Jason Child

Thank you, Rene. Q2 has continued our strong start for fiscal year '25. Total revenue was $844 million, which was above the top end of our guided range. Royalty revenue was $514 million, which grew 23% year-over-year and matched our highest royalty revenue quarter to date. Our Q2 royalty revenue growth was driven by continued Armv9 adoption and the start of CSS deployments.
As with last quarter, royalty revenue from smartphone significantly outperformed smartphone shipments. Smartphone royalties increased approximately 40% year-over-year compared with mid-single-digit increase in the number of smartphones sold, mainly due to smartphone application processors being increasingly Armv9 based with a higher royalty rate. In addition, we continue to gain share in automotive applications and with cloud service providers. However, this growth is partially offset by continued weakness in industrial, given the ongoing inventory correction in that part of the semiconductor industry, as indicated last quarter, and widely reported by many of our semiconductor customers.
Licensing revenue declined 15% year-over-year to $330 million, which was better than our expectations, which was for a 25% decline. License revenue varies quarter-to-quarter due to the normal fluctuations in timing and the size of multiple high-value license agreements and contributions from backlog. Because of this, we recommend that you look at annualized contract value, or ACV, to best understand the underlying license growth rate. ACV in Q2 was up 13% year-over-year, which is consistent with recent quarters. Remaining performance obligations, or RPO, was up 10% sequentially as we had a very strong bookings quarter. Some of this RPO will be recognized as revenue later this year.
Turning now to guidance, I will briefly touch on both third quarter and fiscal year ending March 31, 2025. This guidance reflects our current view of our end markets and our licensing pipeline. For Q3, we expect revenue between $920 million and $970 million, which at the midpoint, represents revenue growth of 15% year-over-year. Investment in our next generation of technologies are on track, and we expect our non-GAAP operating expense to be around $525 million. We expect our non-GAAP EPS to be between $0.32 and $0.36.
Looking out to fiscal year '25, we are reiterating our guidance for revenue, cost and profit. We expect revenue to be between $3.8 billion and $4.1 billion, which represents an 18% to 27% year-over-year increase. At the midpoint of our revenue guidance, this includes full year royalty revenue growth in high teens. We expect that our revenue growth from smartphones will continue to be driven by Armv9 based chips becoming a greater proportion of the mix, with CSS ramping over the next couple of quarters.
We also expect to continue to gain share in cloud and automotive. Feedback from our customers leads us to expect sequential growth in networking in both Q3 and Q4, while IoT is not expected to recover until next year. Strong demand for our latest technologies will continue to drive license revenue for the rest of the year. We have kept the range for full year revenue guidance the same as last quarter at plus or minus $150 million as we have some large licensing deals in funnel.
Although the timing of these deals and the shape of the revenue recognition is not yet clear, we do expect all these deals to close. We do expect non-GAAP operating expenses to be approximately $2.05 billion, which represents a 19% year-over-year increase and is unchanged from our prior guidance. As we continue to invest in R&D to support future growth initiatives, we expect operating expenses to ramp consistently through the year. We expect our full year non-GAAP EPS guidance of between $1.45 and $1.65.
With that, I'll turn the call back to the operator for the Q&A portion of the call.

Question and Answer Session

Operator

(Operator Instructions) Andrew Gardiner, Citi.

Andrew Gardiner

Rene, Jason, I was hoping you could shed a little light on a topic that you didn't touch on in your prepared remarks. There have been myriad press reports regarding your working relationship and litigation with Qualcomm. I believe the case is still due in court next month, and you have reportedly canceled Qualcomm's architecture license recently both, I think, as they plan to ship a greater volume of product based on the Nuvia designs and also ahead of the court case. Can you comment on the veracity of these reports.
And I suppose, specifically on numbers, are there any issue with your revenue recognition and operating expenses into the coming periods related to these actions? Do you have to reduce rev rec, will OpEx rise on increased legal costs? I know it's tricky to talk about legal things, but if you can shed any light on that, I'm sure it would be helpful.

Rene Haas

Sure. Yes, happy to. So I'll address what I can as it is an ongoing litigation. There isn't a great deal I can say on it. But at a base level, contractual consent was required by Qualcomm to sign a Nuvia license, and that consent was not obtained. As a result of not obtaining that consent, they are in breach. And what we did was sent a notification letter regarding cancellation of the architectural license. And to be clear to your question, we have not canceled the license, but we have sent a notification to them.
Now getting consent for an assignment is fundamental for our license agreements. And as a result, we need to ensure fairness and protecting Arm ecosystem who rely on these license agreements. On the financials, I'll let Jason maybe chat on the OpEx. But regarding the revenue, our forecast and guidance has always taken into consideration that we may not prevail in this case so we have essentially taken that forward look. So Jason, i don't know if there's any comment you want to make on the expense side?

Jason Child

No, there isn't any change in, I think, some of the changes you pointed out, rev-rec expenses like that. No, there's no changes because at this point, as Rene said, and as I think we said actually back at IPO and consistent since then, is our forecast assumes that we're going to get paid existing ALA royalty rates. So until something changes, there won't be any increase or change in those rates.

Operator

Harlan Sur, JPMorgan.

Harlan Sur

So now that we're midway through your fiscal year, the upside in your business through the first half been coming from Licensing, right? This is, in my view, the best forward indicator of your pipeline, your growth prospects, and it looks like licensing for the full year will come in better versus your view even 90 days ago. So this quarter, backlog was up percent sequentially; book-to-bill, 1:7, so very strong. I assume you're driving more value uplift per renewal. It looks like you've seen some add-on licensing activity.
As you mentioned, Rene, CSS engagements are strong. Bookings were originally expected to be lower this year versus last year. But just given the strong design activity by your customers, requirements for more compute capability per program, visibility on renewals for the remainder of the year, can the team grow its backlog for this fiscal year?

Rene Haas

Yes. I would say, a couple of things that we're observing in the marketplace. We've talked about the demand for Arm technology being quite strong given the ecosystem and the overall increased demand for Arm technology. I think when we look at what's going on with AI and when you think about AI, it's not just training in the data center, but it's inference in the data center, it's inference across different parts of the overall value chain, the network, the automobile, the PC, the mobile phone, the wearable, which can be kind of what people would call the edge. We're seeing an increased demand for compute resources to run these agents and run these small language models or large language models on top of compute requirements that they already have.
So what that's driving is, I think, an increase for us that we're seeing across the board for R&D and innovation to capture this platform opportunity. So we're seeing pretty broad licensing demand across candidly, all the markets and all the sectors. And you're right, it's stronger than we had, I think, originally communicated and anticipated. It's a very good forward indicator for the strength of the business and also for the strength of the royalties going forward.
In addition, as mentioned, we've doubled the number of CSS licenses now over the year. That's also been, I think, stronger than we anticipated. We did, as I mentioned in the opening remarks, announced MediaTek's first chip design using CSS. So I think it's a combination of increased compute demand, AI and also the CSS. I don't know, Jason, if there's anything you want to add on to that?

Jason Child

Yes. I think you asked, Harlan, if there's a way to increase our backlog. I would say, certainly, based on the strength that we've seen, yes. In fact, if you look at the license revenue kind of implied guidance, I think we're up about 45% higher this year than what was our plan back at IPO. So we have seen strength. Now our forecast now is I don't think there probably will be significant increase to backlog. But again, hard to say.
The large deals that we have coming later this year, there's some impact there. But then, of course, that's going to be offset by some kind of the amortization or recognition of milestones that will be delivered in the next quarter or so. So overall, I wouldn't count on that being the driver. I think the real focus is really the royalty growth and the 23% year-on-year growth that we saw in the quarter, which we're particularly excited about.

Operator

Ross Seymore, Deutsche Bank.

Ross Seymore

Going to stick on the licensing side of things. Rene and Jason, you guys talked about, I think, signing six new ATAs and then I think, more than doubled the CSS side of things. Is there a TAM that is larger now? I think originally, you talked about ATAs having somewhere around 50 potential licensees. It seems like that number must be growing if you're upwards of 40 already.
So how do we think about the potential future growth there? And if that has to slow eventually before it hands over to being such a good precursor for the royalty side of things in the future?

Rene Haas

Well, I think the way to maybe think about it is one of the opportunities we have with ATA is to expand the portfolio and size, if you will. ATA's grant access to Arm technology, in a broad sense, in other words, a broad portfolio of IP and also a broad set of rights in terms of how many chips to build and tape out in a year. We can scale that to different variations, i.e., smaller set of IP and/or reduce number of tape-outs, but still give additional value to end customers because they get a larger suite than they might on a single instance. In theory, I think the majority of our customer base can move to some version of ATA.
And the reason for that is customers love the concept because it addresses a number of things. They know they're going to use the Arm technology. Number two, having their engineers have access to the broad set of IP allows them to do a lot of experimentation and evaluation in a very easy way. And thirdly, they can now essentially fix their costs on an R&D forward cycle since they know they're going to be purchasing the IP anyway.
So what that ends up meaning for Arm is it's pretty much broader upside because less churn on deals because they candidly are more repeatable. And also, by having a broader set of IP available, what we find is engineers end up using more. So to your question, I think the vast majority of our customer base can ultimately go to some version of ATA. And given the broad set of IP that's used in ATA, I think that's going to drive higher royalties in the future.

Jason Child

Yes. Maybe just a couple of things to add. I think in the past, we had said that we probably about 80% of our think of our license revenue and license fees as probably ATA working well for them. I think we said in this last quarter, we're now kind of in the over 50% range. The one thing I think we've said in the past, just to make sure it's clear is with ATA, the annual costs or price increase is roughly 7% per year.
And then those contracts, on average, are somewhere around three-ish years. And so even once everyone is an ATA, you're still going to have renewals and you're still going to have the annual increases. And then of course, as we deploy and release new technologies and expand the product offering, then of course, you can also upsize as well. So I think we're quite a ways from reaching the TAM, and there's multiple kind of points of growth to expand that TAM.

Operator

Vivek Arya, Bank of America Securities.

Vivek Arya

A few questions on the royalty business. I think you have changed this year's contribution to, I think, high teens growth from low 20s and then mid-20s and I imagine that's all the cycle, but if there's any other color, I would appreciate that.
But the other kind of related question is v9 contribution to royalties kind of stalled at 25%. I thought the plan was to continue to expand that every quarter by 5 points. And since smartphone is kind of the biggest contributor of that and your smartphone business is growing very nicely, like over 2 times the pace, I would have thought the conversion would go on. So I was just hoping you could give more color on why did v9 conversion kind of stall in this quarter? And then the overall royalty growth, what is the need to take it down?

Rene Haas

Yes. So let me take the first part of that question, Vivek. I'll address both the v9 transition, and then I want to make an important point regarding how to think about pricing inside the v9 envelope and then let Jason talk about the specifics. But first off, the adoption of v9 is going very, very well. We're seeing very, very strong uptick of it in mobile. All of our NeoVerse products are v9 and we're now starting to see the transition of it in Automotive and IoT.
And I mentioned in the opening comments that we've now seen the first shipments from Apple that are v9. So we're very, very happy with the rate of adoption. And I think if you looked at the adoption rates on a quarter-to-quarter basis, it may not look linear, but it will definitely be increasing quarter-over-quarter-over-quarter when we start looking at it on a multi-quarter basis. That being said, I think there's a very important distinction that I want to make regarding how to think about royalties during the time that the v9 architecture is ramping. In contrast to version 7 and version 8 which, by the way, by way of reference, those architectures tend to have a life of approximately 10 years in terms of their peak run rate.
During version 7 and version 8, once the royalty rates were fixed for that version, there was very little delta throughout the period, meaning that as those versions reached maturity in terms of saturation, the royalty growth would asymptote. That's not going to be the case with v9 and there's two primary reasons for that. Reason one is that generation on generation, we introduced multi year improvements in the technology and multiyear improvements in terms of the product, meaning that when we deliver a product, let's say, for a phone that goes into production in 2025, and then we delivered the v9 version for a phone that goes into production in 2026, because the product is better, 15% better, let's say, on performance and power, we're able to drive better value-based pricing for that solution.
So during v9, even though the royalty rates are increasing from version 8 to Version9, we're going to see continued increases throughout the life of version 9 even as version 9 adoption increases. Secondly, in version 9, we've also introduced CSS, which we've mentioned several times, carries a higher royalty rate, in some cases, double, if not more, of what a standard version 9 core would be. So as a result, as we see growth in version 9 adoption, the royalty growth track higher than it has traditionally. And again, it's for those 2 factors. Number one, the value-based pricing that sees an increase because of the better economics delivered. And secondly, more of a transition to CSS.
With that, Jason, if you have any comments, I think maybe we answered the extent of that question.

Operator

(Operator Instructions) Srini Pajjuri, Raymond James.

Srinivas Pajjuri

My question is on the networking and data center business. Rene, I think that accounted for about 10% of your mix in the last 12 months. Obviously, a lot of momentum on the data center front and then you have CSS products potentially ramping at some point in the next 12 months. So I'm just trying to understand how important of a driver that could be for the next 12 months, both from, I guess, a royalty standpoint and also from a licensing standpoint?

Rene Haas

Yes. So let me talk about the macro, and I'll let Jason address the numbers piece. I think the adoption of NeoVerse in the network and data center is going to mean very, very strong trajectory for Arm, not just in the 12 months, but over the next number of years. Two components to that. For general-purpose compute, we now obviously have had Graviton in production for many years.
And with Microsoft Azure on Cobalt and Google GCP with Axion, general purpose compute, they are now general availability, which means now instances can be purchased by end users. And based upon the type of deployment we've seen with Graviton, we're very excited about the opportunity there, both with Azure and GCP on these chips. So for general-purpose compute, we're expecting to see very, very good results, driven primarily by the fact that the power and efficiency is anywhere between 50% to 60% better than compared to x86. In addition, we are seeing, as I mentioned in the opening, demand for NVIDIA's advanced training and inference chip, Grace Blackwell, which uses the Arm CPU Grace as part of that overall solution.
One of the benefits of that solution being introduced in the data center is that many of the base OS and workloads that are required for a general-purpose cluster are also used on this AI cluster, meaning there is a lot of leverage between using Arm in the data center for general purpose compute and using it for an AI training or inference center. There's a lot of software reuse. So as a result, we believe that from an overall TCO standpoint, that will accelerate adoption of Arm to data center, which we're very excited about. So I'll let Jason talk about the numbers and the mix.

Jason Child

Yes, in terms of the data infrastructure and data center market, we do expect networking to continue to be slow. But the data center side, and specifically the cloud compute market for all the reasons that Rene just mentioned, it has been strong. We do expect it to accelerate throughout the back half of the year as the deployments of both Cobalt and Axion continue to ramp as well as some of the other custom silicon chips from other makers that are also continuing to ramp. So we do expect that to accelerate.

Operator

Lee Simpson, Morgan Stanley.

Lee Simpson

I just wanted to go back to licensing again. And really just trying to understand what contributed in particular to that better-than-expected licensing here. And maybe how does this relate to the RPO number? It looks as though it's down a couple of percentage points year-on-year.
So how do we think about the growth going through the rest of this year? Do we still think we can track to a $1.7 billion number? And have we perhaps pulled forward some of the licensing deals into this quarter?

Rene Haas

Thanks, Lee. So first, I would say, no, we don't foresee any sort of pull forwards. In terms of the upside, I think it was about $35 million higher than the guide, which I think we have guided to a 25% decrease. We actually had a 15% decrease. And again, a decrease is mostly because of a tough comp with a couple of really large deals that were lapped a year ago.
In terms of the reduction year-on-year in RPO, well, we happen to have the all-time record a year ago, and we actually have the second highest quarter was this quarter, because it actually stepped up sequentially 10%. So it's actually a pretty nice increase. And as you know, earlier this year, we had a fairly large amount of RPO that effectively amortized into revenue from a large deal signed last year. So overall, I think the RPO trend has been very, very healthy.
But yes, I wouldn't expect that I think the range that you indicated for end of the year is still probably a reasonable range. And so I think other than what we've included in the guidance on the deals that we expect to close in Q4. At this point, we don't have line of sight to any significant deals beyond those. Obviously, if something changes, we'll let you guys know.

Operator

[Yanko Venta], Arete Research.

Yanko Venta

I just want to get back to the announcement of MediaTek’ and smartphones. How do you see the adoption of CSS in smartphones and trending through this year into next year? Could we see 50% of the market in the next few years?

Rene Haas

Yes. Thank you for the question. I think so because when we think about what the value proposition of CSS, it's really about reducing time to market and increasing overall confidence in the design and performance. One of the hallmarks of the mobile phone market, as I'm sure you know, is the fact that the product cycle is rather relentless. They're on an annual cadence.
It's not very forgiving in terms of when units need to be available, whether it's aligning with MWC or Single's Day in China. So with a very relentless product cycle, combined with these new smartphone chips are becoming more and more complex, these application processors. And then they're also being built in the most advanced geometries at the fabs which have longer manufacturing cycle times. Anything you can do to improve the time it takes the design a chip is welcomed.
So when we introduced CSS, we had a bit of skepticism, whether it would be applicable for the mobile phone market, but what we've actually seen is very, very good adoption. So in summary, yes, can we get to 50% of the market? I believe so because there's real value being delivered on a product that can move months off the development time in what typically is a very, very severe and relentless product cycle.

Operator

Krish Sankar, TD Cowen.

Sreekrishnan Sankarnarayanan

I had a quick question on China. It seems like the smartphone mix is shifting towards the entity there in the Android market. Are you seeing any meaningful impact to your China royalties? And also, what is the penetration of v9 in China today?

Rene Haas

Yes. Thank you for the question. We've definitely seen strength in the overall handset numbers in China due to the reasons that you had just mentioned. I think the local brands there, Xiaomi, Oppo, Vivo are all seeing very, very good growth. Our team just participated in an Oppo product launch last month.
So number one, growth has been strong there. I think there is an allegiance towards the local brands. And in the premium flagship market, as we get into the next year or so, that's all version 9, and that's also going to find its way into the midrange. So China Android yes, strong. And also, the vast majority of the products now are going to be moving to version 9. The high end already is all there.

Operator

John DiFucci, Guggenheim Securities.

John DiFucci

Rene, thanks for that explanation you gave on the v9 mix and the related variables that are going to affect royalties going forward. That was really helpful to think about some of that. But shouldn't we expect to continue to see that v9 mix grow? And if not, what would cause it not to grow? Is it just, I don't know, a temporary macro issue?
AI, you guys are obviously benefiting. NVIDIA, there's some others out there, the cloud vendors, but we really haven't seen it like in a lot of other places. And I'm just curious, could there be, at least, people just stepping back and let's just catch up in our minds like to see how this is going? Or I don't know, I'm just trying to figure out how to model this going forward. Like what's sort of implied in guidance?

Rene Haas

Thank you for the question. No, it's actually going to be quite predictable. As we said, it was 10% a year ago, 25%. Now I can guarantee you, it will not be 25% a year from now. It will be considerably higher than that. So if you step in and say, okay, well, what is really driving that? I think you're going to see more version 9 in PCs as more Windows and Arm products are introduced in the market, as more vendors introduce chips.
That's a given, number one. Number two, what you will see in the mobile phone space is a very classic waterfall of what existed in the flagship will find its way in the midrange and what was in the midrange, will find it's way in the low end. That's a very natural evolution of products. It happens every single time. It's going to happen this time.
So a year from now, I would expect now that the majority of the mobile phones will be version 9. Everything we do in the data center, everything we do in the data center is version 9. So what will really be the slowest of the products to come to market will be the automotive sector, and that's really driven by just the time lag it takes for those products to be introduced, particularly around IVI and ADAS. But we know the pipeline. We've got CSS activity in particular there.
So when we have this call a year from now, hopefully, again, and then we'll look at the numbers. But I'll guarantee you, we'll be way ahead of 25% for the reasons I gave.

Jason Child

John, just to add on modeling complexity of v9 penetration. So if you look specifically at the most recent quarter, where it was flat quarter-on-quarter, the primary reason why was because the mid-market had higher growth in the quarter in smartphones, the mid market doesn't have v9 yet, it hasn't waterfall yet to be v9. And so as a result, it looks like v9 installed and it really has more to do with the kind of where the growth in the market turned out to be in this most recent quarter. So over time, that stuff all shakes out, as Rene said. But you will have some maybe trends like we saw in this last quarter to this quarter, but I don't think going to sustain.

Rene Haas

We tend to have Arm's products anywhere between a 3- to 5-year visibility on when they how up. And that's because the licenses have been done years ago. And we have very good line of sight with the chip vendors in terms of when they're going to increase the products, and we have very good line of sight with the OEMs when they plan to take it.
So don't read too much in terms of the quarter-to-quarter thing. We have extremely good visibility in terms of where they're going to land and the trend is only going forward.

Operator

Toshiya Hari, Goldman Sachs.

Toshiya Hari

Rene, I was hoping to get your thoughts, your forward outlook on the PC and smartphone markets. AI PC so far have been relatively underwhelming relative to expectations at the beginning of the year or maybe at Computex. AI really hasn't catalyzed a replacement cycle in smartphones either.
I know you guys are not so dependent on units because you've got things like v9 and CSS growing your royalty rate. But given your visibility that you just spoke to in terms of what your customers are working on, when do you expect volumes across pieces and smartphones to grow? Is it a '25 dynamic? Is it more of a '26 dynamic? And I think at one of the conferences, you spoke to a 50% market share goal in PCs over the long run. Is that still the aspirational goal for you guys?

Rene Haas

Yes. Sure. Answering your last question first, absolutely, and I stand by that prediction. When we think about the Windows ecosystem, it's really all about filling out a broad lineup of SKUs. If you go into a retailer today, you'll see that there's thin-and-light machines, low-end machines, mid-price machines, gaming laptops, et cetera, et cetera. I think what we're going to need to see, and we will see it because we have great line of sight in terms of people working on multiple chips in multiple areas, is a filling out of that SKUs.
So I think when you start to see thin and light devices at an entry price point based on Arm with a different chipset, you'll see growth. When you see a gaming laptop that has Arm inside, which by the way, will be compelling if and when that comes out, I think you'll see growth. So we're pretty confident in terms of the next couple of years in terms of growth in the PC market.
Now whether that's related to AI demand, I think what's important is that these PCs are equipped with the horsepower required to run the AI applications. I think it's early days in terms of the definition of an AI PC relative to the value prop. But I wouldn't get so hung up on that. What we typically see with these types of cycles is that you've got more technology in there, then you need to take advantage of the applications and then it flips the other way around. So this is being largely future-proofed. But broadly speaking, yes, we're pretty bullish about the demand, and I definitely stick to my prediction on growth of Windows on Arm.

Operator

I will now hand the call back to Rene Haas, CEO, for closing remarks.

Rene Haas

Thank you. And thank you, everyone, for the questions and the interest in what we're doing. Again, to summarize, now our fifth conference call since we've been a public company, and we've been consistently beating the expectations of what we have told you that we were going to do the prior quarter. I think more broadly, though, what you're seeing with Arm and our strategies, as I said in the opening, really coming to life. We've talked many, many times about the resiliency of our business, the growth of v9, the increased royalty rates.
And when you see a royalty growth in a market where units, for example, in smartphones are up 4%, and we're up 40%, that's a great proof point that the strategy we've put in place are working. The future is incredibly bright. For all the questions that came up regarding automotive, the data center, AI, smartphones, PCs, we're very fortunate to be able to talk about exciting opportunities in all those verticals, all of them that are using Arm and all of them are going to grow in the future. So thank you very much, and I appreciate all the questions and comments.

Operator

Thank you. This concludes today's conference call. Thank you for participating. You may now disconnect.