of business. And as Anirudh would say, I mean design activity is very, very strong. So I guess to the extent that we can continue to reduce those lead times and deliver hardware earlier, that certainly a track to the higher end of the guidance range because with hardware, of course, you get revenue upfront. But on the software side, growth takes – revenue takes longer to kind of play through because of the recurring revenue profile.

Andrew DeGasperi

Understood. Thank you.

Operator

We'll take our next question from Harlan Sur with JPMorgan.

Harlan Sur

Hi, good afternoon. Thanks for taking my question. The fact the full-year is playing out as you expected, right, the stronger second half in front of you. Your semiconductor, your systems customers continue to drive strong chip packaging, system design activity and road maps, especially in the areas of accelerated compute and AI. And so when you combine this with your pipeline, your renewal visibility, you guys have already said your second half bookings will be strong. Does the team believe that it can grow its backlog for the full-year?

A - John Wall

So in terms of backlog, I mean, we don't guide bookings. But I don't think it's unreasonable to expect our backlog to end the year higher than here -- where we are here. Our focus is really on growing the annual value of that backlog. So I mean, like a three-year software contract worth $10 million a year, we'll add $30 million to backlog into RPO and a two-year contract worth $12 million a year will add $24 million to backlog in RPO.

Now if it's a choice between those two options with one customer, we'll typically take the $12 million a year because $12 million a year is better than $10 million a year, and there will be a lower booking number, lower ARPU number, but it will show up in the current RPO. And like I said earlier, that -- I mean if you look at our Q2 backlog compared to Q1, it dropped down from 5.4 to 5.3.

But if you look at our Q2 current RPO, that's the amount of our RPO that we expect to turn up in revenue in the next 12 months. At the end of Q1, with a higher backlog number, that was $2.7 billion. At the end of Q2 with a lower backlog number, that's gone up from $2.7 billion to $2.8 billion. So our focus is generally on the annual value. But I think I understand where your question is coming from -- because first half was quite light for bookings. We expect the second half to be stronger. I do think that it's, like I say, not unreasonable to expect RPO and backlog to increase from here.

Harlan Sur

I appreciate that. And then for Anirudh, all of the leading-edge chip companies are moving to this chiplet that die-based packaging architecture. The systems customers are adopting these very, very complex board designs, right? Here you guys benefit with Allegro, Integrity, Clarity, Celsius platforms. Can you just give us a rough sense of the revenue contribution from advanced packaging and board design, simulation, verification? Is it near 10% of revenues? And roughly how fast is this subsegment growing on a year-over-year basis?

Anirudh Devgan

Yes, that's a good observation, of course. I mean there's a lot of chiplet-based design activity and it started in HPC, high-performance computing, AI. And I think it will permeate to other parts -- other verticals as well. And also, all the leading foundries, as you know, are now leading with their 3D-IC kind of solutions. And we -- it shows up in different parts in our business. It definitely shows up in SD&A because SD&A system design and analysis includes both the Allegro and all the system analysis of Clarity, Celsius.

And that's why you are seeing very, very strong growth. I mean that's a very, very strong growth number, I would say, especially on a ratable revenue model, right? So if you're growing 23% this quarter, 20% plus for the last several years, so a lot of it is driven by this 3D-IC chiplet because the system market is interesting, right, the thermal market and the electromagnetic market traditionally. But a lot of the new growth is this intersection of chip and system, which is, by nature, 3D-IC and chiplet.

So we have a lot of customers buying thermal solutions, especially targeted for chiplets because that's a big thing, or electromagnetics targeted for chiplets. So we definitely see that in SD&A. And then we continue to see that in our regular business, like Virtuoso or Innovus and Digital. But I still think this is still early innings. It's still the baseball analogy, still the second, third inning for 3D-IC, and it will become the dominant technology.

I mean, of course, the 2-nanometer and 1.4-nanometer all that will be critical, but they will be all integrated on chiplets and into poses going forward. So -- and I think what is important to remember, of course, a lot of people talk about chiplets and 3D-IC, but we are uniquely positioned. So not only for chip implementation but then also for package design, we provide the leading -- we are the leading software provider for package design with Allegro and then all the thermal and system analysis tool. So I think Cadence is, I believe, very uniquely positioned to benefit from chiplet and 3D-IC.

Harlan Sur

Thank you for the insight.

Operator

We'll take our next question from John Marco Conti with Deutsche Bank.

John Marco Conti

Hi, Anirudh and John. Thanks for taking my questions. And so the first one is, I know you've touched on this a little bit. Could you share some more color on the usage of AI tools in system companies this quarter? Are any of the new wins related to AI tools for system customers? And how can we think about the acceleration of usage, not just leading nodes, but in the broader design space, even at more mature nodes?

I think investors are thinking what next when you taking the current use cases in the next two, three years, where the cost trade-off might not be as imperative as to say in the GPU market? I imagine is this a function of better price recovery matched with higher adoption/proven use cases.

Anirudh Devgan

Yes. I think the application of AI has like, as you know, has both horizontal and vertical components, right? Horizontal is building out the AI infrastructure. So those in the beginning tend to be advanced node GPUs and also system companies. So the system companies we mentioned because one of them is Tesla, right? So that's -- they have a server side chip. They have talked about this, which is Dojo and then also in the car, right, with full self-driving. So that, I think, is a pretty remarkable use of AI that's only going to grow with other companies doing this -- other auto companies doing similar things.

And then we have several customers, which are hyperscalers, which are building their own silicon for AI. I mean this is well publicized also, and you can guess who they are. So we are continuing to work with them. And we had several meaningful expansions in Q2 linked to AI chips done by these large hyperscale companies.

And now on the vertical side, it kind of applies to all of 5 main product portfolios for AI like we talk with built on JedAI. And those -- that application in EDA and SDA is not node dependent because that can be in all nodes. And even the horizontal AI application, I think, it started advanced node, but it will go to mainstream nodes because of Edge computing. The Edge computing doesn't need to be at 3-nanometer or 2-nanometer.

But especially the vertical ones when we put AI in Virtuoso or Allegro or verification, these are -- these can be applied across all the nodes, not just advanced nodes. And the potential -- the benefits that there we are seeing are independent of the nodes. And just to give you an example, like in some cases, it depends on the design. And in some cases, we can get like 10%, 15% PPA benefit from AI.

And if you go from one node to another node, sometimes the benefit is 10% to 15%. So you're getting as much benefit from better AI algorithms as moving to a new node. So it also improves the efficiency at your current node. But I think in terms of applying it to EDA and system design and analysis, that will be node independent.

John Marco Conti

Got it. Just a follow-up on hardware. Could you talk about the renewal cycle for hardware? When do you expect the customers to come back and buy newer generation of Protium or Palladium? and is it in line with the EDA cycle, i.e., 2.5 to three years?

Anirudh Devgan

Yes, that's a good question. So typically, what we see hardware purchases are tied to their design cycles and sizes of the chips, not necessarily tied to our -- when we introduce our new systems. That can help when we go from Z1 to Z2. But typically, any big customer will typically buy hardware every year, because of they're doing more designs or the design size is going up. If the design size goes up, then they need to consume more hardware.

So that's why this hardware business has been so good over the last four, five years because it's not a one and done kind of purchase. You need to buy multiple times as you do bigger and bigger chips and more chips. So -- and then we have multiple lines now, palladium and Protium, so it makes it much less cyclical that you would expect or what it was like a few years ago.

John Wall

Yes, there's a secular demand for hardware capacity or emulation capacity that's -- and it's a pipeline business. So unlike our software business where you know when the existing contract expires and you know the renewal opportunity comes up, typically, we see -- in the past, we used to see a hardware customer come back and purchase a new system, maybe 2.5, three years after the first system, but now they're purchasing new systems regularly because like I said, demand is really, really strong, and there's just a secular trend in that demand for growth in emulation capacity.

John Marco Conti

Yes. Thank you.

Operator

Our final question comes from Joshua Tilton with Wolfe Research.

Joshua Tilton

Hey, guys. Thanks for sneaking my two questions at here. The first one, I want to go back to the full-year guide. I know that you guys mentioned that you expect, I think, hardware to grow 15% year-over-year in the second-half. But what about compared to the first half? Are you still baking in any conservatism in the hardware guide relative to the strength in hardware that you saw in the first-half of the year?

John Wall

So Josh, just to correct that, the expectation for 15% growth in the second-half is comparing second-half versus second half or total revenue at Cadence, all our businesses. We expect all of our businesses to grow by double-digits in the second-half, compared to the second half last year. Some are stronger than others.

In relation to hardware specifically, we've seen strong demand pretty consistently for a couple of years now. In Q1, we ramped up production capacity because lead times had gone over six months. We tried to address those, and it certainly helped with improved demand in the second quarter, and we're seeing more opportunities in the pipeline for the second-half. Now that gives us comfort to raise the guide for the second half of the year. But it is a pipeline business.

Typically, people only put their opportunities into the pipeline about a quarter before they intend to purchase. In some cases, you might get visibility four to six months in, but it's a pipeline business. Right now, we have strong backlog for hardware, and that's spread into our decision to increase our guide for the second half, but there is potential for that for that demand to continue. And if we're going to keep production capacity at the levels we're at right now because we expect that demand to continue into the future.

Joshua Tilton

Thanks for clarifying that. And then just my second one, in the prepared remarks, I think you guys used the word full flow in Cerebrus in the same sentence. Are you seeing a dynamic or maybe your AI tools are driving more of your customers to choose to go full flow with Cadence?

Anirudh Devgan

Yes, absolutely. That's a very good observation. I mean one trend that is driving full flow adoption is advanced nodes when you move from seven to five to three to two. So that -- because things get more complicated and this -- and we architected this years ago to be full flow. So that helps -- and secondly, that helps is AI orchestration or Cerebrus because it can give better results if the AI engine is able to optimize all aspects of the flow.

So there is a natural tie-in to that, which is great for us. This is a fabulous thing and the same thing in other areas as well, like verification. But both these trends are accelerating the deployment of full flow, advanced nodes and AI.

Joshua Tilton

Thank you very much.

Operator

I will now turn the call back over to Anirudh Devgan for closing remarks.

Anirudh Devgan

Yes. Thank you all for joining us this afternoon. It's an exciting time for Cadence with strong business momentum and growing opportunities in the semiconductor and systems industry. With a world-class employee base, we continue to deliver on our innovation road map and delight our customers and partners. On behalf of our Board of Directors, we thank our customers partners and investors for their continued trust and confidence in Cadence.

Operator

Thank you for participating in today's Cadence Second Quarter 2023 Earnings Conference Call. This concludes today's call, and you may now disconnect.