Intel (Minutes): The peak production of 18A will be around 2030.

The following is a summary of the$Intel(INTC.US) Q2 2025 earnings call minutes organized by Dolphin Research. For earnings interpretation, please refer to "Intel: After Massive Layoffs, Is the American 'SMIC' the Best Outcome?"

I. Review of Intel's Core Financial Information

1. Market Situation: Q2 market operated normally, core demand drivers emerged:

a. Client: End of Windows 10 support + aging devices during the pandemic drove PC demand; AI PC share increased.

b. Server: Hyperscale and regular enterprises upgraded CPUs, preferring high-performance, low-power products.

2. Cash Flow: Q2 operating cash flow was $2.1 billion, total capital expenditure was $4.5 billion, net capital expenditure was $3.1 billion, adjusted free cash flow was -$1.1 billion. Cash and short-term investments were $21.2 billion, with plans to start deleveraging in 2025.

3. Q3 Guidance and 2025-2026 Planning:

a. Q3 Revenue: $12.6-13.6 billion (QoQ -6% to +6%), below historical Q3 high single-digit growth levels (due to potential demand overdraft from exceeding expectations in the first three quarters, planning for the second half below seasonal levels).

- CCG expected to be strong; foundry revenue slightly down due to Intel 7 capacity constraints + external packaging demand reduction; other businesses remain stable.

b. At the midpoint of $13.1 billion in revenue, non-GAAP gross margin is approximately 36% (impacted by increased outsourcing product share, early Panther Lake mass production, and tariff costs).

c. Tax rate 12%, EPS breakeven (non-GAAP).

d. 2025 OpEx target $17 billion, 2026 $16 billion.

e. 2025 total CapEx approximately $18 billion, net CapEx $8-11 billion; 2026 CapEx expected to be lower (due to improved utilization of construction in progress).

4. Equity:

a. Raised $900 million through Mobileye equity financing, Q3 will complete Altera equity sale (post-transaction deconsolidation, remaining equity accounted as equity investment).

b. Non-controlling interest (NCI): Q3-Q4 expected to be approximately $250-300 million each (GAAP), significant growth expected in 2026.

II. Detailed Content of Intel's Earnings Call

2.1 Key Information from Executive Statements

1. Organizational Adjustment:

a. Completed a review of all organizations and functions under the CEO's jurisdiction over the past three months, covering dimensions such as employee numbers, skills, and expenses, aiming to create a streamlined and efficient organization;

b. Most measures needed to achieve the year-end target of 75,000 employees have been completed, with about 50% of management levels to be cut. Need to balance personnel scale while retaining core talent and recruiting external elites. Implementing a "return to office work" policy starting September.

2. Foundry Strategy Transformation:

a. Terminated manufacturing projects in Germany and Poland, integrated Costa Rica packaging and testing business into Vietnam and Malaysia plants, slowed Ohio plant construction (retaining acceleration flexibility).

b. Technical Progress: Intel 18A focuses on yield and performance improvement, supporting the next three generations of products; Intel 14A positioned as a foundry node from the design stage, developed in conjunction with internal and external customer needs.

c. Financial Discipline: Intel 14A relies on both internal and external customers to support investment returns, capital is only invested when returns are confirmed.

3. Core x86 Business:

a. Client Business: Launching the first Panther Lake SKU by year-end, expanding the product line in the first half of 2026 to consolidate notebook market share. There is still a gap in the high-end desktop market, plans to improve with Nova Lake by the end of 2026.

b. Server Business: Traditional servers maintain a strong position in AI hosts, storage, etc., Granite Rapids mass production is on schedule, and mature product demand is good. Need to improve per-watt performance for hyperscale workloads, correct multi-threading capability issues, and introduce new leadership.

c. Simplify product architecture and SKU system to improve profitability; all important chip designs require CEO approval to enhance execution efficiency.

4. AI Strategy:

a. Shift from "chip and training-centric" to "system and software + chip" full-stack layout, addressing shortcomings in traditional software fields.

b. Focus on differentiated areas such as inference and agentic AI, first understanding real workloads, then reverse designing software and chips.

5. Financial Optimization:

a. Cash Flow Improvement: Adjusted free cash flow has been negative since 2021, plans to improve through enhancing operational leverage and controlling capital expenditure (due to procurement agreements, 2025 CapEx is difficult to reduce further, continue to cut in 2026).

b. Sold part of Mobileye equity this month, will complete Altera transaction with Silver Lake this quarter, may evaluate more non-core asset disposals in the future.

2.2 Q&A

Q: How fast is the recovery of the x86 business? Does the smooth mass production of the 18A process depend on Nova Lake and Diamond Rapids? Are the related timelines around 2027, 2028? What specific steps need to be taken to build trust in the x86 business to expand the foundry business?

A: We are focusing on advancing the 18A process to repair the x86 business, conducting two reviews with the team each week, with steady progress in technology performance and reliability, and are confident in launching Panther Lake by year-end.

18A is the foundation for at least three generations of client and server products, with mass production progress relying on internal production (such as Panther Lake), not dependent on Nova Lake and Diamond Rapids (no related timelines mentioned), the current focus is on launching Panther Lake within the year, gradually attracting external customers through internal mass production.

Steps to build trust include: collaborating with external ecosystem partners to improve technology; ensuring product reliability, yield, and timely large-scale delivery to support customers; accumulating trust through high-quality performance to attract customers to invest resources in product development based on our foundry business.

Q: What are the specific reasons for the sequential decline in gross margin guidance? Looking ahead to next year, what are the driving and adverse factors for gross margin?

A: The sequential decline in gross margin guidance is mainly due to two reasons: First, Lunar Lake will be mass-produced significantly in the third quarter, its packaging includes memory, and we price it at procurement cost, which will negatively impact gross margin, and the transition from the second to the third quarter may be more pronounced; Second, Panther Lake is in the early stages of mass production, with high wafer costs, which will also bring pressure.

In the long term, factors driving gross margin improvement include: large-scale mass production of Panther Lake, with many wafers returning to internal production to reduce costs, its better cost structure will show advantages as production increases; foundry business gross margin is expected to improve with the mass production of more advanced process nodes, and this trend will continue for several years; on the product side, introducing high-value products to enhance pricing, optimizing cost structure (such as more streamlined design), while improving other factors affecting cost of sales.

Q: Why can the 14A process attract external customer cooperation despite a cautious attitude towards its development? How to balance the cautious attitude towards 14A process development with building an external foundry business? Is this a shift from foundry business to prioritizing product business development?

A: We are not taking a cautious attitude towards the development of the 14A process, but rather focusing on solidifying the foundation, such as technical definition, transistor architecture, and other core aspects, while learning from the lessons of 18A and collaborating with external partners to improve yield, which allows customers to see our progress. At the same time, we involve customers early in the development, clarify technical milestones, and provide complete EDA and IP ecosystem support, which customers highly appreciate and are naturally willing to cooperate.

Balance: 14A development is closely integrated with customer feedback to ensure technology meets needs; at the same time, capital expenditure is not invested until internal and external customer production commitments are seen, ensuring efficient resource utilization and letting customers know that our capacity construction will match their needs, which in turn helps build a reliable foundry business.

This is not a shift from foundry business to product business. We still value the important process node of 14A, but focus more on advancing responsibly—first establishing a solid technical and customer foundation, then gradually investing, ensuring the foundry business can develop steadily and support the product business.

Q: What is the situation with normalized capital expenditure? How much capital expenditure can be reduced next year? Specifically, can total capital expenditure be reduced by about $5 billion next year?

A: Regarding normalized capital expenditure, currently about half of the total capital expenditure belongs to this category, with part of this year's $18 billion total capital expenditure considered as maintenance expenditure.

Regarding capital expenditure reduction next year, the core reason is that a large number of assets have been acquired in recent years, which need to be digested first, providing a basis for capital expenditure reduction. (However, no specific reduction amount was mentioned)

Q: What is causing the tight supply of Intel 7 process until the end of the year? Can the production of Intel 4 products be increased? What measures need to be taken to increase Intel 7 capacity, and is it necessary to wait for its capacity to start?

A: Mainly because Raptor Lake performed well, its price positioning fits the PC procurement needs of many consumers and enterprises, and strong demand has driven tight supply.

However, the product mix is expected to change. Currently in the mass production stage of Granite Rapids, which will further increase the production of Intel Core 3rd generation products, the company is also expanding capacity and increasing wafer output in this area.

Q: Can a large number of construction-in-progress projects bring full value? Will the situation of a large number of construction-in-progress projects continue to exist? Will depreciation be started on construction-in-progress projects at some point?

A: A large number of construction-in-progress projects are advancing towards creating value, such as the Arizona project has completed transformation and fully advanced 18A process mass production, with the scale of construction-in-progress reduced from over $50 billion at the end of the second quarter to around $35-40 billion, and this number will continue to decrease through improved asset utilization, with the potential to realize its value.

The situation of a large number of construction-in-progress projects will not remain high, it will steadily decrease for the rest of the year, with a goal to further reduce next year; however, due to the need to retain the option of idle space in wafer fabs (such as the Ohio project, which is slowed but not stopped), there will still be some construction-in-progress on the balance sheet.

Q: What trend is expected for the server business in the third quarter? What is your view on market share loss in the server field? When might this share loss ease next year?

A: The server business is expected to show slight growth overall in the third quarter, with specific performance depending on the actual situation of the server and CPU business.

In terms of market share in the server field, our product portfolio has not yet met expectations, but market share has been relatively well maintained, thanks to the strong x86 ecosystem and our capabilities in this field; the team is focusing on improving the product portfolio, with subsequent products like Granite Rapids and Diamond Rapids enhancing competitiveness, but reaching the ideal state still requires effort.

Q: Can you elaborate on the AI strategy? When will the AI strategy content be updated? Is it a GPU-centric strategy or another strategy? How should we view your approach to entering the AI market?

A: Regarding the AI strategy, the focus will be on the inference field and the rapidly developing inference side and agentic direction, aiming to establish a foothold in these areas; will comprehensively consider from system software to chip level to enhance performance, combining x86 architecture advantages with accelerators, striving to become the computing platform of the future, with related planning underway. The AI strategy will be gradually updated in the coming months.

In terms of entering the AI market, the team is actively forming, focusing on recruiting software talent, leveraging good communication with customers and its own technical planning, striving to seize opportunities and establish a foothold in this field.

Q: Is Intel's AI strategy a full-stack model? Does it aim to become a company like NVIDIA? How do you position yourself in the AI market where NVIDIA and others have an advantage? Have you considered establishing a more favorable market position by making ASIC chips?

A: Intel's AI strategy is a full-stack model, providing solutions that include system software as a full-stack solution.

In the AI market where NVIDIA and others have an advantage, Intel is currently in a lagging position, planning to find areas to enter, providing differentiated solutions and services, while leveraging x86 architecture advantages, participating in workload allocation and optimization, and focusing on new architectures, integrating new ideas through supporting startups and incubation projects.

Considered establishing a more favorable position by making ASIC chips, willing to cooperate with related companies, providing customizable AI platforms to enhance performance, and will seize this opportunity.

Q: Does this quarter's asset impairment include inventory? Can you explain the specific content of the impairment?

A: This quarter's asset impairment includes inventory, but inventory impairment is not listed separately, it is part of the normal cost of sales accumulation.

The specific content of the impairment mainly includes:

- Equipment impairment: For example, replacing older tools in the production line, as these older tools have a book net value higher than the open market sale price, so they are impaired to saleable value and classified as held-for-sale assets, mainly involving some older and unused tools.

- Approximately $200 million adjustment: Some additional costs originally included in inventory are now changed to be included in period expenses.

Q: Is the timeline for 18A supporting three generations of products from 2026-2028 and 14A launching as early as 2029 correct? If the 14A process fails, will the wafer foundry strategy fail as well? If the 14A process does not meet expectations, can sustainable business operations be maintained solely by relying on 18A's internal capacity and increasing external foundry?

A: The 18A process is crucial for the R&D iteration of our internal products. Once we are ready, we can more confidently move towards the external market and seek customer support. As for the 14A process, like TSMC's A14 process, its timeline is around 2028, 2029, which is consistent with previous statements.

The peak output of 18A will appear in the early next decade (around 2030), it will be used for a long time with good investment returns, mainly for internal use in the short term, and in the future, as product performance and yield improve, it can also attract external customers.

Q: Can next year's capital expenditure be maintained at 50% of the current level (about $9 billion)? Based on this year's pace (about $14-14.5 billion for the whole year), what is a reasonable number for next year's capital expenditure? Is it more likely to be around $9 billion or $14 billion? What are your actual thoughts on next year's capital expenditure?

A: Even considering next year's maintenance capital expenditure, existing construction-in-progress assets are not enough to support all capital expenditures, so next year's plan is not yet determined, it will be locked in around January 2026. The number will be significantly higher than $9 billion, but definitely lower than $18 billion.

Q: Why did the average selling price (ASPs) decrease by 8% compared to last year despite the increase in the number of server cores? From a medium to long-term perspective, ARM claims to occupy half of the server market, how significant is its competitive impact?

A: The company made mistakes in the high-performance server field (such as neglecting the once advantageous simultaneous multithreading technology), and is working hard to close the performance gap.

The company currently holds about 55% market share in the server market, and is striving to regain market share through listening to customer needs, early cooperation, and clear product roadmaps, combined with new products.

Q: Assuming mid-single-digit growth in sales next year, how will the gross margin change considering the 18A product portfolio and outsourcing situation? If revenue grows, is there a simple formula to calculate the change in gross margin relative to the current 36% level?

A: Clearly, the details are crucial here, but a good rule of thumb is that the company's realization rate is about 40% to 60%. If things go well, we hope to approach the upper end of the range.

Q: Is stabilizing and regaining market share dependent on Diamond Rapids? Is the launch time of Diamond Rapids in 2026 or the second half of 2026?

A: The company is focusing on the product roadmap including Diamond Rapids, and is committed to making the data center business competitive again. Its time frame is roughly within a six-month range before and after. As for the next generation of products, the company will obviously evaluate the entire core market, with related work starting in 2028, 2029, ensuring the launch of reliable performance products.

Q: How should the Arizona and Ireland wafer fabs be analyzed? What is the latest situation regarding the previously mentioned adverse impacts (this year $500 million, next year $1.3-1.5 billion, significant increase in 2027)? Are you still considering it this way? Will the numbers rise?

A: These numbers are roughly suitable for prediction. Clearly, by around 2027, 2028, these fabs will gradually reach normal operating levels, with related figures exceeding $1.2-1.3 billion. We will update information as progress is made.

Risk Disclosure and Statement of This Article:Dolphin Research Disclaimer and General Disclosure