Intel Corporation (INTC.US) Q3 Earnings Call: Server CPU Market to Exceed US$200 Billion Over the Next Five Years; PC Downturn May Persist Into Next Year; Cash Flow Could Be Negative in 2027 "Investment Year"

Title context: Intel Corporation (INTC.US) Q3 Earnings Call: Server CPU Market to Exceed US$200 Billion Over the Next Five Years; PC Downturn May Persist Into Next Year; Cash Flow Could Be Negative in 2027 "Investment Year" Text: Since the beginning of this year, one of the core beneficiaries of the global AI agent frenzy, U.S. CPU and chip manufacturing giant Intel Corporation (INTC.US), provided core content from its earnings call showing that AI server demand and manufacturing execution are improving, but PC market weakness, supply bottlenecks tied to chip manufacturing capacity, and a new round of capital investment mean there is still a clear time gap before the trajectory of significant earnings improvement and accelerated realization of free cash flow. Earnings data show that Intel Corporation's third-quarter total revenue reached US$16.5 billion, non-GAAP gross margin was 44.1%, and earnings per share were US$0.52, above the company's previous guidance of US$0.38; Data Center and AI segment revenue was US$7.3 billion, up 17% quarter over quarter and up nearly 80% year over year (the precise figure is 78%). However, fourth-quarter revenue guidance fell to US$15 billion to US$16 billion, non-GAAP gross margin was expected to fall to 38%, and earnings per share were expected to fall to US$0.25. The Israel fab incident is expected to affect fourth-quarter revenue by about US$300 million and gross margin by about 2 percentage points, with the revenue impact in the first quarter of next year potentially similar. Content from the earnings call shows that management expects weak PC demand to persist into 2027, and management clearly distinguished between "easing inventory destocking pressure" and "recovery in end demand." Intel Corporation management expects global PC market shipments to decline about 12% in 2026, including a decline of more than 20% in the second half; in 2027, they will decline further by about 10%. Customer inventory adjustments are mainly concentrated in the fourth quarter of 2026 and the first quarter of 2027, and may also extend slightly into the second quarter. After that, the relationship between company shipments and end consumption is expected to return to normal, but this does not mean that memory supply and pricing pressure have been resolved. The chief financial officer explicitly stated that the memory chip environment is not expected to improve significantly in 2027, and whether it improves still needs to be observed in 2028. The company also expects the server CPU total addressable market to potentially expand from the current US$40 billion to more than US$200 billion within five years. Capital expenditure and cash returns were the most important forward-looking information from this call. Total capital expenditure for 2026 is expected to be about US$20 billion, and it will increase substantially in 2027, with the specific amount not yet determined; the roughly US$30 billion raised by analysts during the Q&A session was only a hypothetical in the question and was not confirmed by management as formal guidance. The chief financial officer defined 2027 as an "investment year," stating that with investments concentrated in accelerating capacity expansion for Intel 3 and 18A advanced process nodes and preparing 14A for production in 2028, it is expected to be difficult to achieve positive free cash flow next year. The company's third-quarter operating cash flow was about US$5.1 billion, and cash and short-term investments at quarter-end were about US$54 billion, plus US$15 billion in revolving credit borrowing capacity, providing an important buffer against heavy investment costs. Intel Corporation management still targets at least a 20% return on investment for the relevant most advanced process nodes (for example, Intel Corporation's Intel 3, Intel 18A, and Intel 14A, involving transistor structures, manufacturing processes, and related production equipment), and has committed to further reducing leverage in 2027, but these are future goals and expectations. Management stated that Intel Corporation's long-term growth logic remains centered on server CPUs, custom chips, advanced packaging, and external wafer foundry. Management expects industry server CPU shipments to grow more than 30% in both 2026 and 2027, with supply constraints continuing into 2028; Intel Corporation management emphasized at the call that, as AI agent-driven workloads expand, the data center server-side high-performance CPU total addressable market may increase from US$40 billion to more than US$200 billion within five years. Granite Rapids has already been accretive to the company's gross margin, and 18A costs have fallen about 40% from the first quarter, but the foundry business still recorded an operating loss of about US$1.9 billion, and the quarter-over-quarter growth in external revenue mainly came from a non-recurring customer project; Granite Rapids is the product codename for Intel Corporation's sixth-generation Xeon (Xeon 6) high-performance-core server CPU, with its core compute chip using the Intel 3 process, mainly used in traditional data centers, cloud computing, and large data centers dominated by AI-related computing tasks. Management is also advancing inference GPUs, modular x86 architecture data center CPUs, advanced packaging, and commercialization of quantum computing. The investment research focus of the entire call therefore rests on three links: whether demand can be converted into actual supply, whether capacity expansion can improve unit costs, and when new investment will sustainable free cash flow. Below is the full transcript of Intel Corporation (i.e., Intel)'s earnings call (translated with the assistance of AI tools): Conference Operator Thank you for standing by, and welcome to Intel Corporation's third-quarter 2026 earnings conference call. Currently, all participants are in a listen-only mode. After management's remarks, there will be a question-and-answer session. (Conference Operator Instructions) Please be advised that today's conference is being recorded. Now, allow me to introduce today's host, Mr. John Pitzer, Vice President of Investor Relations. Sir, please go ahead. John Pitzer, Vice President of Corporate Planning and Investor Relations Thank you, Jonathan, and good afternoon to everyone joining us today. By now, you should have received the third-quarter earnings release and presentation materials. Both documents are available on our investor relations website at intc.com. Those joining via the web today can also view the presentation materials in the webcast window. Joining me today are Chief Executive Officer Lip-Bu Tan and Chief Financial Officer Dave Zinsner. Lip-Bu will first discuss third-quarter results and our progress on our strategic priorities. Dave will then discuss overall financial results, including fourth-quarter guidance, after which we will take your questions. Before we begin, please note that today's presentation contains forward-looking statements based on the environment as we currently see it. As such, these statements are subject to various risks and uncertainties. The presentation will also reference non-GAAP financial measures, which we believe provide useful information to investors. Our earnings release, our most recent Annual Report on Form 10-K, and other filings with the Securities and Exchange Commission provide more information about specific risk factors that could cause actual results to differ materially from our expectations. These documents also provide more information about non-GAAP financial measures, including, where applicable, reconciliations to the corresponding GAAP financial measures. With that, I will turn the call over to Lip-Bu. Chief Executive Officer Lip-Bu Tan Good afternoon, and thank you for joining us. The demand environment in the third quarter was more balanced than in prior quarters. Strong momentum in Data Center and AI drove results above expectations and delivered the segment's strongest revenue growth in 15 years. This growth was partially offset by weakness in the Client business, which was affected industry-wide by rising memory prices and limited IT budgets. Unprecedented demand in superintelligence is causing the entire industry to experience one of the worst supply shortages in history. Against this backdrop, I am pleased with the progress we made in the third quarter to increase server supply. At the same time, I am clear-eyed that we will continue to face challenges in keeping up with server demand in the coming quarters. We successfully completed an equity offering in August, providing the financial resources necessary to capture future growth. In addition to existing shareholders, I am pleased to welcome new sovereign funds and long-term growth investors. We are deploying this capital quickly and prudently. We expect capital expenditures and key component purchase commitments to increase significantly over the coming quarters. I want to thank our substrate, memory, foundry, and equipment suppliers for their close partnership and collaboration as we work together to expand supply and support growing demand. Now to the business units. First, Data Center and AI. In our server business, we are steadily executing our product roadmap. Granite Rapids, built on the Intel 3 process, has become one of the fastest-ramping server products. Clearwater Forest, our first Intel 18A server product, has begun shipping and recognizing revenue, with considerable demand for 2027, especially in CPU-intensive infrastructure for large-scale agents. Diamond Rapids, our first Intel 18A-P server product, is currently sampling and remains on track to ramp in 2027, with strong customer interest in its use for inference and agentic workloads. We are pleased to have recently announced an agreement with Oracle Corporation cloud infrastructure to deploy our latest-generation CPUs to support its AI infrastructure buildout. I also want to be clear about what is limiting our progress in meeting this extraordinary AI-driven demand. First, Intel 3 yields and cycle times continue to improve, but I am still demanding significantly better execution from the team. Second, even as our Intel 3 wafer output is ramping quickly, the industry-wide shortage of advanced organic substrates used for EMIB [inaudible] packaging is also affecting our ability to expand output. Finally, earlier this month, an unfortunate and tragic accident caused a prolonged shutdown at our facility in Israel. Although the team acted quickly to respond, this will affect fourth-quarter shipments of 10nm server products. Custom chips remain our fastest-growing business segment, having tripled in size over the past three years, and are expected to potentially triple again in the coming years. We see further growth opportunities in working with existing customers as well as potential new customers, including AI labs. We are also expanding our intellectual property portfolio to meet the needs of a broader customer base. Finally, we recently launched Crescent Island GPU, optimized for inference. Initial feedback has been encouraging, with good performance progress. For us, this is a new market, and scaling will take time, but engagement with customers is moving toward clear commitments. Now to the Client business. In recent weeks, customers began signaling to us that the PC market is slowing due to memory prices and supply conditions. Year to date, system prices have risen significantly, putting additional pressure on already limited IT budgets. Over the past two months, the pace of this PC market correction has accelerated. Our OEM partners are actively managing inventory to align with weaker end sales. We expect this environment to persist until memory supply and pricing begin to normalize. Although the PC market will face challenges over the coming quarters, our product roadmap continues to strengthen, and the long-term outlook remains positive. We have successfully driven Panther Lake, built on the Intel 18A process, into volume ramp, while demand for Wildcat Lake exceeds supply. Nova Lake, our first Intel 18A-P client product, is ahead of schedule and is expected to secure twice the number of design wins at launch compared with [the comparison object is missing in the original text]. We look forward to sharing more at CES in January. As the market normalizes, we expect this product portfolio to drive significant market share gains. Execution in Intel Corporation's foundry business continues to improve. In the quarter, yields improved again for both Intel 3 and Intel 18A, with Intel 18A yield improvement reaching double digits. Wafer output for both processes exceeded expectations. Intel 18A-P remains on track to launch in internal products by year-end, and engagement with external customers continues to increase. We are working with several highly strategic customers, and I am increasingly confident that Intel 18A-P will be an attractive process node for external foundry customers. [The process node name is not listed here in the original manuscript] also remains on track. In October, we delivered on our commitment by releasing an industry-standard version 0.9 process design kit to selected customers. This is our highest-quality version 0.9 process design kit, meeting the yield and performance metrics required by industry standards at release. As is customary in the foundry industry, I have consistently refrained from disclosing customer details. Nevertheless, the progress we have made with multiple potential customers on 14A continues to encourage me. In advanced packaging, we are pleased with the progress of EMIB and EMIB-T on internal products, and both technologies continue to attract strong interest from external customers. First, we continue to improve execution and work with the supply chain to put in place the capacity necessary to meet advanced packaging customer demand. We also see solid progress in preparing hybrid bonding interconnects, a technology that will support 3D chip stacking in our upcoming server products. Across the manufacturing network, additional capacity is coming online. Intel 3 tool installation in Ireland is progressing well, as is the 18A ramp in Oregon and Arizona, including Fab 62, our second high-volume 18A manufacturing facility. In New Mexico and Malaysia, we are adding capacity to support strong customer ramps in advanced packaging next year. In New Albany, Ohio, the team recently completed installation of the first large structural truss, marking an important construction milestone. We are also continuing to evaluate a unique and highly differentiated opportunity to accelerate the Ohio project, and hope to share more soon. Looking ahead, I have consistently emphasized that rebuilding Intel Corporation will be a multi-year journey, and my commitment to it is unwavering. We are only a year and a half into this journey. In these 18 months, we have made tremendous progress, including significantly resizing the workforce, reducing middle management and bureaucracy, streamlining the executive team, and building an almost entirely new leadership team. We have re-established an engineering-centric culture, streamlined the product portfolio, and built a more competitive roadmap for Intel Corporation products and Intel Corporation's foundry business. We still have much work to do. The renewed growth in CPU demand is providing important support for our transformation journey. Energy efficiency is increasingly becoming the most important metric for AI infrastructure buildout. GPUs excel at thinking-type tasks, while CPUs with deterministic execution characteristics are more efficient at lookup-type tasks. A single GPU inference query may consume several watt-hours of electricity, while a single CPU query may require only energy measured in joules. Even with conservative estimates for the total number of digital agents and the number of agents served per CPU, it is easy to foresee that over the next five years the CPU total addressable market will be very large in both dollar terms and unit terms. Based on our estimates, driven mainly by workloads, the server CPU total addressable market may expand to five times its current size within five years, from US$40 billion to more than US$200 billion. This presents a huge opportunity for Intel Corporation. Of course, we are also clear that intense competition is underway, with nearly every fabless chip company racing to launch its own products to meet this unprecedented demand. But Intel Corporation is uniquely positioned to benefit from this demand in three ways. First, as the [missing in original text] of x86x86 is the world's most widely deployed instruction set architecturewe are doubling down on x86 innovation and have an exciting CPU innovation roadmap using a "tick" alternating approach. In addition, for the first time, we are offering customers a fully modular, validated x86 CPU implementation optimized for custom workloads. Second, as one of only three advanced-process semiconductor foundries in the world, and the only U.S. one among them, we are well positioned not only to meet our own demand but also to support the broader fabless chip industry. We are also willing to manufacture any CPU at scale for customers, regardless of the underlying instruction set architecture. Third, modern computing platforms are increasingly evolving into complex system-in-package designs, and our leadership in advanced packaging can provide highly differentiated solutions, especially to meet the needs of current large-scale AI infrastructure buildout. That is why, even with work still to do, I remain optimistic about the opportunity ahead for Intel Corporation. Finally, while focusing on near-term priorities, we must not lose sight of what comes next. Intel Corporation has a deep technology and intellectual property portfolio and the ability to drive the next wave of computing. We will continue to invest in breakthrough technologies such as silicon photonics, neuromorphic computing, advanced materials, and quantum computing, which we believe will unlock important new opportunities over time. This week, we announced plans to acquire a quantum software vendor headquartered in Israel [the company name is missing in the original manuscript], advancing our quantum business from basic research toward commercialization. This company's [company name transcription unclear in the original manuscript] hardware-agnostic software platform layer provides more customers with an entry point to explore quantum computing and complements our already strong internal quantum hardware and software stack, extending our computing platform offerings beyond CPUs, GPUs, and ASICs to quantum processors. Upon completion of the acquisition, Intel Corporation will become a quantum computing company with a full stack and revenue generation. In summary, despite our good third-quarter performance, it is clear there is more work to do. I am disappointed that we could not meet strong customer demand this quarter. We face near-term challenges, including manufacturing supply shortages, an unprecedented memory shortage, and the resulting weakness in PC demand. Still, I remain confident in the road ahead. We need to address these challenges with urgency without losing focus on the enormous opportunity in front of us. I want to thank our employees, partners, and suppliers for their tireless efforts this quarter. I will now turn the call over to Dave for more detail. David Zinsner, Executive Vice President, Chief Financial Officer, and Principal Financial and Accounting Officer Thank you, Lip-Bu. We delivered a strong quarter, with revenue, gross margin, and earnings per share all exceeding our guidance. This marks our eighth consecutive quarter of exceeding our own financial expectations. Third-quarter results further highlight the important role our server CPUs play in superintelligence or AI infrastructure buildout. The results also show that execution is improving as we increase supply to capture strong AI compute demand. Third-quarter revenue was US$16.5 billion, driven primarily by record year-over-year growth in Data Center and AI. Excellent execution across our factory network and external supply chain drove better-than-expected server CPU output. Client weakness partially offset this strength, as the memory shortage and limited IT budgets are affecting PC consumption industry-wide. Third-quarter non-GAAP gross margin was 44.1%, about 200 basis points above the second quarter and our guidance. The sequential improvement was driven by Intel 3 and Intel 18A cost improvements and a more favorable product mix. Third-quarter non-GAAP earnings per share were US$0.52 versus our guidance of US$0.38. The beat was driven primarily by higher revenue, stronger gross margin, and disciplined spending management, which together delivered better-than-expected operating leverage. Operating cash flow was US$5.1 billion. At quarter-end, we held about US$54 billion in cash and short-term investments, and the newly extended revolving credit facility provides an additional US$15 billion of borrowing capacity, up from US$10 billion previously. In early August, we further strengthened the balance sheet through a US$23 billion equity offering. The additional interest income from this offering contributed about US$0.01 to third-quarter earnings per share, which was not included in our original guidance. Now to segment results. Data Center and AI revenue was US$7.3 billion, up 17% quarter over quarter and nearly 80% year over year. Revenue significantly exceeded expectations as our factory network and external suppliers supported higher output. The work the team has done to increase supply is producing tangible results. Based on the current outlook, Data Center and AI full-year 2026 revenue is expected to be nearly US$10 billion above our expectations at the beginning of the year. Third-quarter growth was broad-based across hyperscalers, enterprises, and new cloud customers. The custom chip business also continued its strong momentum, with revenue up more than 50% quarter over quarter and more than doubling year over year. Despite our progress in expanding capacity, demand for Data Center and AI products still significantly exceeds what we can supply. Data Center and AI operating profit was US$3 billion, or 42% of revenue, up about US$600 million quarter over quarter. This improvement was driven primarily by higher revenue and lower operating expenses. With output continuing to increase at the Ireland facility, Granite Rapids has now become the largest product family in our server portfolio. Earlier this month, at the Open Compute Project Global Summit, we demonstrated certain Clearwater Forest configurations scaling to more than 46,000 cores per rack, with total rack power of about 133 kilowatts, delivering industry-leading performance per watt. We also introduced the Intel Corporation GPU D series, formerly codenamed Crescent Island, designed to improve the economics of sustained memory-intensive AI workloads. Customer interest in both product families has exceeded initial expectations. Client revenue was US$8.1 billion, down 9% quarter over quarter and below our expectations. We previously noted that the memory shortage could pressure the total PC market size in the second half. That pressure intensified in the third quarter, especially late in the quarter, and led to a marked shift in customer purchasing patterns. Despite near-term market weakness, we still see several positive developments. With total Client revenue down 5% year over year, AI PC revenue grew 35% year over year. Client products built on the Intel 18A process continue to ramp smoothly. Wildcat Lake, our cost-optimized client CPU built on the Intel 18A process, saw shipments increase about 70% quarter over quarter, and the product is sold out. Finally, our Edge business was more resilient, supported by continued design wins in edge AI and physical AI. Client operating profit was US$2.3 billion, or 28% of revenue, roughly flat quarter over quarter, driven primarily by lower inventory-related charges. Now to Intel Corporation's foundry business. Intel Corporation foundry revenue was US$6.7 billion, up 16% quarter over quarter, driven primarily by better-than-expected growth in Intel 3 and Intel 18A. External foundry revenue was US$411 million, up US$118 million quarter over quarter, mainly from a non-recurring customer project. Intel Corporation foundry recorded an operating loss of US$1.9 billion, narrowing by US$236 million quarter over quarter, even as we continue to invest aggressively to prepare Intel 14A for risk production next year. Higher yields, shorter cycle times, and increased factory utilization continue to drive wafer cost reductions. Intel 3 costs improved about 15% from first-quarter levels, with output up mid-teens quarter over quarter. Intel 18A costs have fallen about 40% since the first quarter, reflecting continued improvements in yield and cycle time. Intel 18A yields continue to improve toward industry-standard levels, ahead of internal forecasts. Since the first quarter, the revenue share of Intel 18A client CPUs has increased from less than 5% to nearly one-third of Client CPU revenue currently. These results represent substantial progress, but we still have significant work to do to make the foundry business sustainably profitable. Now to the outlook. Demand signals from data center customers continue to strengthen. We expect industry supply constraints to persist through 2027 and into 2028. As Lip-Bu said, AI adoption is broadening the role of CPU computing. AI workloads are shifting from prompt-based responses to multi-step reasoning, task orchestration, and autonomous, continuous action. Our CPUs play a central role in supporting these workloads. As a result, we continue to see strong demand for Data Center and AI products. We now expect industry server CPU shipments to grow more than 30% in both 2026 and 2027. If the industry can provide more supply, growth in both years would be even higher. At the same time, the PC market is weakening after remaining resilient in the first half of 2026. We expect total PC market shipments to decline about 12% in 2026, including a decline of more than 20% in the second half. We do not expect a sustained recovery until memory supply improves and prices begin to normalize. For 2027, we are planning for total PC market shipments to decline about 10%. We expect the year-over-year decline to be more pronounced in the first half, with possible improvement in the second half due to easier comparisons. Our long-term view of the PC market is unchanged. Traditional PCs will play an important role in supporting AI applications and driving strong returns on AI infrastructure spending. In addition, we are only just beginning to see new agentic systems enter the market. As applications spread across the economy, we also continue to see attractive long-term opportunities in edge AI and physical AI. Combining these factors, we are guiding fourth-quarter revenue to US$15 billion to US$16 billion. At the midpoint of US$15.5 billion, we expect gross margin of 38%, a tax rate of 11%, and earnings per share of US$0.25, all on a non-GAAP basis. Based on our preliminary assessment of the Fab 28 incident, our guidance includes an estimated revenue impact of about US$300 million and a gross margin impact of about 200 basis points. By business, we expect Data Center and AI revenue to grow mid-teens quarter over quarter, with custom chips growing faster than the segment overall. Please note that the US$300 million impact from Fab 28 relates almost entirely to 10nm server products. We expect Client revenue to decline more than 20% quarter over quarter as customers rapidly reduce inventory weeks and align purchases with their expectations for 2027 end demand. The quarter-over-quarter decline in fourth-quarter gross margin mainly reflects the decline in overall revenue while the fixed cost base remains, as we adjust client wafer starts to current demand. In addition, revenue mix, especially the expected strong growth in custom chips, also makes the quarter-over-quarter comparison more difficult. Continued improvements in Intel 3 and Intel 18A yields and costs will partially offset these headwinds. We expect fourth-quarter non-GAAP operating expenses of about US$4.4 billion and about US$16.3 billion for the full year. We expect fourth-quarter non-controlling interest to be about US$250 million, and currently expect about US$1.1 billion in both 2027 and 2028. Finally, we expect fourth-quarter fully diluted shares of 5.44 billion, reflecting a full quarter of the additional shares from the August equity offering. Now to capital expenditure. Total capital expenditure for 2026 remains on track at US$20 billion, and we expect total capital expenditure to increase substantially in 2027. We have made good progress securing the equipment needed to support our growth plans, with firm purchase orders to equipment suppliers up nearly 40% quarter over quarter. We are also accelerating cleanroom construction, including Fab 62 in Arizona, and exploring opportunities to accelerate the Ohio project. Despite higher capital expenditure and working capital investment, we remain committed to reducing leverage in 2027. We will remain disciplined in our investments and will not spend even US$1 of capital unless we are confident in the health of our products and process technologies and see clear evidence of customer demand. Our confidence in all three areas is increasing. Finally, I want to summarize in a few sentences the future opportunity we see. Our core server CPU market is growing at an unprecedented rate; despite the challenging current market environment, edge AI and physical AI still present important incremental opportunities for client-class CPUs. Importantly, as product execution and roadmap improve, we have an opportunity to regain market share and grow faster than our markets. Beyond our core x86 business, we are investing in three large opportunity areas: custom chips, advanced packaging, and external foundry. We have made clear progress in these areas, but revenue is still relatively limited today. Each represents a tens-of-billions-of-dollars new market, and our technology, manufacturing assets, and customer relationships provide differentiated advantages for participating in them. Taken together, we can already see a path to long-term double-digit revenue growth, significant operating leverage, and improved margins. There is still a great deal of work ahead. However, a stronger market and the cultural changes already underway to improve execution, transparency, and discipline are increasing our confidence in Intel Corporation's long-term profitability and free cash flow growth and laying a solid foundation for creating sustainable shareholder value. I will now turn the call back to John to begin the Q&A session. John? John Pitzer Thank you, Dave. As a reminder, to give as many callers as possible a chance to ask questions, please ask one question and one brief follow-up. So, Jonathan, shall we begin with the first question? Below is the Q&A session of the earnings call (AI-assisted summary and translation): Conference Operator Certainly. The first question comes from Timothy Arcuri of UBS Group AG. Please go ahead. Q: Analyst Thank you very much. Dave, given the various puts and takes and the changes taking place in the PC market, I wanted to ask whether you could give us some rough ranges, perhaps for next year? I mean, it looks like Data Center and AI revenue could increase by US$10 billion to US$15 billion year over year, while the PC business, I guess, could decline slightly. So could you give us some rough full-year framework and how we should think about the cadence through the year? A: David Zinsner Okay, thanks, Tim. Of course, I will not give guidance for next year now, but I can provide some qualitative information. Overall, we believe next year will be a growth year. We believe the momentum in the data center business this quarter, excluding the impact of the Israel incident, would have continued into the fourth quarter. We expect that momentum to continue into next year as well. The only small caveat is that we are still somewhat constrained in the first half of next year. We talked about adding 40% substrate capacity, but a substantial portion of that will only show up in the second half of next year. In addition, Intel 3 remains capacity constrained. As Lip-Bu said, we will work to shift some supply to Oregon, but it takes time from fab output to ultimately reaching products. Still, we are very optimistic and very excited about the growth opportunity in the data center business next year. Also, we talked about the custom chip business within the Data Center and AI segment. That business has tripled over the past three years, and we expect it to triple again from the current level over the next three years. So we are very positive on its full-year 2027 performance. Obviously, the Client business will be the hardest part to judge. As I said, in unit terms, we expect the total market size to decline about 12% in 2026 and about 10% next year. That is one factor. There is also inventory adjustment. We are currently in the middle of that adjustment. We felt some impact in the third quarter and will feel a lot in the fourth quarter. We believe this will continue at least into the first quarter. After that, I think the inventory adjustment component will begin to fade, and the business will again depend more on changes in the overall market size. As I said, we expect the market to decline about 10%. In addition, next year our product lineup will include Nova Lake, which we believe will be the best set of products we have had in a long time. We believe this set of products covers the range of needs required to compete in the Client market. We are very excited about it. We believe that next year and in the years after, our Client business will be in an environment where it can gain market share. So unfortunately, we have the memory challenge and its impact on Client, but we are very confident in the products themselves. Also, although you did not ask this, I want to extend the view beyond this year. Despite these challenges, we do believe there will be a PC renaissance. As AI workloads begin to move across devices, the client space will need to carry a lot of workloads, whether on PCs or on some kind of agentic device. So we think this market is quite exciting. Of course, it first needs to get through this difficult period, but over the longer term, we are very confident in this market. A: John Pitzer Tim, do you have a brief follow-up? Q: Analyst Yes. Yes. Dave, you did say capex will increase substantially next year. I think many of us are thinking around US$30 billion. But I think the real question is: are you committing to positive free cash flow next year, or do you think the capex burden will make that difficult to achieve? Thank you. A: David Zinsner Yes, good question. We have not finalized the specific number yet. We typically do not finalize it until closer to the end of the year, but we already have a directional view of what will broadly happen. Given lead times, we have already placed purchase orders with equipment suppliers, so we have some understanding of what comes next. I think next year is an investment year for us. On a full-year basis, next year will probably not be a year in which we can achieve positive free cash flow. But that is because we believe this is a huge opportunity to capture growth through investment. We have already seen that if we invest correctly and ramp process nodes in the right way, we can achieve profitability and earn a return on invested capital. When we did the equity financing in early August, we talked about this: we believe investing in these process nodes should earn at least a 20% return on investment. However, next year will be a year in which multiple things happen at once. We are ramping 18A, also ramping Intel 3 to some extent, and at the same time making appropriate investments to prepare 14A for production starting in 2028. That means a large amount of investment will be concentrated in the same year. So I do not think we will be able to achieve positive cash flow next year. By the way, that is exactly why we raised capital in early August: to put the company in a position to make these investments. Over the longer term, however, these businesses can generate substantial cash flow. Take the current Intel 7, or 10nm/Intel 7 process, as an example; it is already a node that generates profitability for us. Intel 3 will also be profitable next year. So we know we can make these businesses profitable, know we can generate good cash flow from them, and know we can earn a 20% return. We just need to get through 2027, which is basically a bottleneck period, to achieve these goals. A: John Pitzer Thanks, Tim. Jonathan, please take the next question. Conference Operator Okay. The next question comes from Ben Reitzes of Melius. Please go ahead. Q: Analyst Okay, thanks very much. I wanted to ask about some of the fourth-quarter issues you mentioned, which may also extend into the first quarter. Could you quantify the impact relative to prior expectations and say when they will improve? You mentioned execution issues. Obviously, the Israel facility was just mentioned at US$300 million, and I want to know how large the impact will be in subsequent quarters and when it will be resolved. By the way, I am sorry that this happened. And then there is the Client business. Given that guidance is about US$1.8 billion below Wall Street expectations, could you talk about each of these three factors and how much each contributed to that US$1.8 billion gap, and when they will improve? That would be very helpful. Thank you very much. A: David Zinsner Okay. That was the Israel facility, Client, and what was the third one? A: John Pitzer The gap versus guidance and Wall Street expectations? A: David Zinsner Yes. Okay. Yes. [The end of the original transcript is unclear] Q: Analyst Execution, Dave. You made some comments about execution. I want to know how much that factor contributed to the guidance cut. A: David Zinsner Okay. On the Israel facility, we had about a two-week shutdown. We believe we should be able to begin resuming production by the end of this week. As we resume production, we are evaluating all wafers in the line to determine which materials can still be used and which cannot. So this is still a preliminary assessment for both the fourth quarter and the first quarter. However, the current judgment is that the fourth quarter will be affected by about US$300 million, and the first-quarter impact will likely be similar. We believe most of the wafer write-offs or scrapping will likely occur in the fourth quarter. That is basically what causes the 200 basis point impact. So I do not expect additional wafer scrapping in the first quarter. But keep in mind, it is still early. We are working through these items one by one and will of course update everyone as the work progresses. On Client, obviously, the business will reset to the current level. I think inventory adjustments are usually very similar over two consecutive quarters. So I expect the overall situation to be broadly similar, although first-quarter Client will also be affected by seasonality. So whatever level the fourth quarter ultimately reaches, the next quarter will likely see some seasonal decline as well. Beyond that, from an execution perspective, I need to think about what the negative factors are, but much of the execution progress is actually very positive. I mean, [name transcription unclear] and the team have done a great job improving Intel process yields, improving 18A yields, and working with substrate suppliers. Lip-Bu Tan was in Japan meeting with multiple substrate suppliers to ensure we can get appropriate supply allocation. So I think we are in as good a position as anyone in terms of delivering supply. Supply is indeed tight, will be tight next quarter, and will be tight in the first half of next year. But I think we are well positioned, and after that, the data center business should begin to show fairly healthy growth. A: John Pitzer Ben, do you have a brief follow-up? Q: Analyst Yes. On the server business growth rate, despite shortages and other issues, do you know when you will be able to increase output so that growth can keep pace with a market heading toward more than US$200 billion? Based on the comments, is it more likely in the second half of next year? Thank you. A: David Zinsner Yes, good question. I... A: Lip-Bu Tan I think, Ben, first, the server product that generates the most revenue for us is Granite Rapids on Intel 3. Obviously, we are continuing to improve yields, thereby improving efficiency. Second, we are also working to expand capacity. I think Dave just mentioned Oregon. We plan to bring Intel 3 capacity there into production in the second half of next year to meet some customer demand. So overall, this should help address demand on Intel 3. Next, we have Clearwater Forest on 18A, which has already begun