Arm (ARM.US) Q2 conference call: AGI CPU commercialization confidence is “strongest so far” and data center royalties will maintain three-digit growth in the next few years

The Zhitong Finance App learned that the British chip design company Arm (ARM.US) released financial results for the second quarter of the 2027 fiscal year ending September 30, 2026 after the US stock market on Wednesday. According to financial reports, Arm Q2's revenue increased 27% year over year to US$1,443 million, exceeding analysts' expectations of US$1.38 billion and recording the best second-quarter results in history. This is the fifth consecutive quarter in which Arm's revenue grew by more than 20% year over year.

In terms of earnings, adjusted net profit of US$566 million, analysts expected US$514.2 million; adjusted operating profit of US$638 million, analysts expected US$575.7 million; adjusted earnings per share were US$0.52, and analysts expected $0.47.

Driven by the full explosion in demand for artificial intelligence (AI) computing power, Arm's licensing and other revenue for the quarter reached US$773 million, a sharp increase of 50% year over year; royalty revenue was US$670 million, up 8% year over year, with data center royalty revenue doubling year on year.

The strong increase in licensing revenue is mainly due to increased customer demand for Arm intellectual property (IP), as well as changes in the timing of signing multiple high-value licensing agreements and revenue recognition. The company disclosed that the annualized contract value (ACV) increased 12% year over year to $1.79 billion. Arm said that as customers integrate higher-performance CPUs, GPUs and NPUs in chip designs, system complexity and power consumption constraints continue to increase, driving customers to increase their investment in Arm technology.

According to reports, Arm's share in the data center market is rapidly increasing. The share of Arm-based servers of leading hyperscale cloud vendors has climbed from about 15% in the 2024 calendar year to more than 50% in the 2025 calendar year, and the company expects this share to exceed 60% by the end of 2026.

Royalty revenue is mainly based on actual shipments of chips using Arm technology by customers. According to Arm, the increase in royalty revenue this quarter was driven by increased adoption of higher-rate technologies such as Armv9 architectures and Arm computing subsystems (CSS), and the expansion of data center chip deployment, but was partially offset by the impact of rising memory prices on the smartphone market.

This difference also forms an important highlight of this quarter's results: customers' upfront investment in Arm technology is accelerating, but the increase in licensing revenue has yet to be fully translated into a simultaneous increase in royalty revenue. Arm management believes that the strong momentum of the licensing business is an important leading indicator of future royalty revenue growth, and royalty revenue is expected to accelerate as new designs enter mass production.

However, the financial report also showed hidden concerns. Due to the impact of rising memory prices squeezing the smartphone market, mobile phone-related businesses are under pressure. Free cash flow for the quarter was US$110 million, down 73% year on year (free cash flow remained stable at US$1,096 million over the past 12 months). According to US GAAP, the company's net profit for the quarter was US$130 million, a year-on-year decrease of 45%, mainly due to a one-time revenue of US$131 million from business disposal in the same period last year.

In terms of guidance, Arm expects third-quarter revenue of US$1.54 billion (fluctuating up and down US$50 million), slightly higher than analysts' average expectations of US$1.53 billion, but lower than the US$1.7 billion forecast by some analysts; adjusted earnings per share are expected to be US$0.56 (up and down US$0.04).

The core focus of this earnings call was the commercialization prospects of Arm's self-developed chip business Arm AGI CPU, and the explanation that royalty revenue fell short of expectations. CEO Rene Haas said that demand has continued to strengthen since the release of Arm AGI CPUs in March, and the company has received purchase orders from many customers for the 2027 calendar year and signed supply agreements with supply chain companies. “I can tell you today that confidence (in achieving the $2 billion target) is the strongest so far, definitely stronger than 90 days ago,” and the first commercial deliveries will continue as planned in the fourth quarter of fiscal year 2027; Meta is the chief customer, “but there will be many customers other than Meta”. In response to questions about supply assurance, he emphasized that he has had many years of cooperative relationships with partners such as TSMC, Micron, Samsung, and SK, and sees this as a “5 to 10 year business”.

As for royalty revenue falling short of expectations, management attributed the increase in memory prices to suppressing mid- and low-end smartphone shipments, and edge AI royalties were basically flat year over year; however, data center royalties doubled year on year and physical AI royalties increased by 50%, and the company expects data center royalties to maintain three-digit growth in the next few years.

The licensing business set a record, and long-term agreement signing accelerated. Licensing and other revenue for the second fiscal quarter was US$773 million, a record for the second fiscal quarter, which was about US$100 million higher than the company's own expectations, due to heightened customer interest in long-term licensing of Arm technology. During the quarter, 3 new Arm Total Access (ATA) agreements were signed (the total number of active ATA subscriptions exceeded 60, including the first time for a major customer), and 4 new computing subsystem (CSS) licenses (2 for data center chips and 2 for smartphones) were signed. Among them, technology licensing and design service agreements with SoftBank contributed US$209 million, an increase of 17% over the previous year. The company expects this revenue to remain near this level for the rest of the year. Annualized contract value (ACV) increased 12% year over year. CFO Jason Child said the licensing pipeline was “unexpectedly strong” and that licensing beyond expectations usually translates into design wins and more royalties after 12 to 24 months.

The royalty structure is fragmented, data centers are doubled, and mobile phones are under pressure. Data center royalties doubled year over year, driving volume from Google Axion, AWS Graviton, Microsoft Cobalt, and NVIDIA platforms; Rene Haas said that in addition to increasing share, they are also seeing a trend where head-node CPUs connected to accelerators are shifting from x86 to ARM — Google has announced that the TPU V8 will switch to Axion connectivity, and NVIDIA Vera Rubin also uses an Arm architecture. Physical AI royalties increased 50% year over year, driven by advanced driver assistance and in-vehicle infotainment; the company also launched an Arm Total Design program for physical AI, bringing together more than 80 companies. Edge AI royalties were generally flat year over year: the increase in memory prices suppressed the shipment of mid- and low-end smartphones, but Armv9 (which has a penetration rate of over 50%) and the higher royalty rate brought about by CSS partially offset this headwind. For next year, management expects mobile phone industry shipments to remain flat (IDC predicts about -1%), but with v9 and CSS penetration, the company's mobile business royalties are still expected to grow by more than double digits. In the current quarter, partners shipped more than 10 billion Arm-based chips for the first time in a single quarter, an increase of 1.5 billion over the previous year.

AGI CPU commercialization path, OCI will provide instances in 2027. Oracle Cloud Infrastructure (OCI) currently provides ARM-based Nvidia Vera CPUs, and plans to provide Arm AGI CPU instances in 2027; Supermicro (Supermicro), Lenovo, and ASRock (ASRock) are building a complete cabinet for this product and deploying it to data centers such as OCI. Rene Haas said that agentic workloads are inherently CPU-limited — each agent running a specific task usually takes up one CPU core in the form of a virtual machine, which will drive the CPU's evolution from 128 cores to 256 cores and above 500 cores, thereby increasing the average cost of a chip. CFO Jason Child added that the company's confidence in achieving a maximum of 2 billion US dollars of AGI CPU revenue in the 2027-2028 fiscal year increased. The original plan was to reach about 2 billion US dollars in the next year and about 5 billion US dollars in the next year. The next update will be presented at next year's Arm Everywhere event; the gross margin of the first generation of AGI CPUs is expected to be in the 30% to 40% range, and the next generation products are expected to approach the market standard of more than 50%.

Respond to x86 competition. An analyst asked about AMD's claim at its conference call that Venice CPUs provide the best performance per watt. Rene Haas responded that Signal65's independent testing at the OCP summit showed that the performance of the Arm AGI CPU per rack was about twice that of the AMD EPYC 9755, and compared to Venice's previous generation products, “we are very confident in these numbers”. At the same time, he emphasized that the differentiation of the Arm model is that it allows customers to make high customizations around their own systems — “what end customers can do, standard products will never reach”.

Here are the details of Arm's second quarter earnings call.

I. Opening remarks and management statements

Moderator: Hello everyone, thank you for your patience. Welcome to Arm's Q2 FY2027 online video results conference call. Please note that today's meeting is being recorded. Let's leave the conference to today's first speaker, Ian Thornton, VP of Investor Relations. Please speak.

Ian Thornton (Vice President of Investor Relations): Thank you very much and welcome to our FY2027 second quarter results conference call. Today's conference was attended by Arm CEO Rene Haas and CFO Jason Child. Today's call will include forward-looking information about the company and its financial results. While these statements represent our current best judgment, our business faces many risks and uncertainties, and actual results may vary significantly. Significant risk factors that may affect our business and future financial results are set out in our 20-F Annual Report to the US Securities and Exchange Commission. Arm is under no obligation to update any forward-looking statements. We will also cite non-GAAP financial measures, which are reconciled with the most directly comparable GAAP measures, and discussions of some predictive non-GAAP measures that cannot be adjusted without unreasonable effort are included in our shareholder letter and supplementary financial information. Results materials are available at investors.arm.com. Let's leave the meeting to Rene.

Rene Haas (CEO): Thanks Ian and good afternoon everyone. Arm's second-quarter results were a record: revenue reached $1440 million, up 27% year over year. This is our fifth consecutive quarter of growth of more than 20%. Licensing and other revenue grew 50% to $773 million, driven by increased partner investments in Arm computing platforms. Strong licensing momentum is an important leading indicator of future royalties — we expect royalties to accelerate as these new designs enter mass production.

Second-quarter royalty revenue increased 8% to $670 million, with data center royalties doubling year over year. Non-GAAP earnings per share increased 33% to $0.52, while we continued to increase our investment in R&D. AI is driving unprecedented demand for computing power in the cloud, terminal devices, and autonomous systems, and Arm is the computing platform of choice.

Today, our business is driven by two major areas of rapid growth: the IP business — partners are increasing their investment in Arm computing platforms and deploying a record number of Arm architecture chips; the second is our new chip business — where demand for Arm AGI CPUs is extremely strong. Both rely on the same Arm computing platform and software ecosystem. We clearly see that these two major growth engines are working simultaneously in the data center field: partners are expanding the deployment of Arm architecture infrastructure, and demand for Arm's own chips is strong.

Oracle Cloud Infrastructure (OCI) is a prime example of two major businesses reinforcing each other. Currently, OCI provides Nvidia Vera CPUs based on ARM architecture; in addition, it plans to provide Arm AGI CPU instances in 2027. Since the release of Arm AGI CPUs in March, customer demand has continued to strengthen. We have received purchase orders from a number of customers continuing into the 2027 calendar year, and the first commercial deliveries will continue as planned in the fourth quarter of fiscal year 2027. Over the past 90 days, with additional supply guarantees and customer portfolios, our confidence in achieving up to $2 billion in ARM AGI CPU revenue in the 2027 and 2028 fiscal years has been further strengthened.

We remain focused on expanding production capacity to meet growing customer demand. The power supply makes performance and energy efficiency even more important. Morgan Stanley predicts that the global data center power gap will reach 38 gigawatts by 2028. Arm is helping customers use existing electricity to obtain more effective computing power to address this constraint. At the OCP summit, independent testing of Signal65 showed that the performance of Arm AGI CPUs per rack was about twice that of the AMD EPYC 9755. Such improvements could cut the projected global electricity gap in half.

Our data center royalties have doubled year over year, and the market share of Arm-based servers among leading hyperscale cloud service providers has more than tripled — from 15% in calendar year 2024 to over 50% in calendar year 2025. Based on anticipated volumes from Google Axion, AWS Graviton, Microsoft Cobalt, and Nvidia platforms, we now expect this share to exceed 60% by the end of the 2026 calendar year. Outside of data centers, Arm's role in edge AI is also growing as AI agents become more powerful and more continuously online.

Arm-based PCs currently account for 20% of global shipments. Google recently released GoogleBook, a new category of AI-native notebooks built for Gemini intelligence, and Arm is the main computing platform for many of these models. The NVIDIA RTX Spark system, which is now on the market, further expands Arm into local AI, content creation, and gaming. On the smartphone side, we recently released CSS for Mobile 2, which increases the performance of smart AI workloads by up to 70% through SME2.

Our first AI-native Mali GPU can achieve up to 4 times the frame rate at the same continuous power consumption. These capabilities allow more AI and richer experiences to run directly on the device without sacrificing battery life. Arm is the platform of choice for physical AI: royalty revenue increased 50% year over year, driven by advanced driver assistance and in-vehicle infotainment. We have also launched an Arm Total Design program for physical AI, bringing together more than 80 companies to accelerate the development and deployment of complete physical AI systems.

This quarter, our partners shipped more than 10 billion Arm-architecture chips for the first time in a single quarter, an increase of 1.5 billion over the previous year, behind a software ecosystem of more than 30 million developers. Demand for ARM technology has never been so strong — AI is driving more complex chip designs and greater demand for energy efficient computing power. Our IP business generates licensing revenue at the start of design and royalties when it is put into production; now, Arm AGI CPUs, as a new direct chip business built on the same computing platform and software ecosystem, have enabled Arm to have two strong growth engines, and we expect them to drive faster revenue growth in the future. Let's leave it to Jason.

Jason Child (CFO): Thank you Rene. We handed over another strong quarter: total revenue increased 27% year over year to $1440 million, a record for the second fiscal quarter, and the second-highest revenue in a single quarter in our history, reflecting continued strong business. Licensing and other revenue was $773 million, up 50% year over year. It also set a record for the second fiscal quarter and the second highest in history. Licensing revenue was approximately $100 million higher than our expectations, as interest from customers looking to sign long-term commitments to license Arm technology is heating up.

This quarter, we signed 3 new Arm Total Access (ATA) agreements, with a total of more than 60 active ATA subscriptions, including a major customer signing an ATA license for the first time. We have also signed 4 new Compute Subsystem (CSS) licenses, two for data center chips and two for smartphones. For existing customers, these agreements expand and deepen their rights to our future technology roadmap, laying an important foundation for Arm's future royalty growth.

Additionally, we're seeing that ATA and CSS are becoming an accelerated path for new customers — these customers previously had limited use of Arm technology, and several signed ATA or CSS licenses this quarter. Of the $773 million licensing revenue, our technology licensing and design services agreement with SoftBank (SoftBank) contributed $209 million, an increase of 17% over the previous year. We expect quarterly operating levels to remain close to this level for the rest of the year. Annualized contract value (ACV) increased 12% year over year. As we discussed previously, licensing revenue can fluctuate greatly from quarter to quarter due to when large agreements are signed and when accounting is processed, so ACV is still an effective measure of underlying licensing trends.

Now let's look at royalties. Royalty revenue of $670 million, up 8% year over year, is also our all-time high for Q2. Cloud AI is still the biggest driver of growth: data center royalties have once again doubled year over year, and ARM architecture CPUs are being invested in AI infrastructure and general cloud workloads. Physical AI royalty revenue increased 50% year over year, driven by increased adoption of Arm technology in the automotive and robotics sectors.

Edge AI royalty revenue remained roughly the same year over year. Higher memory prices continue to depress smartphone shipments, especially mid-range and low-end models. This headwind was partially offset by higher royalty rates brought about by the adoption of the latest Armv9 and CSS technology. As conditions in the smartphone market return to normal, we expect edge AI royalty revenue to accelerate again. As Rene discussed, customer demand for Arm AGI CPUs is still very strong.

Last quarter, we expressed increased confidence in achieving more than $1 billion in revenue in the 2027 and 2028 fiscal years. Over the past 90 days, we have signed purchase orders with a number of customers and supply agreements with supply chain companies, which further strengthened our confidence in achieving up to $2 billion in revenue. We'll provide an update on developments when the fiscal third quarter results are released — visibility will be better at that time.

Now let's look at operating expenses and profits. Non-GAAP operating expenses were $774 million, up 19% year over year as we continued to expand our engineering capabilities and invest in next-generation platforms, computing subsystems, and the Arm AGI CPU product family. Non-GAAP operating profit was $638 million, up 37% year over year. The non-GAAP operating margin was 44.2%, up 300 basis points from the previous year, and the revenue growth rate exceeded the growth rate of operating expenses. Non-GAAP earnings per share were $0.52, up 33% year over year. Importantly, even with this high level of investment, we have generated $1.1 billion of free cash flow over the past 12 months, enabling us to continue investing in long-term growth.

Now look at the guidelines. For the third fiscal quarter, we expect revenue of $1.54 billion, fluctuating up and down by $50 million. The median value corresponds to a year-on-year increase of about 24%. We expect licensing and other revenue to continue to grow strongly, growing slightly above 50% year over year, supported by strong customer demand and licensing pipeline visibility. In terms of royalties, we expect cloud AI and physical AI to continue to be strong, but this is offset by continued pressure from edge AI due to weak smartphone demand, and the year-on-year base is high. As a result, we expect double digit month-on-month growth and single-digit year-over-year growth in royalty revenue. We expect non-GAAP operating expenses to be approximately $825 million, and non-GAAP earnings per share of $0.56, fluctuating up and down $0.04. Looking further, we continue to see strong customer demand in various business lines. Combined with the expansion of our technology portfolio and the deepening of strategic customer relationships, we are confident about the long-term growth prospects. Let's return the meeting to the host for a question-and-answer session.

II. Q & A session

MODERATOR: Today's first question comes from Gary Mobley of StoneX Group.

Gary Mobley (StoneX Group): I think people are most concerned about Arm AGI CPUs. You're clearly sending a signal — revenue is closer to $2 billion in the five quarters ending the 2028 fiscal year. I wonder what gave you this confidence, particularly in locking in more supply with different partners? I'm guessing most of the initial revenue will be used to support Meta, but can you talk about whether you've signed binding contracts with other customers?

Rene Haas (CEO): OK, thanks Gary. I'll answer that question, and Jason can add if I'm missing something. Regarding the volume of this product, we carefully manage supply and demand expectations every quarter. Demand for Arm AGI CPUs is extremely strong overall, and it's no secret — what really drives it is everything we've seen in intelligent workloads, and the CPU is at the core of it. Today's demand is even stronger than what we talked about when we hosted the Arm Everywhere event in March. Since then, the needs of many customers — Meta is the lead customer — have also exceeded our expectations. That being said, we've been working very hard over the past few quarters to lock in everything we need to meet this demand in our supply chain — involving multiple links: wafer supply, substrates, test equipment, memory, etc. We want to be confident when sending a signal to the outside world that “additional demand can be met with supply in the next few quarters”. So on Arm Everywhere Day, we were talking about the visibility of 1 billion dollars, and we soon saw that it became a figure of about 2 billion dollars. I can tell you today that in terms of locking in supply, the current confidence is the strongest so far, definitely stronger than it was 90 days ago — and that covers all of these aspects. But we didn't just start doing this yesterday or last quarter; we've actually been working on this for a few quarters and have really good partners supporting us. As you mentioned, Meta is the lead customer, but there will be multiple customers in addition to Meta. Is there anything else you would like to add, Jason? If not, thanks Gary for asking.

Gary Mobley (StoneX Group): Can I...

Moderator: Please go ahead, Gary.

Gary Mobley (StoneX Group): Sorry, I just wanted to ask a quick question. You provided enough information in your shareholder letter to calculate the year-on-year change in the single-chip royalty rate, which appears to be rising by at least a mid-single digit percentage. I wonder if this is mainly an accident caused by the product structure, or should it be viewed as a long-term trend of increasing single-chip royalty rates?

Jason Child (Chief Financial Officer): Okay. This is due to product structure factors, but in addition, we shipped 10 billion chips this quarter, so the average rate will fluctuate up and down depending on the specific quarter. I would say that the average rate for mobile services is definitely lower than for servers and physical AI, because the specifications and values of the chips are different, so this will have some impact. I wouldn't recommend using an overall average of 10 billion chips as the best way to predict the future. I think it should be broken down into the three major business units: the mobile business is growing quite slowly; server growth is very strong at double digits, and the number of cores is increasing, and there are more CSS revenue inflows this quarter — this trend will continue, which is why we are highly confident about the three-digit growth (data center royalties) over the next few years. The same goes for physical AI. We're talking about a 50% year-on-year increase, which will also bring some benefits to the overall royalty rate and overall royalty growth. But once again, I recommend looking at these three categories rather than averaging all businesses together.

Rene Haas (CEO): Gary, let me add one more detailed point about data center growth. I think there are two drivers independent of Arm AGI CPUs in terms of royalties: One is that Arm gains market share. I mentioned it earlier, and now we think it will reach 60% by the end of this year. This increase in share comes not only from increased shipments, but also from smart AI driving an increase in platform design projects such as Graviton, Cobalt, and Axion. In addition to this, we also saw that head node (head node) CPUs connected to accelerators are moving from x86 to ARM. We already know Vera Rubin, and Google has announced that the TPU V8 will switch from x86 to Axion connectivity. So back to Jason's point, that's why we have high confidence in the three-digit growth in data center royalties over the next few years.

Moderator: The next question comes from William Blair's Sebastien Naji.

Sebastien Naji (William Blair Research): I'd like to ask how ARM can beat x86 in AI data centers. At the OCP summit, the smart device AI benchmarks you announced showed that the performance of an AGI CPU per rack is twice that of the x86 competitor. But last night AMD said during its earnings call that its Venice CPU provides the best performance per watt. Rene, can you talk about where do you think Arm CPUs have the most meaningful chance against x86 in the data center?

Rene Haas (CEO): Okay. I'm answering from two perspectives. The first perspective is taking the position of Amazon, Google, or Microsoft — they build custom CPUs for their own implementations, racks, and systems. In these designs, especially when designing SoCs, end customers have a high degree of freedom to innovate around their own systems — they can make various trade-offs between data handling, DPU, and storage within the SoC. At this level, you almost have to measure performance per rack rather than a single chip. At the chip level, Arm not only gave them the most efficient CPU core on the planet, but also enabled them to design the most efficient SoC and thus create the ultimate most efficient system. We've seen how fast these customers are moving designs from standard x86 to ARM. As for the Arm AGI CPU benchmark test we ran — which Signal65 did at OCP — compared to Venice's previous generation. We are very confident in these numbers and in the choices the customer made after a positive comparison with this design. Venice is a great product, but we're also very confident about the demand for Arm AGI CPUs and the road map for the future. In summary, there are two dimensions: one is the energy efficiency, performance per dollar, and performance per watt of an AGI CPU; second, at the customization level, what end customers can do, standard products will never reach — this is the beauty of our business model.

Moderator: The next question comes from TD Cowen's Krish Sankar.

Sreekrishnan Sankarnarayanan (TD Cowen Research): Rene or Jason, you introduced the AGI CPU reference server design on OCP. I was wondering, are you working with ODMs or server manufacturers? Is the goal an enterprise private cloud? In line with this topic, what is the biggest driver for AGI's target of $15 billion in CPU revenue for fiscal year 2031? Is it an increase in the number of cores, or an increase compared to the first generation product ASP?

Rene Haas (CEO): OK, thanks for the question. Speaking of our partners — they are Supermicro (Supermicro), Lenovo (Lenovo), and ASRock (ASRock). They built an entire cabinet for us and then deployed it to a data center like OCI. They are our construction partners. As for the revenue outlook, ask Jason to discuss the exact figures and timing. What is clear is that the number of cores in these SoCs will definitely accelerate growth. This is largely driven by intelligent workloads — specifically, an agent that runs a specific task, usually runs in the form of a virtual machine on a CPU core, and software that is managed across single or multiple cores is quite complicated, so being able to run on a single core is a better choice. So, have I seen CPUs move from 128 cores to 256 cores to 500 cores or more? Definitely yes. As you can imagine, the average price of chips with more cores will definitely be higher. But I think we'll see a faster increase in core numbers, which will help increase revenue. Jason, is there anything you would like to add on the numbers?

Jason Child (Chief Financial Officer): Okay. We said at the Arm Everywhere event in March of last year that by the end of fiscal year 2027 and fiscal year 2028, we have 1 billion dollars in visibility. Of course, now we have raised our confidence to reach $2 billion. As for the next year, we haven't updated it, but the original plan was to reach around $2 billion (per year) one year after that, and around $5 billion in the next year. So if we successfully reach more than $2 billion, you can expect that future numbers may still have room to rise. We plan to provide updated targets at the next Arm Everywhere event — the timing is yet to be determined, probably in spring or the first half of next year.

Rene Haas (CEO): Let me add: When we held Arm Everywhere Day in March, the general view in the market was that it was a $60 billion market and wasn't going to grow much. But our own calculations came to a different conclusion. The figure we gave that day was $100 billion by 2030. I think this figure was later raised by various parties — Lisa (note: AMD CEO Su Zifeng Lisa Su) even said more than $200 billion. I think this is mostly true, and maybe even underestimated, who knows? What is obvious is that the computing power requirements of intelligent AI tasks are limited by the CPU. The more agent workloads you run, the more CPU you'll need. So this will drive higher core count devices, which is very beneficial to us — because at the core level, what matters is performance per watt, having the most efficient core on the planet, and that's what we're good at.

Moderator: The next question comes from Susquehanna's Mehdi Hosseini.

Mehdi Hosseini (Susquehanna Financial Group Research Department): I have two questions. Back to the updated $2 billion revenue target for AGI chips — should we assume any revenue contributions for the September or December quarter of this fiscal year (FY2027)?

Jason Child (Chief Financial Officer): Let's assume our shipments begin in the fourth fiscal quarter. If there are a few in the third fiscal quarter, we'll tell everyone, but if there are, there won't be many. What we're really targeting is the fourth fiscal quarter.

Mehdi Hosseini (Susquehanna Financial Group Research Division): OK. Just a quick question: Can you think of data centers as a share of royalty revenue approaching one-third (30% to 35%) this fiscal year? Is it likely to rise to around 45% by the next fiscal year (2028)?

Rene Haas (CEO): Sorry, you're asking...

Jason Child (Chief Financial Officer): Percentage of total royalties?

Rene Haas (CEO): Data centers account for a percentage of total royalty revenue, is that a problem?

Mehdi Hosseini (Susquehanna Financial Group Research Division): Yes. Can we assume that the 2027 fiscal year will reach about one-third, and that the 2028 fiscal year will reach another level and move towards 45%?

Jason Child (Chief Financial Officer): OK — I mean, we're looking at a few pieces of business independently. Edge AI royalties will be fairly flat this quarter, and we said the fourth fiscal quarter will begin to see some recovery. When modeling, you can consider this: high-single-digit growth in edge services, double-digit growth in data centers — as we said, data centers were roughly close to 20% of total royalties last year — and physical AI grew in the range of about 20% to 30%. You should model this way. I haven't calculated the exact ratio. I don't think the data center will reach one-third this year, but it will be there soon. I'm just not sure if it will reach it before the end of this year.

Moderator: The next question comes from Raymond James' Simon Leopold.

Simon Leopold (Raymond James Research): I'd like to ask about the customer concentration of the cloud and AI product portfolio—can we have a better quantitative concept? For example, can we guess that about 60% of this portion of royalties comes from 3 customers? Any quantitative information would be helpful.

Rene Haas (CEO): I can say this -- and Jason can add: Amazon, Google, and Microsoft are the biggest contributors, then NVIDIA is now getting started fast. These four are the customers in the data center that contribute the vast majority of royalties.

Moderator: The next question comes from Harlan Sur of J.P. Morgan Chase.

Harlan Sur (J.P. Morgan Chase Research): Your first generation of AGI CPUs got off to a strong start, and I'm sure customers are already looking forward to the next generation platform. Can you update the progress of the second-generation AGI? I remember the plan was to release the film (tape out) by the end of this year or the beginning of next year, for the 2028 calendar year. Perhaps just as important: as you continue to add more power to your computing chip portfolio, does it make sense to add more networking and high-performance I/O capabilities? We are increasingly seeing that computing and networking are tightly coupled to determine system performance. Will the team add more networking capabilities to the mix?

Rene Haas (CEO): OK, thank you, Harlan, all great questions. Without disclosing the exact schedule for subsequent chips, you can estimate according to the market's mass production pace of about 18 to 24 months. The first generation is entering mass production, and you can calculate the approximate launch time for the second generation yourself. As for anything else that can be added to the “toolbox” to improve performance — whether it's general connectivity, specific I/O, networks, or other aspects of the system — we're definitely looking into it. We developed these products with overall rack system performance in mind. Chips and CPUs are an important part of it, but as you asked, how to move data into and out of a system is as important as any other field. We have a great deal of work going on in all of these areas. There was nothing to announce on today's call, but we are fully aware that this is an area where more differentiation can be made.

Moderator: The next question comes from Lee Simpson at Morgan Stanley.

Lee Simpson (Morgan Stanley Research): If I could, I would ask two quick questions. First for Jason: The authorized revenue of $773 million is impressive, congratulations to the team. You've talked about Total Access, and you've also talked about the CSS protocol—specifically smartphones and data centers. Is there an architecture license agreement (architecture license) in it? Can we view more than $770 million as a new structural level? Ask Rene: Following the previous few questions, will the dollar value of a single core increase as CPUs with higher ASP and higher number of cores appear in the future? In other words, would the associated capabilities mentioned in the previous question — network, I/O — drive up the dollar value of a single core?

Jason Child (CFO): OK, thanks Lee. Answer the authorization question first. The main thing we're talking about now is ATA and CSS, which account for the vast majority of our future sales. Architectural licensing is already very, very rare. So you can think that the main ones are ATA and CSS — these are the absolute subjects of future sales. As for the absolute amount, $773 million was indeed a fantastic quarter. But if you do the math, I said it will go up in the next quarter and reach another level in the fourth fiscal quarter. We haven't talked about next year—whether this will become the new normal every quarter, and I don't want to say that. But what I'm saying is that the pipeline is surprisingly strong, even stronger than we expected at the beginning of the year. It's a bit like the past few years: there are some deals in the pipeline that aren't renewals — these are new deals that are new and hard to predict in advance. We are now in the same situation again. Interestingly, at the beginning of the year, we thought this year would mainly rely on royalties, but we thought most of the licensing growth associated with AI investment was over. But the reality is that as AI, workloads continue to change, and now everyone is dealing with intelligent workloads, the need for additional licenses, additional product variants, and different technologies is not slowing down at all. This is particularly exciting. As you know, any licensing business that exceeds expectations — ATA and CSS in particular — is likely to turn into a design win order, and most likely to turn into more royalties after 12, 18, or 24 months. This lays a good foundation for next year and beyond. As for next year, it will be a few more quarters before I can talk about it. But the line looks really strong. Even though memory issues have brought a difficult year-on-year base for the mobile business — these effects are basically over until the end of our third fiscal quarter, and we will begin to overcome these headwinds in January and February next year, and the year-on-year base will become easier. Overall, we think things are going very well for the rest of the year and next year.

Rene Haas (CEO): One more thing about pricing. The problem is rather delicate: since we released Arm AGI CPUs, the price of memory has clearly increased a lot; data center construction has also become more and more complicated — memory prices, labor, etc. Considering these external factors, we want to be restrained in pricing and don't want to put too much pressure on them; but it's fair to say that we may have some pricing flexibility we can use on these larger, more core chips. We'll obviously take all the factors into consideration, but there should definitely be some upside.

Moderator: The next question comes from Vivek Arya of Bank of America Securities.

Vivek Arya (Bank of America Securities Research Division): One person. Jason, assuming another double-digit decline in smartphone shipments next year, do you think royalty performance will be similar to this year? Or are there other dynamics at play? Rene, about the AGI platform: Assuming we recognize its technical advantages, my question is, does Arm have enough ecological scale to lock in supply? For example, when AMD released its results yesterday, it raised its server CPU business by several billion dollars, which shows that they have the scale to seek incremental supply. How do you respond to this question — given your confidence in the technical advantages of the product, does Arm have enough scale to obtain supply and become a meaningful player in the CPU market?

Rene Haas (CEO): OK, thank you, Vivek. I'll answer the second half of your question, and then Jason talks about the smartphone scenario you described. Your question is reasonable. My opinion is: Arm AGI CPU is our first product. AMD has been working for many years, has many SKUs, procures a large amount of production capacity from TSMC, and has a deep relationship established over decades. This is our first product. But then again, this isn't our first collaboration with partners like TSMC, nor is it our first time working with memory manufacturers such as Micron, Samsung, and SK — they've known us for years, even decades, and we've been sharing our long-term plans with them very seriously. I think supply will be tight across the industry. As Jason mentioned, we'll talk more about the outlook for the next few quarters. But we're going to be working very, very hard to lock in the supply needed to meet the expansion demand because we think this demand is very, very, very large. Our partners know this too. We're not communicating with them on a weekly or monthly basis, but it's probably been over a year — right when we planned this platform. We see it as a long-term business, not a single project for a year, but a 5 to 10 year career requiring partnerships across the supply chain. We are very, very serious about making sure that all of this is in place.

Jason Child (CFO): Thank you Vivek. Regarding the mobile business, let's look at some external forecasts - IDC predicts that next year's mobile phone shipments will remain flat, about -1%. We have communicated with our partners, and in general, everyone's expectations have remained the same. The impact of memory limitations and shipping downgrades have already occurred, and everyone expects these restrictions to remain relatively stable — that is, flat, not worsen. At the same time, the adoption rate of v9 is still increasing and is now over 50%, so this part of the tailwind effect is weakening. However, the adoption of CSS continues to grow, and we also have many partners working together on the next generation of CSS, and their royalty rates and chip prices are slightly higher. All of this means that our overall business will grow much faster than the market — think next year will be at least double digits. This is our best judgment right now. If there are any changes, of course we'll let everyone know. But looking at it now, the industry and our partners all see it that way.

Moderator: Now onto today's last question, Joe Quatrochi from Wells Fargo.

Joseph Quatrochi (Wells Fargo Securities Research): When you talk about locking in more supply for AGI, I'm assuming that means higher costs. Can you talk about your ability to hedge against these costs — how do you transfer costs to customers, and the long and short factors of gross margin once you start shipping?

Rene Haas (CEO): Sorry, can we repeat that?

Joseph Quatrochi (Wells Fargo Securities Research): OK. I'm assuming that the incremental supply costs you are seeking for AGI are higher, so I'd like to ask you to talk about the ability to hedge these higher costs, the ability to transfer pricing to customers, and the long and short factors of gross margin.

Rene Haas (CEO): OK, thank you. As I mentioned earlier, memory is probably the single biggest factor in the overall cost of the system, and in addition, other aspects of the supply chain will also be affected — you're right, this will definitely put pressure on gross margins. That being said, we have some flexibility and room for increases in pricing — I think we already demonstrated this ability when the project started. So in terms of hedging supply constraints and costs, and downstream management, I'm actually less concerned. Everyone has seen other companies in this field publicly comment on similar practices. We will follow the trend of the market to ensure that we do not lag behind common sense, but at the same time do not cause much pain to end customers. This is our first product, and we want to ensure a win-win situation for all. I'm not that worried about this part's impact on gross margins.

Jason Child (Chief Financial Officer): Let me add that for the first generation products, we expect gross margins to be roughly in the range of 30% higher to around 40% lower; the first generation is expected to be slightly lower. As we accumulate experience, we may gradually do more self-research, rely less on partners, and move closer to 50% or more market standards for a long time, like other manufacturers in the market. But I don't expect the first generation to reach it; that's more of a next-gen product. So for this generation, please model in the range of 30% to 40%. Once the fourth fiscal quarter starts shipping, we will update you on a quarterly basis.

Moderator: This concludes today's questions. Let's return the meeting to Rene for her concluding remarks.

Rene Haas (CEO): Thank you for all the questions — they were all very timely and wonderful. To sum it up in a nutshell: It's been a fantastic quarter for the company. The 50% increase in the licensing business means there will be a lot of royalty growth in the future, and we feel very good about that. We are also very, very proud that our data center business continues to prosper. We've been talking about Arm reaching 20% of the data center share for years; the truth is that we reached 50% last year and will reach 60% by the end of this year — and we're also incredibly proud. But most importantly, I think we're now beginning to see that what was once a great business is turning into two great businesses. The IP business continues to perform very, very well, and we are confident of its growth; now, as the new AGI CPU enters mass production and our confidence in achieving the 2 billion dollar target we have been talking about increases, I am extremely excited to be able to talk about both businesses that will embark on an amazing growth path at the same time. I am extremely proud of what the company has achieved and thank our employees, partners, and customers for making it all possible. Thanks again for all the questions and comments, and congratulations to the Arm team for an excellent quarter.

Moderator: Thank you. This concludes today's call. Thanks for participating, I can hang up.