Arm Holdings PLC

NASDAQ:ARM

Good day, and thank you for standing by. Welcome to the Arm Third Quarter Fiscal Year 2025 Webcast and Conference Call. [Operator Instructions] Please be advised that today's conference is being recorded.

I would now like to turn the conference over to your first speaker today, Jeff Kvaal, Head of Investor Relations. Please go ahead, sir.

Thank you, and welcome to our earnings conference call for the third quarter of fiscal '25, which ended December 31, 2024. On the call today are Rene Haas, Arm's Chief Executive Officer; and Jason Child, Arm's Chief Financial Officer.

During the call, Arm will discuss forecasts, targets, and other forward-looking information regarding the company and its financial results. While these statements represent our best current judgment about future results and performance as of today, our actual results are subject to many risks and uncertainties that could cause actual results to differ materially.

In addition to any risks that we highlight during the call, important risk factors that may affect our future results and performance are described in our registration statement on Form 20-F filed with the SEC. Arm assumes no obligation to update any forward-looking statements.

We will refer to non-GAAP financial measures during the discussion. Reconciliations of certain of these non-GAAP financial measures to their most directly comparable GAAP financial measures as well as a discussion of certain projected non-GAAP financial measures that we are not able to reconcile without unreasonable efforts and supplemental financial information can be found in our shareholder letter. The shareholder letter and other earnings-related materials are all available on our website at investors.arm.com.

And with that, I'll turn the call over to Rene. Rene?

Thank you, Jeff, and good afternoon, everyone. AI demand continues to drive strong momentum for the Arm ecosystem, which is built on the world's most pervasive compute platform. We're pleased to report record total revenue and an all-time record royalty revenue for the third quarter of fiscal 2025. Total revenue grew 19% year-on-year to an all-time high that exceeded the high end of guidance.

Our royalty revenue grew 23% year-on-year to a new record. This was driven by adoption and deployment of v9 and strong demand for Arm's CSS. Licensing remains strong as our partners make long-term commitments to more of our advanced technology to take AI everywhere. AI growth requires significantly more compute across all of our end markets from smartphones with better chat features, to autos with better driving and parking assist, to IoT microprocessors with embedded NPUs. This is driving continued adoption of our v9 and CSS technologies.

New flagship smartphones from OPPO and vivo take advantage of MediaTek's new Dimensity 9400 system-on-chip. This SoC is based on our CSS for Client, which includes the Arm Cortex-X925 CPU and the Immortalis-G925 GPU. Increased chip complexity is driving the top hyperscalers to customize silicon on the latest Armv9 and CSS. We are gaining share in the data center with AWS Graviton, Microsoft Cobalt, Google Axion and NVIDIA's Grace Arm-based chips.

AWS recently announced more than 50% of new CPU capacity installed over the past 2 years was on Graviton. Over 90% of AWS' top 1,000 EC2 customers use Graviton. We have more than 20 million developers, the world's largest developer community, and we continue to increase investment in our ecosystem. NVIDIA also announced Project DIGITS, which combines the Arm-based Grace CPU and the Blackwell GPU into the new GB10 Superchip to powers the world's smallest AI supercomputer.

The Grace CPU is based upon Arm's CSS for Client and includes 20 power-efficient Arm cores, 10 Cortex-X925 and 10 Cortex-A725. With Project DIGITS, developers, data scientists and AI researchers will be able to more easily able to build inference models before deploying them into the cloud.

I'd also like to call out 2 key projects that further cement Arm's position at the very center of the AI ecosystem. Along with SoftBank Group and OpenAI, Stargate for AI infrastructure deployment, and advanced enterprise AI called Cristal intelligence to develop AI agents for knowledge work. We strongly believe that the advances in AI, both for training and inference, are going to increase the demand for compute in the AI cloud. We expect Arm's solutions to address the needs from the cloud to the edge to power growth in the world's most popular compute ecosystem for decades to come. And with that, I will hand over to Jason.

Thank you, Rene. Q3 was another record quarter as we continued to deliver strong growth. Total revenue was $983 million, which was above the top end of our guided range. Royalty revenue was a record $580 million and grew 23% year-on-year and was above our expectations. This growth was driven by continued Armv9 adoption and initial shipments of chips based on our compute subsystems as well as the growth in revenues from custom silicon going into the data center.

Royalty revenue from chips for smartphones, the data center, networking equipment and automotive were all within expectations, while royalty revenue from IoT showed signs of recovery after multiple quarters of weakness. Based on the most recent royalty reports, revenues from smartphones continue to grow much faster than the overall smartphone market. This was helped by chips based on both Armv9 and CSS such as MediaTek's Dimensity 9400, which is being deployed in flagship smartphones from both OPPO and vivo.

Licensing revenue increased 14% year-on-year to $403 million, which was better than we forecasted. License revenue varies quarter-to-quarter due to normal fluctuations in timing and size of multiple high-value license agreements and contributions from backlog. As always, we recommend that you look at annualized contract value, or ACV, to best understand the underlying licensing growth rate. ACV in Q3 was up 9% year-on-year, which was a little lower than the recent run rate of low teens but is above our long-term plan.

Remaining performance obligations, or RPO, was down slightly sequentially as Arm delivered products that released revenue from backlog into the P&L. As you know, Arm's revenues today come from technology developed years or even decades ago, and our costs today are investments for future revenue streams. To maximize our future revenue opportunity, we want to maximize our investment in R&D today.

And in the third quarter, heightened R&D spending led our non-GAAP operating costs to their highest level at $522 million, which was in line with our expectations. At the same time, we also delivered near-record levels of non-GAAP operating profit at $442 million. Going forward, we will continue to balance increasing investments for the long-term growth of the business and near-term profitability.

Turning now to guidance, I'll briefly touch on both fourth quarter and fiscal year ending March 31, 2025. This guidance reflects our current view of our end markets and our licensing pipeline. For Q4, we expect revenue of between $1.175 billion and $1.275 billion. At the midpoint, this represents revenue growth of 32% year-on-year. We have left the revenue guidance range slightly wider than in prior quarters as we have some large license deals in play.

Although we have high confidence of deal closure, the timing of deals could be hard to forecast and some may slip into the next fiscal year. As previously mentioned, revenue growth today enables us to increase our investments in the R&D essential for our long-term success. We are accelerating the investments in our next generation of technologies.

We now expect our Q4 non-GAAP operating expense to be approximately $590 million. And for Q4, we expect non-GAAP EPS to be in the range of $0.48 to $0.56. For fiscal year '25, we are increasing the midpoint of full year guidance to around $4 billion. This midpoint represents about a 24% year-on-year growth, which is ahead of our long-term target of 20%. With this, we expect full year royalty revenue growth rate in the high teens year-on-year, which is consistent with our previous guidance.

We expect our full year license revenue to grow around 30% year-on-year. We expect non-GAAP operating expenses to be about $2.1 billion, which represents a 21% year-on-year increase. We therefore expect our full year non-GAAP EPS to be between $1.56 and $1.64.

With that, I will turn the call back to the operator for the Q&A portion of the call.

[Operator Instructions] And your first question comes from the line of Lee Simpson from Morgan Stanley.

Maybe I could just kick off by asking about where we see the strength in licensing as we go into the subsequent year. And I guess I'm really asking this really as it relates to some of the recent news flow that we've seen in the broader AI landscape but also really touches on Arm's capability.

So if we look 2 weeks back, we had the announcement of Stargate in the U.S. with budgets ranging over $100 billion. And in the last day or so, we've heard of your new joint development between SoftBank and OpenAI called Cristal intelligence. And we're just trying to understand where it seems Arm sits here because interestingly, it looks like both an enabler and even having first-mover advantage as a kind of AI adopter in an era of agentic AI. So maybe if I take a step back, maybe just trying to help us, can you outline just more clearly what the opportunity set is for Arm, including the sell-in of new products to these new and large projects? And what the long-term benefit is to Arm in the earnings perspective?

Thank you for the question. There's a lot there so I'll try to simplify a bit. So Stargate, a project that was announced a few weeks back, which is an extremely significant infrastructure project in the United States, where $100 billion will be invested immediately and $500 billion over time. And this is a partnership with OpenAI and Oracle and SoftBank, technology partners being those companies in addition to ourselves, Microsoft and NVIDIA.

For Arm, we are extremely excited to be the CPU of choice for such a platform. Combined with the Blackwell CPU with Grace, Arm will be the CPU of choice for the initial configurations. And going forward, there'll be huge potential for technology innovation around that space. So incredibly exciting project, which we think will be transformational for the industry.

Cristal intelligence, which was discussed earlier this week, is really about agentic AI and agents moving across every node of the hardware ecosystem. So if you think about the smallest devices such as earbuds all the way to the data center, this is really about agents increasingly being the interface and/or the driver of everything that drives AI inside the device.

For Arm, it's a significant opportunity because AI workloads will run on every one of those endpoints that I mentioned. Additionally, given that Arm is the world's most pervasive compute platform, those AI workloads will run on Arm and through Arm. And through our KleidiAI libraries, we will make it very easy for developers to target and optimize to the Arm platform running these agents. So both of those announcements are very significant, both in terms of their impact to the industry and represent significant opportunity for our company.

And your other question on kind of the licensing upside components. So in the quarter, it was up about $27 million versus our kind of guidance are up 14% year-on-year. Certainly, the drivers have been the same components over the past few quarters, which is AI and the need to continue to access v9 technology to prepare for all of the needs that go into AI chips, and then, of course, also CSS. As Rene mentioned, there's additional CSS contracts that we've sold and we continue to be working on.

In terms of the guidance for this next quarter, we're actually guiding that license revenue can grow in the order -- kind of in the order of around 60% year-on-year. And so we've said since early in the year, we have a number of big deals that we had planned on for Q4. Those look to be kind of mostly on track and as expected. And those deals very much are in the same kind of bucket of very much being AI- and CSS-driven.

Your next question comes from the line of Joe Quatrochi from Wells Fargo.

You mentioned that the partner demand for CSS is stronger than initially anticipated. I guess how should we think about the contribution to royalty revenue ramping as we enter fiscal '26 versus your prior expectations?

Yes, so we're not ready to talk about '26 yet. But in terms of exit rate, we feel very good about the momentum that we've delivered. For example, this last quarter at $500 million of royalty revenue, that's about 13% higher than our previous record of $514 million. Certainly no surprise that, that also is in the first quarter that we actually have material CSS revenue this quarter led mostly by the Dimensity 9400 chip that we talked about. Also Cobalt CSS starting to see shipments and deployments and that's flowing into royalty revenue now. As we continue to see more CSS deployments over the next couple of quarters, we do expect there to be continued momentum. In terms of quantifying the momentum, we're going to have to wait until next quarter.

Your next question comes from the line of Charles Shi from Needham & Company.

Just want to touch upon since we just had a lot of news coverage on the trial last month with one of your large customers. There seems to be, based on the press reporting, there seems to be some difference between what you think the contract will expire versus what they think. They think it's expiring in '28. I do think it's expiring in '25. I know this is all calendar year commentary, but what's the current view there [indiscernible] revenue maybe you're expecting to happen, haven't quantified maybe that won't happen in the next calendar year. I just want to get some thoughts there.

This is Jason. I think the question was about the impact of the Qualcomm lawsuit and on revenue.

Correct, correct. Yes.

There's no impact. We had forecasted really all the way back at IPO and continue to forecast as though we were not going to prevail in that lawsuit. The primary reason for the lawsuit very much was around defending our IP and that's important. But from a financial perspective, we had assumed that we'll continue to be receiving royalties that basically the same rates that they've been paying for -- in the past and will continue to pay.

Got it. Maybe another question, just a little bit of clarification. I don't see an RMB contribution number from the shareholder letter. Maybe I missed it somewhere. But if you have the number, can you tell us what the number is for the past quarter?

Yes. As a percentage of total royalties in the quarter, it was 25%, so consistent with the prior quarter.

Okay. Okay. The 25% looks like it's been a couple of quarters of 25%. Do you expect it to go up again in coming quarters? I know the answer is probably yes, but -- and also, can you also help us understand why it's a little bit stalling for a couple of quarters already at the 25%?

Let me just maybe cover just some of the math piece, and then I'll let Rene kind of talk about the overall kind of larger view of how v9 adoption is occurring and how it will continue to occur. So I would just say, first of all, the math is, it's a percentage of total, and so we saw 23% growth in the quarter so it's flat as a percentage of total.

But if you actually look just at the v9 dollars, the rate went from about 15% a year ago to about 25% this quarter. So the absolute dollars grew by, I would say, triple-digit rate. And so the fact that it's relatively flat actually is a great indication of the fact that you have -- we have a longer runway for future growth. We still expect that you will see v9 grow to probably 67% to 70% of total royalties. And the fact that we're able to kind of meet or exceed our royalty growth rates while it's not been accelerating should provide more confidence about our ability, as that rate goes higher, our ability to drive further royalty growth in the next quarters and beyond.

Yes. And maybe just to give you a sense of how to think about those transitions, they're largely driven by the transitions of OEM products and when they get introduced. So take for example, the MediaTek 9400, which has been designed into OPPO and vivo phones, those are now in their ramp for production. So you'll see a spike up as a ramp and then you get to a steady state.

But as the next versions are released, a couple of things happen. Broader adoption across the high end of the segment, and then the high to mid-range to mid-range products start to move from v8 to v9. So the transition rate that we're seeing is completely expected and very, very consistent with how we expected the overall ramp to be. So very happy about where we are. And as you said, room for expansion but it's largely driven by OEMs shifting their chip mix as opposed to licensing new companies.

[Operator Instructions] And your question comes from the line of Timm Schulze-Melander from Redburn Atlantic.

Maybe just to key off the question on the royalty growth. Can you maybe just talk a little bit more specifically about FY '26 and kind of what momentum and what kind of mix evolution we should expect on v9 to contribute to the full year?

Yes, this is Jason. We're -- we'll provide guidance on '26 next quarter. We typically -- just like last year, we do it when we announce Q4 results. So we'll go into more detail then. I would say what we've said in the past is that we would expect royalties to grow probably somewhere in the mid-20% growth rate. That's what we've said in the past. We haven't -- we're not providing an update to that today.

The way that we're going to get there is through the combination of v9 adoption as well as CSS. And so what you saw in this most recent quarter is v8 actually had some really strong growth. That's great, but overall, you should expect to see v9, over time, continue to grow. But also remember that v9 is just 1 input into growth. CSS adoption is a, I would say, even a larger input to growth because the royalty rate on CSS is roughly double that of v9, and of course, v9 is roughly double the royalty rate of v8. So overall, you have to look at all the components, but the momentum that we just showed in this last quarter and what we expect over the next quarter, we think, sets us up for a very good '26.

Yes. And maybe just to expand further on that and make sure Jason doesn't kill me, but the royalty rate on a CSS of '26 is not necessarily the same as the royalty rate for CSS in '25. Those CSS rates change year-on-year when the new solution is offered to the market. So those are variable over time and they increase over time.

That's great. Super helpful. And then maybe just one quick follow-up. The licensing revenue, clearly it's super hard to predict the exact week or day or month. But is there anything that's sort of structurally changing in terms of the complexity of your client agreements? That means that these are just structurally going to take longer and longer to get over the line. Is that something we should expect to see in coming years?

This is Jason. I would say no. But you need to kind of look at the contracts primarily in 2 buckets. There's brand new contracts with someone we haven't done work with before. That probably takes a while. There's a lot longer approval process versus a renewal of maybe an existing ATA, and that's probably a shorter cycle. But nonetheless, the larger the deal is, the longer the time frame.

Typically, when these get in the hundreds of millions of dollars, they probably involve Boards of Directors and all these different approvals that can just take many, many months. But in general to your question, is anything changing now or next year versus where it's been this last year? No, no. It's all the same factors.

Your next question comes from the line of Vijay Rakesh from Mizuho.

Just a quick question on the -- it looks like on AWS, you're having very good success, almost 90% of the top 1,000 customers using Graviton and the Arm IP. Can you talk to how the Cobalt 100 is progressing, how that's ramping? And I have a follow-up.

Yes. I'll defer to Microsoft making statements about the shipments of Cobalt. To be clear, we've taken the AWS comments directly from AWS. But in general, what we're seeing inside of Microsoft with Cobalt and more broadly with all of our partners is not only increased momentum for deployment of those products into the cloud.

But because of the transition away from an x86 plus H100 to a Grace Blackwell GB200, that is also an accelerant for Arm in the data center because these AI data centers, the primary CPU being Grace, running everything relative from a host control standpoint is a good tailwind or kicker, if you will. So we're very happy with the momentum in general. I'd defer again to Microsoft on specific commentary on Cobalt, but the momentum has been terrific.

Got it. And then on the related party side, it looks like a very nice traction sequentially, I saw it up 48%. Just wondering, long term, are you seeing any changes there? Does that ramp? As you look out longer term, how does that progress?

I would expect it to be pretty consistent. I mean related party, the largest component is Arm China, and so I would expect that will be kind of pretty consistent over time. Arm China, probably, we expect it to probably be a smaller percentage of revenue. I think this quarter, it was in the 25-ish percent range. I've said in the past, we expect it probably to fall to mid-teens over time. But over the next few quarters, probably not going to change a whole lot.

Your next question comes from the line of Andrew Gardiner from Citi.

I had one on M&A. There continues to be speculation in the press about M&A, whether it's Arm or perhaps related to SoftBank above you, and in particular, sort of the potential for you to move beyond a pure traditional IP model and get further into the silicon space. Can you -- without sort of saying anything specific on M&A, I know you won't be able to really do that, but just in terms of the discussions you're having internally and perhaps with your customers and what they're needing, why -- what is happening that is causing you to perhaps lean further in that direction as you look out sort of beyond the current year?

Yes, thank you for the question. And as you surmised, there isn't very much we can say about speculation and/or rumors. We spend a lot of time, most of my time personally, thinking about growth and thinking about the future. One thing I can say that we're seeing that I'm sure all of us in the market and in our space are trying to reflect upon is at the rate of which AI is evolving and the rate at which the software models change, it puts a tremendous pressure on our ecosystem to develop products faster, better, sooner, more efficient.

And as Arm is the heart of all of that, we look very hard in terms of how to solve customer problems. But unfortunately, I can't give you much more detail than that and certainly can't speculate on any rumors on M&A.

Your next question comes from the line of Harlan Sur from JPMorgan.

Maybe as a follow-on to that question, Rene, I mean, the team has always led with a system-level strategy, which then pulls demand for your compute IP solutions, right? AMBA was a good example of that. Your move to CSS is another good example of that. Now it seems like the team is like scaling the strategy to attack the complex SoC market that is rapidly moving towards this more sort of heterogeneous, sort of chiplet-based strategy.

You guys are building an ecosystem around your CSA or this Chiplet System Architecture. Like what's the progress like so far? Are you seeing an acceleration of activity around CSA? And more importantly, like how is the Arm team going to monetize this strategy?

Yes. Thanks, Harlan. A lot of good points there. We've absolutely seen an acceleration of demand for CSS, as Jason mentioned. Customers clearly see the benefit of it. We're now involved with CSS in just about all the major markets that we engage. There's just significant demand for that product. We've also kicked off something we call the Arm Total Design partners. And this allows end partners, design houses to take our CSSs and potentially develop chiplets and very, very strong demand for that program as well.

To your point in terms of what's really driving all of this, I go back to the previous answer I gave. It is really the fact that these designs are incredibly complex to build, getting even more so. There's a very, very strong link between the hardware and the software. And as Arm is the heart of everything that takes place in any one of these chips relative to the software ecosystem, there's just huge demand on asking us to do more sooner and help products get to market faster.

Because we're in a phenomenal time in our industry where the compute demands are outpacing the silicon to serve it. We get lots of questions about the smartphone market and the AI capabilities to harness what's going on inside there. You have to remember that these smartphones were -- the chips for these smartphones were designed 2, 3 years ago, the memory subsystem, the power, everything was predefined.

So to be able to fit these small language models or anything that goes inside the phone is quite a challenge given the fact that you still have to run a display, you still have to run an operating system, you still have to run apps. So what helps address that, getting products to market faster. And that is really what we're very focused on.

Your next question comes from the line of Vivek Arya from Bank of America.

I wanted to go back to this v9 adoption question, and I appreciate that in absolute numbers, right? Of course, it represents more dollars of -- as a percentage of your royalty. But at the start of the fiscal year, I think you said that adoption would grow by -- or percentage contribution would grow by 5 points or 500 basis points every quarter, but it has stalled at 25%.

So I was hoping you could explain what's happening versus the assumption you had at the start of the year. And since we did not see it as much in this last fiscal year, does this create an acceleration or an upside driver for next year? Or is that not the right way to look at it?

Yes, this is Jason. I'll take that one. So I would say -- I mean, ultimately, what you're trying to get to is, does this set us up for better growth next year? I would say it could. I just don't know if it's going to be next year and the year after or what exactly is the time frame. If you go back to, as you said, our assumptions, we had seen 3 straight quarters when it went from 10% of total to 15% and then went to 20%, then it went to 25%.

And now a few quarters where it's leveled out at 25%. The reason we provide the metric is to help people understand the adoption of v9. The fact that it's slowed as a percentage of total, again, as I said, is actually a good thing because that just gives us, as you said and already indicated, it gives us, I think, a better view into the growth because we still believe it's going to reach the 60% to 70% as a percentage of total.

Why do we believe that? Well, we've already sold the contracts, and we already have insight into the products that are going to come to market on v9. So as a result, we have high confidence. It's just a question of what's the timing. The fact that it has not grown over the past few quarters is a little bit surprising to us. And the reason why is because you've actually seen stronger v8 growth than we'd previously expected.

That's great. If all generations are going to be able to have strong growth, that's good. But what's coming online, again, already under contract, and we know it's coming. It's just a question of when do those shipments occur and what's the mix. So ultimately, those are the drivers. Again, the reason we provided this number was more as an indicator to try to help understand the leading indicators of royalty growth.

And so the fact that we've been able to meet or exceed our royalty numbers while having a slightly slower growth of v9 mix again is a good thing. But I think you should focus on the overall royalty growth. We gave those metrics back when royalties were going much slower than they are now. And so that's why I think I would focus more on overall royalty growth than I would v9 adoption.

Your next question comes from the line of John DiFucci from Guggenheim Securities.

My question is for Rene. It's kind of a high-level question, I think. There's been a lot of investor discussion around training optimization and a lot of discussion around -- in regards to DeepSeek. Assuming this is all valid, how does that change your view on Arm's opportunity as it pertains to AI, if at all? And I guess, Rene, more broadly, I'd just like to hear your thoughts on this topic even beyond Arm.

Well, a few things. Broadly speaking, if you think about the achievements of DeepSeek, both V3 sort of their generic model, if you will, and R1, their reasoning model, there was obviously a lot of work that was done on top of existing work, AKA a frontier model that was created by someone else, and then DeepSeek doing some very creative work to build a model on top of that, that was very, very efficient in terms of inference.

I think that's great, to be honest with you and I think it's great for a couple of reasons. It's great for the industry because it drives efficiency. It lowers the cost. And by doing that, it expands the demand for overall compute. So just from a general standpoint, it's a good thing. When you think about the application to Arm, given the fact that AI workloads will need to run everywhere and lower cost of inference, more efficient inference makes it easier to run these applications in areas where power is constrained, the amount of compute you can put down is constrained.

It's terrific for Arm. As wonderful a product as Grace Blackwell is, you'd never be able to put it in a cell phone, you'd never be able to put it into earbuds. You can't even put it into a car. But Arm is in all those places. And as a result, I think when you drive down the overall cost of inference, it's great.

I also think that we're also nowhere close to good enough because if you look at the expenditures being announced by the large players in this market, I think Google just announced a $60 billion spend in their call. Satya has talked about $80 billion. Meta's talked about $40 billion to $50 billion. No one is pulling back. And the reason for that is we're nowhere near the capabilities that could be transformational in terms of what AI can do. So I'm actually -- I think it's a great thing and a good thing, and I think it will actually increase overall compute demand. And for Arm, even better because it allows us play in areas where efficiency is key, and that's our sweet spot.

Your next question comes from the line of Krish Sankar from TD Cowen.

Jason, just wanted to follow up on the ACV. You mentioned the growth decel. I know you will give full year guidance later on. Just kind of curious how to think about the ACV on a go-forward on a longer-term basis. Are there any big data center or mobile programs that are coming that could ramp it back to the teens or double-digit growth? And how much of the ACV is coming from the Arm China business today?

Sure. So let's see. So in terms of the -- I guess the forecast and where we think it should land, if you go back to way back when we went public, we basically provided ACV with the idea that it probably should be growing in the mid- to high single digits. And the reason for that is if you treat all of our deals as ratable, knowing that the vast majority of our license revenue and ACV is actually coming from ATA deals.

And nearly all of our ATA deals typically have a 7% annual escalator. So it should be -- and then all the non-ATA deals won't have that. So it's probably plus or minus in the 7% range. So that's kind of what our forecast has been. Then over the last I guess, now 6 quarters, we've seen -- because of AI, we've seen acceleration in licenses, mostly around adopting v9 as well as now CSS as well. So that's provided upside that took us from that kind of 7-ish percent, plus or minus, all the way up to, what was it, 14% or 15% at a high.

Will we get back to the mid-teens range? I don't know. That's not what we never really forecasted. It certainly could happen. When I provide guidance next quarter, I'll try to give you a little more insight on that. But in terms of kind of the longer-term growth model, you should expect the vast majority of the growth really to come from royalties. And so certainly with v9, at a much higher royalty rate. And as Rene said earlier, both v9 and CSS, those are not onetime increases in royalty rate.

They're a step change, but each year, there's still, with new versions, typically going to be increases annually as well. So overall, you should expect the majority of the growth coming more from royalties in the future than from license.

Got you. And Jason, just how much of ACV is Arm China?

Yes. The Arm -- sorry, the Arm China portion, it's somewhere in kind of the 20-ish percent range, so pretty close to kind of the overall mix. This last quarter, it was at 25% so it's kind of somewhere in that range.

We will now take our final question for today, and your final question comes from the line of Mark Lipacis from Evercore.

It's Mark Lipacis from Evercore. A CSS question for Jason and Rene. Jason, the CSS license activity, is that mostly data center or is it balanced between data center and handsets? And then Rene, when you -- I just want to make sure I understood your comments that you made earlier. As AI comes into smartphones, does that necessarily mean that those processors are going to be kind of chiplet architected? And so then longer term, there's likely a very high attach rate of CSS to the smartphones?

And then a similar question on the IoT side. Does the earbud become a CSS device also? Or is that like a monolithic die with just more processing power on the Arm chip?

Yes. So maybe I'll take the first part of that and then Jason sort of can address the numbers. I think every endpoint that you just described, earbuds maybe not just because they're so, so tiny. But certainly, this chiplet approach is going to be pervasive across just about every SoC, if you will. In other words, inside the package, you'll have a number of small die everywhere. It exists on the high end today almost as a standard.

But I think you'll see that everywhere, which is a gigantic opportunity for us because not only can we provide the compute CSS from a CPU standpoint, but it allows us to have the right mix of whether it's an NPU or combined with a GPU and the right CPU combination to maximize performance.

You sort of hit on a very, very key point in terms of demand driver, which is why we're seeing really strong CSS type of demand across all those end markets. I'll let Jason handle the last piece.

Yes. On the CSS mix, we said in the past that auto, we've announced that it's coming, not here yet. So in terms of the roughly dozen that we've sold, assume that it's basically about 50-50 between infrastructure and our client business.

I will now hand the call back for closing remarks.

Thank you, and thank you, everyone, for all your questions. As always, we very much appreciate the interest in what Arm is doing and very, very good questions. As summarized at the beginning, the quarter was just phenomenal, a record quarter. We've never been close to $1 billion before in revenue, and we just about got there. Royalty's a record at $580 million, and we're now guiding to well north of $1 billion in the next quarter, which is something obviously the company has never done before.

So being able to share with you record revenues for the quarter just ended and a projection to beat that by a healthy margin in the next quarter is just something we're so proud of at Arm. So we are very excited about the future, whether it's about the near-term execution of our strategies with v9 and CSS to all the opportunities that Stargate and Cristal intelligence bring us, fantastic time to be with Arm. So thank you all for your questions and interest, and we will speak to you next quarter.

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