Intel Corp (INTC) — Q3 2025 Earnings Call Transcript

Thank you, Jonathan, and good afternoon to everyone joining us today. By now, you should have received a copy of the Q3 earnings release and earnings presentation, both of which are available on our Investor Relations website, intc.com. For those joining us online today, the earnings presentation is also available in our webcast window. I am joined today by our CEO, Lip-Bu Tan; and our CFO, David Zinsner. Lip-Bu will open up with comments on our third quarter results as well as provide an update on our progress implementing strategic priorities. Dave will then discuss our overall financial results, including fourth quarter guidance before we transition to answer your questions. Before we begin, please note that today's discussion contains forward-looking statements based on the environment as we currently see it and, as such, are subject to various risks and uncertainties. It also contains references to non-GAAP financial measures that we believe provide useful information to our investors. Our earnings release, most recent annual report on Form 10-K and other filings with the SEC provide more information on specific risk factors that could cause actual results to differ materially from our expectations. They also provide additional information on our non-GAAP financial measures including reconciliations where appropriate to our corresponding GAAP financial measures. With that, let me turn things over to Lip-Bu.

Thank you, John, and let me add my welcome this afternoon. We delivered a solid Q3 with revenue, gross margin, earnings per share above guidance. This marks the fourth consecutive quarter of improved execution driven by the underlying growth in our core markets and the steady progress we are making to rebuild the company. While we still have a long way to go, we are taking the right steps to create sustainable shareholder value. We significantly improved our cash position and liquidity in Q3, a key focus for me since becoming CEO in March. This includes accelerated funding from the United States government, important investments from NVIDIA and SoftBank Group, and monetizing a portion of Altera and Mobileye. The actions we took to strengthen the balance sheet give us greater operational flexibility and position us well to continue to execute our strategy with confidence. In particular, I'm honored by the trust and confidence President Trump and Secretary Lutnick have placed in me. Their support highlights Intel's strategic role as the only U.S.-based semiconductor company with leading-edge logic R&D and manufacturing. We are fully committed to advancing the Trump administration's vision of restoring semiconductor production and proudly welcome the U.S. government as our essential partners in our efforts. We also made tangible progress to improve our execution this quarter. We remain on track not only to rightsize the company by year-end, but also to evolve the talent mix, reestablish the engineering-first mindset and optimize the executive and management levels across the organization. We are seeing a significant increase in day-to-day energy and collaborations as our employees return to the office after a sustained period of remote and hybrid work. Let me dive deeper into our underlying business trend. Over the course of my career, I have had the privilege of contributing in multiple ways to disruptive innovation. But I can't recall a time when I have been more excited about the future of computing and the opportunities in front of us. We are still in the early stage of the AI revolution, and I believe Intel can and will play a much more significant role as we transform the company. Let's start with our core x86 franchise, which continues to play a critical role in the age of AI. AI is clearly accelerating demand for new compute architectures, hardware models, and algorithms, while at the same time fueling renewed growth of traditional compute as the underlying data and the resulting insights continue to rely heavily on our existing products from cloud to edge. AI is driving near-term upside to our business and it is a strong foundation for sustainable long-term growth as we execute. In addition, with unmatched compatibility, security, and flexibility by virtue of being the largest installed base of general purpose compute, x86 is well positioned to power the hybrid compute environment that AI workloads demand, particularly for inference edge workloads and agentic systems. It is a great starting point from which to rebuild our market position, revitalizing and rejuvenating the x86 ISA, and positioning for the new era of computing with great products and partnerships. Our collaboration with NVIDIA is a prime example. We are joining forces to create a new class of products and experience spanning multiple generations that accelerate the adoption of AI for the hyperscale, enterprise, and consumer markets. By connecting our architectures to NVIDIA NVLink, we combine Intel CPU and x86 leadership with NVIDIA unmatched AI and accelerated computing strength, unlocking innovative solutions that will deliver a better customer experience and provide a big hit for Intel in the leading AI platform of tomorrow. We need to continue to build on this momentum and capitalize on our position by improving our engineering and design execution. This includes hiring and promoting top architectural talent as well as reimagining our core roadmap to ensure it features best in class. To accelerate this effort, we recently created the Central Engineering Group, which will unify our horizontal engineering functions to drive leverage across foundational IP development, test chip design, EDA tools and design platforms. This new structure will eliminate duplications, improve time to decision-making, and enhance coherence across all product development. Additionally, and just as importantly, the group will spearhead the build-out of our new ASIC and design service business to deliver purpose-built silicon for a broad range of external customers. This will not only extend the reach of our core x86 IP but also leverage our design strength to deliver an array of solutions from general purpose to fixed function computing. In client, we are on track to launch our first Panther Lake SKU by year-end, followed by additional SKUs in the first half of next year. This will help us solidify our strong position in the notebook segment across both consumer and enterprise with cost-optimized products across our full PC stack from our entry-level offering to our mainstream core family, up through our highest-performing Core Ultra family. In high-end desktop, competition remains intense, but we are making steady progress. Arrow Lake shipments have increased throughout the year, and our next-generation Nova Lake product will bring new architecture and software upgrades to further strengthen our offerings, particularly in the PC gaming space. With this lineup, we believe we will have the strongest PC portfolio in years. In traditional servers, AI workloads are driving both the refresh of the installed base and capacity expansion, fueled by rapid growth in tokenization, the increased demands around data storage and processing, and the need to elevate power and space constraints. We remain the AI head nodes of choice with strong demand for Granite Rapids, including instances across every major hyperscaler. We are listening to what customers need, and strong performance per watt and TCO are top of mind. As I shared with you last quarter, the key part, including improving our multi-treading capabilities, as we close existing gaps and work to regain shares. Finally, on our AI accelerator strategy, I continue to believe that we can play a meaningful role in developing compute platforms for emerging inference workloads driven by agentic AI and physical AI. This will be a far larger market than that for AI training workloads. We will work to position Intel as a compute platform of choice for AI inference, and we look to partner with a range of incumbents as well as emerging companies that are defining this new compute paradigm. This is a multi-year initiative, and we will strike partnerships when we can deliver true differentiation and market-leading products. In the near term, we will continue delivering AI capabilities to Xeon, AI PCs, Arc GPUs and our open software stack. Looking ahead, we plan to launch successive generations of inference-optimized GPUs on an annual cadence that features enhanced memory and bandwidth to meet enterprise needs. Turning to Intel Foundry. Our momentum continues. We are making steady progress on Intel 18A. We are on track to bring Panther Lake to market this year. Intel 18A yields are progressing at a predictable rate and Fab 52 in Arizona, which is dedicated to high-volume manufacturing, is now fully operational. In addition, we are advancing our work on Intel 18AP, and we continue to hit our PDK milestones. Our Intel 18A family is the foundation for at least the next three generations of client and server products. I will work with the U.S. government within the secure enclave and other committed customers. It is a critical node that will drive wafer volumes well into the next decade and generate a healthy return on our investment. On Intel 14A, the team continued to focus on technology definition, transistor architecture, process flow, design enablement and foundation IPs. We remain active, engaged with potential external customers, and are encouraged by the early feedback which helps us to drive and inform our decisions. Lastly, our advanced packaging activities continue to progress well, especially in the areas like EMIB and EMIB-T, which we have two differentiations. Like our Intel products, my conviction in the market potential for Intel Foundry continues to grow. The rapid expansion of critical AI infrastructure is fueling unprecedented demand for wafer capacity and advanced packaging services that present a substantial opportunity demanding multiple suppliers. Intel Foundry is uniquely positioned to capitalize on this unprecedented demand as we execute. As I mentioned last quarter, our investment in foundry will be disciplined, and we will focus on capability and scalability, giving us the flexibility to ramp quickly, and we will only add capacity when we have committed external demand. Building a world-class foundry is a long-term effort founded on trust. As a foundry, we need to ensure that our process can be easily used by a variety of customers, each with a unique way of building their own products. We must learn to delight our customers as they call on us to build wafers to meet all their needs for power, performance, yield, cost and schedule. It is only by doing this that they can rely on us as a true long-term partner to ensure their success. This requires a change of mindset that I'm driving across Intel Foundry as we position this business for long-term success. As we look ahead, my focus remains firmly on the long-term opportunity across every market we serve today and those we will enter tomorrow. Our strategy is crystallized around our unique strength and value proposition, supported by the accelerating and unprecedented demand for compute in the AI-driven economy. Our leadership continues to strengthen. Our culture is becoming more accountable, collaborative, and execution-oriented. My confidence in the future grows stronger every day. I look forward to keeping you updated as we advance our journey. I will now turn it over to Dave for details on our current business trends and financials.

Thank you, Lip-Bu. In Q3, we delivered the fourth consecutive quarter of revenue above our guidance, driven by continued strength in our core markets. Although we remain vigilant regarding macroeconomic volatility, customer purchasing behavior and inventory levels are healthy and industry supply has tightened materially. Furthermore, we are increasingly confident that the rapid adoption of AI is driving growth in traditional compute and reinforcing momentum across our businesses. In client, we are five years post the COVID pull forward and are benefiting from the refresh of a larger installed base. Enterprises continue to migrate to Windows 11, and AI PC adoption is growing. In data center, the accelerating build-out of AI infrastructure is positive for server CPU demand from head nodes, inference, orchestration layers and storage. We are cautiously optimistic that the CPU TAM will continue to grow in 2026 even as we have work to do to improve our competitive position. Third quarter revenue was $13.7 billion, coming in above the high end of our guidance range and up 6% sequentially. Capacity constraints, especially on Intel 10 and Intel 7, limited our ability to fully meet demand in Q3 for both data center and client products. Non-GAAP gross margin was 40%, four percentage points better than our guidance on higher revenue, a more favorable mix and lower inventory reserves, partially offset by a higher volume of Lunar Lake and the early ramp of Intel 18A. We delivered third quarter earnings per share of $0.23 versus our guidance of breakeven EPS driven by higher revenue, stronger gross margins and continued cost discipline. Q3 operating cash flow was $2.5 billion with gross CapEx of $3 billion in the quarter and positive adjusted free cash flow of $900 million. One of our top priorities for 2025 was shoring up our balance sheet. To that end, we executed on deals to secure roughly $20 billion of cash, including three important strategic partnerships. We exited Q3 with $30.9 billion of cash and short-term investments. In Q3, we received $5.7 billion from the U.S. government, $2 billion from SoftBank Group, $4.3 billion from the Altera closure and $900 million from the Mobileye stake sale. We expect NVIDIA's $5 billion investment to close by the end of Q4. Finally, we repaid $4.3 billion of debt in the quarter and we will continue prioritizing deleveraging by paying maturities as they come due in 2026. Moving to segment results for Q3. Intel products revenue was $12.7 billion, up 7% sequentially and above our expectations across client and server. The team executed well to support upside in the quarter given the current tight capacity environment, which we expect to persist into 2026. We are working closely with customers to maximize our available output, including adjusting pricing and mix to shift demand towards products where we have supply and they have demand. CCG revenue was $8.5 billion, up 8% quarter-over-quarter and above our expectation due to a seasonally stronger TAM, Windows 11-driven refresh and a stronger pricing mix with the ramp of Lunar Lake and Arrow Lake. Within the quarter, CCG further advanced its relationship with Microsoft through a collaboration with Windows ML and the deep integration of Intel vPro manageability with Microsoft Intune enabling secure cloud connected fleet management for businesses of all sizes. The team also met all key milestones in support of launching Core Ultra 3, code-named Panther Lake. We expect the client consumption TAM to approach 290 million units in 2025, marking two straight years of growth off the post-COVID bottom in 2023. This represents the fastest TAM growth since 2021, and we're prudently preparing for another year of strong demand in 2026 as Core Ultra 3 ramps into a healthy PC ecosystem. PC AI revenue was $4.1 billion, up 5% sequentially, above expectations, driven by improved product mix and higher enterprise demand. The strength in host CPUs for AI servers and storage compute continued in the quarter even as supply constraints limited additional upside. Our latest Xeon 6 processors, code-named Granite Rapids, offer significant benefits, including up to 68% TCO savings and up to 80% less power as compared to the average server installed today. It is increasingly clear that CPUs play a critical role today and will going forward within the AI data center as AI usage expands, especially as inference workloads outpace that of training. Some data center customers are beginning to ask about longer-term strategic supply agreements to support their business goals due to the rapid expansion of AI infrastructure. This dynamic, combined with the underinvestment in traditional infrastructure over the last couple of years should enable the revenue TAM for server CPUs to comfortably grow going forward. Operating profit for Intel products was $3.7 billion, 29% of revenue and up $972 million quarter-over-quarter on stronger product margin, lower operating expenses and a favorable comparison due to period costs in Q2. Before discussing Intel Foundry, I want to acknowledge the tireless effort of the NVIDIA and Intel teams. There's a lot of work in front of us, but the collaboration we announced this quarter was the culmination of almost a year of hard work with a company that cuts no corners and prioritizes engineering excellence above all. The x86 architecture has been the foundation of the digital revolution that powers the modern world. AI is the next phase of that revolution, and we're on a path to ensure x86 remains at the heart of it. Engagements like this one with NVIDIA are critical to this effort. Moving to Intel Foundry. Intel Foundry delivered revenue of $4.2 billion, down 4% sequentially. In Q3, Intel Foundry delivered Intel 10 and 7 volume above expectations, met key 18A milestones and released hardened 18A PDKs to the ecosystem. Foundry also advanced the development of Intel 14A and continues to make progress expanding its advanced packaging deal pipeline. Intel Foundry operating loss in Q3 was $2.3 billion, better by $847 million sequentially primarily on favorable comparison due to the approximately $800 million impairment charge in Q2. As Lip-Bu discussed, our confidence in the long-term foundry TAM continues to grow, bolstered by accelerating deployment and adoption of AI and the growing need for wafers and advanced packaging services. Projections are calling for a greater than 10x increase of gigawatts of AI capacity by 2030, creating significant opportunities for Intel Foundry with external customers, both for wafers and our differentiated advanced packaging capabilities like EMIB-T. We continue the work to earn the trust of our customers, and our improved balance sheet flexibility will allow us to quickly and responsibly respond to demand as it comes. Turning to all other. Revenue came in at $1 billion, of which Altera contributed $386 million and was down 6% sequentially due to the intra-quarter closure of Altera. The three primary components of all other in Q3 were Mobileye, Altera, and IMS. Collectively, the category delivered $100 million of operating profit. Now turning to guidance. For Q4, we're forecasting a revenue range of $12.8 billion to $13.8 billion. At the midpoint, and adjusting for the Altera deconsolidation, Q4's revenue is roughly flat quarter-over-quarter. We expect Intel products up modestly sequentially but below customer demand as we continue to navigate the supply environment. Within Intel products, we expect CCG to be down modestly and PC AI to be up strongly sequentially as we prioritize wafer capacity for server shipments over entry-level client parts. We expect Intel Foundry revenue to be up quarter-over-quarter on increased Intel 18A revenue, and its external foundry revenue should increase due to the deconsolidation of Altera. For all other, which now excludes Altera, we expect revenue to decline consistent with Mobileye's guidance, partially offset by sequential growth in IMS. At the midpoint of $13.3 billion, we forecast a gross margin of approximately 36.5%, down sequentially due to product mix, the impact of the first shipments of Core Ultra 3, which usually have higher costs in the early stages of a new product ramp, and the deconsolidation of Altera. We forecast a tax rate of 12% and EPS of $0.08, all on a non-GAAP basis. We expect non-controlling income to be approximately $350 million to $400 million in Q4 on a GAAP basis, and we forecast an average fully diluted share count of roughly 5 billion shares for Q4. Moving to CapEx. We continue to anticipate 2025 gross capital investment will be approximately $18 billion, and we expect to deploy more than $27 billion of CapEx in 2025 versus $17 billion deployed in 2024. I'll wrap up by saying we exit Q3 with a significantly stronger balance sheet, solid demand in the near term, and growing confidence in our core x86 franchise as well as the longer-term opportunities in foundry, ASICs, and accelerators. We also recognize the work we need to reach our full potential. We continue to add external talent and unlock our workforce to improve our execution across product and process development as well as manufacturing. We will closely manage what's in our control, react quickly as the environment evolves, and focus on delivering long-term shareholder value. At this time, I'll turn it back to John to start the Q&A.

Thank you, Dave. We will now transition to the Q&A portion of our call. With that, Jonathan, can we take the first question, please?

Congratulations on the strong results. Lip-Bu, the first one for you is going to be on the foundry side. You guys announced a ton of collaborations in the quarter. You very much strengthened your balance sheet. And the tone you took in your preamble sounds much more confident on the progress you're making in foundry. Do any of these collaborative announcements or equity investments go into that increased confidence? Or are there some sort of technical merits that you're seeing that are raising your optimism in that part of your business?

Yes, Ross, thank you so much for the questions. A couple of announcements we made are clearly more on the product side. Also, one is the SoftBank investment because they are building up all the infrastructure for AI. That definitely will need more capacity on the foundry side. So I think that would be the answer. But meanwhile, I've been saying that we made tremendous progress on 18A and 14A. Panther Lake will depend on it. Clearly, we see the yield in a more predictable way. I visited Fab 52, and it is fully operational for the 18A. With 14A, we are engaging with multiple customers in terms of milestone basis, driving some of the yield and performance, reliability that we are seeing improvement. Additionally, we also see important demands from some key customers for advanced packaging from both the cloud and enterprise sides. Overall, we are looking quite excited to build this long-term trust with some customers and scale it. We also focus on hiring some of the top talent and driving process technology improvement.

Ross, do you have a follow-up question?

Yes, I do. One for Dave on the gross margin side of things. You talked through the upside in the third quarter and the sequential downside in the fourth. But could you just walk us through some of the pluses and minuses as we think about 2026, just kind of directionally? Where I'm going is it seems like the biggest improvement has to come on the foundry gross margin side of things. Is that the biggest driver? What drives it? And as those gross margins go up, does that have any impact on the Intel products gross margin?

Yes, I can provide some insights, though we're not providing guidance for 2026. First, it's important to note that Altera will not contribute to the numbers in 2026, whereas they were a significant part of our 2025 figures, leading to a margin challenge for us since they positively impacted our gross margins. Despite this, we still anticipate a 40% to 60% fall-through for margins, although this is a general range influenced by product mix. Lunar Lake will play a significant role, especially in the first half of the year, but it is also a dilutive product for us. Panther Lake is expected to be beneficial in the long run, but it will come with higher initial costs. We do expect to see improvements in gross margins on the foundry side, primarily due to economies of scale, and as we shift toward a leading-edge mix, including 18A, Intel 4, and 3, these products should offer better pricing and cost structures, contributing positively to margins. However, the extent of improvement will depend on how the product mix evolves throughout the year.

I was really interested in a lot of the prepared remarks around the sort of differences in your approach to foundry, and you talked about this last quarter and this quarter that you're sort of looking for customer commitments before you make the investment. Can you just talk about how those conversations are going? And I certainly can see the trade-off from a customer standpoint. They're making a commitment to you. Do they expect that capacity to be built ahead of time? Just is there a bit of a chicken and egg aspect to these investments? And just how are you approaching those conversations?

Yes, Joseph, thank you so much for the question. I think on the foundry side, clearly, we are engaging with multiple customers. To build trust with the customer, we need to really show the yield improvement, reliability, and we also need to have all the specific IP that they require. It's a service industry. You need to have all the right IP. That's why I formed the Central Engineering to get all the right IP matching with customer requirement. The best way is to show the performance, the yield. Then, we can test chips so that customers can rely on us for their most important revenue wafers to drive success for them. These are very important. In terms of potential investment and collaboration, we are working with different customers with different requirements. But more importantly, it's about getting their commitment to the foundry.

Yes, maybe just to add on, I would say that I think customers understand that it takes time from the time you deploy capital to the time where you have output. Our expectation is we will get those commitments firmed up in time to deploy the capital to meet the demand. We're in a reasonably decent position given the CapEx investments we've already made. We have a lot of the assets on the books and what we call assets under construction. We've made a lot of investments around the shelf space. We do see a line of sight to driving a reasonable amount of supply for our external foundry customers with our existing footprint and reuse of equipment that we have on the books today.

Joe, do you have a follow-up question?

I do, yes. Regarding the supply constraints in server CPUs and other CPUs, we are noticing them in the market; however, your growth was 5% sequentially and single-digit growth year-on-year. Where is the shortage originating from? Is it due to higher demand that you are unable to fulfill? Is it linked to some transitions you have managed? I definitely see the tightness in the marketplace, so I'm curious about the source of that shortage and how it might be resolved.

The shortage is prevalent throughout our business. We are facing tightness in Intel 10 and 7, and we are not planning to increase capacity in those areas. With rising demand, we are constrained and relying on our existing inventory. We are also attempting to guide customers towards alternative products. There are additional shortages beyond our specific challenges at the foundry, including widely recognized substrate shortages. Overall, there is a noticeable caution regarding demand as we enter the year. However, indications suggest that demand will strengthen this year and likely continue into next year as companies navigate these challenges.

I guess a follow-up on the current outlook for demand outpacing supply into 2026. Curious if that's a comment largely focused on server or also including clients and I guess depending on your thoughts there. How should we be thinking about Q1 trends versus normal seasonality, which typically would be down high single digits, low double digits?

Yes. It's both. Although, as we said, we are yielding a bit of the small core market and client to fulfill customer requirements more broadly in the client space and more specifically in the server space. So that's how we're managing it. As you look into Q1, obviously, again, this is something we'll probably give you a lot more color around in January. I would just say we may actually be at our peak in terms of shortages in the first quarter because we've lived through Q3 and Q4 with a little bit of inventory. We probably won't have as much of that luxury in Q1. So I'm not sure we'll buck the trend on seasonality given we're going to be really, really tight in the first quarter. After that, I think we'll start to see some improvements, and we can get ourselves caught up as we get through the rest of the year.

C.J., do you have a follow-up question?

I do, John. I guess given the investments from the U.S. government and NVIDIA, SoftBank, et cetera, I'm curious, with that improved cash position and liquidity, how has your thinking evolved in terms of investments in either CapEx or other investments in your product businesses?

Yes, we are in a strong position. Our primary focus with this cash is to reduce our debt. When Lip-Bu joined, his main priority was the balance sheet. This quarter, we eliminated $4.3 billion in debt. All the maturities due next year should be settled, and we will pay those off. Regarding capital expenditures, we have flexibility, but we intend to be very disciplined with our spending. We will carefully monitor demand, and if it warrants, we will increase our capital spending accordingly. Beyond that, we will see how things progress. We aim to be disciplined about our operating expenses as a percentage of revenue while maximizing leverage. We also recognize opportunities to invest that could provide significant returns for our shareholders, and we are willing to pursue those.

I had two. Just on the CapEx, I think you reiterated $18 billion, but I think you spent, I guess, less than I was modeling in Q3. So is that really still the number? And I'm just kind of curious, as you start to ramp this 18A in Arizona, is there a way to think about the timing of when you add capacity there?

On the $18 billion, yes, I think that's still the number. CapEx can be unpredictable and depends on when all the requirements for paying the invoice are fulfilled, which is when we would make the payments. We expect to be within that range, but there may be some variation. For 18A, we still need to ramp this up. I wouldn't anticipate significant increases in capacity in the near term, as we are not at peak supply for 18A. We won't reach that until the end of the decade, and we believe this node will be long-lasting for us. We will continue to invest in 18A over time, but I don't expect the supply to change significantly from our current expectations.

Blayne, do you have a follow-up?

Yes. Just I wanted to follow up on the gross margin trajectory as 18A layers in. I know comparing it to the prior couple nodes is not a great comparison, but maybe to a successful one. When you say yields are in a good spot and improving, is there a way to think about where those 18A yields are versus the successful products that you've seen in your history and how that layers in the first half?

Yes. Generally, I'm not sure yields in older nodes have been a big focus for us. We're blazing a new trail. I'd say, the yields are adequate to address the supply, but they are not where we need them to be in order to drive the appropriate level of margins. By the end of next year, we'll probably be in that space. On 14A, we're off to a great start. If you look at 14A in terms of its maturity relative to 18A at that same point of maturity, we're better in terms of performance and yield. We just need to continue that progress.

I wanted to revisit the issue of supply constraints. You mentioned that AI is increasing demand for servers and PCs, yet it seems customers prefer your older products, as they can't seem to get enough of them. I assume you have adequate supply for Granite, Meteor, and even Lunar Lake. So, how do you plan to encourage customers to transition away from the older products, given they haven't shown any interest in doing so despite the current constraints? Additionally, how do we view the transition for those customers, considering you have indicated that you are not increasing the capacity for the older products and have even taken some offline?

Yes. Good question, Stacy. I would say it's a misnomer to say AI hasn't done well. It was sequentially up double digits quarter-over-quarter, and we projected about shipments of 100 million units by the end of this year with AI PCs, and we're in that range. I think it's going pretty well. Clearly, though, the older nodes have also done well, and that was probably the part that was more unexpected. We've just got to ensure that the ecosystem drives enough applications for AI in the PC space. We're working with ISVs to drive that. They're getting there. It starts relatively immature and builds out over time. In our company, we're starting to find uses for AI PCs. In fact, our IR is developing one that we'll be using. I think it's just a matter of time. That said, the Windows refresh is happening more significantly than we expected. That's not necessarily an AI PC story. Raptor Lake is also a product that addresses that. We're seeing upside in that part of the market as well.

Dave, I want to follow up on two things that I think I heard you say on 18A. I thought I heard you say, number one, the yields would not be in a great place at least until the end of next year. And then I thought I also heard you say that you were not going to be adding a lot of 18A capacity next year. Did I hear those wrong? I mean, how can such the latter one, how can that be true if you're ramping Panther? Or is that like a...

We're obviously at our infancy. What I'm saying is relative to the CapEx plan, it's not like we're going to incrementally add supply for 18A next year. But we are going to be ramping the volume over the course of the next year. I wouldn't say 18A yields are in a bad place. They're where we want them to be at this point. We had a goal for the end of the year, and we're going to hit that goal. But to be fully accretive in terms of the cost structure of 18A, we need the yields to improve. That takes time. It's going to take all of next year to get to a place where that's the case.

I wanted to ask about some comments Lip-Bu made in the prepared remarks about fixed function computing and potentially supporting more ASICs. Could you provide more context on the scope of this? Is this for potential foundry customers? Or are these products? If it's products, what types of applications do you expect to be supporting with custom silicon?

Good question. First of all, I just mentioned about the Central Engineering. We are driving the ASIC design, and that will enhance our reach of the core x86 IP and drive purpose-built silicon for some of our systems, cloud players, and customers. This AI will drive a lot of growth, especially in doubly down on Moore's Law. This will help us a lot in our x86 uplift. Building the entire ASIC design will serve some customer requirements.

Josh, do you have a quick follow-up?

Yes. On the last quarter, obviously, the disclosure that you may decide to abandon 14A got a lot of attention. I just wanted to ask, given your balance sheet is in a lot different spot than it was three months ago, has anything changed from that regard? I don't think the Q is out, so I haven't seen if any of the language changed there. But was curious if anything had moved around since last quarter given all the changes in your balance sheet.

Yes. Since the last quarter, our engagement with customers for 14A has increased, and we are working heavily with customers on defining the technology, process, yield, and IP requirements to serve them. They clearly see the tremendous demand they need from Intel to be strong on 14A. We are more confident and delighted about this prospect moving forward.

Lip-Bu, can we get an update on the NVIDIA relationship, timing of products? Have you gotten any feedback from customers in terms of your ability to articulate on the materiality of the relationship and in terms of timing and materiality or any other color you want to give us on that?

Sure. Thank you. This is a very important collaboration with NVIDIA. It's a great company, as you guys know. I've been known as a friend of Jensen for more than 30 years. We are very excited about this effort of Intel CPU x86 leadership, and their unmatched AI and accelerated computing. By connecting with their NVLink, we're really creating a new class of product in the multigenerational space. This is a very heavy engineering-to-engineering engagement. It will drive new products for custom data centers and PC products, optimizing for the AI era. This will involve years of engagement addressing the market, and we're excited to drive this requirement for AI infrastructure.

Just to add, what makes this really special for us is it's not attacking our existing TAM. It's an incremental opportunity for us to expand the TAM. So these are great opportunities for us.

Ben, do you have a follow-up?

Yes. Lip-Bu, you mentioned a little bit about your AI strategy now to attack the inference market and that you see room for Intel solutions. Does this strategy focus more on partnerships? Is there specific Intel IP for inference that you're particularly excited about? Or is it more of a Switzerland approach where you could partner with a lot of the existing players to address more of the TAM?

Yes, good question. First of all, I think with AI driving a lot of growth, we want to play a significant role in this market. This is just the early innings, and it presents an opportunity for us. We focus on revitalizing our x86 architecture to really tailor to purpose-built CPU and GPU requirements for the new AI workload. This will address the efficiently managed power from agency workloads and be a new compute platform of choice, applying to system and software. We will partner with some incumbents and also emerging companies driving these changes.

Dave, it’s not common to see high fixed cost businesses that have gross margins under 40%. I understand a lot of this is due to the wafer costs for Intel 10 and 7, along with the low yields on 18A. However, I’m curious if you could give us an idea of what the gross margin might look like if we moved away from 10 and 7 and switched to 18A. Is there a way you could help us visualize that?

Yes. You've definitely been listening in on some of my conversations with the team here because that's definitely something I've been making the point of. I would say there's two dynamics: one of which you're hitting on, the high cost of older processes versus the better cost structure for the newer processes, which is meaningful. We're in negative gross margin territory for foundry, and that would make a meaningful improvement if you move into the positive territory. The other aspect of our gross margin is a function of just the product quality. We're in reasonably decent shape on client in terms of product performance and competitiveness, with a few exceptions not where we need to be on a cost basis. We have that on the roadmap. The team recognizes it, but that's a multi-year process to get there. It's more pronounced on the data center side. We don’t have the right cost structure nor the competitiveness to really get the right margins from our customers. Lip-Bu and the team have pulled in a hyper-focused effort to get great products at the right cost structure for better gross margins. Improvements on the foundry side will come, mixing higher and higher to Intel 3, 4, and ultimately 18A, and the cost structures of those will be similar. It will simply be the value provided by those leading-edge nodes that will significantly drive gross margins up.

I do. Yes. Lip-Bu, you didn't give an update last call on Diamond Rapids' launch date. I know the whole roadmap is under review, but you did sound fairly optimistic about Coral Rapids. Can you give us an update on the data center roadmap here?

Yes. Thank you for the good question. The feedback for Diamond Rapids is getting better from hyperscale, and we are focusing on the new product, Coral Rapids, which will include SMT and multithreading to drive higher performance. We're in the definition stage and will work out the road map as we execute.

I have a couple of quick questions. Regarding the NVIDIA relationship, I understand the announcement was related to the NVLink Fusion strategy and its integration with the x86 ecosystem. However, I noticed some recent reports about possibly using Gaudi for dedicated inference workloads within an NVIDIA stack. Is this relationship just a starting point? Should we anticipate more integration opportunities beyond NVLink in the future?

Yes, let me answer that. I think NVLink is more like the hub connecting the x86 and GPU. Regarding our AI strategy, we are defining what we call Crescent Island. We also have a new product line that focuses on agentic, physical AI, and additional developments in inference. Stay tuned for updates.

Aaron, do you have a quick follow-up?

Yes. For '26, we're looking at somewhere in the $1.2 billion to $1.4 billion range as a good estimate for non-controlling interest. We are focused on that and will work to minimize it as much as possible.

With that, I want to thank everyone for joining us today. We are on the journey of rebuilding Intel, and we have a lot of work ahead of us, but we are making solid progress in Q3. I look forward to seeing many of you throughout the quarter and providing you another update in January.