AMAT (Trans): Q4 guidance up 51%; GPM held in check

Trans of AMAT FY26 Q3 earnings call compiled by Dolphin Research

I. Key Financial Recap

1) Shareholder returns: $860mn returned this quarter, including $420mn in dividends and $440mn in buybacks. Remaining buyback authorization stands at $12.8bn, and the company maintains its policy to return 80%–100% of FCF to shareholders.

2) Q4 (FY26) guide

Total revenue: $10.25bn (+/– $0.5bn), +51% YoY.

non-GAAP EPS: $4.02 (+/– $0.20), +85% YoY.

By segment: Semiconductor Systems approx. $7.9bn (+62% YoY); AGS approx. $1.84bn (+22% YoY); Other approx. $0.51bn (mainly Display).

Margins and opex: non-GAAP GPM ~50.4% (+230bps YoY); non-GAAP opex ~ $1.58bn.

3) Q3 (FY26) key metrics

Total: revenue of $9.1bn (record), +15% QoQ/+25% YoY. non-GAAP GPM 50.4% (+40bps QoQ/+150bps YoY), the 13th straight quarter of YoY expansion; non-GAAP OPM 34% (record), +190bps QoQ/+330bps YoY. non-GAAP EPS $3.50, +22% QoQ/+41% YoY.

Semiconductor Systems: revenue $7.0bn (record), +18% QoQ/+27% YoY. Segment non-GAAP GPM 55.4% (+190bps YoY); non-GAAP OP $2.7bn (record), +45% YoY.

AGS and Other: AGS revenue $1.8bn (record), +22% YoY, with GPM 35.6% (+180bps YoY) and OPM 30.1% (+280bps YoY). Other revenue $294mn, in line with expectations.

Cash flow: operating cash flow topped $3.0bn (record). Capex $707mn; FCF $2.3bn.

4) Operating leverage and regional mix: revenue growth far outpaced opex, taking opex as a % of revenue to a near four-year low, with G&A as a % of opex at a company record low. China accounted for 26% of Semiconductor Systems + AGS revenue; for the calendar year, China revenue is now expected to grow YoY, driven by 28nm foundry logic investments.

5) Modeling updates

Other revenue (mainly Display): model at an Avg. ~$400mn per quarter through 2027.

14-week quarter: FY27 Q1 is a 14-week quarter, so Q1 opex will rise more than usual QoQ.

Tax rate: non-GAAP tax rate ~11% for this year, rising to ~13% in 2027 as the global minimum tax phases in.

Definition: references to 'calendar 2026' on this call denote FY26 Q2 to FY27 Q1.

II. Call Details

2.1 Management highlights

1) Demand and visibility All tracked leading indicators strengthened across the board in the past quarter, with hyperscalers continuing to step up AI infrastructure investments and many already realizing positive returns. Customers announced 10+ new fab projects this quarter; most advanced logic and DRAM fabs are running full, utilization is broadly moving up, and ICAPS is lifted by AI-related markets such as power and optical comms chips.

Our largest customers are now providing longer-term commitments with rolling eight-quarter forecasts, and some discussions extend to 2030; on this basis, we see 2027 as another strong growth year. Over the last three months, customers found new ways to ease cleanroom space constraints, significantly raising 2026 demand pull-ins; we have further raised our in-year revenue outlook and still expect to outgrow the market for the full year.

2) AI’s industry reshape and two races First, the technology race: AI data center returns are driven by tokens/sec and total cost of ownership dominated by power, and the improvement in tokens/sec/W largely depends on innovation in semiconductor devices and systems. Second, the capacity race: demand for advanced chips far exceeds supply, and while fabs are building new capacity quickly, they are highly focused on boosting output and yield at existing sites. Advanced logic foundry, DRAM, and advanced packaging together are expected to account for ~80% of WFE incremental spend in 2026–2027, which are also AMAT’s strongest leadership areas.

3) Advanced packaging We are the overall leader in this market and hold strong positions in HBM and 3D chiplet stacking; we now expect total packaging revenue to grow over 70% in calendar 2026. For future architecture transitions and larger panel formats, we have built a next-gen portfolio spanning digital lithography, deposition, etch, and eBeam inspection.

4) New product cadence We launched six new systems this past quarter: the Centura Prime epitaxy platform for high-performance DRAM, and the Producer Avila to support higher-performance, higher-layer HBM. We also introduced the next-gen plating system Nokota VMax, the Opta Quad CMP for advanced packaging, and two eBeam systems for advanced packaging that extend our front-end eBeam advantage into the package flow.

5) AGS services Over 37,000 chambers are now connected to our in-house AIx software, delivering AI-driven monitoring, diagnostics, and predictive analytics that directly lift yields. We expect AGS to grow over 20% in calendar 2026, with sustainable long-term growth in the mid-teens. This quarter we added over 1,000 customer support engineers to AGS and are using AI and warehouse automation to contain the cost of growth.

6) Process Diagnostics & Control (PDC) The most advanced logic and DRAM devices require more eBeam steps with sub-nm resolution for high aspect-ratio structures, and we are the sole-technology leader with a share lead in this market. We are also ramping new optical inspection products to expand application share, and we expect PDC to grow over 50% in calendar 2026 with a pipeline to support 2027 and beyond.

7) Output innovation We are developing a new class of products aimed at increasing wafers processed per square foot of cleanroom space. For example, our new DRAM epitaxy system improves device performance while using ~20% less cleanroom space than the prior generation; several other output-innovation products are in customer qualification.

8) EPIC strategy progress Since last quarter, Broadcom joined EPIC as an innovation partner to co-accelerate advanced packaging for next-gen AI systems; we also signed EPIC agreements with SCREEN and UC Berkeley. Announced EPIC collaborations now total 11, spanning systems companies, leading chipmakers, top research universities, and innovation partners. Our Silicon Valley EPIC Center will move in its first R&D tool next week, operations will ramp in the coming months, the unveiling is on Oct 12, and an investor breakfast will be held on Oct 13 at Yerba Buena Center in San Francisco.

9) Capacity expansion and supply chain Our new Singapore manufacturing center came online this quarter, and together with global expansions, manufacturing floor space has nearly doubled over the past few years. With customers’ long-cycle demand signals, we are hiring and training manufacturing and customer support teams, targeting by 2028 the ability to double quarterly system output vs. today; this quarter, global manufacturing and AGS customer support headcount rose by over 1,500 net. We are also planning the next wave of manufacturing capacity expansion to preserve options to support higher demand into 2030.

10) Value creation and pricing We institutionalized value-based pricing three years ago, reassessing and repricing every tool; company-level GPM has topped 50%, Semiconductor Systems over 55%, with higher prices and margins on both new and existing products. R&D has increased annually since 2012, broadening from equipment innovation to materials engineering solutions, and in recent years we significantly stepped up investment in advanced packaging (supplemented by two small acquisitions), with current focus shifting to technologies that improve fab economics. We continue to offset higher input costs with cost improvements while raising prices.

2.2 Q&A

Q: Last quarter you guided Semiconductor Systems growth at 30%+, what is the framework now, and how do you see calendar 2027?

A: Demand strengthened again this quarter, with customers adding factory projects and lifting capex outlooks, and hyperscaler capex prints remaining strong; what we flagged as '30%+' is now higher, but we are not guiding FQ1, so we will leave it at that. For 2027, the AI-driven demand function persists, and we see another strong year; we also expect share gains this year. Structurally, the fastest-growing areas are advanced logic foundry, DRAM, and advanced packaging, which we said comprise ~80% of WFE spend growth this year and should look similar in 2027, and these are precisely where our leadership is strongest; we first said 20%+ YoY in Feb, then 30%+ in May, and now see even stronger growth, with eight-quarter rolling forecasts indicating robust multi-year demand and customers flexibly expanding space and pulling in tools, while our teams execute to outperform the market.

Q: How exactly do you implement value-based pricing, and what does higher like-for-like pricing mean for FY27 and beyond GPM?

A: About 300bps of company-level GPM expansion over the past three years was driven by value-based pricing, instituted after the pandemic and supply-chain crisis as input costs rose; we reassess and reprice each tool and have embedded this process, which we see as necessary in a dynamic cost environment. Looking forward, we expect further GPM improvement, with Semiconductor Systems already above 55%, and value pricing will remain a contributor. On products, our ability to create value for customers is at a historic high: the race to bring new AI architectures to market is every customer’s focus, and we hold the most enabling technologies in the fastest-growing markets, making our products and services more valuable, which supports >20% growth in services and >50% in PDC, and our new-product pipeline carries higher margins, providing a tailwind.

Q: Many memory customers now have 3–5 year LTAs with volume/price terms; beyond your eight-quarter visibility, what is your longer-term visibility?

A: We essentially see the roadmaps for major customers, with roughly a five-year view on the largest accounts, though for planning we request eight-quarter granularity to cascade into the supply chain; POs now have longer lead times with details locked earlier, and mechanisms like cancellation and expediting fees have emerged to coordinate in this environment. The supply-demand gap is evident to all, and as AI extends from training to inference to agentic and physical AI, memory demand rises, especially DRAM; this year is very strong for DRAM with a skew to 2H, and into 2027 DRAM remains strong, with our DRAM share having expanded significantly in recent years. We are the largest process equipment supplier in DRAM, lead in CMOS periphery upgrades that improve performance/power, see strong Epi growth, lead in HBM packaging across materials deposition for wiring and patterning, conductor etch, and eBeam, and have deep collaborations on future DRAM architectures like 6F² and 4F² with strong positioning for 3D DRAM.

Q: From calendar 2021–2025 your GPM was in line to slightly ahead of U.S. peers, but you now trail by ~150bps; why, and can you catch up?

A: Cross-company comparisons are hard to explain, but our own GPM progress is substantial; beyond value pricing, the real foundation is mix, as focusing R&D and co-development with customers on the highest-value inflections strengthens the portfolio, and pricing reflects that value, a path we expect to continue. We also have other pieces such as Display, where faster growth can impact consolidated GPM differently than peers given different business mixes, but overall we still expect to improve GPM and grow value. We have delivered notable margin gains—13 straight quarters of YoY expansion and +190bps in Semiconductor over the past year—and we are confident in continuing to push margins higher.

Q: With such a large revenue step-up, why guide GPM roughly flat—mix from Display, or something else?

A: Display growth is one factor, but the real headwind is ramp costs—adding a large number of customer service engineers and resources within Semi as we scale; so while we benefit from favorable mix and scale, the ramping costs are embedded in the outlook. At the company level, holding GPM flat in the guided quarter is acceptable to us, and over a longer horizon we still expect to drive margins higher.

Q: How long will ramp costs persist, will headcount still grow in coming quarters, and does the ~10bps per quarter improvement cadence still hold?

A: We previously characterized the cadence as 'gradual', and that remains the right framing; over the long cycle we do expect improvement. We will continue to add employees over the next few quarters, but as revenue grows, this headwind should gradually fade.

Q: Even at 40%, Dec-quarter growth would decelerate (Jul +18% QoQ, Oct +12%, Jan +6%), while a peer cites ~38% WFE growth; will you commit to 40%+ Systems growth?

A: Our commitment is to outperform the market, which we believe we have done year-to-date and expect to continue regardless of where the final growth number lands; the directional dynamics you cite are right—initially 20%+, then 30%+, now higher—but we will not provide a number because we are not guiding the next-next quarter. We do expect FQ1 (calendar Q4) to grow QoQ, but we are not providing guidance at this time.

Q: Should we take 'doubling manufacturing capacity' literally—if Jul-quarter recognized revenue is ~$7.0bn, could you do $14.0bn at some point in calendar 2028?

A: Interpret 'capacity' literally, but it is not a revenue forecast; long-cycle investments like cleanrooms must be built ahead when economically justified to meet a range of potential demand outcomes. We are also signaling to suppliers that we are laying in capacity to support a wide range of output needs by 2028 and beyond; again, this is not revenue guidance, but indicates the scale we are preparing for.

Q: As customer dialogs shift from annual price talks to time-to-market deliveries with two-year visibility, does that increase pricing freedom or add costs to build capacity and the supply chain; and with two-year visibility, why not give at least a qualitative Jan-quarter view?

A: We have very strong visibility with large customers, and you are right that on-time delivery against plan is a growing focus, which is why they are collaborating on order specificity and providing eight-quarter visibility that we pass on to suppliers; this mechanism vastly improves our planning environment versus last year.

Q: Conversations extending to 2030—are they about technology or capacity expansions to meet demand?

A: Primarily technology; with large or mature customers we know which fab projects are on the roadmap and which technologies are planned, and even for new nodes we know our positioning well enough to plan five years in detail; by the eight-quarter window, discussions get down to nodes and tool SKUs to cascade to suppliers. I have joined many CEO-level meetings at our largest customers, who provide detailed eight-quarter visibility and, beyond that, signal strong multi-year demand and want us to prepare supply accordingly—which takes time—so beyond eight quarters our visibility is mainly on capacity needs. Technology discussions can extend 10 years because AMAT most strongly enables key architectural inflections—new transistors, wiring, DRAM architectures, and new packaging—and if you list the top 5–10 technologies needed to enable them, the majority sits with AMAT; taking these innovations to market takes time, so in deep co-creation relationships, technology visibility extends beyond five years.

Q: You guided global ICAPS to flattish to slightly up this year, but ex-China auto/industrial/analog power/MCU customers are seeing a strong cyclical recovery with tighter supply and higher utilizations; are those customers stepping up spend, and will ICAPS grow for the year?

A: ICAPS dynamics are indeed changing, and rising utilizations at those customers are positive signals; in China we expect growth this year and next, and China is a large part of ICAPS. Overall, we think ICAPS will grow this year and next as the digestion phase on the equipment side ends and growth resumes; ex-China, we expect positive growth next year with bright spots in power and photonics, much more constructive than the past two years.

Q: AGS OPM is 30%, a two-year high, with recent incremental margins clearly above 40%; could AGS approach ~40% GPM and ~35% OPM as revenue scales and advanced services mix rises?

A: Like Semi, we do expect long-term margin improvement in services, enabled by digitization from AI that lets us develop new offerings and run services more efficiently on a growing installed base; we also benefited this year from sharply higher utilizations that lifted spares growth, which helps margins—but utilizations only rise to 100% once, so that tailwind will moderate. Structurally, services margins still have room to improve. From the customer perspective, with supply constrained, optimizing output and yield is critical and will remain so for some time, greatly increasing the value of yield-optimization services; with 37k+ chambers linked to AIx and AI apps for preventive maintenance and chamber matching, we see faster revenue and contract growth and better value capture, making us more optimistic than ever on AGS revenue and profit.

Q: NAND doubled QoQ this quarter off a small base—what are you seeing, and does Oct keep growing?

A: In percentage terms NAND is growing well this year off a small base, and we see a strong year for NAND; next year the overarching AI dynamic remains—advanced logic, DRAM, and advanced packaging are the fastest growers, ICAPS should return to growth, and while NAND should grow, it will likely be the slower-growing piece next year.

Q: On spending for capacity aimed at 2030, what is your opex outlook for Oct and next year; previously EPIC spend was expected to roll off and opex to decline, but with business so strong, how are you thinking about it?

A: With the business this strong, we have more attractive investments to make, which fall under capex—we will place tools within EPIC and make other investments; so we still expect a capex year above normal, but as we enter 2027, capex as a % of revenue should decline.

Q: For FY27 or calendar 2027, can you rank the incremental strength across foundry logic, DRAM, and advanced packaging?

A: We do not differentiate—AI’s system-level pull is similar across these end markets, and differences are not sufficient to rank; advanced logic will be strong, DRAM will be strong, and advanced packaging will be strong, with a new positive being that we now expect ICAPS to grow next year as well. As we said, these three account for ~80% of WFE growth in 2026 and should look similar in 2027—possibly even better—and frankly they will remain the fastest growers for years.

Q: Given the 1H/2H split in calendar 2026, is there reason not to expect faster revenue growth in 2027, and what could constrain it?

A: Many ask what constrains growth, and over time the constraint from our perspective is cleanroom availability; customers keep adding cleanroom projects, which is why we raised our outlook this year, and some of that space becomes additive next year, with even more as we look further out since these projects take years. At a high level, cleanroom availability determines how much everyone can ship next year.

Q: Leaving aside near-term trends, in a $150–$175bn WFE scenario, what are your target revenue, GPM, and OPM?

A: We will provide more specifics at our Oct investor event, but today we see a multi-year strong growth driver from compute demand and leadership in the fastest-growing, highest-value parts of the market—advanced logic foundry, DRAM, and advanced packaging—where we can continue to gain share; thus, we have multiple drivers to outperform in 2026 and sustain strong growth thereafter. At the same time, we are delivering more value with the most unique and connected portfolio, enabling customers’ new chips and packaging architectures and supporting further margin expansion; coupled with yield and output innovation—every customer is focused on more chips per square meter—this raises services value, and we are targeting innovations directly at those concerns, leaving us optimistic on both revenue and margins, with more color in Oct. As you describe this year’s scenarios, we have provided the building blocks—services growth >20% in the calendar year, Semi growth above '30%+', GPM outlook, Display framing, and share gains regardless of the final WFE number—enough to inform a view.

Q: NAND has been the smallest piece of Semi, driven more by layer upgrades than wafer adds—when will the industry add true wafer capacity to offset capacity lost to 300+ layer transitions?

A: Our view has been NAND wafer starts continue to decline, so most projects serve layer upgrades, and we expect this dynamic to persist for the next few years; the only potential exception is new projects in China, while for large customers the focus remains adding space to accommodate layer upgrades.

Q: When will panel-level packaging become mainstream, how big is the opportunity vs. today’s advanced packaging, and hybrid bonding seems finally near—what are you seeing?

A: Packaging is one of the most important domains for improving AI compute performance and power, with multi-chip interconnect and data movement being a major focus for both incumbents and new-architecture innovators, supporting >70% growth this year; our technology portfolio is far stronger than peers, with recent acquisitions further enhancing our packaging capabilities, and we are highly confident in sustained, significant multi-year growth. On new substrates, everyone wants to connect as many logic and memory chips as possible at the best performance and power, and we are in deep co-creation with these companies; each is racing to bring new architectures to market given the performance/power value, and our share modeling shows strong share-gain opportunities, with new capabilities expanding our SAM as new architectures are adopted, and while we will not give specific timing, panel revenue should grow meaningfully next year with a robust ramp thereafter and a stronger position for us. For hybrid bonding, shortening interconnect lengths to improve performance and power is critical across advanced logic foundry, DRAM, and HBM, making this a key inflection for all, and AMAT has strong technology in hybrid bonding and large adjacent positions in surrounding steps, plus the only integrated R&D facility co-validating these new packaging architectures with customers; hybrid bonding will be a meaningful long-term growth driver alongside our other advanced packaging technologies, with >70% growth this year and strength into 2027 and beyond.

Q: Without FQ1 guidance, how does the 14-week quarter affect QoQ revenue?

A: Historically, in 14-week quarters, services benefit roughly proportionally, while equipment does not see a clear linear benefit because production planning is done at the quarterly cadence; on the expense side, as you would expect, most costs are recognized during the quarter, including payroll.

Risk disclosure and disclaimer:Dolphin Research Disclaimer and General Disclosure