ASML Holding (ASML)
100% EUV Monopoly. €630B Market Cap. The Picks-and-Shovels Play for AI.
Last updated: 14 March 2026 · YK Research
Contents
Company Snapshot
Market Cap
~€630B
Revenue (2024)
€28.3B
Net Income (2024)
€7.6B
Gross Margin
51%
Litho Market Share
83%
EUV Market Share
100%
EUV System Price
~€200M
High-NA Price
~€350M
Employees
43,395
Founded
1984
HQ
Veldhoven, NL
CEO
C. Fouquet
The Moat: No One Can Copy This
ASML doesn't have a competitive advantage. It has a physical impossibility barrier. No other company on Earth can build EUV lithography machines.
- 100% EUV monopoly: Canon and Nikon quit. ASML is the sole supplier for chips below 7nm. Every leading-edge chip on the planet runs through an ASML machine.
- 30 years of physics: EUV development started in the 1990s. Tin plasma, 13.5nm light, Zeiss mirrors, wafer patterning. A competitor would need 15-20 years and tens of billions to replicate it.
- Irreplaceable supply chain: Zeiss builds the mirrors. Trumpf builds the lasers. 5,000 tier-1 suppliers feed the machine. No one else can assemble this network.
- Installed base flywheel: 140+ EUV systems shipped. Each one needs servicing, upgrades, and consumables. High-margin recurring revenue.
- Customer lock-in: TSMC, Samsung, and Intel invested billions in EUV process flows. Switching cost is infinite.
Revenue Growth
Margin Trajectory
Revenue Breakdown
2025-2030 projections use management guidance and consensus estimates. EUV includes both standard and High-NA from 2025.
AI Compute Supercycle
Source: Dylan Patel (SemiAnalysis) on Dwarkesh Podcast. “ASML will be the #1 constraint for AI compute scaling by 2030.”
“ASML will be the #1 constraint for AI compute scaling by 2030. TSMC is not raising prices, but until TSMC or ASML break out and say 'No, we're going to charge a lot more,' the margin stays downstream.”
Dylan Patel, CEO of SemiAnalysis
The Numbers
Big 4 CapEx 2026
$600B
Total Supply Chain
~$1T
New GW This Year (US)
~20 GW
Anthropic Revenue Run
~$60B/yr
Why ASML Wins
- Every GPU needs lithography. NVIDIA's Blackwell and Rubin run on TSMC N3/N2. Both require EUV. No ASML machines means no new GPUs.
- NVIDIA locked up 70%+ of N3 capacity through 2027. TSMC needs more EUV tools to serve everyone else: Google TPU v7, Amazon Trainium 3, AMD.
- H100s are worth more today than 3 years ago. GPT-5.4 generates more revenue per H100 than GPT-4 did. GPU value goes up with model capability. Chip demand accelerates. So does lithography demand.
- 20 GW of new data center capacity deploying in the US this year. Each GW needs thousands of leading-edge chips. Each chip needs EUV lithography.
- Labs are capacity-constrained. Anthropic needs 5+ GW by year-end. OpenAI is in the same position. They cannot get enough compute. The bottleneck sits upstream at the fab and lithography level.
“An H100 is worth more today than it was three years ago. What is the maximum TAM for GPT-4 tokens? Maybe tens of billions. For GPT-5.4, that number is probably north of a hundred billion.”
Dylan Patel on why GPU demand only grows
High-NA EUV: The Next S-Curve
System (EXE:5000)
~€350M
NA Increase
0.33→0.55
Resolution
8nm features
First Shipments
2023-24 (R&D)
- 75% price increase per system vs current EUV. €200M to €350M. Higher ASP means higher revenue per unit even on flat volume.
- Required for sub-2nm nodes. TSMC N2, Intel 18A/14A, and Samsung 2nm all need High-NA for critical layers.
- Intel received the first R&D unit in December 2023. TSMC got theirs in late 2024. Production ramp targets 2026-2028.
- More EUV layers per chip: Advanced chips need 15-20+ EUV layers vs 5-8 for first-gen EUV. More passes means more revenue per chip.
Risk: High-NA Adoption Speed
- Original EUV took 15 years longer than planned to reach volume production. High-NA could face similar yield challenges.
- At €350M/unit, the cost barrier is real. Smaller fabs may stick with multi-patterning on standard EUV.
- Intel's foundry execution risk: if Intel 18A stumbles, a major High-NA customer gets delayed.
Risk Matrix
| Risk | Severity | Probability | Impact on Thesis | Mitigant |
|---|---|---|---|---|
| China export controls tighten further | HIGH | MEDIUM | China was 30% of revenue, mostly DUV. Each tightening costs revenue. The Dutch government now restricts DUV exports too. | China revenue was mostly DUV and lower margin. Rest-of-world reshoring fills the gap. Export controls actually strengthen the monopoly long-term. |
| AI capex cycle peaks / pulls back | HIGH | LOW (near-term) | ASML's order book gives 1-2 years of visibility. A prolonged downturn still hurts. Revenue dropped 30% in 2009. | $600B Big 4 CapEx in 2026. AI revenue growing faster than expected. Pullback unlikely before 2028. |
| High-NA ramp slower than expected | MEDIUM | MEDIUM | Delays the ASP uplift and margin expansion. Original EUV was years late. | Standard EUV revenue continues growing regardless. High-NA is upside, not base case dependency. |
| Customer concentration (TSMC ~35-40%) | MEDIUM | LOW | If TSMC cuts capex from a geopolitical shock, ASML loses its biggest customer. | TSMC capex follows downstream demand from Apple, NVIDIA, and others. They won't cut while AI demand is this strong. |
| China builds domestic EUV | HIGH | VERY LOW | Would break the monopoly. SMEE making progress on DUV, but EUV is 10-15 years away minimum. | EUV requires Zeiss mirrors, Trumpf lasers, decades of know-how. No viable alternative before 2035. |
| Valuation compression | MEDIUM | MEDIUM | 30-35x forward P/E means a lot is priced in. A small guidance miss causes 20-30% drawdowns. See October 2024. | Quality premium justified by monopoly and growth. Long-term holders win by buying drawdowns. |
| Zeiss mirror bottleneck / supply chain | MEDIUM | LOW | Zeiss is a single point of failure for mirrors. Any disruption means ASML can't ship. | ASML holds equity in Zeiss SMT. Both are incentivized to invest in capacity. |
| Talent constraints (Netherlands) | LOW | MEDIUM | Need to scale from 43K to 50K+. Eindhoven housing, immigration policy, PhD competition. | ASML expanding in the US, Germany, and Taiwan. No longer dependent on Eindhoven alone. |
Valuation and Scenarios
DCF Assumptions
Discount Rate (WACC)
9.5%
Terminal Growth
3.0%
Shares Outstanding
~390M
Current Price
~€730
Scenario Analysis (5-Year Target, 2030)
Bear Case
€550
-25% from current
AI capex plateaus in 2027. China revenue gone. High-NA delays 2+ years. P/E compresses to 22x.
Base Case
€1,100
+51% (~9% CAGR)
Revenue hits €44-50B by 2030. Gross margins expand to 56%. High-NA ramps on schedule. P/E sustains at 28x.
Bull Case
€1,600+
+119% (~17% CAGR)
AI supercycle extends through 2030+. ASML flexes pricing power. Revenue €55B+. GM 58-60%. P/E 32x.
DCF Sensitivity: Fair Value per Share (€)
| WACC ↓ / Terminal growth → | 2% | 2.5% | 3% | 3.5% | 4% |
|---|---|---|---|---|---|
| 8% | €920 | €1,010 | €1,120 | €1,260 | €1,450 |
| 8.5% | €850 | €930 | €1,030 | €1,150 | €1,310 |
| 9% | €790 | €860 | €950 | €1,050 | €1,180 |
| 9.5% | €735 | €800 | €880 | €970 | €1,080 |
| 10% | €690 | €745 | €815 | €895 | €990 |
| 10.5% | €645 | €695 | €760 | €830 | €910 |
| 11% | €605 | €650 | €710 | €770 | €845 |
Outlined cell = base case (9.5% WACC, 3% terminal growth → €880/sh). Red cells sit below the current price of ~€730 — the market is pricing roughly an 11% WACC or sub-2% terminal growth.
Investment Framework
Answer These Before Sizing
- Time horizon: Over 3-5 years, ASML compounds. Over 1 year, cyclical risk is real. Best held through drawdowns.
- AI capex view: If AI spend sustains for 5+ years, ASML is the picks-and-shovels play. If you see a pullback coming, wait for entry.
- China risk tolerance: Each export control escalation is a short-term headwind. It strengthens the long-term monopoly.
- Entry price discipline: October 2024 dip to €600 was a 25x P/E entry. Current levels are richer. Have a buy-zone range.
- Position sizing: One company, one country, one technology. Cap at 5-10% of portfolio even if very bullish.
The Bull Case
- Deepest moat in tech. Literal monopoly on the machines that make all advanced chips.
- The critical bottleneck for AI.
- High-NA extends the growth runway through 2030.
- Dylan Patel / SemiAnalysis: “ASML will be the #1 constraint for AI compute scaling by 2030.”
- Pricing power largely untapped. TSMC and ASML haven't flexed pricing yet.
The Bear Case
- The moat probably doesn't break. But the cycle could turn and you overpay. China creates headline volatility. High-NA could disappoint.
- 30-35x P/E means a lot of the AI supercycle is priced in.
- Semis are cyclical. ASML dropped 30% in 2009 and 40%+ in October 2024.
“The structural position is almost unassailable for long-term holders. The real question is entry price and patience through cyclical drawdowns.”
Strategy
- ACCUMULATE Build position in tranches. Start small, add on drawdowns.
- BUY ZONE €580-650 (P/E 22-25x). Aggressive accumulation zone.
- FAIR VALUE €700-800 (P/E 27-30x). Hold. Small adds only.
- EXPENSIVE €900+ (P/E 34x+). Take partial profits. No new buys.
- HOLD PERIOD 5+ years minimum. Capture the High-NA ramp and AI capex supercycle.