ASML (Trans): Supply will remain tight at least through 2028; price hikes to start in 2027

Compiled by Dolphin Research: $ASML(ASML.US) Q3 2026 earnings call summary

I. Key financial takeaways

1. Shareholder returns:

a. In Q3, declared the first 2026 interim dividend of EUR 1.88 per share; the second interim dividend of EUR 1.88 per share will be paid on Nov 4.

b. Repurchased approx. EUR 4.1bn of shares in Q3 under the 2026–2028 buyback program.

2. Guidance:

a. Q4: total revenue of EUR 14.2–15.2bn with GPM around 57%. Installed Base Mgmt revenue of ~EUR 3.7bn; R&D of ~EUR 1.2bn and SG&A of ~EUR 0.4bn.

b. Full-year: total revenue of EUR 44–45bn (+~36% YoY) with GPM around 56%. Effective tax rate ~17%.

c. FY mix: EUV up >45%; non-EUV systems up ~20%; Installed Base Mgmt up ~50%. Logic system sales up ~10%, memory up ~75% (capacity adds plus EUV penetration at advanced DRAM nodes); China to remain ~20% of total revenue.

3. Q3 key metrics:

a. Total: record revenue of EUR 11.7bn, at the high end of guidance. System sales EUR 8.4bn, including EUV EUR 4.1bn (incl. 1 High NA) and non-EUV EUR 4.3bn; logic/memory mix 54%/46%.

b. Installed Base and margins: Installed Base Mgmt revenue EUR 3.3bn, beating guidance by ~EUR 0.4bn on robust demand for EUV/DUV capacity upgrades. GPM 57.5%, above guide, driven by contributions from very high-margin parts within Installed Base.

c. Profit and cash: net income EUR 5.0bn, including ~EUR 0.5bn from revaluing the initial Mistral AI investment. EPS EUR 12.95; effective tax rate 16.1%.FCF EUR 11.4bn, supported by strong orders and related prepayments; cash and short-term investments EUR 12.7bn at quarter-end.

II. Call details

2.1 Management highlights

1. Demand and orders:

a. AI build-out remains strong, with shortages at advanced logic and memory nodes widening in scale and duration. Customer demand signals strengthened again this quarter.

b. Leading customers have signed more long-term commitments with their downstreams, improving visibility. They expect product demand to outstrip supply at least through 2028, accelerating expansions and preparing for further capacity beyond 2028.

c. Customer focus has shifted to 2027 and beyond. Orders pointing to 2028 are rising, and clients have indicated interest for 2029–2030 as well.

2. Capacity expansion:

a. Low-NA EUV output in 2026 around 65 systems, up nearly 50% vs 44 systems shipped in 2025. DUV immersion, originally planned below last year, is now expected to be flat vs 2025.

b. As flagged last quarter, evaluating higher Low-NA EUV and immersion capacity for 2027 and studying a further 30% increase for 2028. Current view is that the 2027–2028 output plans will not fully meet demand, and a >30% uplift in 2028 is under review.

c. Upgrades: boosting output on the same footprint with immediate effect. The upgrade product portfolio will keep expanding over the next few years.

d. New products: next-gen Low-NA EUV NXE:3800F to start small-batch introduction in early 2027 at 260 wph (vs 230 wph for NXE:3800E). Next-gen critical-layer immersion DUV will reach 330 wph.

3. Technology progress:

a. High NA: Intel has exposed over 1mn wafers on High NA with performance at or above Low-NA counterparts. Samsung and TSMC plan HVM in 2028 and 2030, respectively; fleet availability is near 90%, with adoption expected to rise node by node.

b. Launched with TSMC an industry initiative for 12-inch large-size EUV masks. Target to have a pilot line by 2031 and HVM by 2033.

c. Metrology: shipped the first YieldStar 550B overlay metrology system for logic A14 and DRAM 1d nodes, with >60% capacity vs the prior generation. YieldStar demand is expected to more than double vs 2025 over the next few years.

2.2 Q&A

Q: Beyond the ~80+ Low-NA EUV capacity indicated for 2027 last quarter, how much more can be added?

A: No update on 2027–2028 capacity figures; incremental opportunities are mainly in 2028. We have expanded materially this year and will keep pushing over the next two years as customers still want more tools. The further out we go, the easier it is to realize incremental capacity, but it is too early to cite numbers now.

Q: How will value-based pricing/price increases phase through 2027–2028?

A: Pricing is agreed; no impact in 2026, starts in 2027, and fully reflected from 2028. Negotiations kicked off in Jul and are complete, covering products most critical to customers. From 2027, pricing will progressively lift revenue, ASP and margins.

Q: How do you see margin trajectory from here?

A: We will guide on 2027 margins in Jan; drivers skew positive, with upgrades the largest swing factor. The ~56% this year is not solely from upgrades (~EUR 0.4bn incremental this quarter), but also from volume and strong contributions from Low-NA EUV and immersion.Positives for 2027: higher volume (capacity +30%, with controlled expansion costs improving fixed-cost absorption); mix (incremental units from higher-margin immersion and Low-NA EUV, with dry DUV also growing); pricing; internal EUV mix improving as D-models largely exit and a small number of F-models appear.Watch items: ramp costs; the durability of upgrade revenue — many installed systems remain un-upgraded, and new upgrade products will open additional serviceable markets.

Q: Is there a hard ceiling to EUV output (e.g., Zeiss capacity)? Could you reach 140–150 units in 2028?

A: No specific unit number; Zeiss is not a bottleneck today, and we are discussing >30% expansion in 2028. Three levers: optimize existing floor space to prioritize shipments.Shorten cycle time, seen as the largest lever — as NXE:3800E matures, we reduce time per step and, with customers, remove some early validation steps; and the supply chain, where we have coordinated broadly in recent years. Each expansion brings new challenges, and we will keep a tight grip; we feel good about 2027 and still have time to mine potential for 2028.

Q: Are customers inflating orders or forecasts to secure slots?

A: We believe the demand signals are real and not amplified by stockpiling. Discussions go down to tool-in dates, which we cross-check against customers' fab build and ramp plans.

Q: Do the price increases cover systems, service and parts? Will existing backlog be repriced?

A: No specifics disclosed; pricing targets high-volume HVM products with the highest value to customers, with upgrade pricing following suit. We did not directly address repricing of existing backlog, only reiterating that the impact starts in 2027 and is fully reflected from 2028.

Q: For >30% expansion in 2028, how will the supply chain keep up? Are large prepayments to suppliers needed?

A: The key challenge is whether suppliers can lift output by 30% per year, and every disclosed figure has been validated with the chain. EUV output is up ~50% this year, demonstrating capability; plans for +30% in 2027 and >30% in 2028 are built on that confidence and cover both EUV and DUV.We continue to onboard new suppliers for non-critical parts, while alignment with key partners like Zeiss is strong. We did not directly comment on large advance payments to suppliers.

Q: With revenue up sharply, why is Q4 GPM guided down QoQ?

A: Mainly ramp costs, which offset fixed-cost absorption from higher volume.

Q: Do price increases all drop through to margin?

A: They lift margins meaningfully, but part of the increase covers extraordinary ramp costs in 2027 at ASML and in the supply chain. Ramp costs were one rationale for pricing with customers; some suppliers need our support to keep pace, but a substantial portion of the increase still accrues directly to margins.

Q: In a shortage, are you tightening contract terms, e.g., higher prepayments and earlier firm orders?

A: Current orders are mutual hard commitments; customers have requested 2029–2030 orders, which we have not accepted yet. Placing orders requires sizable, pain-inducing down payments, and cancellations cost more than the down payment.We are holding off to first aggregate all customer demand signals for that period and match them with capacity, to avoid allocating to A/B/C first and later telling a long-standing customer D that we are sold out. Once we see the full picture, we will accept orders and commit to deliveries.

Q: As High NA enters HVM, could 2028 shipments step up materially, even to double-digits?

A: No unit numbers for 2028; we are still in the early ramp, with shipments growing year by year. The 12-inch mask initiative, led by TSMC with broad customer participation, implies expectations for sustained High NA usage; otherwise no one would invest in 12-inch masks.Uncertainty is whether, amid large-scale expansions, customers view High NA as additional capacity or slow its insertion, which could affect the roll-out pace either way.

Q: Non-current liabilities rose by ~EUR 5.0bn QoQ; can we infer Q3 orders from that?

A: Non-current liabilities include down payments and can serve as a lagging proxy for orders. Down payments typically arrive after POs are signed; we do not disclose order data, and this is a decent reference.

Q: Are the 2029–2030 order discussions mainly from one new entrant, potentially crowding out incumbents?

A: Interest is not limited to a single customer; it is far broader. We continue to see new Low-NA EUV customers, especially in DRAM, and it is not just the one widely cited externally; this is an added variable in the 2028–2030 capacity discussion. A similar trend is visible for High NA.

Q: With immersion DUV at 330 wph, what does that imply for 2028 pricing? Will we see an EUV-like ASP step-up?

A: Value rises with throughput, and EUV pricing logic applies to DUV as well. All immersion variants segmented by overlay requirements will move to 330 wph early next year, a short-term capacity lever. This is an opportunity for value-based pricing and margin, and it enables more capacity with the same tool count, but it does not mean customers will buy fewer tools.

Q: Has incremental demand shifted from memory-only to a more balanced logic and memory mix? What is the driver?

A: Memory share continues to rise, but we still expect logic to exceed memory in 2027. Memory system sales are up ~75% this year, mainly DRAM, and the mix will keep tilting toward memory; logic is up ~10% overall, with advanced logic up >25% expected. Both are tightly coupled in delivering AI compute, making the final split hard to pin down.

Q: Are customers constrained by cleanroom space? What does that mean for 2029?

A: We do not see cleanrooms as the main constraint on demand, aside from tactical cross-quarter moves for a few tools. Customers are building cleanrooms very quickly; in tight supply, we confirm site and facility readiness before allocating tools, keeping shipment delays to days or weeks. That said, the 2027–2028 build-out is massive and requires careful execution, which we will track closely to prevent execution issues from becoming constraints.

Q: With China ~20% this year, can that be sustained next year?

A: No quantified guide; domestic demand in China is strong, and conditions remain robust into 2027 and beyond. There is solid local demand in mature logic and memory, with many fabs under construction entering production.

Q: If tightness persists, can you raise prices every year?

A: No plan to raise prices annually; this is the first increase in many years. Given the ecosystem-wide effort to expand, we believe we can seek a higher step-up within a value-based framework, and customers accept that AI is shifting value allocation across the chain.We insist on value-based pricing to avoid being pressured to cut prices in down-cycles; the principle is to hold pricing aligned to next-gen tool value even if the market softens, so we will not price opportunistically.

Q: If China launches low-end immersion around 2029, would that erode pricing power and margins in China?

A: As long as market access persists, we see low odds of major competition arising in China. We do not disclose share, but our share in China is very high, especially in immersion, and customers prefer to work with us; if some tools cannot be obtained, self-development motivation rises, but the gap remains large. Claims that 'China is about to launch competitors' have circulated for 25 years, and facts today are unchanged vs a year or even five years ago.

Q: For 2029–2030, is demand led by memory or balanced between logic and memory?

A: Trajectories are broadly similar across logic and memory, with DRAM expansions slightly more aggressive. AI chips require both, and the downstream signals are the same for both customer groups.

Q: Is DRAM demand emerging in the US, Korea and also China?

A: We were referring mainly to EUV demand, excluding China.

<End of text>

Risk disclosure and statement:Dolphin Research disclaimer and general disclosures