Intel Earnings Call: Strategic collaboration with NVIDIA to open up market opportunities via NVLink technology; tight chip production capacity expected to persist until 2026.

Intel exceeded revenue expectations in the third quarter of 2025, but capacity bottlenecks in mature process nodes are constraining growth.

According to a previous report by Wall Street News, Intel's third-quarter revenue reached $13.7 billion, surpassing the upper limit of its guidance range, with a 6% sequential increase. Non-GAAP earnings per share were $0.23, far exceeding analyst expectations of $0.01 and the company’s own break-even guidance, marking the first net profit since the end of 2023.

However, Chief Financial Officer Dave Zinsner pointed out that Intel chips are currently in "short supply," particularly due to capacity constraints in mature process nodes such as Intel 10nm and Intel 7nm, which limit the company’s ability to fully meet demand for data center and client products. This situation is expected to continue until 2026.

To address these challenges and seize AI opportunities, Intel took a series of measures in the third quarter to strengthen its balance sheet, including securing $5.7 billion in U.S. government funding, receiving a $2 billion investment from SoftBank Group, monetizing part of Altera’s equity, and selling shares of Mobileye. NVIDIA’s $5 billion investment is expected to be completed in the fourth quarter. These actions have provided Intel with approximately $20 billion in cash, significantly enhancing its financial flexibility and confidence in executing its strategy.

Intel CEO Lip-Bu Tan and CFO David Zinsner engaged in discussions with analysts during the subsequent earnings call.

Zinsner acknowledged during the conference call that capacity limitations, "especially in Intel 10nm and Intel 7nm," impacted third-quarter results. The company is addressing this by adjusting pricing and product mix to steer demand toward products with available supply. Lip-Bu Tan emphasized that building a world-class foundry operation is a long-term endeavor based on trust, and Intel is driving a fundamental shift in its entire foundry business mindset.

Intel expects that tight capacity conditions will persist until the end of fiscal year 2026.

In response to this challenge, Intel is actively managing its supply chain, prioritizing wafer capacity allocation for server products, and accelerating the transition to Intel 18A and more advanced process nodes. The company expressed satisfaction with the progress of Intel 18A and is engaging in deep discussions with potential customers regarding the even more advanced Intel 14A node.

Below is a summary of key points from the call:

AI-Driven Demand and Strategic Collaboration

Intel's revenue growth has been primarily driven by the boost in traditional computing from the accelerated construction of AI infrastructure. In the client segment, the adoption rate of AI PCs is rising, with shipments expected to reach approximately 100 million AI PC units by the end of this year. In the data center segment, AI servers’ head nodes, inference, orchestration layers, and storage are all driving demand for server CPUs.

CEO Lip-Bu Tan highlighted the strategic collaboration with NVIDIA during the earnings call. He stated that both companies would connect their architectures via NVIDIA NVLink, combining Intel CPUs with NVIDIA’s AI-accelerated computing advantages to jointly develop multi-generational new products for hyperscale cloud, enterprise, and consumer markets. He emphasized that this partnership does not erode existing markets but instead opens up entirely new incremental opportunities for Intel.

In addition, Intel has established the Central Engineering group to unify horizontal engineering functions and spearhead the expansion into new ASIC and design services businesses, providing customized silicon solutions to external customers and thereby extending the influence of its x86 IP.

Capacity Constraints and Capital Expenditure Strategy

To meet unprecedented demand, Intel is facing significant capacity challenges. CFO David Zinsner noted that capacity constraints are affecting the company across all business lines, particularly on the Intel 10 and Intel 7 nodes. The company is utilizing inventory and attempting to guide customers toward other available products through “demand shaping.”

Despite strong demand, Intel remains prudent in its capital expenditure strategy. The company reiterated its total investment for 2025 at approximately $18 billion, emphasizing that future investments will be disciplined and capacity increases will only occur after securing clear customer commitments. Zinsner stated that the company’s existing assets and construction capacity provide considerable flexibility to address external foundry demand.

On the advanced process front, Intel 18A is on track to launch this year, with Fab 52 in Arizona now fully operational for high-volume manufacturing. Development of the next-generation Intel 14A has also shown positive progress, with better performance and yield starting points at the same maturity level compared to 18A.

Addressing Investor Concerns on Gross Margin and Execution

Investors are closely monitoring Intel's profitability amid capacity constraints and substantial investments. The non-GAAP gross margin for the third quarter was 40%, exceeding guidance by four percentage points, mainly due to higher revenue, an improved product mix, and reduced inventory reserves. However, the fourth-quarter gross margin is expected to decline to approximately 36.5%, partly due to the initial ramp-up costs of the new Core Ultra 3 (Panther Lake) product and the impact of the Altera spin-off.

When asked about the long-term path to improving gross margin, Sinsner pointed out that the key lies in product competitiveness and cost structure. He acknowledged that in the data center sector, Intel still has work to do in terms of product cost and competitiveness, which is central to boosting gross margins. Meanwhile, as the product mix shifts towards more advanced nodes such as Intel 3, 4, 18A, and 14A, along with the gradual decline in startup costs for new processes, gross margins will see structural improvement.

Regarding the yield rate of Intel 18A, Sinsner stated that it is currently at a level "sufficient to meet supply demands," but has not yet reached the height necessary to drive ideal profit margins. It is expected to improve to an ideal state by the end of 2025.

New Partnerships and Financial Impact

Strategic collaborations with NVIDIA, SoftBank Group, and government support have significantly bolstered Intel's cash position and liquidity. As of the third quarter, the company held $30.9 billion in cash and short-term investments.

These partnerships not only provide financial support but also bring strategic business opportunities. For example, the collaboration with SoftBank stems from its substantial investment in AI infrastructure, which will translate into demand for foundry capacity. When asked about confidence in the foundry business, Lip-Bu Tan responded that beyond the capital injection, the company’s solid technical progress on the 18A and 14A processes, improvements in yield rates, and milestone-based cooperation with customers are the primary sources of his confidence.

Below is the full transcript of the conference call:

CEO Lip-Bu Tan:

We delivered solid results in the third quarter, with revenue, gross margin, and earnings per share all exceeding guidance. This marks our fourth consecutive quarter of improved execution, driven by underlying growth in core markets and steady progress in rebuilding the company. While there is still a long road ahead, we are taking the right steps to create sustainable shareholder value. We significantly improved our cash position and liquidity in the third quarter, which has been a key focus since I became CEO in March.

We made tangible progress this quarter in improving execution. Not only are we on track to complete the company’s scale optimization by the end of the year, but we are also continuously adjusting our talent structure, reestablishing an engineering-first mindset, and optimizing execution and management levels across the organization.

Let me delve deeper into the trends underpinning our core business. Throughout my career, I’ve had the privilege of engaging with disruptive innovation in various ways. But I have never been as excited about the future of computing and the opportunities before us as I am now. We are still in the early stages of the AI revolution, and I believe Intel can and must play an even more significant role in the company’s transformation process.

This begins with our core x86 business, which continues to play a critical role at the edge of AI. AI is clearly accelerating the demand for new computing architectures, hardware, models, and algorithms. At the same time, it is also driving renewed growth in traditional computing, as fundamental data and the insights derived from it still heavily rely on our existing products from cloud to edge. AI has brought near-term upward momentum to our business and serves as a solid foundation for sustainable long-term growth as we execute our strategy.

Moreover, with unparalleled compatibility, security, and flexibility stemming from being the largest installed base for general-purpose computing, x86 is well-positioned to power the hybrid computing environments required for AI workloads, particularly for inference and edge workloads and systems. This presents an excellent starting point for rebuilding our market position, revitalizing the x86 ISA, and positioning it through superior products and partnerships for a new era of computing.

Our collaboration with NVIDIA exemplifies this perfectly. Together, we are creating multi-generational new products and experiences to accelerate the adoption of AI across hyperscale cloud, enterprise, and consumer markets. By connecting our architectures via NVLink, we combine Intel’s CPU and x86 leadership with NVIDIA's unmatched AI and accelerated computing strengths, unlocking innovative solutions that deliver better customer experiences and securing a place for Intel in leading AI platforms of the future.

We need to continue building on this momentum by leveraging our position through improved engineering and design execution. This includes hiring and promoting top architectural talent and reimagining our core roadmap to ensure it features best-in-class capabilities. To expedite these efforts, we recently established a Central Engineering Group that will unify our horizontal engineering functions to drive synergies across foundational IP development, test chip design, EDA tools, and design platforms. This new structure will eliminate redundancies, improve decision-making timelines, and enhance consistency across all product development.

Furthermore, equally important, this group will spearhead the creation of a new ASIC and design services business, offering customized silicon solutions to a wide range of external customers. This will not only extend the reach of our core x86 IP but also leverage our design strengths to deliver a spectrum of solutions ranging from general-purpose computing to fixed-function computing.

In the client segment, we are on track to launch our first SKU by the end of the year and additional SKUs in the first half of next year. This will help us solidify our strong position in both consumer and enterprise notebooks with a full-stack cost-optimized product lineup spanning from entry-level offerings to the mainstream Core family and up to the highest-performing Core Ultra family.

In the high-end desktop space, competition remains intense, but we are making steady progress. Shipments of Arrow Lake have continued to grow throughout this year, and our next-generation Nova Lake products will introduce new architectures and software upgrades to further strengthen our offerings, especially in the PC gaming sector. With this product lineup, we believe we will have the strongest PC portfolio in years.

In the traditional server space, AI workloads are driving updates to the installed base and capacity expansion, fueled by the rapid growth of tokenization, increasing data storage and processing needs, and the necessity to address power and space constraints. Xeon remains the preferred choice for AI head nodes, with robust demand for Granite Rapids, including instances from all major hyperscale cloud providers.

We are listening to our customers, for whom strong performance per watt and total cost of ownership (TCO) remain top priorities. As I shared with you last quarter, key areas of focus include improving our multithreading capabilities to close existing gaps and working to regain market share.

Finally, regarding our AI accelerator strategy, I remain confident that we can play a significant role in developing computing platforms for emerging inference workloads driven by AI and physical AI. This represents a much larger market than AI training workloads. We will strive to position Intel as the preferred computing platform for AI inference and look forward to collaborating with a range of established companies and emerging firms defining this new computing paradigm. This is a multi-year initiative, and we will build partnerships when we are able to deliver truly differentiated and market-leading products.

In the short term, we will continue to deliver AI capabilities through Xeon, AI PCs, ARC GPUs, and our open software stack. Looking ahead, we plan to introduce several generations of inference-optimized GPUs on an annual cadence, featuring enhanced memory and bandwidth to meet enterprise demands.

Turning to Intel Foundry Services (IFS), our momentum continues. We are making steady progress on Intel 18A and are on track to bring it to market this year. The Assembly Units (AUs) for Intel 18A are progressing predictably, and Fab 52 in Arizona, dedicated to high-volume manufacturing, is now fully operational. Additionally, we are advancing work on Intel 18AP and will continue to meet our PDK milestones.

Our Intel 18A family forms the foundation for at least three generations of future client and server products. Our collaboration with the U.S. government and other committed customers positions this as a critical node that will drive wafer output into the next decade and deliver healthy returns on our investments.

On Intel 14A, the team remains focused on technology definition, transistor architecture, process flow, design enablement, and foundational IP. We are actively engaging with potential external customers and are encouraged by early feedback that informs and drives our decision-making.

Finally, our advanced packaging initiatives continue to progress well, particularly in areas where we have true differentiation, such as EMIB and Foveros. Just as with our Intel products, my belief in the market potential for Intel Foundry Services grows daily. The rapid expansion of key AI infrastructure is driving unprecedented demand for wafer capacity and advanced packaging services, creating substantial opportunities that require multiple suppliers. Intel Foundry Services is uniquely positioned to capitalize on this unprecedented demand as we execute our plans.

As I mentioned last quarter, our investments in foundry operations will be disciplined, focusing on capability and scalability that allow us to ramp production quickly, and we will only add capacity when there is committed external demand. Building a world-class foundry is a long-term endeavor built on trust. As a foundry, we must ensure our processes can be easily adopted by a wide range of customers, each with their unique approach to building products. We must learn to satisfy customers who depend on us to manufacture wafers that meet all their requirements for performance, power, yield, cost, and schedule. Only by doing so can they rely on us as a true long-term partner to ensure their success. This requires a mindset shift, and I am driving this change within Intel Foundry Services to position the business for long-term success.

Looking forward, my focus remains firmly on the long-term opportunities presented by every market we serve today and those we will enter tomorrow. Our strategy is crystallizing around our unique strengths and value proposition, supported by the unprecedented demand for compute acceleration in an AI-driven economy. Our leadership continues to strengthen, and our culture is becoming more accountable, collaborative, and execution-focused. My confidence in the future grows daily, and I look forward to keeping you updated as we advance on our journey.

I will now turn the time over to Dave for a detailed review of our current business trends and financial performance.

Chief Financial Officer David Zinsner:

In the third quarter, we exceeded our revenue guidance for the fourth consecutive quarter, driven by continued strength in our core markets. While we remain vigilant about macroeconomic volatility, customer purchasing behaviors, and inventory levels, industry supply has tightened materially. Moreover, we are increasingly confident that the rapid adoption of AI is driving growth in traditional computing and reinforcing momentum across our businesses.

In the client domain, we have passed five years of pre-pandemic demand and are now benefiting from updates to a larger installed base. Enterprises continue to migrate to Windows 11, and adoption rates of AI PCs are increasing. In the data center domain, the accelerated construction of AI infrastructure is positively impacting server CPU demand (from head nodes, inference, orchestration layers, and storage). We are cautiously optimistic that, even as we work to improve our competitive position, the CPU TAM (Total Available Market) will continue to grow into 2026.

Third-quarter revenue was $13.7 billion, above the high end of our guidance range, with a 6% sequential increase. Capacity constraints, particularly on Intel 10 and Intel 7, limited our ability to fully meet third-quarter demand for both data center and client products.

Non-GAAP gross margin was 40%, 4 percentage points higher than our guidance due to higher revenue, a more favorable product mix, and reduced inventory reserves, partly offset by higher production volumes for Lunar Lake and early ramp-up costs for Intel 18A. Our third-quarter earnings per share were $0.23, compared to our breakeven guidance, driven by higher revenue, stronger gross margins, and ongoing cost control. Third-quarter operating cash flow was $2.5 billion, capital expenditure for the quarter was $3 billion, and adjusted free cash flow was positive at $900 million.

One of our top priorities for 2025 is to strengthen our balance sheet. To this end, we executed transactions to secure approximately $20 billion in cash, including three significant strategic partnerships. We ended the third quarter with $30.9 billion in cash and short-term investments. During the third quarter, we received $5.7 billion from the U.S. government, $2 billion from SoftBank Group, $4.3 billion from the completion of the Altera transaction, and $900 million from the sale of Mobileye shares. We expect NVIDIA’s $5 billion investment to close by the end of the fourth quarter. Finally, we repaid $4.3 billion in debt during the quarter, and we will continue prioritizing repayment as 2026 maturities approach to reduce leverage.

Next, let's look at the segment results for the third quarter. Intel Products Group revenue was $12.7 billion, up 7% sequentially, exceeding our expectations in both client and server segments. Given the current tight capacity environment, the team executed well to support the quarter’s outperformance, and we expect this trend to continue into 2026. We are working closely with customers to maximize our available output, including adjusting pricing and mix to shift demand toward products where we have supply and they have demand.

Client Computing Group (CCG) revenue was $8.5 billion, up 8% sequentially and exceeding expectations, driven by a stronger seasonal TAM, Windows 11-driven upgrades, and improved pricing mix as Lunar Lake and Arrow Lake ramped up. During the quarter, CCG further strengthened its relationship with Microsoft through collaboration with Windows ML and deep integration of Intel vPro manageability with Microsoft Intune, enabling secure cloud-connected fleet management for businesses of all sizes. The team also achieved all key milestones to support the launch of Core Ultra 3, codenamed Panther Lake. We project the client consumption TAM for 2025 to approach 290 million units, marking two consecutive years of growth since the post-pandemic low in 2023. This represents the fastest TAM growth since 2021, and we are prudently preparing for another year of strong demand in 2026 as Core Ultra 3 enters a healthy PC ecosystem.

Data Center and AI Group (DCAI) revenue was $4.1 billion, up 5% sequentially and exceeding expectations, driven by an improved product mix and higher enterprise demand. Although supply constraints limited additional upside, demand for AI server host CPUs and storage compute remained strong during the quarter. Our latest Xeon 6 processor, codenamed Granite Rapids, delivers significant advantages, including up to 68% TCO savings and up to 80% power reduction compared to today’s average servers. It is becoming increasingly clear that as AI adoption expands, particularly as inference workloads grow faster than training, CPUs play a critical role in both today’s and future AI data centers.

Some data center customers have begun inquiring about long-term strategic supply agreements to support their business goals, driven by the rapid expansion of AI infrastructure. This dynamic, combined with underinvestment in traditional infrastructure over the past few years, should enable comfortable growth in server CPU revenue TAM going forward.

Intel Products Group’s operating profit was $3.7 billion, or 29% of revenue, increasing by $972 million sequentially due to stronger product margins, lower operating expenses, and favorable comparisons to second-quarter period costs.

Before discussing Intel Foundry Services, I would like to thank the NVIDIA and Intel teams for their relentless efforts. There is much work ahead of us, but the partnership announced this quarter is the result of nearly a year of hard work with a company that places engineering excellence at the forefront. The x86 architecture has been foundational to driving the digital revolution of the modern world. AI represents the next phase of this revolution, and we are committed to ensuring that x86 remains at its core. Collaborations like this one with NVIDIA are essential to this effort.

Turning to Intel Foundry Services (IFS). Intel Foundry Services generated revenue of $4.2 billion, a 4% decline quarter-over-quarter. In the third quarter, Intel Foundry Services exceeded expectations for Intel 10 and 7 node production, achieved a key 18A milestone, and released a hardened 18AP PDK to the ecosystem. The foundry business also advanced the development of Intel 14A and continued to make progress in expanding its advanced packaging deal pipeline.

Intel Foundry Services recorded an operating loss of $2.3 billion in the third quarter, improving by $847 million quarter-over-quarter, primarily due to a favorable comparison with approximately $800 million in impairment losses in the second quarter.

As previously mentioned, our confidence in the long-term Total Addressable Market (TAM) for foundry services continues to grow, driven by the acceleration of AI deployment and adoption as well as increasing demand for wafer and advanced packaging services. Projections indicate that AI capacity (measured in gigawatts) will grow more than tenfold by 2030, presenting significant opportunities for Intel Foundry Services with external customers—whether in wafer services or our differentiated advanced packaging capabilities such as EMIB and Foveros. We continue to work on earning customer trust, and our improved balance sheet flexibility will enable us to respond quickly and responsibly as demand arises.

Turning to all other businesses. Revenue was $1 billion, with Altera contributing $386 million; revenue declined 6% quarter-over-quarter due to Altera completing its transaction and spin-off during the quarter. The three main components of all other businesses in the third quarter were Mobileye, Altera, and IMS. This category collectively contributed $100 million in operating profit.

Now turning to guidance. For the fourth quarter, we forecast revenue in the range of $12.8 billion to $13.8 billion. At the midpoint and adjusting for the impact of the Altera spin-off, fourth-quarter revenue is expected to be roughly flat quarter-over-quarter. We anticipate slight growth in Intel’s Product Group quarter-over-quarter but below customer demand as we continue to navigate a tight supply environment. Within Intel’s Product Group, we expect CCG to decline slightly quarter-over-quarter, while DCAI will experience strong quarter-over-quarter growth as we prioritize wafer capacity for server shipments over entry-level client components.

We expect Intel Foundry Services revenue to grow quarter-over-quarter, driven by increased Intel 18A revenue and higher external foundry revenue following the Altera spin-off. For all other businesses (now excluding Altera), we anticipate revenue declines consistent with guidance, partially offset by sequential growth in IMS.

At the midpoint of $13.3 billion, we forecast a gross margin of approximately 36.5%, declining quarter-over-quarter due to product mix, the impact of initial Core Ultra 3 shipments (which typically carry higher costs at the start of new product ramps), and the Altera spin-off.

We forecast a tax rate of 12% and earnings per share of $0.08, both on a non-GAAP basis. We expect non-controlling interests (on a GAAP basis) for the fourth quarter to be approximately $350 million to $400 million, and we project the fully diluted average share count for the fourth quarter to be approximately 5 billion shares.

Turning to capital expenditures, we continue to expect total investments for 2025 to be approximately $18 billion, and we anticipate deploying over $27 billion in capital expenditures in 2025, compared to $17 billion deployed in 2024.

Finally, I would like to note that we ended the third quarter with a significantly strengthened balance sheet, solid near-term demand, and growing confidence in the long-term opportunities for our core x86 business as well as foundry, ASIC, and accelerator segments. We also recognize there is work to be done to realize our full potential. We are continuing to bring in external talent and unlock the potential of our workforce to improve execution in product and process development as well as manufacturing. We will closely manage what is within our control, respond swiftly as conditions evolve, and remain focused on delivering long-term shareholder value.

At this point, I will turn the time back to John to begin the Q&A session.

Q&A Session

Q1: Ross Seymore (Deutsche Bank): First on the foundry side. You've announced a number of partnerships this quarter and your balance sheet has significantly strengthened. The tone in your opening remarks seemed more confident regarding the progress of the foundry business. Are these partnership announcements or equity investments boosting your confidence? Or are you seeing specific technological advantages that are increasing your optimism about the business?

Lip-Bu Tan:

I think several of the announcements we made were clearly more product-focused, and one of them was with SoftBank Group, as they are building all their AI infrastructure, which will certainly require more foundry capacity. So I think that's part of the answer. But at the same time, I would say, based on what I’ve seen with 18A and 14A, we have made tremendous good progress. The steady progress on 18A, and Panther Lake will depend on it. Then obviously, we see yields improving in a much more predictable manner. I visited Fab 52, which is now fully operational for 18A. And on 14A, we are clearly engaging in milestone discussions with multiple customers, and we're really pushing improvements in yield and performance reliability, which is even more exciting. On advanced packaging, we also see strong demand from key cloud and enterprise customers within the foundry space. So overall, we believe establishing long-term trust with some customers and expanding that is quite exciting. We are also focused on hiring top talent to drive improvements in certain process technologies.

Q2: Joseph Moore (Morgan Stanley): I'm very interested in how your approach to the foundry business differs. You mentioned last quarter and again this quarter that before making investments, you’re looking for customer commitments. Can you talk about how those conversations are progressing? Of course, I can see the trade-offs from the customer perspective when they make commitments to you. Are they expecting capacity to be ready earlier? Is there a bit of a 'chicken-and-egg' problem with these investments? How are you handling these discussions?

Lip-Bu Tan:

I think in terms of the foundry business, we are clearly engaging with multiple customers. In the process of building customer trust, you need to truly demonstrate improvements in yield, reliability, and you need to have all the specific IPs that they require. This is a service industry, and you need to have all the right IPs. That’s why I established the Central Engineering Group, to acquire all the right IPs to match customer needs. Then, I think the best approach is to truly demonstrate performance and yield, and then we can get test chips so they can actually start working on them. Then they can begin deploying their most important revenue wafers relying on us, so we can drive success for them. So I think these things are very important.
Regarding potential investments and partnerships, I think different customers have different requirements, and we are working with them. But more importantly, we are gaining their commitment to the foundry and support. I think building trust is the more significant positive factor.

David Sinsner:
To add a bit more, I would say that customers understand that there is a time lag from when you deploy capital to when you see output. So our expectation is that we will finalize these commitments in a timely manner so that we can deploy the capital as needed. I would also say that, given the capital expenditure investments we have already made, we are in a fairly good position. So we already have a significant amount of assets on our books, as well as what we call work-in-progress, and we have made substantial investments in plant space. Therefore, we do see the prospect of leveraging our existing footprint to drive reasonable supply volumes for external foundry customers, and frankly, by utilizing work-in-progress and redeploying equipment that is currently on our books.

So we have flexibility. Clearly, if things go better, we may look to invest more and move faster, but we have reasonable confidence in our ability to navigate through this situation.

Q2: Joseph Moore (Morgan Stanley): Additionally, regarding supply constraints for server CPUs and other CPUs, we observe those markets—I guess—but your growth was 5% quarter-over-quarter and single-digit year-over-year. I am curious where the shortages are coming from? Is it simply due to stronger-than-expected demand that you cannot fulfill? Or is it related to some transitions you are undergoing? I certainly see the market tightness, so I am not disputing that, but I am just curious about your view on where these shortages are coming from and how they might be resolved?

David Zinsner, Executive Vice President and Chief Financial Officer:
Yes. I mean, the shortages are pretty much across our business, I would say. We are definitely constrained on Intel 10 and 7. Clearly, we are not planning to build more capacity there. So as demand increases, we are limited. To some extent, we are relying on inventory to sustain us. We are also trying to guide customers toward other products through what we call 'demand shaping.' Even beyond the specific challenges in our foundry business, there are shortages. For example, I think there are widely reported substrate shortages. So obviously, I think demand—everyone was cautious at the beginning of the year, which I believe was true across industries—but now it looks like this year will be stronger, and it might extend into next year. I think everyone is trying to manage through this.

CJ Muse (Cantor Fitzgerald):
Okay, good afternoon, and thank you for taking my question. I would like to follow up on the current outlook that demand will continue to exceed supply through 2026. I am curious whether this is primarily focused on servers or also includes clients. Based on your thoughts there, how should we think about the trends in the first quarter compared to normal seasonality, which I assume would typically be a high single-digit to low double-digit decline quarter-over-quarter? Given the investments from the U.S. government, NVIDIA, SoftBank Group, etc., I am curious how your thinking has evolved regarding capital expenditures or other investments in your product businesses amid improving cash flow and liquidity.

Unspecified Speaker:
Yes, both. But as we said, we are ceding some portion of the small-core market and client segment to more broadly meet customer needs, specifically in the client space and, more pointedly, in the server space. That’s how we are managing it. Looking ahead to the first quarter, clearly, we might provide more information in January. I would just say that in the first quarter, we might actually be at the peak of the shortage because we relied on some inventory to help us through the third and fourth quarters, and we were just running the factory as hard as possible. We might not have as much of that convenience in the first quarter.
So I’m not sure we can reverse the seasonality trend because we will be very, very tight in the first quarter. After that, I think we will start to see some improvement, and we can catch up over the rest of the year.

David Zinsner, Executive Vice President and Chief Financial Officer:
Clearly, we are in a good position. I would say that when we think about this cash, our top priority is deleveraging. I mean, this was one of the things Lip-Bu Tan really wanted to do when he took over because he wasn’t satisfied with the balance sheet. So we have done a lot of work on it and improved the situation for him. We repaid $4.3 billion of debt this quarter, and all maturities for the next quarter or for next year should be repaid. I think it puts us in a flexible position regarding capital expenditure, but we want to be very disciplined in that area. So we will absolutely monitor demand. He has been very direct with us on this point.
He wants to see clear customer demand that allows us to believe in that demand. If it exists, we will certainly increase capital expenditures as needed. When you think about investments, we still believe that a $16 billion operating expense investment next year is appropriate, although Lip-Bu Tan and I continue to study how to allocate that $16 billion to drive the best possible growth and returns for investors, and we will make those adjustments. Beyond that, we will see how things evolve. We aim to maintain discipline in keeping our operating expenses as a percentage of revenue while driving leverage, but we do see opportunities where we can invest, and I think those investments will bring significant returns to shareholders, so we are not afraid to do so.

Bryan Curtis (Jefferies): Hey, everyone. Thank you. I have two questions. Regarding capital expenditures, I think you reiterated $18 billion, but I feel like your spending in Q3 was less than what I modeled. Does that number still hold? I am a bit curious about when you start ramping up capacity in Arizona for this (unclear: possibly referring to 18A), how are you thinking about the timing of increasing capacity?

David Zinsner, Executive Vice President and Chief Financial Officer:
Regarding the $18 billion, yes, I think that number still holds. Clearly, capital expenditures may not be evenly distributed. It depends on when things are completed — when all requirements related to invoice payments are completed, that’s when we make the payment. So we expect to stay within that range. Clearly, there is some margin of error. It could be slightly less or slightly more than that number. As for 18A, yes, we still need to expand capacity. I don’t expect a significant increase in capacity in the short term. But as we’ve said, we haven’t yet reached the peak supply for 18A.
In fact, we won’t reach it until the end of this decade. We do believe this node will be quite long-lasting for us. So over time, we will continue to invest in 18A. There will be capital expenditure investments next year, but I don’t expect supply — at least capacity — to change significantly relative to our current expectations.

Bryan Curtis (Jefferies): I just wanted to follow up on the gross margin trajectory after 18A starts being phased in. I know it may not be a good comparison to previous nodes, but perhaps it can be compared to a successful node. When you say yields are in a good state and improving, is there a way to think about how these yields for 18A compare to what you’ve seen historically for successful products, and how they will be gradually incorporated in the first half?

David Zinsner, Executive Vice President and Chief Financial Officer:
Yes, I would say, overall, I’m not sure — honestly, yields for older nodes aren’t our focus. So we’re charting new territory here. The yields are — I would say sufficient to meet supply needs, but they haven’t yet reached the level we need them to be at to drive an appropriate level of profitability. By the end of next year, we might reach that level. Certainly, the following year, I think they will reach industry-acceptable yield levels. I can tell you that for 14A, we had a good start. If you look at 14A compared to 18A at the same maturity, we performed better in terms of both performance and yield. So we had an even better start with 14A. We just need to keep progressing.

Stacy Rasgon (Bernstein Research):
Hi, everyone. Thank you for taking my question. I want to go back to the question about supply constraints. You’ve talked a lot about how AI is driving significant demand for servers and PCs. But at the same time, it seems that customers don’t really want your AI products. In fact, they can't get enough of the older products. So I assume you definitely have sufficient supply even for (unclear: possibly referring to older products such as Raptor Lake, Meteor Lake) or even Lunar Lake. So how do you plan to make customers move away from the older products since so far they haven’t shown any willingness to do so, even under these constraints? I mean, how should we think about this transition for those customers because obviously—well, you yourselves have said—you are no longer increasing the old capacity. In fact, you've actually shut some of it down, right?

David Zinsner, Executive Vice President and Chief Financial Officer:
Yes, good question. I think it's a misunderstanding to say AI isn't performing well. I mean, it has grown double digits quarter-over-quarter. We mentioned a figure — we will ship around 100 million AI PC units by the end of this year, and we are roughly on track for that. So I think the progress has been quite good. However, clearly, the older nodes have also performed well, which might be the more unexpected part.
Yes, we — I think we just need to engage in ensuring the ecosystem drives enough AI applications in the PC space. We regularly work with ISVs to push this forward. They are making progress. Like any market, it starts relatively immature and then builds over time. But even within our company, we're starting to find uses for AI PCs. In fact, the Investor Relations department is preparing one, and we will use it. So I think it’s only a matter of time. That said, what is clearly happening now is that the Windows refresh cycle is much stronger than we anticipated. And this isn’t necessarily an AI PC story. So Raptor Lake is also addressing this demand. So we're seeing upside in that segment of the market as well.

Stacy Rasgon (Bernstein Research): I would like to follow up on two things I think I heard you mention regarding 18A. I think I heard you say that first, the yield won’t be in a great state at least until the end of next year. And then I think I also heard you say that you’re not going to increase 18A capacity much next year. Did I hear that wrong? I mean, especially the latter — how is that possible? If you’re ramping up, if you’re proceeding along a ramp-up path, how does that hold true? Or is it something like…? (multiple people talking simultaneously)

David Zinsner, Executive Vice President and Chief Financial Officer:
We are clearly still in the early stages. I mean, relative to the capital expenditure plan, we’re not going to incrementally increase 18A supply next year. But yes, of course, we will gradually ramp up production throughout next year. Yes, I wouldn’t say the yields for 18A are in a bad state. I mean, they are where we expect them to be at this point in time. We set a target for the end of the year, and we will achieve that target. But to fully contribute to the cost structure of 18A, we need yields to improve further. I mean, this happens with every process. I think it will take all of next year to truly reach that point.

Joshua (TD Cowen): I wanted to ask a follow-up on Lip-Bu Tan’s prepared remarks regarding fixed-function computing and the potential support for more ASICs. Is this — maybe you could provide more background on this scope? Is this aimed at potential foundry customers, or are these products? And if they are products, what types of applications do you expect to support with custom silicon? Thank you.

Lip-Bu Tan, Chief Executive Officer:
Good question. So I think first of all, as I mentioned earlier regarding the Central Engineering Group driving ASIC design, this will actually be — it's a great opportunity to expand the coverage of our core x86 IP and drive some customized silicon for certain systems, cloud vendors, and customers. And certainly, foundry and packaging are also helping us meet their needs. So in summary, I believe this AI will drive significant growth, especially in areas like doubling down on Moore's Law, which will greatly help enhance our x86 capabilities. This is an opportunity for us to build out the entire ASIC design to serve specific customer demands.

Joshua (TD Cowen): Yes. Clearly, there was a lot of attention last quarter regarding your potential decision to discontinue 14A disclosures. I just wanted to ask if there have been any changes in that area given that your balance sheet situation is quite different compared to three months ago. I think the queue (possibly referring to the 10-Q filing) hasn't come out yet, so I haven't seen whether the wording has changed. I'm curious, given all the changes in the balance sheet, whether there have been any updates since last quarter? Thank you.

Lip-Bu Tan, Chief Executive Officer:
Yes. Since last quarter, it’s clear that our engagement with customers on 14A has increased. We are deeply engaging with them to define the technology, process, yield, and IP requirements to serve them. They clearly see strong demand and need Intel to remain strong on 14A. So we are pleased and more confident. At the same time, we’ve attracted key talents in process technology who can really drive success, which is why it gives me more confidence to push forward on this initiative.

Ben Reitzes (Melius Research):
Hey, thanks, everyone. Lip-Bu, could you provide an update on the NVIDIA relationship, product timelines? Are you getting any feedback from customers about your ability to articulate the importance of this relationship, its timelines, and significance, or any other information you’d like to share with us? Thank you very much.

Lip-Bu Tan, Chief Executive Officer:
Certainly. Thank you. This is a very significant collaboration with NVIDIA. As you know, they are a great company, and Jensen Huang and I have been friends for over 30 years. We are very excited about combining Intel's leadership in x86 CPUs with their unparalleled AI and accelerated computing strengths. Through this partnership, it will truly create new product categories spanning multiple generations. It is a deep engineering-to-engineering engagement that will drive some new custom data center and PC products optimized for the AI era. In summary, I believe this will be a multi-year collaboration entering a market that excites us while also driving some demand for AI infrastructure.

David Zinsner, Executive Vice President and Chief Financial Officer:
Perhaps to add a little more, what is particularly special about this for us is that it is not attacking our existing TAM but rather expanding our TAM with incremental opportunities. So these represent huge opportunities for us.

Ben Reitzes (Melius Research):
Yes. Thank you, John. Lip-Bu, you briefly mentioned that your current AI strategy is to target the inference market and that there is — you see space for Intel's solutions. It sounds like there will be a lot of collaboration in this area. Is this strategy more about collaboration? Or is it more about — is there a specific Intel IP inference technology that excites you, or is it more of a 'Switzerland' model where you can partner with many existing players to capture more of the TAM? Thank you.

Lip-Bu Tan, Chief Executive Officer:
Yes, good question. So I think, first of all, as AI is driving significant growth, we definitely want to participate. I think it’s still very early days, so I think this is an opportunity for us. One area I think we are focused on is revitalizing our x86 and really tailoring purpose-built CPUs and GPUs for new AI workloads, then truly addressing energy efficiency issues and managing all different proxies. This is a new approach to preferred computing platforms, which will also apply to systems and software. They will tell you, I think we will — partner with some existing companies as well as emerging companies that are driving these changes.

Timothy Arcuri (UBS Group):
Thank you very much. Dave, we don’t often see high fixed-cost businesses with gross margins below 40% when constrained. I certainly understand that this is primarily due to Intel 10 and 7 wafer costs, and that you still have low yields on 18A. But I wonder — this might be a difficult question to answer, but I wonder if you could give us a bit of a forward-looking view and say what the gross margin would look like if you were rid of 10 and 7 and fully on 18A? Is there any way to give us a normalized estimate?

David Zinsner (Executive Vice President and Chief Financial Officer):
Yes. I mean, obviously, you may... you might have heard some of my internal discussions with the team because this is something I’ve been emphasizing. I think there are two dynamics at play here, one of which you mentioned: the contrast between the high cost of older nodes versus the better cost structure of newer nodes. That is clearly important. I mean, our foundry business currently has negative gross margins. Even just moving it into positive territory would represent a significant improvement. But the other aspect of our gross margin relates to the quality (competitiveness) of the products themselves.
In the client business, our product performance and competitiveness are generally okay, although there are some exceptions, we are not yet at an ideal level on a cost basis. So, we must make improvements in this area. We have already planned for this in our roadmap, and the team is aware of it, but it will take a multi-year process to achieve. In the data center segment, the issue is even more pronounced. Not only do we lack the appropriate cost structure, but we also lack sufficient competitiveness to secure desirable margins from customers. Therefore, we still have work to do in this area. This is precisely the area where Lip-Bu and the team are highly focused right now: building great products with the right cost structure to drive better gross margins. In my view, this is the key to everything.
However, I believe the improvements in the foundry business will come naturally. Our product mix will gradually shift to a higher proportion of Intel 3 and 4, followed by 18A, and eventually 14A. The cost structures of these (advanced) nodes are actually quite similar. At that point, the increase in gross margin will mainly be driven by the significantly higher value provided by these leading-edge nodes, which will materially boost gross margins.
I would also like to mention that, given the rapid pace at which we have been introducing a series of new processes in quick succession, we are currently bearing a significant amount of initial ramp-up costs. As we move into the 14A phase, our cadence of technology rollouts will become more normalized. Consequently, you will not see as many ramp-up costs stacked together, which is currently impacting our gross margin by billions of dollars. So I believe that, in a few years, these costs will gradually decrease, which will be beneficial.

Timothy Arcuri: Lip-Bu, you didn’t provide an update on the release timing for 'Diamond Rapids' during the last conference call. I know the entire roadmap is under review, but you seem — the company appears quite optimistic about the 'Rapids' series. Could you give us an update on the data center roadmap? Thank you.

Lip-Bu Tan (Chief Executive Officer):
Alright, thank you. That's a good question. I think it’s clear that 'Diamond Rapids' is receiving stronger feedback from hyperscale customers. At the same time, we are also focused on the new product 'Rapids,' which will include SMT (Simultaneous Multithreading), enabling higher performance. We are currently in the definition phase, after which we will finalize the roadmap and execute it going forward.

Aaron Rakers (Wells Fargo & Co):
Alright. Thank you for the opportunity. Just a couple of very quick questions. I want to return to the relationship with NVIDIA. I understand this announcement primarily relates to the NVLink Fusion strategy and its integration with the x86 ecosystem. But I think there have also been some recent reports suggesting the possible use of Gaudi for handling certain specialized inference workloads within NVIDIA’s tech stack. How do you view this — is this partnership a starting point? Should we expect to see more potential integrations beyond the current scope in the future?

Lip-Bu Tan (Chief Executive Officer):
Alright. Let me address that. I believe NVLink serves more as a hub connecting x86 and GPUs. In terms of AI strategy, we are clearly defining what we’ve referred to as 'Crescent Island,' and we also have a new product line aimed at addressing physical AI and earlier-stage (AI) domains. So, I think, stay tuned. We will provide updates on this subsequently.

Aaron Rakers (Wells Fargo & Co):
Alright. Very briefly. Could you give us an update on the Non-Controlling Interest (NCI) spending to help us understand the situation for the remainder of this year and beyond? I recall you previously provided some comments on how to think about the situation in 2026? Thank you.

David Sinsner (Executive Vice President and Chief Financial Officer):
Yes. I think for 2026, a range of approximately USD 1.2 billion to USD 1.4 billion might be a reasonable estimate. Clearly, we are monitoring this closely and will do our best to minimize it.