---
title: "INTC (Trans): 2026 CapEx raised to $20bn+"
type: "Topics"
locale: "en"
url: "https://longbridge.com/en/topics/42904412.md"
description: "Below is Dolphin Research's recap of $Intel(INTC.US) FY26 Q2 earnings call Trans. I. Core results review. 1) Q2 results beat guidance across the board, marking a seventh straight beat.Revenue was $16.1bn, $1.8bn above the midpoint of guidance. AI-driven businesses grew over 70% YoY and contributed ~70% of revenue, including record growth in the data center. Non-GAAP GPM was 41.8%, ~280bps above guide, helped by higher revenue and yield improvements..."
datetime: "2026-07-23T22:41:20.000Z"
locales:
  - [en](https://longbridge.com/en/topics/42904412.md)
  - [zh-CN](https://longbridge.com/zh-CN/topics/42904412.md)
  - [zh-HK](https://longbridge.com/zh-HK/topics/42904412.md)
author: "[Dolphin Research](https://longbridge.com/en/news/dolphin.md)"
---

# INTC (Trans): 2026 CapEx raised to $20bn+

**Dolphin Research's** $Intel(INTC.US) **FY26 Q2 earnings call Trans**

For the full commentary, see '[Intel: CPU price hikes help, but the real pivot hinges on foundry](https://longbridge.com/zh-CN/topics/42907951)'.

**I. Key takeaways**

1\. **Q2 beat across the board, seventh straight quarter above guidance**: Q2 revenue of $16.1bn, $1.8bn above the midpoint. AI-driven lines grew over 70% YoY, contributing ~70% of revenue, including record data center growth.Non-GAAP GPM was 41.8%, ~280bps above guide, helped by higher revenue, yield gains, and ASP uplift from mix and pricing. Non-GAAP EPS was $0.42 vs. $0.20 guided, with revenue growth the strongest in 15 years.

2\. **Q3 guide**: Revenue of $15.8–16.8bn (mid $16.3bn). Non-GAAP GPM 42%, tax rate 11%, EPS $0.38.Full-year Non-GAAP opex is ~$16.5bn. NCI is ~$250mn in Q3 and Q4; 2027/2028 NCI ~$1.1bn per year (GAAP). **Management reiterated keeping quarterly GPM firmly above 40% as the top FY target.**

3\. **Cash flow and balance sheet solid**: Q2 operating cash flow was $7.0bn. Cash and short-term investments were ~$30bn, plus a $10bn revolver, for ~$40bn liquidity.The company proactively delevered to secure investment-grade; ~$10bn of non-core assets remain monetizable with no near-term plan. If business outperforms, tapping capital markets is on the table.

4\. **CapEx raised materially**: On strong demand signals, **2026 CapEx was lifted to $20bn+, up ~$3bn vs. early-year view**, with tooling up ~40% vs. 2025.2027 CapEx will be significantly above 2026, with the vast majority in the U.S. Cumulative U.S. plants and equipment investment in 2021–2026 is near $100bn, above any peer over the same period.

**II. Call details**

**2.1 Management highlights**

1\. **Strategy and supply/demand**

a. Demand continues to outstrip supply: the industry faces one of its worst supply squeezes across advanced logic nodes, silicon wafers, memory and substrates, unlikely to ease near term.The company benefits from three strategic assets — x86 CPU, advanced packaging, and a large foundry network.

b. As AI moves from training to inference, then toward agentic and multi-agent, server CPU core density keeps rising. Core server CPU growth is at a new high.

c. Deepening collaboration with $Alphabet - C(GOOG.US) Cloud to accelerate an AI-first internal pivot. Operational discipline implemented 15 months ago is showing results with better efficiency, faster decisions, and closer customer engagement.

2\. **Intel Foundry and advanced nodes**

a. Q2 foundry revenue was $5.8bn, +6% QoQ. 18A output ran ~25% above target and grew \>50% QoQ; external foundry revenue was $293mn.Foundry OP loss was $730.8mn, narrowing QoQ on yield gains, cycle-time improvements, and wafer cost declines as Intel 4/3 and 18A scale up.

b. **18A**: Now in volume across multiple commercial and consumer products, with yields exceeding plan. Panther Lake single-SKU cost is down ~50% year-to-date, with another ~20% cut expected this year and further declines in 2027.18AP has entered risk production, offers ~5% performance improvement vs. 18A, remains IP/design compatible, and targets external customers.

c. **14A**: PDK 0.5 completed; PDK 0.9 targeted for Oct, a key milestone. 256 SRAM yield, defect density and transistor performance are ahead of 18A at the same stage.Internal products are slated for H2 2027 risk production, with a Q2 decision to fully ramp in 2028. External interest has risen visibly after seeing PDK 0.9 and yields.

d. **Advanced packaging**: EMIB-T continues to see strong customer interest and a growing backlog. Yield and reliability are on target, with a goal to ramp for customers in 2027.

3\. **Data Center & AI**

a. Q2 revenue was $6.3bn, +24% QoQ and +59% YoY, well above expectations on hyperscaler and enterprise demand. OP was $2.5bn with 40% OPM, up ~$1.0bn QoQ.Server YoY growth set a record in Q2.

b. Purpose-built chip revenue grew ~20% QoQ and was nearly 3x YoY.

c. Launched the first 18A server-grade product, Xeon 6+ (Clearwater Forest). Announced rack-scale and disaggregated inference with SambaNova and Foxconn, and new controllers/adapters supporting 10–200Gb Ethernet.

d. Server roadmap includes Clearwater Forest, Diamond Rapids, and Coral Rapids with SMT. Granite Rapids on Intel 3 sees very strong demand and is supply constrained.

4\. **Client Computing & Physical AI (CCPG; ex-PC)**

a. Q2 revenue was $8.9bn, +15% QoQ and above plan. OP was $2.3bn with 26% OPM, down ~$173mn QoQ due to an inventory write-down tied to fab network optimization.

b. AI PC revenue grew 26% QoQ and now accounts for roughly two-thirds of client revenue. Edge deployments are ~10% of CCPG.

c. 18A is fully ramped; Series 3 has 400+ designs. Arc iGPU has 40+ designs, with Arc G launched for handheld gaming.vPro activations surged 1,500% over the past four quarters. Edge AI secured 130 Series 3 designs, including robotics brain and control deployments.

d. The rename to CCPG reflects edge AI opportunities. Management believes edge and physical AI TAM can match client TAM over time.

5\. **ASIC/design services and new opportunities**

a. Design services revenue rose nearly 3x YoY. Management estimates ASIC TAM at $100bn+.With end-to-end design, IP breadth, and advanced packaging/foundry capabilities, the company is expanding from networking chips into compute and ultimately accelerators.

b. Announced a partnership with Fortinet for next-gen security processors, a key step in the ASIC strategy. Intel IPU is shipping to some hyperscalers.

c. Multi-year work with SambaNova advances heterogeneous AI and disaggregated inference.

6\. **Memory strategy**: Memory is now a major supply bottleneck, and the company is working closely with the three leading memory makers to secure supply as the top priority.It recently hired the former CEO of $SK Hynix(SKHY.US). The firm is exploring compute-memory integration, stacking, and efficiency, with internal R&D including Z-Angle Memory and cross-batch memory.

**2.2 Q&A**

**Q: What does the CapEx raise (up ~$3bn this year and significantly more next year) signal for foundry customers? Any firm 14A or 18AP orders? What about packaging?**

A: This is a broad investment covering advanced packaging. We are optimistic on EMIB-T and will keep investing, but front-end fabs are far more costly than packaging, so the tilt remains toward front-end.For customers, the scale-up reflects confidence in demand across business lines, especially those under long-term agreements, giving enough visibility to pre-build capacity for the next few years. We stay highly disciplined, investing only where returns are clear; new fabs are a cash outflow at first, which is why next-year CapEx rises, but long-term ROI is attractive.

As we shift to longer node lifecycles, returns should be compelling. We already see this on Intel 10 and Intel 7.

**Q: A competitor guides CPU TAM to $220bn by 2030 with ~45% CAGR. Do you agree, and can you meet that?**

A: We will not provide a specific number, but we agree it is a strong, fast-growing market. **We have said CPU-to-GPU ratios are rising and now near parity at ~1:1, and shipments could tilt even more toward CPU.**We have a solid position and an opportunity to capture substantial share. The number will be large, and while precise forecasting is hard, based on client spend, LTAs and visibility, we expect very robust growth.

**Q: Does having internal fabs help regain server share this year? Over the next five years vs. AMD and ARM, how do you view share recovery?**

A: Demand is very strong, and the near-term challenge is expanding supply to meet it. In server/DC, our roadmap is strong — Clearwater Forest, Diamond Rapids, and Coral Rapids adding SMT, with ongoing single-thread and multi-thread gains for competitiveness.On ARM, we are strong partners, including on ASIC foundry with IP collaboration. Overall we feel competitive with a strong roadmap; we are catching up quickly where behind and investing heavily to leapfrog in parts of the CPU architecture.

**Q: Is there a net vs. gross CapEx distinction? How do you allocate between internal vs. foundry, and will you disclose the external portion?**

A: There is a gross-to-net difference primarily from AMIC (advanced manufacturing investment tax credit, ITC), currently in the low tens of billions of dollars (~$10–40bn) over time and growing in share as it ramps.In the U.S., each $1 invested gets ~$0.35 back via credits, and most CapEx is U.S.-based. There is timing lag — fabs must be built and production-ready to claim equipment credits, followed by IRS filings, so cash out and refunds are offset in time.

We plan by wafer starts at each node, investing front-end accordingly, and place orders based on evolving demand drivers while staying flexible. Packaging has a sizable backlog and needs expansion; part will be in-house, and part requires substrate purchases that need prepayments, so this portion moves faster.

**Q: Is the strong client performance purely pricing-driven? With a potentially softer H2 end market, what is the client outlook?**

A: Client beat was mainly ASP uplift, partly from mix and partly from same-spec price increases, as we passed through cost inflation. YoY the market is down given the Windows refresh in 2025 created a high base, and memory cost/availability is a headwind.We are skewing product mix higher, supporting ASPs and performance. Next quarter, client revenue should be roughly flat; CCPG may be up slightly on edge strength, with client flat.

Underneath, the market is soft due to memory supply/demand, and we expect the end market down this quarter. But CPU inventories were kept relatively tight by customers, so Q3 may see some CPU restocking, with Q4 we will start to feel it.This is positive for us, **as data centers urgently need CPU and supply falls well short; we will redirect capacity to data center CPU to narrow the gap.**

**Q: What is the client inventory write-down, and how large? Excluding it, does the GPM guide imply a QoQ decline?**

A: Some products were not fully completed from a match-set perspective, and economically we prefer to shift capacity elsewhere rather than finish them, creating stranded inventory and write-downs.The guide is roughly flat QoQ: the Q2 write-down will not repeat, which helps, but Panther Lake and Granite have improving unit costs while rising in mix share and still below corporate average GPM early in their lifecycles, which is a drag. Net-net, GPM is guided flat.

As 18A yields improve further, both effects turn into tailwinds, with Panther Lake exceeding corporate GPM and lifting the whole. Our top goal this year is keeping GPM above 40% every quarter; we did so in H1 and guided Q3 similarly, and will seek further GPM gains after that.

**Q: With substantial capacity coming online by quarter-end, does that mean Q4 revenue will jump QoQ? If the Sept supply gap is similar to last quarter (\> $1bn), Q4 looks strong — is that fair? What are the puts/takes?**

A: We guide only one quarter ahead. That said, **if inventory or supply starts improving late Q3 or early Q4, it will help, but even with improvements we will not fully meet demand, and Q4 remains supply constrained.**Internal wafers are tight as well. Supply comprises internal wafers plus advanced packaging elements like substrates, glass and memory, and those purchases are bottlenecked, with some of the tightest points in the chain.

Front-end improvements are more linear, while back-end is chunkier. As some chokepoints start to ease by late Q3, this quarter should be more flattish, with upside potential in Q4.

**Q: Mar and Jun drop-throughs were strong; by guide, drop-through will be just above 50%, still within the 40–60% range you mentioned. Is that framework still valid, and what are next year's puts/takes?**

A: Over the longer term, **we expect drop-through to stay in the 40–60% range.** Each quarter has unique dynamics that place it at the low, high or midpoint of that range, but the range is a reliable rule of thumb.

**Q: When will external foundry confidence turn into customer announcements? How much will next-year CapEx rise, and how is it split between external and internal?**

A: On confidence: **18AP is in risk production and will be ready by year-end**, with ~5% performance uplift over 18A. 18A yields and output are strengthening, with Panther Lake ramping visibly.More importantly, 14A: PDK 0.5 is done, PDK 0.9 is targeted for Oct as a key milestone, and 256 SRAM yield, defect density and performance all beat rigorous targets. 14A internal products will enter risk production in H2 2027, with a 2028 mass production commitment.

External customer feedback is very positive. After seeing PDK 0.9 and yields, they are assessing product candidates and capacity, which are strong signals.Our principle is to invest CapEx only when yields perform, IP is ready to serve, and customer engagement is at the right stage. On CapEx mix: since 2026 space has been well invested over recent years, **we need relatively limited incremental space and will spend mainly on tooling, up ~40% vs. 2025**, focused on Intel 3, 18E and 18AP.

We intentionally did not give the 2027 figure as we are finalizing scale, and industry practice is to disclose early-year. The 2027 number will be higher, with funds going to both internal and external lines.We take a holistic view of wafer demand across customers and build matching capacity accordingly.

**Q: How do you view the balance sheet as you invest for H2 and next year? Will CPU success fund this, or will other sources be needed?**

A: The balance sheet is in good shape: \>$30bn cash plus a $10bn revolver for ~$40bn liquidity. We proactively delevered to keep investment-grade.Revenue, profitability and EBITDA are expanding, driving operating cash flow. We also have ~$10bn of non-core assets that can be monetized if needed.

Customers are willing to co-invest, with prepays helping unlock capacity. If business is very successful, which we are driving toward, we may raise more in capital markets and will update shareholders.

**Q: As capacity ramps this and next year, what is the server revenue cadence? How do shipments and ASP evolve in H2 2026 into 2027?**

A: Server wafers are almost entirely internal (aside from some ASIC). **We are investing heavily to expand wafer starts at key nodes, especially Intel 3** for Granite Rapids, which has extremely strong demand and is very tight.We are gradually expanding at Intel 3, with chunky progress, but plans for the rest of this year and next year look solid.

The challenge is expanding both front-end and back-end, hitting tight points like substrates, which we are addressing. We talk shipments, and unit growth looks strong.Weighted-average core counts per product are rising, and pricing in this market is typically per-core ASP, so higher core counts help ASP. This will be a key revenue driver and supports well-above double-digit CAGR over the next few years.

**Q: Do you have a CapEx framework to balance meeting demand and FCF targets? How should we model reasonable CapEx, or do you build capacity upon customer sign-ups?**

A: We are more cautious. On base business operating cash flow, even with higher CapEx, AMIC credits make cash flows quite healthy.We likely need back-end investments, including third-party, which could pressure next year's cash flow and make FCF turning positive more challenging, but all carry attractive ROI.

As long as we have confidence in growth rates, pricing and cost structures, and know node lifecycles are long with strong ROIC, we will invest. We avoid large bets before customer commitments — a major shift under new management.Conversely, our confidence for next year implies high customer certainty; otherwise we would not be placing POs now.

**Q: ASIC is at a ~$1.2bn run rate and growing well. How do you view diversification, growth prospects, and margin structure as scale expands?**

A: This is a massive opportunity, with TAM likely \>$100bn. Our differentiation is advanced design (CPU, XPU), a strong IP portfolio, and unique advanced packaging enabling radical integration, which new AI technologies require alongside advanced silicon.That creates opportunities for purpose-built chips across many customers. For example, our announced Fortinet security ASIC will deliver higher-performance next-gen security processors, and Intel IPU for some hyperscalers is another large opportunity.

The business is up ~3x YoY. In scale terms, operating revenue is near $2.0bn run rate today, and we expect to reach ~$4.0bn run rate not far out.Facing a $100bn TAM, our IP and integrated capabilities should secure a substantial share.

**Q: With changes in memory tiers and architecture (e.g., Z-Angle Memory, cross-batch memory), will Intel play a bigger role in memory? Are there internal R&D and long-term opportunities?**

A: First, memory is a major supply constraint, and we are working closely with the three leading memory makers as a top priority to serve customers. Second, Intel has deep memory heritage and recently hired SK Hynix's former CEO. Memory is becoming a bottleneck across AI infrastructure and a pain point for customers.We are researching compute-memory integration, stacking, and more efficient memory use. There is a lot underway, and we will keep updating.

<End of text\>

**Risk disclosure and statement:**[**Dolphin Research Disclaimer and General Disclosure**](https://support.longbridge.global/topics/misc/dolphin-disclaimer)

### Related Stocks

- [GOOG.US](https://longbridge.com/en/quote/GOOG.US.md)
- [GOOGL.US](https://longbridge.com/en/quote/GOOGL.US.md)
- [601138.CN](https://longbridge.com/en/quote/601138.CN.md)
- [INTC.US](https://longbridge.com/en/quote/INTC.US.md)
- [04335.HK](https://longbridge.com/en/quote/04335.HK.md)
- [GOOGN.US](https://longbridge.com/en/quote/GOOGN.US.md)
- [SKHY.US](https://longbridge.com/en/quote/SKHY.US.md)