--- title: "INTC (Trans): Server shortages persist, PC destocking drags Q4" type: "Topics" locale: "en" url: "https://longbridge.com/en/dolphin/post/44645150.md" description: "DCAI revenue rose nearly 80% YoY. Full-year revenue is on track to beat the initial outlook by almost $10bn.However, an Israel fab shutdown and PC destocking drove Q4 guidance down to $15–16bn and GPM to 38%. The company expects the server CPU TAM to top $200bn within five years and is ramping 2027 capex." datetime: "2026-10-29T23:26:47.000Z" locales: - [en](https://longbridge.com/en/dolphin/post/44645150.md) - [zh-CN](https://longbridge.com/zh-CN/dolphin/post/44645150.md) - [zh-HK](https://longbridge.com/zh-HK/dolphin/post/44645150.md) author: "[Dolphin Research](https://longbridge.com/en/dolphin.md)" generator: "portal-rs" --- # INTC (Trans): Server shortages persist, PC destocking drags Q4 **Dolphin Research compiled the Trans of** $Intel(INTC.US) **FY26 Q3 earnings call** **I. Core Takeaways** 1\. Capital structure and shareholder returns a. Completed a $23 bn equity raise in Aug, with incremental interest income contributing approx. $0.01 to Q3 EPS (not in prior guide). Q4 diluted share count is approx. 5.44 bn, reflecting a full quarter of the new issuance. b. End-of-period cash and ST investments were approx. $54 bn, and the revolving credit facility was upsized from $10 bn to $15 bn. Despite higher capex and working capital, the company reaffirmed deleveraging by 2027. No dividends or buybacks were addressed this time. 2\. Q4 guidance a. Revenue of $15.0–$16.0 bn; at the midpoint of $15.5 bn, non-GAAP GPM of 38%, tax rate of 11%, and EPS of $0.25. b. The Israel Fab 28 incident is initially expected to reduce revenue by approx. $300 mn (almost entirely 10nm server), and compress GPM by approx. 200 bps. c. Segments: DCAI to grow mid-teens QoQ, with purpose-built silicon outpacing the segment overall. CCPG to decline over 20% QoQ as customers rapidly cut weeks of inventory. d. QoQ GPM contraction is mainly from lower revenue diluting fixed costs (client wafer starts reduced), plus a higher mix of lower-margin purpose-built silicon. Yield and cost improvements on Intel 3 and 18A partially offset the pressure. 3\. Opex and capex a. Non-GAAP opex is approx. $4.4 bn in Q4 and approx. $16.3 bn for the year. NCI is approx. $250 mn in Q4 and approx. $1.1 bn in each of 2027 and 2028. b. 2026 total capex remains at $20 bn, with a sharp step-up in 2027. Firm POs to equipment vendors rose nearly 40% QoQ, and the company is accelerating cleanroom build-out at Arizona Fab 62 while evaluating a faster Ohio timeline. c. Investment discipline: deploy only when there is confidence in product health, process technology, and customer demand. Confidence across all three is improving. 4\. Key metrics this quarter a. Total: revenue of $16.5 bn; non-GAAP GPM of 44.1%, approx. 200 bps above Q2 and guidance (from Intel 3/18A cost-down and better mix). EPS of $0.52 (vs. guided $0.38). Operating cash flow of $5.1 bn. b. DCAI: revenue of $7.3 bn, +17% QoQ and nearly +80% YoY. OP of $3.0 bn (OPM 42%), up approx. $600 mn QoQ. Full-year revenue is tracking nearly $10 bn above the initial plan. c. CCPG: revenue of $8.1 bn, -9% QoQ and -5% YoY, below expectations. OP of $2.3 bn (OPM 28%), roughly flat QoQ due to lower inventory-related charges. d. Foundry: revenue of $6.7 bn, +16% QoQ. External revenue of $411 mn, up $118 mn QoQ, mainly from a one-off customer project. Operating loss of $1.9 bn, narrowing by $236 mn QoQ. **II. Earnings Call Details** **2.1 Management highlights** 1\. Data Center (DCAI) a. Demand: Q3 delivered the strongest growth for the segment in 15 years, spanning hyperscale, enterprise, and emerging cloud customers. Demand still materially exceeds supply. b. Industry view: supply constraints will run through 2027 and into 2028. Industry server CPU shipments are expected to grow 30%+ in both 2026 and 2027, with even higher growth once supply loosens. c. Product: Granite Rapids (Intel 3) is one of the fastest ramps and has become the largest server product family. Clearwater Forest (first 18A server product) is shipping and generating revenue, with substantial demand in 2027. d. Diamond Rapids (first 18A-P server product) is sampling, with volume in 2027. An agreement with Oracle Cloud Infrastructure will deploy the latest-gen CPUs. e. At OCP, Clearwater Forest demonstrated over 46k cores per rack with approx. 133 kW rack power. Inference GPU Crescent Island has been renamed Intel GPU D Series. f. Customer interest in both products is above initial expectations. GPU is a new market and will take time to scale, while customer dialogues are moving toward firm commitments. g. Bottlenecks: Intel 3 yield and cycle time still need improvement. Advanced organic substrates for EMIB packaging are in industry-wide shortage, and the Israel incident caused a prolonged shutdown that will impact Q4 10nm server shipments. h. Purpose-built silicon has tripled over the past three years and could triple again over the next few years. Potential new customers include AI labs, and the IP portfolio is expanding. 2\. Client (CCPG) a. Memory price hikes and shortages lifted system prices, while constrained IT budgets drove faster PC adjustments over the past two months, with OEMs actively destocking. This should continue until memory supply and pricing normalize. b. Industry PC TAM is down approx. 12% in 2026 (H2 down 20%+), and planning assumes approx. 10% down in 2027. Declines are steeper in H1, with potential improvement in H2 as comps ease, while the long-term view is unchanged. c. Product: Panther Lake (18A) is ramping smoothly. Cost-optimized Wildcat Lake shipments rose approx. 70% QoQ and are sold out. 18A client CPU revenue mix rose from under 5% in Q1 to nearly one-third. d. Nova Lake (first 18A-P client product) is ahead of schedule, with expected design wins at 2x Panther Lake at launch, and more to be disclosed at CES in Jan. AI PC revenue grew 35% YoY, with edge businesses showing greater resilience. 3\. Foundry and advanced packaging a. Yields on Intel 3 and 18A continue to improve, with 18A advancing at a double-digit pace, reaching industry-standard levels and beating internal timelines. Wafer output on both nodes is ahead of expectations. b. Unit cost: Intel 3 is down approx. 15% vs. Q1, with shipments up mid-teens QoQ. 18A unit cost is down approx. 40% vs. Q1. c. 18A-P will be enabled across internal product lines by year-end. Engagements with several highly strategic customers are underway. d. 14A released the 0.9 PDK to select customers in Oct as promised, meeting industry-standard yield and performance thresholds, with risk production next year. As usual, customer specifics were not disclosed, but discussions with multiple potential customers are progressing well. e. EMIB and EMIB-T are seeing strong external interest. HBI readiness for future server 3D stacking is advancing. f. Capacity: tool installs for Intel 3 in Ireland are progressing. 18A is ramping in Oregon and Arizona, including the second high-volume 18A fab, Fab 62, while packaging capacity is expanding in New Mexico and Malaysia. In Ohio, the first major structural truss is installed and an acceleration is under evaluation. 4\. Long-term strategy a. Server CPU TAM, driven primarily by agentic workloads, could grow from approx. $40 bn to over $200 bn in five years. This assumes CPUs are far more power-efficient than GPUs for deterministic tasks. b. Three advantages: an x86 roadmap that, for the first time, opens validated modular x86 CPU instances to customers. As the only US-based advanced-node foundry, the company can scale CPUs on any ISA for customers, and it brings advanced packaging. c. Purpose-built silicon, advanced packaging, and external foundry each represent new markets in the tens of billions. Management sees a path to double-digit long-term revenue growth, operating leverage, and margin expansion. d. The company plans to acquire Israel-based quantum software firm Classiq, shifting quantum from fundamental research to commercialization. Post-close, it would be a full-stack quantum computing company with revenue. **2.2 Q&A** **Q: Can you frame 2027, and how DCAI growth offsets PC declines?** A: **No 2027 guide, but qualitatively a growth year: DC momentum continues, though H1 remains supply constrained.** Ex-Israel, DC strength carries into next year. Substrate capacity rises 40%, but most lands in H2; Intel 3 stays tight, and shifting to Oregon takes time to ship. Purpose-built silicon is expected to triple again over the next three years. The client side is hardest to call: beyond TAM declines, inventory adjustments showed in Q3 and peak in Q4, lasting at least into Q1 before normalizing to TAM cadence. Nova Lake is the strongest lineup in years, with share gains from next year. Longer term, AI workload migration to endpoints could revive PCs or agentic devices. **Q: With capex surging in 2027 (street at approx. $30 bn), do you still commit to positive FCF?** A: **2027 is an investment year, and positive FCF is unlikely.** Final capex will be set by year-end, but orders are placed given equipment lead times, so the direction is clear. Next year, the company will ramp existing nodes while investing for 14A production in 2028, concentrating spend in one year, with the Aug raise pre-positioning for this. Each node targets at least 20% ROI. Intel 7 is already profitable, Intel 3 will be profitable next year, and these businesses should be cash-generative over time. **Q: For Q4, how much impact from Israel, client, and execution, and when does it improve?** A: **Israel reduces Q4 revenue by approx. $300 mn, with a similar impact in Q1, and most wafer scrap booked in Q4.** The plant has been down about two weeks and should restart by this weekend, while the usability of in-line wafers is still being assessed, so estimates are preliminary. Q4 GPM takes approx. a 200 bps hit from scrap, with no additional scrap expected in Q1. Client will reset to current levels, with inventory drawdowns in Q1 similar to Q4, plus seasonal QoQ decline. Execution is mostly positive: Intel 3 and 18A yields are improving, and the CEO is in Japan securing substrate allocations. Supply remains tight in Q4 and H1 next year, after which DC growth should normalize to a healthier pace. **Q: When can servers grow in line with the market, is it H2 next year?** A: **Intel 3 capacity in Oregon comes online in H2 next year, easing server supply.** Granite Rapids is the largest revenue driver in servers, with yield improvements and capacity adds progressing in parallel. Roadmap: Clearwater Forest on 18A should have ample capacity. Diamond Rapids in 2027 is on 18A-P, and Coral Rapids in 2028 is on 14A. **Q: Can weaker PCs free wafers for servers and speed the transition?** A: **Yes: reallocate 18A capacity from Panther Lake to Intel 3.** This frees capacity to build more DC wafers, but constraints are not only wafers; substrates must also scale in tandem to meet demand. **Q: Peers are guiding +70% and ARM camps are growing fast; how does server share evolve next year?** A: **Share depends primarily on supply; Coral Rapids will be competitive vs. AMD.** The next-gen products will lead AMD, and CPUs remain the core business. The CFO added that the more relative supply improves, the better share should perform, without quantifying share. **Q: How does GPM evolve in 2027 (destock and utilization, higher depreciation, 18A/ASIC mix vs. server mix and pricing)?** A: **Hard to quantify, but they lean on the rule-of-thumb that incremental revenue expands margins.** Israel's GPM impact is mainly in Q4, with no large scrap expected in Q1. Positives: Xeon demand is strong and Granite Rapids is accretive to margins; Intel 3 and 18A continue to lower structural costs while improving yields and output, with Q4 expected to top Q3. Negatives: 18A under-load could drag as volume is below plan, partially mitigated by moving Intel 3 into Oregon. Purpose-built silicon has lower early-stage margins than the company average, though the focus is on growing absolute GP and it could dilute percentage margins next year. Flow-through is approx. 40%–60%, not necessarily linear QoQ but within that range YoY. **Q: Was there channel stocking in Q2, and why believe memory supply will improve?** A: **Not optimistic: memory likely will not improve in 2027, and next year's PC TAM is down approx. 10%.** They did not directly address Q2 stocking. Destocking is concentrated in Q4 and Q1 and could extend into Q2, with a larger company impact than TAM declines per se, after which shipments should align with consumption and modestly recover. Whether 2028 improves remains to be seen. **Q: Q4 implies DC revenue of approx. $8.5 bn, or ~$34 bn annualized — is that a 2027 floor, and is ASP moving up?** A: **They view it as a reasonable starting point and expect strong growth in 2027, though wafers and substrates are tighter in H1.** More Intel 3 output, yield-for-volume trade-offs, and investments with substrate vendors should release capacity in H2. Purpose-built silicon, booked in DCAI, is a key driver of Q4 growth and extends into 2027, while ASP was not addressed. **Q: How are external foundry engagements progressing, and when do packaging and wafers (18A-P vs. 14A) tape out or generate revenue?** A: **Two packaging customers to be announced within two quarters, and the 14A 1.0 PDK will be out in Q1 next year.** EMIB-T is on track, with focus on securing substrate capacity and yields and shortening cycle times, led by the former SK Hynix CEO, driving a large substrate capacity increase over the next two years. Disclosure timing is up to customers. 18A-P: 18A family is in volume on Panther Lake and Wildcat Lake, with external customers proactively engaging and satisfied with yields, while current focus is on die size and IP readiness. 14A 0.9 PDK has drawn strong responses from select customers, with performance, yields, defect density, and variability improving daily; after 1.0 PDK, product onboarding will proceed. No timing was given for modeling external revenue. **Q: Can DCAI grow QoQ in Q1?** A: **No Q1 DC revenue guide, only that H1 supply is tighter than H2.** CPU demand is strong through next year, and the company will likely be unable to fully meet it. The CEO added they announced an Oracle Cloud collaboration and that several customers are interested in LTAs to lock CPU supply. **Q: Has Panther Lake GPM surpassed the low-to-mid 40% range?** A: **Panther Lake margins have improved meaningfully, but remain below the company average.** Continued yield, output, and scale improvements are needed, with long-term accretion expected. **Q: Is the capex step-up more for tools or buildings?** A: **Primarily tools at this stage, with heavier tool spend in Q4; only Ohio remains in heavy construction.** Tool spend rose sharply in Q3. Most other fabs are structurally complete, with final fit-out ongoing. **Q: How will next year's PC product mix affect pricing?** A: **Mix should be favorable: Nova Lake fills the high-end desktop gap, entering as a share gainer next year.** Panther Lake demand is relatively resilient, and cost-optimized Wildcat Lake is growing strongly and is timely. Pricing impact was not quantified. \ **Risk disclosure and statements:**[**Dolphin Research disclaimer and general disclosures**](https://support.longbridge.global/topics/misc/dolphin-disclaimer) ### Related Stocks - [LAKE.US](https://longbridge.com/en/quote/LAKE.US.md) - [ORCL.US](https://longbridge.com/en/quote/ORCL.US.md) - [INTC.US](https://longbridge.com/en/quote/INTC.US.md) - [ORCL-D.US](https://longbridge.com/en/quote/ORCL-D.US.md) - [04335.HK](https://longbridge.com/en/quote/04335.HK.md) - [ORCX.US](https://longbridge.com/en/quote/ORCX.US.md) - [ORCU.US](https://longbridge.com/en/quote/ORCU.US.md) - [ORCS.US](https://longbridge.com/en/quote/ORCS.US.md) - [INTW.US](https://longbridge.com/en/quote/INTW.US.md) - [LINT.US](https://longbridge.com/en/quote/LINT.US.md) ## Comments (2) - **lyx · 2026-10-31T00:05:24.000Z**: then better and most idea then beplz and it is then - **YAKYBOBO · 2026-10-30T00:03:41.000Z**: Good analysis --- > **Disclaimer: This article is for reference only and does not constitute any investment advice.**