The Zhitong Finance App learned that the core content of the US CPU and chip manufacturing giant Intel (INTC.US) performance conference call, which has been one of the core beneficiaries of the global AI intelligence frenzy since this year, shows that AI server demand and manufacturing execution are improving, but there is still a clear time gap between the weak PC market, supply bottlenecks associated with chip manufacturing capacity, and a new round of capital investment, resulting in a significant improvement in profit trajectory and accelerated free cash flow.
According to performance data, Intel's total revenue for the third quarter reached 16.5 billion US dollars, with non-US GAAP gross margin of 44.1%, and earnings per share of 0.52 US dollars, higher than the company's previous guideline of 0.38 US dollars; data center and artificial intelligence business revenue was 7.3 billion US dollars, up 17% month-on-month and nearly 80% year-on-year increase (78% accurate data). However, fourth-quarter revenue guidance falls to $15 billion to $16 billion, non-GAAP gross margin is expected to fall to 38%, and earnings per share are expected to fall to $0.25. The Israeli factory accident is expected to affect fourth-quarter revenue of about 300 million US dollars and gross margin of about 2 percentage points. The revenue impact for the first quarter of next year is likely to be similar.
According to the results conference call, management expects weak PC demand to continue until 2027, and management clearly distinguished between “reduced pressure to remove inventory” and “recovery in terminal demand.” Intel management expects global PC market shipments to drop by about 12% in 2026, with a drop of more than 20% in the second half of the year; a further decline of about 10% in 2027. Customer inventory adjustments were mainly concentrated in the fourth quarter of 2026 and the first quarter of 2027, and may extend slightly into the second quarter. After that, the relationship between the company's shipments and terminal consumption is expected to return to normal, but this does not mean that the pressure on memory supply and price has been lifted. The chief financial officer made it clear that the memory chip environment is not expected to improve significantly in 2027, and whether it improves remains to be seen in 2028. The company also anticipates that the potential market for server CPUs could expand from the current $40 billion to more than $200 billion within five years.
Capital expenditure and return on cash are the most important forward-looking information in this call. The total capital expenditure is expected to be around US$20 billion in 2026, and the exact amount is yet to be determined; the approximately US$30 billion proposed by analysts during the questioning session is only an assumption in the questioning process and has not been confirmed as official guidance by management. The Chief Financial Officer defined 2027 as an “investment year,” saying that it is expected that it will be difficult to achieve positive free cash flow next year in the face of intensive investment such as the accelerated expansion of production capacity of Intel 3 and 18A and 14A to prepare for production in 2028.
The company's operating cash flow for the third quarter was approximately US$5.1 billion, with cash and short-term investments of approximately US$54 billion at the end of the quarter, and an additional US$15 billion revolving credit borrowing capacity, which provided an important buffer for the heavy pressure on investment costs. Intel management still aims for a return on investment of at least 20% for relevant cutting-edge process nodes (such as Intel's Intel 3, Intel 18A, and Intel 14A, involving transistor structures, manufacturing processes, and supporting production equipment), and promised to continue reducing leverage in 2027, but these are future goals and expectations.
Management said that Intel's long-term growth logic still revolves around server CPUs, dedicated chips, advanced packaging, and external foundry. Management expects industry server CPU shipments to increase by more than 30% in 2026 and 2027, and supply constraints will continue until 2028; Intel management emphasized during the conference call that with the expansion of workloads dominated by AI agents, the potential market size of data center server-side high-performance CPUs may increase from 40 billion US dollars to more than 200 billion US dollars within 5 years.
Granite Rapids has increased the company's gross margin. The cost of 18A decreased by about 40% from the first quarter, but the foundry business still recorded an operating loss of about 1.9 billion US dollars, and the month-on-month increase in external revenue mainly came from non-recurring customer projects; Granite Rapids is the product code name for Intel's sixth-generation Xeon (Xeon 6) advanced performance core server CPU. The core computing chip uses Intel 3 processes and is mainly used in traditional data centers, cloud computing, and large-scale data centers dominated by AI-related computing tasks. Management is also promoting commercialization of inference GPUs, modular x86 data center CPUs, advanced packaging, and quantum computing. As a result, the investment research focus of the entire conference call was on three aspects: whether demand can be converted into actual supply, whether expanding production can improve unit costs, and when additional investment can be converted into sustainable free cash flow.
The following is the full text of the Intel (i.e. Intel) earnings call (artificial intelligence tool assisted translation):
conference operator
Thank you for your patience and welcome to Intel's third quarter 2026 results conference call. Currently, all participants are in listen-only mode. Following the management's presentation, there will be a question and answer session. (Conference operator instructions) I would like to remind everyone that today's meeting will be recorded. Now, allow me to introduce today's presenter, Mr. John Pitzer, VP of Investor Relations. Please start, sir.
John Pitzer, Vice President of Corporate Planning and Investor Relations
Thank you, Jonathan, and I would like to say good afternoon to everyone attending today. By now, you should have received the third quarter results announcement and presentation materials. Both documents are available on our investor relations website intc.com. Everyone attending the conference online today can also view the presentation materials in the webcast window.
Joining me today are CEO Chen Liwu and Chief Financial Officer Dave Zinsner. General Liwu will first introduce the third quarter results and our progress in various strategic priorities. Afterwards, Dave will discuss overall financial results, including guidance for the fourth quarter, and then we'll answer your questions.
Before we begin, please note that today's presentation contains forward-looking statements based on our current observations of the environment. As a result, these statements are subject to various risks and uncertainties. The presentation will also mention non-GAAP financial indicators, which we believe can provide investors with useful information.
Our earnings announcements, recent Form 10-K annual report, and other documents submitted to the U.S. Securities and Exchange Commission provide more information on specific risk factors that may cause actual results to differ materially from our expectations. These documents also provide more information on non-GAAP financial measures, including reconciliation instructions with corresponding US GAAP financial measures, where applicable. Next, I'll hand over the phone to Liwu.
CEO Chen Liwu
Good afternoon, and thank you all for attending the meeting. The demand environment in the third quarter was more balanced than in previous quarters. Strong momentum in the data center and artificial intelligence business drove performance beyond expectations and enabled the business to achieve its strongest revenue growth in 15 years. This increase was partially offset by a weak client business, which was affected by the industry as a result of rising memory prices and limited IT budgets. The unprecedented demand for superintelligence is causing the entire industry to experience one of the worst supply shortages in history. In this context, I am pleased with the progress we have made in increasing our server supply in the third quarter. At the same time, I am aware that we will continue to face challenges in order to keep up with server demand over the next few quarters.
We successfully completed the equity issuance in August, providing the necessary financial resources to seize future growth. In addition to current shareholders, I am excited to welcome new sovereign funds and long-term growth investors. We are deploying this capital quickly and judiciously. We expect capital expenditure and key component procurement commitments to increase significantly over the next few quarters.
I want to thank our substrates, memory, foundry and equipment suppliers for their close partnership and collaboration as we work together to expand supply and support growing demand.
Let's talk about each business unit. The first is the data center and artificial intelligence business. In the server business, we are steadily advancing our product roadmap. Using the Intel 3 process, Granite Rapids has become one of the fastest growing server products in production. Clearwater Forest is our first Intel 18A server product. Shipments have now begun and revenue has been confirmed. Demand for 2027 is impressive, especially in the field of CPU-intensive infrastructure for large-scale agents.
Diamond Rapids, our first Intel 18A-P server product, is currently being sampled and is scheduled to expand shipments in 2027. Customers are showing keen interest in its application in inference and intelligent workloads. We are very happy that we have recently announced an agreement with Oracle Cloud Infrastructure to deploy our latest generation CPU to support the construction of its AI infrastructure.
I would also like to explain what negative factors are limiting our progress in meeting this extremely high demand driven by AI. First, the yield and production cycle of Intel 3 continues to improve, but I'm still asking the team to significantly improve execution effectiveness. Second, even though our Intel 3 wafer production is rapidly increasing, the entire industry is experiencing a shortage of advanced organic substrates for EMIB (inaudible) packaging, which is affecting our ability to expand output.
Finally, earlier this month, an unfortunate and tragic accident caused our factory in Israel to shut down production for a long time. Despite the team's quick response, the incident will still affect shipments of 10nm server products in the fourth quarter. Specialized chips are still our fastest-growing business segment. The business scale has tripled in the past three years, and is expected to triple in the next few years.
We see that there are still opportunities for further growth in partnership 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 wider customer base.
Finally, we recently released Crescent Island GPUs optimized for inference. Initial feedback has been encouraging, and performance is progressing well. For us, this is a new market, and scaling up will take time, but engagement with customers is moving in the direction of a clear commitment. Let's talk about client business. In recent weeks, customers have begun to signal to us that the PC market is slowing down due to memory prices and availability. Since this year, the price of complete terminals has risen significantly, putting additional pressure on the already limited IT budget.
Over the past two months, the pace of this round of PC market adjustments has accelerated. Our OEM partners are actively managing inventory to accommodate weakening end sales. We expect this environment to continue until memory supply and prices begin to return to normal. Despite the challenges the PC market will face in the coming quarters, our product roadmap continues to strengthen, and the long-term outlook remains positive. We have successfully promoted mass production of Panther Lake using the Intel 18A process, and demand for Wildcat Lake has exceeded supply. Nova Lake is our first Intel 18A-P client product. It is currently progressing ahead of schedule, and it is expected that the number of design imports received at launch will reach twice that of [original text lacks comparison objects]. We look forward to sharing more information at CES in January. As the market returns to normal, we expect this product portfolio to drive a significant increase in market share. Intel's foundry business execution performance continues to improve. This quarter, the yield of both Intel 3 and Intel 18A increased again. Among them, the yield increase of Intel 18A reached double digits. Both processes surpassed expectations in wafer output. The Intel 18A-P is still on schedule to launch in the internal product portfolio before the end of the year, and engagement with external customers continues to increase. We are working with several highly strategic customers, and I am increasingly confident that the Intel 18A-P will be an attractive process node for outside foundry customers. [The original does not specify the name of the process here] is also still progressing according to plan. In October, we delivered on our promise and released the 0.9 process design kit in line with industry standards to selected customers. This is our highest-quality 0.9 process design kit, which met the yield and performance specifications required by industry standards at the time of release. As is customary in the foundry industry, I always insist on not disclosing customer details. Nevertheless, I am still very encouraged by the progress we have made on 14A with a number of potential clients. In terms of advanced packaging, we are pleased to see the progress made by EMIB and EMIB-T on internal products, and these two technologies also continue to attract strong interest from external customers. First, we continue to improve execution and cooperate with the supply chain to implement the production capacity necessary to meet the needs of advanced packaging customers. We have also seen solid progress in preparations for hybrid bonding interconnection. This technology will support our upcoming server products to achieve 3D chip stacking. New production capacity is being launched one after another throughout the manufacturing network. The installation of Intel 3 equipment in Ireland is progressing smoothly, as is the 18A capacity climb in Oregon and Arizona, including our second 18A mass production plant, Fab 62. In New Mexico and Malaysia, we are increasing production capacity to support our customers' strong volumes in advanced packaging next year. In New Albany, Ohio, the team recently completed the installation of the first large-scale structural truss, achieving 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 information soon. Looking ahead, I've always emphasized that rebuilding Intel will be a journey that will take many years, and my commitment to this is unwavering. We only embarked on this journey for a year and a half. Over these 18 months, we've made tremendous progress, including drastically resizing our workforce, reducing middle management and bureaucratic processes, streamlining the executive team, and forming an almost brand new leadership team. We've re-established an engineering-centric culture, streamlined our product portfolio, and established a more competitive roadmap for Intel products and Intel's foundry business. We still have a lot of work to do. Demand for CPUs has once again increased, providing a significant boost to our transformation journey. Energy efficiency is increasingly becoming the most important measure for AI infrastructure construction. GPUs are good at processing thinking tasks, while CPUs with deterministic execution characteristics are more efficient in processing search tasks. A GPU inference query can consume as much as several watt-hours of power, while a CPU query may only require energy measured in joules. Even with conservative estimates of the total number of digital agents and the number of agents served by each CPU, it is easy to predict that the total potential market size of CPUs in terms of value and quantity over the next five years will be huge. According to our estimates, the total potential market size of server CPUs, mainly driven by workloads, could increase fivefold within five years, from $40 billion to more than $200 billion. This presents a huge opportunity for Intel. Of course, we also know that fierce competition is unfolding, and almost every waferless chip company is racing to launch their own products to meet this unprecedented demand. But Intel is uniquely positioned to benefit from this demand in three ways. First, as x86 [original text missing] — x86 is the most widely deployed instruction set architecture in the world — we are doubling down on x86 innovation and have an exciting roadmap for CPU innovation using an “tick-and-tick” approach. Additionally, 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 semiconductor foundry companies in the world, and the only American company among them, we have favorable conditions to not only meet our own needs, but also support the broader needs of the fabless chip industry. We are also willing to manufacture any CPU on a large scale for our customers, without being limited by the underlying instruction set architecture. Third, modern computing platforms are increasingly evolving into complex system-level packages, and our leading position in the field of advanced packaging can provide highly differentiated solutions, particularly to meet the current needs of large-scale AI infrastructure construction. That's why even though there's still work to be done, I'm still optimistic about the opportunities that lie ahead for Intel. Finally, while we focus on implementing our immediate priorities, we must not lose sight of upcoming developments. Intel has a deep technology and intellectual property portfolio that has the power 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 (company name missing here) to move our quantum business from basic research to commercialization. This company [original company name not clearly transcribed] is a software platform layer that targets a variety of hardware and does not rely on specific hardware, providing an entry point for more customers to explore quantum computing. It also complements our already very powerful internal quantum hardware and software technology stack, further expanding our computing platform products to quantum processors in addition to CPUs, GPUs, and special integrated circuits. Once the acquisition is completed, Intel will become a quantum computing company with a complete technology stack and generating revenue. All in all, despite our strong performance in the third quarter, there is clearly more work to be done. I'm disappointed that we haven't been able to meet strong customer demand this quarter. We are facing recent challenges, including manufacturing supply shortages, an unprecedented shortage of memory, and the resulting weak demand for PCs. Nevertheless, I'm still confident about the path ahead. We need to urgently address these challenges without losing focus on the huge opportunities that lie ahead. I want to thank our employees, partners, and suppliers for their tireless efforts this quarter. Next, I'll hand over the phone to Dave and ask him to provide more details.
David Zinsner, Executive Vice President, Chief Financial Officer and Head of Finance and Accounting
Thank you, Liwu. We had a strong quarter, with revenue, gross margin, and earnings per share exceeding our guidelines. This is the eighth consecutive quarter that we have surpassed our financial expectations. The third quarter results further highlight the important role that our server CPUs play in the construction of super intelligent or AI infrastructure. The results also show that execution performance is improving as we increase supply to capture strong demand for AI computing power.
Third-quarter revenue was $16.5 billion, and record year-on-year growth in the data center and artificial intelligence business was the main driver. Excellent execution of the factory network and external supply chain drove server CPU output to be better than expected. This strong performance was partially offset by weak client services, and memory shortages and limited IT budgets are affecting PC consumption across the industry. Non-GAAP gross margin for the third quarter was 44.1%, about 200 basis points higher than the second quarter and our guidance. The month-on-month improvement was driven by cost improvements in Intel 3 and Intel 18A, as well as a more favorable product portfolio.
Third-quarter non-GAAP earnings per share were $0.52 compared to our guidance of $0.38. Exceeding expectations was mainly due to higher revenue, stronger gross margins, and strict expenditure management. Together, these factors brought better-than-expected operating leverage. Operating cash flow was $5.1 billion. At the end of the quarter, we held approximately $54 billion in cash and short-term investments. In addition, the newly extended revolving credit facility also provided $15 billion in borrowing capacity, up from the previous $10 billion.
At the beginning of August, we further strengthened our balance sheet through a $23 billion stock offering. The additional interest income from this offering contributed approximately $0.01 per share to third-quarter earnings. This portion of earnings was not included in the initial guidance. Segment performance is described below. Data center and artificial intelligence business revenue was $7.3 billion, up 17% month-on-month and nearly 80% year-on-year. Revenue clearly exceeded expectations as our factory network and external suppliers supported higher output.
The team's work to increase supply is yielding real results. According to the current outlook, the annual revenue of the data center and artificial intelligence business in 2026 is expected to be nearly $10 billion higher than our expectations at the beginning of the year. Growth in the third quarter came mostly from hyperscale cloud service providers, enterprises, and new cloud customers. The special-purpose chip business also continued to maintain a strong momentum, with revenue growth of more than 50% month-on-month and more than doubling year-on-year.
Despite our progress in expanding production capacity, demand for data center and artificial intelligence products still clearly exceeds what is available. The operating profit of the data center and artificial intelligence business was 3 billion US dollars, accounting for 42% of revenue, an increase of about 600 million US dollars over the previous month. This improvement was mainly driven by increased revenue and lower operating expenses. As production from the Irish factory continues to increase, Granite Rapids is now the largest product line in our server product portfolio.
Earlier this month, at the Open Computing Project Global Summit, we showcased some Clearwater Forest configurations. Each rack can be expanded to over 46,000 cores, and the total rack power consumption is approximately 133 kilowatts, achieving best-in-class performance per watt. We've also launched the Intel GPU D series, previously codenamed Crescent Island, to improve the economics of continuously running memory-intensive AI workloads. Customer interest in both product lines surpassed initial expectations.
Client business revenue was $8.1 billion, down 9% month-on-month, below our expectations. We previously indicated that the shortage of memory may put pressure on the overall PC market size in the second half of the year. This pressure intensified in the third quarter, particularly at the end of the quarter, and led to significant adjustments in customer procurement models.
Despite short-term market weakness, we have seen several positive developments. While overall client revenue declined 5% year over year, AI PC revenue increased 35% year over year. Client products using the Intel 18A process continue to climb smoothly. Wildcat Lake is our cost-optimized client CPU using the Intel 18A process. Shipment volume increased by about 70% month-on-month, and the product was sold out.
Finally, our edge business performance is more resilient, and the edge AI and physical AI fields continue to receive design introduction to support them. The operating profit of the client business was US$2.3 billion, accounting for 28% of revenue, which was basically flat from month to month, mainly due to lower inventory-related expenses.
Let's talk about Intel's foundry business. Intel's foundry revenue was $6.7 billion, up 16% month-on-month, mainly driven by higher-than-expected growth of Intel 3 and Intel 18A. External foundry revenue was $411 million, an increase of $118 million over the previous month, mainly from a non-recurring customer project. Intel Foundry recorded an operating loss of 1.9 billion US dollars, narrowing by 236 million US dollars from month to month, even though we are still actively investing to prepare for risky trial production of the Intel 14A next year.
Higher yields, shorter production cycles, and increased plant capacity utilization continue to drive wafer costs down. The cost of Intel 3 is about 15% higher than the level of the first quarter, and the month-on-month increase in production is in the middle of about 10%.
Intel 18A costs have dropped by about 40% since the first quarter, reflecting continued improvements in yield and production cycles. The Intel 18A yield continues to rise towards industry standards, and progress is ahead of internal forecasts. Since the first quarter, Intel 18A client CPU's share of revenue has risen from less than 5% to now close to one-third of client CPU revenue. These results represent substantial progress, but we still have a lot of work to do to make the foundry business sustainable and profitable.
Let's talk about the outlook. Demand signals from data center customers continue to grow. We expect industry supply constraints to run through 2027 and continue until 2028. As Liwu said, the application of AI is broadening the role of CPU computation. AI workloads are shifting from cues-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 artificial intelligence products. We currently expect the industry's server CPU shipments to increase by more than 30% in 2026 and 2027. If the industry can provide more supply, the increase in these two years will be even higher. Meanwhile, the PC market is weakening after remaining resilient in the first half of 2026. We expect overall PC market shipments to drop by about 12% in 2026, with shipments falling by more than 20% in the second half of the year. We don't expect a continued recovery until memory supply improves and prices begin to return to normal.
For 2027, we are planning based on a decrease in overall PC market shipments of about 10%. We expect the year-on-year decline in the first half of the year to be more pronounced, while the second half of the year may improve due to a lower comparison base. Our long-term view of the PC market has not changed. Traditional PCs will play an important role in supporting AI applications and driving strong returns on AI infrastructure spending.
Furthermore, we are only just beginning to see new smart 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, our fourth-quarter revenue guidance is $15 billion to $16 billion. Based on a median value of $15.5 billion, we expect gross margin of 38%, tax rate of 11%, and earnings per share of $0.25, all of which use non-US GAAP measures. Based on an initial assessment of the Fab 28 incident, our guidance has taken into account the revenue impact of approximately $300 million and the gross margin impact of approximately 200 basis points. By business, we expect the data center and artificial intelligence business revenue to be in the middle of about 10% month-on-month, and the special chip business will grow faster than the sector as a whole. Note that almost all of the $300 million impact caused by Fab 28 involved 10nm server products. We expect client business revenue to drop by more than 20% month-on-month as customers are rapidly reducing inventory weeks and matching procurement expectations with 2027 terminal demand. Gross margin declined sequentially in the fourth quarter, mainly reflecting a decline in overall revenue when we adjusted client wafer delivery volumes based on current demand, while the fixed cost basis still existed. Furthermore, the expected strong growth in the revenue mix, particularly in the dedicated chip business, made the quarter-on-month comparison even more unfavorable. Continued improvements in yield and cost of the Intel 3 and Intel 18A will partially offset these negative factors. We expect non-GAAP operating expenses for the fourth quarter to be approximately $4.4 billion, and approximately $16.3 billion for the full year. We estimate that the amount corresponding to non-controlling equity in the fourth quarter will be approximately US$250 million, and currently we expect to be approximately US$1.1 billion in 2027 and 2028. Finally, we expect the number of fully diluted shares to be 5.44 billion in the fourth quarter, reflecting that the additional shares brought about by the August stock issuance will be included in the full quarter. Let's talk about capital expenditure. Total capital expenditure in 2026 is still at $20 billion as planned, and we expect a significant increase in total capital expenditure in 2027. We have made good progress in obtaining the equipment needed to support our growth plans, increasing confirmed purchase orders to equipment suppliers by nearly 40% month-on-month. We're also speeding up cleanroom construction, including Arizona's Fab 62, and exploring opportunities to speed up the Ohio project. Despite increased capital expenditure and working capital investment, we are committed to reducing leverage in 2027. We will maintain investment discipline and will not invest even $1 in capital unless we are confident in the health of our products and process technology and see clear evidence of customer demand. Our confidence is growing in all three areas. Finally, I would like to summarize in a few words the future opportunities we see. Our core server CPU market is growing at an unprecedented rate; despite the current challenging market environment, edge AI and physical AI present significant incremental opportunities for client-level CPUs. Importantly, as product execution and roadmap improvements improve, we have the opportunity to regain market share and grow faster than in our own market. In addition to our core x86 business, we are also investing in three large areas of opportunity: specialty chips, advanced packaging, and external foundry. We have made significant progress in these areas, but our revenue scale is still relatively limited. Each sector represents tens of billions of dollars in new markets, and our technology, manufacturing assets, and customer relationships provide a differentiating advantage to participate in them. Taken together, we can already see a path to double-digit growth in long-term revenue, release significant operating leverage, and improve profit margins. We still have a great deal of work ahead of us. However, a stronger market, and cultural changes that have been initiated to improve execution, transparency, and discipline, have continued to increase our confidence in Intel's long-term profitability and free cash flow growth, and have laid a solid foundation for creating sustainable shareholder value. Next, I'll return the phone to John and begin the question and answer session. John?
John Pitzer
Thanks, Dave. As a reminder, to give as many callers as possible a chance to ask questions, please ask each one question and ask a brief question. So Jonathan, can we start with the first question?
The following is the performance conference Q&A session (AI-assisted summary and translation):
conference operator
Of course. The first question came from UBS's Timothy Curry. Please ask questions.
Q: Analyst
Thank you very much. Dave, considering all the positive and negative factors and the changes that are taking place in the PC market, I was wondering if you could give us some general scope of judgment, maybe talk about next year? I mean, it looks like the data center and artificial intelligence business revenue may increase by 10 billion to 15 billion US dollars over the same period last year, while the PC business, I think, may have declined slightly. So, can you give us some general judgments for the whole year and how should we understand the pace of development throughout the year?
Answer: David Zinsner
OK, thanks, Tim. Of course, I'm not going to give next year's guidance now, but I can provide some qualitative information. Overall, we think next year will be a year of growth. We believe that the growth in the data center business this quarter would have continued into the fourth quarter if the impact of the Israeli incident were excluded. We expect this trend to continue into next year. The only minor detail to be aware of is that we will still be subject to certain restrictions in the first half of next year. We've talked about increasing board production capacity by 40%, but quite a bit of it won't be reflected until the second half of next year.
Additionally, Intel 3 is still subject to capacity constraints. As Liwu said, we will try to move some of our supply to Oregon, but it will take some time from the fab to finally enter the product. However, we are very optimistic and excited about the growth opportunities for the data center business next year.
Additionally, we mentioned that there is also a dedicated chip business within the data center and artificial intelligence division. The business has tripled in the past three years, and we expect it to triple the current level in the next three years. Therefore, we are very optimistic about its 2027 full year performance. Obviously, the client business will be the hardest part to determine.
As I said, in terms of shipment volume, we expect the overall market size to drop by about 12% in 2026 and around 10% next year. This is one of the influencing factors. There are also inventory adjustments. We are currently in the process of making this adjustment. The third quarter has already felt some impact, and the fourth quarter will have a big impact. We think this will continue until at least the first quarter.
After that, I think the impact of inventory adjustments will begin to weaken, and the business will once again depend more on changes in overall market size. As I said, we expect the market to drop by around 10%. Additionally, we'll include Nova Lake in our product lineup next year, and we think this will be the best set of products in a long time.
We believe that this set of products covers all the requirements needed to compete in the client market. We're very excited about that. We believe that next year and the next few years, our client business will be in an environment where we can increase our market share. So unfortunately, we have experienced memory challenges and their impact on clients, but we are very confident in the product itself.
Also, even though you haven't asked this question, I'd like to extend my perspective beyond this year. Despite these challenges, we do think PC will see a revival. As AI workloads begin to shift between devices, the client sector will need to take on a large workload, whether it's a PC or some kind of smart device. So we think this is a market worth looking forward to. Of course, it will first need to get through these difficult times, but in the longer term, we are very confident in this market.
Answer: John Pitzer
Tim, do you have a quick question?
Q: Analyst
Yes. Yes. Dave, you did say that capital spending will increase dramatically next year. I think the size many of us are considering is about $30 billion. But I think the real question is: do you promise to achieve positive free cash flow next year, or do you think capital expenditure burdens will make this goal difficult to achieve? Thank you.
Answer: David Zinsner
Yes, that's a good question. We have yet to finalize the exact numbers. We usually don't know until the end of the year to decide, but we already have a directional judgment on what will happen. Considering the delivery cycle, we have placed a purchase order with the equipment supplier, so we have a certain understanding of the upcoming situation.
I think next year will be an investment year for us. Looking at the whole year, next year will probably not be a year where positive free cash flow can be achieved. But that's because we believe the present is a huge opportunity to seize growth through investment. We have seen that as long as the right investment is made and the process node mass production climbs in the right way, profits can be achieved and return on invested capital can be achieved. When equity financing was carried out in early August, we talked about it: we believe that by investing in these process nodes, we should be able to obtain a return on investment of at least 20%.
However, next year will be a year where many issues are being promoted simultaneously. We are advancing mass production of the 18A, and to a certain extent, of the mass production of the Intel 3. At the same time, we also need to make appropriate investments to prepare for the 14A to begin production in 2028. This means that significant investments will be concentrated in the same year.
As a result, I don't think we'll be able to achieve positive cash flow next year. Incidentally, that's exactly why we raised capital in early August: to equip the company to make these investments. However, in the longer term, these operations can generate significant cash flow. Take the current Intel 7, or 10 nanometer/Intel 7 process, as an example. It is already a node that can bring us profit.
Intel 3 will also be profitable next year. So we know we can make these businesses profitable, we know we can generate good cash flow through these businesses, and we know we can get a 20% return. We only need to get past 2027, which is basically a bottleneck period, to achieve these goals.
Answer: John Pitzer
Thanks, Tim. Jonathan, please schedule the next question.
conference operator
OK. The next question comes from Melius's Ben Reitz. Please ask questions.
Q: Analyst
OK, thank you so much. I'd like to ask that some of the fourth quarter questions you mentioned might continue into the first quarter. Can you quantify the impact of these issues compared to previous expectations and indicate when they will improve? You mentioned implementation issues. Obviously, the Israeli factory situation was just mentioned at $300 million, and I wonder how much the impact will be in the next few quarters and when it can be addressed. BTW, I'm sorry that something like this happened. There is also a client-side business. Considering that the current guidance is about 1.8 billion US dollars lower than Wall Street expectations, can we discuss these three factors separately, how much each contributed to this $1.8 billion gap, and when will it improve? That would be very helpful. Thank you very much.
Answer: David Zinsner
OK. What I just mentioned - Israeli factory, client business, what is the third one?
Answer: John Pitzer
Is that the gap compared to guidelines and Wall Street expectations?
Answer: David Zinsner
Yes. OK. Yes. [The end of the original text is unclear]
Q: Analyst
It's the executive side, Dave. You have just made some comments on implementation. I would like to know how much of this factor contributed to the reduction in the guidelines.
Answer: David Zinsner
OK. Regarding the Israeli factory, we have been out of production for about two weeks. We think we should be able to resume production by the end of this week. While resuming production, we are evaluating all wafers on the production line to determine which materials can still be used and which cannot be used. Therefore, 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 300 million US dollars, and the impact of the first quarter is likely to be similar.
We believe most wafer write-offs or scraps will likely occur in the fourth quarter. This is basically what caused the 200 basis point impact. As a result, I don't expect additional wafer scraps in the first quarter.
Keep in mind, though, that it's still early days. We are currently addressing these issues one by one and will of course update you as the work progresses.
On the client side, it is clear that the business will be readjusted to the current level. I think inventory adjustments are usually very similar for two consecutive quarters. As a result, I expect the overall situation to be roughly similar, although the client business in the first quarter will also be affected by seasonal factors. So no matter what level the fourth quarter ends up at, the next quarter is likely to experience some seasonal decline.
Other than that, from an implementation perspective, I need to think about the negative factors, but a lot of implementation progress has actually been very positive. I mean, [name transcribed clearly] and the team have all done a good job in improving Intel's process yield, improving 18A yield, and cooperating with board suppliers. Chen Liwu previously communicated with a number of substrate suppliers in Japan to ensure that we were able to obtain proper supply distribution. So I think we're in a better position than any other company when it comes to delivering supplies. Supply is indeed tight. The next quarter will be tight, and the first half of next year will also be tight. But I think we're ready, and after that, the data center business should start growing quite healthily.
Answer: John Pitzer
Ben, do you have a quick question?
Q: Analyst
Yes. Regarding the growth rate of the server business, although there are problems such as shortages, do you know when production can be increased to keep pace with the growing market of more than 200 billion US dollars? Based on the comments above, is it more likely to be in the second half of next year? Thank you.
Answer: David Zinsner
Yes, great question. I...
Answer: Chen Liwu
In my opinion, first of all, the server products that brought us the most revenue were Granite Rapids using the Intel 3 process. It is clear that we are continuously improving yield and thereby improving efficiency. Second, we are also working to expand production capacity. I think Dave just mentioned Oregon. We plan to put the Intel 3 production capacity there into production in the second half of next year to meet some customer needs.
So overall, this should help address the needs of Intel 3. Next, we also have Clearwater Forest, which uses an 18A process, and has now begun shipping and revenue confirmation. We should have production capacity to support this product. Then there's the 2027 Diamond Rapids, and our 18A-P will be ready by then. Of course, Rapids products using the 14A process will also be launched in 2028. So I think these aspects will all be in pretty good shape.
Answer: John Pitzer
Thanks, Ben. Jonathan, please schedule the next question.
conference operator
OK. The next question comes from Morgan Stanley's Joseph Moore. Please ask questions.
Q: Analyst
OK. Along the same lines, does the weak PC business provide you with an opportunity to reallocate wafer production capacity to solve some of these problems and accelerate this transformation?
Answer: David Zinsner
The answer is yes. In order to introduce Intel 3 in Oregon, the original production capacity we are discontinuing is 18A wafer production capacity, and these wafers are usually used to produce Panther Lake products. As a result, as some capacity is freed up, we see an opportunity to use it to produce more wafers for data center products. Keep in mind, though, that we're not only limited on the wafer side, but also on the substrate side. Therefore, increasing the supply of substrates to the required scale is just as important to meet demand.
Answer: John Pitzer
Joe, do you have any follow-up questions?
Q: Analyst
Yes. What do you guys think of next year's server market share? I remember you mentioned a 30% increase in shipping volume, and AMD gave guidance for a 70% increase in revenue. The Arm camp also seems to be growing quite fast. I understand that you are constrained by supply, but what do you think of changes in market share?
Answer: Chen Liwu
Of course, I think this is a good question. Obviously, we have our own product roadmap. We are optimistic about the Rapids series products, which can compete with AMD. And in terms of [product name not clearly transcribed], we will be ahead of AMD. So stay tuned as we continue to drive product leadership. CPUs are indeed our core business, and we are determined to succeed in this field. I think this is a great opportunity for us. Maybe I can add one more thing: I do think next year's market share will depend on supply. As a result, the more our supply improves compared to our peers, the better our market share performance will be.
Answer: John Pitzer
Jonathan, please schedule the next question.
conference operator
OK. The next question comes from Stacy Rasgon of Bernstein Research. Please ask questions.
Q: Analyst
Hi everyone, thanks for answering my questions. I'd like to discuss gross profit margins further. I understand that your guide is 38%; if it weren't for the unfortunate incident at the Israeli factory, I'd guess 40%. But how should I understand the overall drivers as we look to next year? Obviously, the client business needs to clean up inventory, and there may also be capacity utilization issues, which is a negative factor. Increased capital expenditure may also drive up depreciation, which is also a negative factor.
As mass production of 18A and Intel 3 climbs, I think yield rates are improving, but maybe you can tell us if the gross margin of these products is still lower than [the original text transcribed as “10.7%,” which indicates 10 nanometer/Intel 7 products, which cannot be confirmed here]. I would like to know if this change in the product portfolio is a negative factor, as well as various businesses such as special purpose integrated circuits and packaging. However, on the other hand, the product portfolio's inclination towards servers may be an advantageous factor, and pricing may also help. I just wanted to combine all of these positive and negative factors to understand how gross margin would evolve in this environment. There really seem to be a lot of influencing factors, so I hope you can help us sort them out.
Answer: David Zinsner
That's a great question, Stacey. It's true that a lot of factors are changing. In my opinion, from a gross margin perspective, the impact of the Israeli factory incident is likely to be mainly concentrated in the fourth quarter. As far as wafer scrapping is concerned, I don't expect the first quarter to have a big impact. Usually, as soon as we can identify it, we scrap it for the season. Demand for Xeon processors is still very strong, and I think I've mentioned that Granite Rapids is now able to increase the company's overall gross margin.
As a result, Granite Rapids' growth helped improve gross profit margins. We'll also continue to focus on Intel 3 and 18A. Regarding their structural costs, we can further focus on improving various aspects such as yield and production throughput. We did a great job in the third quarter, and we expect to do better in the fourth quarter, with the goal of continuing these improvements throughout the year. You're right, as far as 18A is concerned, the capacity load will not reach the level we want. We are partially solving this problem by importing Intel 3 into Oregon, but the 18A will still have a certain degree of underutilization of production capacity, which will have some adverse effects.
Another downside — though, I think this is actually a good thing from the right perspective — is our specialty chip business. It is still in the early stages of development, and profit margins are below the company average. But we saw huge room for growth, and I did decide to focus on driving gross profit growth. Therefore, we will focus on increasing gross profit, which in turn increases profit. This may put some pressure on operating margin percentages next year, but I think it will benefit the final profit.
In my opinion, putting all of these factors together, the impact has pros and cons, and is so numerous that it's difficult to quantify. I still prefer a 40% to 60% incremental revenue to gross profit conversion ratio. This is probably what we're trying to achieve. So if revenue grows next year, we should be able to see an increase in gross margin according to this rule of thumb. This isn't necessarily exactly the case every quarter, but I think if you look at it on a year-over-year basis, you should be in this range.
Answer: John Pitzer
Stacey, do you have any follow-up questions?
Q: Analyst
Yes. I'd like to ask about the PC business. Judging from the month-on-month performance, the third quarter was probably the worst PC market third quarter I've seen in at least 20 years, or more. So it sounds like the customer probably built up inventory in the third quarter. Can you guys confirm this? Also, it sounds like they'll be drastically digesting inventory in the fourth quarter. I get the gist of it—probably the first quarter as well. It sounds like you think the situation will improve in the second half of the year; I'm guessing this is related to the supply of memory chips. What exactly makes you believe that the supply of memory chips will actually improve? If you listen to storage vendors, they don't seem to think there will be any improvement.
Answer: David Zinsner
Yes, let me be clear. That's a great question, Stacey. Let me be clear. You're absolutely right; we think the fourth quarter and the first quarter are likely to be the two most concentrated quarters for inventory adjustments, which will affect us more than the overall decline in market size. As a result, we'll see this impact — in the fourth quarter, the first quarter, and possibly a little further into the second quarter. After that, the situation will return to a level where, in theory, our confirmed sales volume roughly matches terminal consumption.
As a result, we'll get some level of recovery from it. But don't get me wrong; as far as memory chips are concerned, we're not optimistic about next year. We believe this will have an impact on the overall market size. We expect the total market volume to drop by 10% next year in terms of shipment volume, and then see if there is any improvement in 2028. However, our expectation is that 2027 will not improve.
Answer: John Pitzer
Thanks, Stacey. Jonathan, please connect to the next questioner.
conference operator
OK. The next question comes from Bank of America Securities's Vivek Alia.
Q: Analyst
Thank you guys for answering my questions. First, if you follow your implicit guidance for the fourth quarter data center business, I think the revenue was around $8.5 billion. If this level is annualized, then as a conceptual goal for next year, it would be around $35 billion. So I'm asking if this is an achievable lower limit? Can you guys do better? Considering that demand exceeds supply, is there any room for improvement in the average selling price? Or what other factors need to be considered when considering the size of your data center business in 2027?
Answer: David Zinsner
Yes, I think this is a great place to start and the right way to think. Our judgment is that the supply of wafers and substrates may be more tight in the first half of 2027. But we've already deployed a lot of work, including increasing Intel 3 wafer output, which I just mentioned, increasing product output by improving yield, and cooperating with suppliers to invest in substrate production capacity to release this production capacity in the second half of the year. So we do expect next year to be a year of strong growth. Also keep in mind that we have also classified the special chip business, that is, the special purpose integrated circuit business, into this division. As a result, this was one of the factors driving significant growth in the fourth quarter. We expect this trend to continue in 2027.
Q: Analyst [The original document combines the tracking tips with the analyst's questions]
Vivek, do you have any follow-up questions?
Yes, thank you. I would like to hear about the latest developments in your foundry business dealing with external customers, including packaging and, more importantly, wafer manufacturing. Liwu, do you expect 18A-P to get external customers, or should we assume that all customer engagements revolve around 14A? When do you think Intel will have enough confidence to move these projects from an evaluation phase to a streaming or substantial revenue phase? In this way, we can determine when Intel's external customer revenue should be included in the model in the next few years. Thank you.
Answer: Chen Liwu
OK, Vivek. Thank you so much for your question. Your first question is about encapsulation. Our EMIB is making a lot of progress. We have always attached great importance to substrate production capacity to ensure that the required production capacity can be obtained, and pay attention to the yield level corresponding to the supply of substrates and materials. Second, we're focusing on yield, and we've made significant improvements. The third focus is the production cycle.
Currently, under the leadership of Mr. Lee — who was the CEO of SK Hynix and an expert in process integration — the team is doing an excellent job and is very engaged. I think in the next two years, we will gain a large amount of production capacity in substrate materials, and yield will continue to improve. Overall, please stay tuned. Over the next two quarters, we will have several customers preparing to announce relevant news.
Of course, we always want to wait for the customer to make a decision first. If they want to make an announcement, we'll respond. Second, in terms of 18A-P, we have made great progress in mass production, [the original text is unclear], and Wildcat Lake. I think we are already using [original text not clear] and are quite satisfied with the performance.
Currently, external customers have also actively expressed interest, and we are working closely with them, and they are very satisfied with the yield. We are now focusing on chip die sizing to ensure delivery to customers while ensuring that the relevant intellectual property modules are ready. So stay tuned for 18A-P, we are making good progress.
On the 14A side, we have made tremendous progress. We have announced and released version 0.9 of the Process Design Kit to selected customers. They have responded and have been very active in working with us. On the other hand, whether it's performance, yield, indicators, defect density, or process variations, we continue to improve every day. So far, we are very satisfied. We should be able to launch version 1.0 of the process design kit in the first quarter of next year and begin collaborating with customers around the products they want our fabs to produce. So stay tuned. Also, we will announce some of these customers in due course when they are ready.
Answer: John Pitzer
Thanks, Vivek. Jonathan, please connect to the next questioner.
conference operator
OK. The next question comes from Harlan Suhl of J.P. Morgan Chase. Please ask questions.
Q: Analyst
Thank you. Good afternoon everyone, and thank you for answering my questions. A number of new GPU and XPU platforms were launched in the second half of this year. Given the strong demand for supporting CPUs on these platforms, many of which are still based on Intel x86 CPUs, most new platforms will continue to expand until the first half of next year.
Also, overall supply is still tight, but as you said, the team is increasing production capacity and improving yield. With all this in mind, should we expect the data center and artificial intelligence business to continue to grow month-on-month in the first quarter?
Answer: David Zinsner
Yes, I wouldn't want to give too much guidance on that quarter right now. I think your overall judgment is correct: CPU demand will be driven by strong factors. We expect this trend to continue throughout next year. We are working on the supply side to ensure that we can meet customer needs, but it is likely that we will still not be able to fully meet their requirements.
However, I think the challenges in the first half of the year will be greater than in the second half. This may be an additional point of information, and I hope you take it into account when modeling. Other than that, I'd rather wait another quarter to see how the situation develops before providing guidance for the first quarter's data center business.
Answer: Chen Liwu
Let me add one more point. Everyone has seen our announcement with Oracle Cloud, and we also have multiple customers. They are all interested in signing long-term agreements with us to ensure we can provide sufficient supply for their CPU deployments. I think these are all positive signs of strong CPU demand.
Q: Analyst [The original document combines the tracking tips with the analyst's questions]
Haran, do you have any follow-up questions?
Yes. Regarding the factors the team can control and the effects of continuous improvements in yield, reliability, and manufacturability, I remember that Panther Lake and Rapids' gross margins in the second quarter were below the company average. However, as Intel 3 and 18A further improved yield in the third quarter, you mentioned that Xeon 6 Granite Rapids using Intel 3 already increased the company's gross margin in the third quarter. Did the team also raise the gross margin of Core's third-generation Panther Lake to over 40% to mid-40% in the third quarter?
Answer: David Zinsner
We have indeed significantly improved Panther Lake's gross margin. It hasn't reached or surpassed the company average, but we've made tremendous progress. This requires continuous improvement in yield and output efficiency, forming an appropriate scale effect, etc. Over time, these products will increase the company's gross margin, but they have yet to fully reach this level.
Answer: John Pitzer
Thanks, Haran. Jonathan, we still have time to answer that last question.
conference operator
OK. Today's last question comes from Cantor Fitzgerald's CJ Muse. Please ask questions.
Q: Analyst
Thank you guys for answering the questions. The first question would go back to capital expenditure. You said capital expenditure would increase significantly. I want to know how do you consider the distribution of expenses between wafer manufacturing equipment and plant construction? I remember you said in the past that most of the additional expenses would go to the factory. I'd like to confirm if it's still moving in this direction?
Answer: David Zinsner
We have seen a significant increase in equipment spending this quarter, and the increase is expected to be even more significant next quarter. As a result, most of our spending now is actually on equipment. We have already made related investments; in many cases, we are only finishing up some plant space, and most of the expenses are going to equipment. The only exception is the Ohio project, where we are still in the large-scale construction phase. Other than that, the main thing is equipment investment.
A: John Pitzer [The original document combines tracking tips, analyst questions, and confirmation questions]
CJ, do you have another quick follow-up question? Yes. Regarding the PC business, I know you said that the entire industry's shipments will drop 10% next year. I'd like to ask you to talk about how you feel about the changes in your product portfolio and how this might affect pricing. Thank you very much. The change in product mix you mean changes between low-end and high-end products, etc. Is that what you mean?
Answer: David Zinsner
Yes, that's exactly right. Yes. In comparison, we're seeing strong demand in Panther Lake. Obviously, the entire market is falling, but Panther Lake has maintained its performance. We've also put a lot of effort into a product called Wildcat Lake, which is a cost-optimized chip. We've seen significant growth at Wildcat Lake. It was launched at just the right time because it was optimized for cost, and we've seen a lot of attention from the market.
Heading into next year, we'll launch Nova Lake. I think the important thing about Nova Lake is that there was a gap in our high-end desktop portfolio before, and Nova Lake filled that gap. As a result, I expect our product portfolio to perform quite well, as we also deserve a better market response in the high-end desktop sector. We'll see how it really is. We do think we can be a market share winner after entering next year with the Nova Lake product portfolio. Of course, we will continue to monitor the market to see how the situation develops.
Answer: Chen Liwu
Finally, thank you all for attending the conference today. We've made good progress this quarter on our path to driving Intel's transformation, but there's still a lot of work ahead. I look forward to meeting many of you during the quarter and reporting back to you all on the latest developments in January.
conference operator
Ladies and gentlemen, thank you all for joining today's conference call. This concludes the meeting. The cable can now be disconnected. I wish everyone a happy day.