---
title: "INTC (Trans): 18A yield beats internal expectations; server CPU to maintain double-digit growth"
type: "Topics"
locale: "en"
url: "https://longbridge.com/en/topics/40159675.md"
description: "Below is Dolphin Research's transcript summary of $Intel(INTC.US) FY26 Q1 earnings call. For our take, please see 'Intel: Leaving the darkest hour behind as the old champion mounts a counterattack.'1) Q1 beat across the board. Revenue was $13.6bn, $1.4bn above the midpoint of guidance. Non-GAAP GPM came in at 41%, ~650bps ahead of guide.Non-GAAP EPS was $0.29 vs. breakeven guided. This included approx. $0.06 from interest and other one-off gains..."
datetime: "2026-04-24T03:40:03.000Z"
locales:
  - [en](https://longbridge.com/en/topics/40159675.md)
  - [zh-CN](https://longbridge.com/zh-CN/topics/40159675.md)
  - [zh-HK](https://longbridge.com/zh-HK/topics/40159675.md)
author: "[Dolphin Research](https://longbridge.com/en/news/dolphin.md)"
---

# INTC (Trans): 18A yield beats internal expectations; server CPU to maintain double-digit growth

**Below is Dolphin Research's transcript of**$Intel(INTC.US) **FY26 Q1 earnings call. For our take on the print, see** [**Intel: Turning the Corner, Legacy Leader Strikes Back**](https://longbridge.cn/zh-CN/topics/40156400?channel=SH000001&invite-code=294324&app_id=longbridge&utm_source=longbridge_app_share&locale=zh-CN&share_track_id=6f9c7d66-63bd-4702-8267-8d4e3aacade7)**.**

**I. Key takeaways from Intel's results**

1\. **Q1 beat across the board**: revenue of $13.6bn, $1.4bn above the guide mid-point. Non-GAAP GPM at 41%, ~650bps above guide, and non-GAAP EPS at $0.29 vs. breakeven guided, including ~$0.06 from interest and other one-offs. OCF was $1.1bn, capex $5.0bn, and adj. FCF at -$2.0bn.

2\. **Q2 guide**: revenue of $13.8bn–$14.8bn (mid-point $14.3bn), non-GAAP GPM at 39%, tax rate 11%, and EPS of $0.20. **Seq. GPM decline is mainly due to a sharp mix shift toward Intel 18A (Panther Lake shipments up 6–7x QoQ) still early in ramp, plus the absence of Q1's one-time inventory benefit**.

3\. **FY outlook**: **PC units expected to fall by low-teens YoY in 2026**; **server CPU units to grow double-digit with momentum into 2027**. **Factory output to keep rising in 2H but at a slower pace given an above-plan 1H base**. **Revenue seasonality between 1H/2H to align with the past decade**. OpEx originally targeted at $16bn, but may run higher on inflation, variable comp, and targeted investments. **Capex to be flat YoY (vs. prior flat-to-down)**. **Ex-Fab 34 buyback, adj. FCF still expected to be positive for the year**.

4\. **Fab 34 buyback**: reacquired 49% of the Ireland Fab 34 JV via ~$7.7bn cash plus $6.5bn of new debt. NCI impact expected at ~$250mn per quarter in Q2–Q4 and ~$1.1bn per year in 2027–2028 (GAAP). The company committed to repay all debt maturing in 2026 ($2.5bn) and 2027 ($3.8bn).

**II. Earnings call details**

**2.1 Management highlights**

1\. **CPU's role in the AI era**

a. AI is moving from training to inference, and from inference to agentic. CPUs are reasserting themselves as the orchestration layer and control plane for the AI stack — this is based on client feedback, not wishful thinking. b. The CPU/GPU ratio is shifting from 1:8 in training toward ~1:4 in inference, trending toward parity and potentially flipping. c. Distributed inference, agentic AI, physical AI, robotics, and edge AI are emerging workloads that create structural tailwinds for CPUs. d. AI-driven revenue now accounts for 60% of total, up 40% YoY.

2\. **Product updates**

a. **Client**: Core Ultra Series 3 on Intel 18A launched as the strongest client rollout in five years, offering better perf/W, stronger iGPU, and richer on-device AI. Core Series 3 brings latest IP and all-day battery to the mainstream for the first time. AI PC revenue rose 8% QoQ, now over 60% of client CPU mix.

b. **Data center**: **Xeon 6 on Intel 3 is ramping at full speed**, with both SKUs among the fastest ramps in five years. Xeon 6 was chosen as the host CPU for NVIDIA DGX Rubin NVL8. A long-term agreement was signed with Google. Multi-year collaboration with SambaNova to co-design next-gen heterogeneous AI inference (SambaNova RDU + Intel Xeon 6).

c. **ASIC**: revenue up 30%+ QoQ and nearly doubled YoY, with an annualized run-rate now above $1bn.

3\. **Intel Foundry**

a. **18A yields are tracking ahead of internal plans, with year-end yield goals likely achieved by mid-year**. **Intel 14A is surpassing 18A at the same point in time on maturity, yield, and performance**. b. 14A now has a 0.5 PDK available and is advancing a 0.9 PDK, upon which customers decide product and capacity needs. Early design commitments are expected to start in 2H26 and extend into 1H27. c. Advanced packaging backlog keeps growing, with annual demand in the 'multi-bn dollars' range, alongside a multi-year expansion of Malaysia back-end facilities. d. Terafab partnerships with SpaceX/xAI/Tesla aim to reinvent silicon process and boost manufacturing efficiency.

4\. **Supply and capacity**

a**. Demand continues to outstrip supply, especially for Xeon server CPUs; revenue lost due to shortages starts with a 'B' (bn)**. b. All three mid-nodes (Intel 10/7, Intel 3, 18A) are increasing wafer starts, with larger increases at EUV nodes. c. Multi-foundry strategy: TSMC remains a key partner, with product teams choosing the optimal foundry mix. d. Capex mix is shifting: space spend is falling sharply while tool spend is up ~25% YoY.

**2.2 Q&A**

**Q: What is the structure of the Google LTA and other LTAs? How do they improve visibility?**

A: Google is one of several long-term deals signed in Q1, spanning products like Xeon and IPUs, underscoring strong CPU and ASIC demand. **Most agreements include volume and pricing terms and run three to five years**. Google's deal is one both sides wanted to disclose, while others prefer confidentiality, which we respect. It is win-win — we gain volume baselines for supply planning and customers secure supply and price expectations. More deals will be disclosed when appropriate.

**Q: Does capex include investments for external foundry customers? When will we hear more?**

A: **We now expect capex to be flat YoY, raised from prior 'flat to down'**, reflecting the current demand backdrop. Structurally, space build-out absorbed outsized capex in recent years, and we have sufficient space reserved. Thus, space spend will drop sharply, while **tool spend is up ~25% YoY to close the supply gap**. Regarding **external foundry customers, our view has been consistent: we expect clearer customer signals in 2H this year into early next year**. Combined with rising internal demand, we will set multi-year supply plans and investments. We maintain close ties with tool vendors, with Lip-Bu and Naga in regular CEO-level dialogue, enabling agile capacity allocation when needed.

**Q: Is the server CPU strength driven by market demand or product differentiation?**

A: Both. First, customers are clear: as workloads shift from training to inference, CPUs are more efficient as the orchestration and control plane managing agent data. **The CPU/GPU ratio is moving from 1:8 to 1:4 or better**, driving strong demand. Second, on competitiveness, we have refined the roadmap over the last year, optimizing CPU architectures by workload. Critically, we offer a unique combo — CPUs plus advanced packaging and foundry — letting us respond faster to diverse workload needs. Beyond CPUs, we are quietly building GPU capabilities with new leadership hires, and we are expanding accelerators to cover edge and physical AI use cases. Also, while Granite Rapids is early in its cycle, initial feedback is very positive.

**Q: As you move from 'survival' to 'expansion', how do the biz. model and spending change? Is capex structurally higher?**

A: On spending, we drove major efficiency gains over the past year and flattened management layers. **Now the focus is deep workload understanding — improving efficiency end-to-end from architecture through tape-out**. On foundry, 18A yield improvements are significant. The 14A 0.5 PDK is available and 0.9 PDK is underway — the key decision point for customers on product design and capacity. Besides yields, we are cutting cycle times to meet customer ramp schedules.

**Q: Q2 segment outlook and GPM drivers? How much will higher 18A mix weigh on margins?**

A: Breaking down Q2 GPM: **pricing will improve more meaningfully vs. Q1; mix is roughly flat**, as data center grows faster but not enough to swing the total. **The main drag is 18A — Panther Lake units will be 6–7x QoQ, and while margins are improving sequentially, they remain below the company avg.** That magnitude of mix shift will weigh on GPM. Overall, it should land around Q1 levels, which does not worry me. **In 2H, yield and product cost structures will improve, but higher substrate, ABF glass, and memory costs will partially offset**. Longer term, client cost structures are on the right path, foundry is improving, and data center needs more work. **Margin expansion remains the top priority**.

**Q: How will client trend through the year? Impact from a low-teens decline in industry PC unit TAM?**

A: Distinguish industry sell-through vs. our sell-in — channel inventory changes create gaps. **We will be less exposed to the industry decline, partly on better pricing and partly on client inventory replenishment**. From a modeling standpoint, **Q2 revenue is a good proxy for quarterly run-rate in 2H**, i.e., roughly flattish from Q2 onward.

**Q: How is Terafab structured? Traditional foundry or could a whole fab be operated by the customer?**

A: On 14A, we have made strong progress on yields and cycle time and are deeply engaged with multiple customers. My approach is to under-promise and over-deliver; unless customers ask for disclosure, we will not name them. For Terafab, Elon and I share a view that the global semi supply chain is falling behind accelerating demand. We will co-learn and explore process innovations together. It is a broad partnership, and we will update as it progresses. It is a very exciting customer, and several others are in active discussion.

**Q: Can you quantify unmet demand due to constraints? How much higher would Q2 be without limits?**

A: I will not quote a number, except to say it starts with a 'B' and is substantial.

**Q: What is server CPU TAM growth this year? How much ASP expansion? How has this changed vs. six months ago?**

A: Six months ago we expected server units to decline; now they will rise meaningfully. I will leave precise figures to industry analysts. **We have raised prices to offset recent cost inflation, but ASP is not the primary revenue driver — unit growth is**. Note that core counts are rising materially in data center CPUs. As cores increase, per-CPU ASP naturally lifts, which is a meaningful contributor.

**Q: How is the competitive landscape in server CPUs? View on AMD share? How to address ARM longer term?**

A: From a roadmap perspective, new silicon typically takes 12–18 months from design to mass production, and we are accelerating execution. Coral Rapids will introduce SMT, improving our competitive stance vs. AMD, and is being fast-tracked. We are also tuning CPU and GPU architectures and have brought in top talent. On ARM, we know it well and respect Renee. ARM continues to raise royalty rates and now offers reference designs with its internal chip team. Amazon and Google using in-house ARM is not new. The good news is OEMs are partnering closely with us, and we have long-standing ties with key customers. Our roadmap from Granite Rapids to Diamond Rapids to Coral Rapids is solid. We also have a dataflow architecture with SambaNova, plus hires like Srini from Cadence (deep ARM optimization expertise) and Kevork, who led ARM data center server CPUs. Beyond products, foundry creates multiple opportunities — advanced packaging and wafers for customers across a broad front.

**Q: What will drive output growth in 2H? Contributions from yields, cycle-time, new tools, and TSMC outsourcing?**

A: First, we are increasing wafer starts across Intel 10/7, Intel 3, and 18A, with larger step-ups at EUV nodes. Second, Lip-Bu has pushed the team to add supply the traditional way — **better yields and faster output** — which drove much of the Q1 upside and should continue through the year. Third, **external foundry is a key part of the plan, and TSMC is a very important partner**. We will flex capacity under a multi-foundry approach.

**Q: Status of advanced packaging? Backlog size, revenue targets, customer count?**

A: We are very pleased with advanced packaging progress. Frankly, I initially thought the opportunity was in the hundreds of millions, but actual demand is in the multi-bn per year range. This will be a major contributor to foundry revenue in the back half of the decade. The good news is advanced packaging is a point of differentiation for us, allows larger reticle sizes, and delivers real customer value, supporting attractive pricing. We expect packaging GPM to be at least in line with foundry or corporate averages.

**Q: How far have 18A yields improved? When will it stop weighing on margins?**

A: **18A yield metrics are sensitive and we will not disclose specifics**. That said, Lip-Bu's year-end target now looks achievable by mid-year, with excellent execution by the team. **By year-end, on a combined product-plus-foundry basis, Panther Lake margins should be in a good place**. Foundry margins will take several more quarters to reach foundry averages, but the trend is better than expected. We have invested heavily in yield — new talent and external partners, including in metrology — and results started to show in Q1.

**Q: ASIC biz. composition and growth strategy? Targeting classic XPU-type ASICs or adjacent IPU areas?**

A: We call this 'purpose-built silicon' tailored to specific workloads. I spend a lot of time with customers to understand workloads and then customize. Intel is uniquely positioned — beyond CPUs and XPUs, we have advanced packaging and leading-edge nodes to optimize end-to-end. We are engaged with multiple customers with very positive feedback. This will be a high-growth area over the next five years. You may be surprised how big it already is — annualized run-rate is above $1bn, and Lip-Bu and Srini are only getting started.

**Q: How to model long-term CPU demand from agentic workloads? Is ARM's $100bn CPU TAM plausible?**

A: **The key metric is CPU/GPU ratio. In training, it is 7–8 GPUs per CPU; in inference, about 3–4:1**. **In agentic, multi-agent scenarios, it could reach 1:1 or even tilt toward CPUs**. As AI evolves, CPUs will be a larger share of AI TAM. And this is not just the data center — clients are migrating to AI PCs, and edge/physical AI may grow even faster given CPU perf/W advantages. To grasp the full opportunity, think full-stack — how CPUs co-optimize with foundation models and leverage data to power agentic AI at scale. The inference market will be far larger than training, and physical AI is another big market. We will keep updating as visibility improves.

**Q: Are tight constraints yielding near- to mid-term share gains? How critical is in-house capacity to win long-term customers?**

A: **The vast majority of current supply is still internal**, but we expect to win external customers over time.

**Q: How binding are supply constraints in Q2 guidance? Will 2H improve? Any prior mix shifts from client to data center?**

A: Supply increases every quarter. **Q1 was the tightest quarter of the year vs. demand**. We did an important exercise in Q1 — we scrubbed **finished goods inventory and identified down-binned or legacy parts once thought unsellable, then worked with customers on use-cases**, which helped a lot. Q2 may not have the same inventory upside, so shipment growth will rely more on real supply increases.

**Q: Server CPU roadmap cadence? From Granite Rapids to Diamond Rapids to Coral Rapids, when does SMT arrive?**

A: Demand is strong and we keep refining the roadmap. After Granite Rapids comes Diamond Rapids, followed by Coral Rapids which brings SMT. We are focused on execution. Meanwhile, we use the ASIC business to deliver custom silicon in the near term, which is a major opportunity given our unique assets. 2026 is the year of execution — yields, productivity, and cycle times are all improving to ensure supply keeps up with demand.

<End\>

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

### Related Stocks

- [INTC.US](https://longbridge.com/en/quote/INTC.US.md)
- [GOOG.US](https://longbridge.com/en/quote/GOOG.US.md)
- [GOOGL.US](https://longbridge.com/en/quote/GOOGL.US.md)
- [TSLA.US](https://longbridge.com/en/quote/TSLA.US.md)
- [NVDA.US](https://longbridge.com/en/quote/NVDA.US.md)
- [04335.HK](https://longbridge.com/en/quote/04335.HK.md)

## Comments (2)

- **LipBu Tan · 2026-04-24T05:58:51.000Z · 👍 1**: Nothing surprising, why did it rise so much
  - **Dolphin Research** (2026-04-24T07:38:51.000Z): Earnings guidance double beats expectations + AI inference logic explodes + foundry turning point confirmed