Verdict: Keep

ASML is 0.67% of the selectable universe (rank 27 of 50). The stock is down 12.9% over the past month in a sector-wide tariff selloff that has hit every semicap name in QQQ. The drawdown is not ASML-specific. The structural thesis — sole supplier of EUV lithography with no competitor on earth — is intact. WFE spending is not peaking. At 0.67% weight, removing ASML captures approximately 3 basis points of filtration alpha per 5% of underperformance. The expected value of the remove decision is approximately zero, and the research time is better spent on higher-weight names.

Factor Profile

Trailing 250-day regression against SPY, XLK, MTUM:

Trailing alpha: +45.0% annualized. Idiosyncratic volatility: 28.0%. R-squared: 54.6%.

Idio variance at 45% is well below the 75% target. ASML is a tech sector bet first, stock bet second — 54% of variance is explained by XLK alone. For basket filtration, the QQQ short neutralizes broad tech beta, so only the idiosyncratic component drives filtration P&L. But the idio signal is noisy because it's only 45% of the total.

The orthogonal Sharpe is 1.61, which exceeds the 1.5 threshold. This flags a latent factor problem: the 3-factor model doesn't capture "AI infrastructure capex" as a thematic factor. ASML, LRCX, AMAT, and KLAC all load on this latent factor, which has been in bull regime. What appears to be 45% idiosyncratic alpha is partly a missing thematic factor. If that factor reverses — via tariff-driven capex slowdown or AI demand plateau — the "alpha" evaporates and takes ASML with it.

The positive MTUM loading (+0.22, 8.2% variance) reflects ASML's +94% 1-year return. Momentum exposure accumulates naturally when a stock performs well. This is modest but worth noting: a momentum reversal creates additional headwind beyond the fundamental picture.

The Structural Thesis

ASML is a monopoly in the literal economic sense. There is one company on earth that manufactures EUV lithography systems, and it is ASML. The competitive landscape:

Low-NA EUV (0.33 NA, 13nm pitch): Sole supplier. Current-generation workhorse for 5nm and below. Productivity increased 40% to 220 wafers per hour on the NXT:3800 platform. Shipped more tools in 2025 than any prior year, with 2026 guided as "another big year."

High-NA EUV (0.55 NA, 8nm pitch): First high-volume manufacturing tool accepted by a customer in Q4 2025. This is the next-generation platform enabling sub-3nm nodes. There is no competing tool from any company, anywhere.

Hyper-NA: Pre-investment underway. Extends the monopoly another decade. "We could do it more, but there's no need for it tomorrow."

China's best effort: A Huawei/SiCarrier consortium has a functional EUV prototype using LDP technology, validated late 2025. Target: functional chips by 2028, realistic 2030. SMEE's best production tool is 28nm DUV. China's prototype is comparable to ASML's technology from approximately 2006. The gap is widening, not narrowing.

This monopoly is not disputed by anyone. There are no analyst debates about whether ASML might face competition. The question is never "will ASML sell EUV tools?" — it is "how many and at what price?"

WFE Cycle Is Not Peaking

The semiconductor equipment spending cycle is the key macro risk for ASML. If WFE rolls over, ASML's revenue rolls over regardless of the monopoly. The current data:

The cycle is bifurcated. TSMC and Samsung are ramping aggressively on AI-driven logic and memory demand. Intel is cutting hard. ASML's exposure to Intel is small relative to TSMC and the memory customers. Lam Research projects WFE at $135B for 2026 (+23% YoY), second-half weighted. No public company has disclosed customer pushbacks or order cancellations.

The CEO's framing on the Q4 2025 call: DRAM is the current AI bottleneck, not logic. Memory customers "moving very aggressively" to build capacity. DRAM prices "going up significantly." For 2026, memory demand "will be very, very tight." NVIDIA's wafer intensity is going from 2.5 to 10 wafers per product by 2027 — a 4x increase in two years. This is what underpins ASML's "EUV up significantly" guidance for 2026.

Financials

FY2025 Results (from 20-F, filed February 25, 2026)

Balance Sheet

EUR 12.9B in cash, EUR 19.6B shareholder equity, EUR 50.6B total assets. Goodwill stable at EUR 4.6B. No going concern language. This is a fortress balance sheet generating EUR 12.7B in operating cash flow annually.

Capital Return

EUR 5.95B in share buybacks in 2025 (up from EUR 0.5B in 2024 — 12x increase). New EUR 12B three-year buyback program announced. Dividend EUR 7.50/share (+17% YoY). Total shareholder return of EUR 8.5B in 2025 — two-thirds of net income returned to shareholders.

FY2026 Guidance

Revenue EUR 34-39B (midpoint +12% YoY). Gross margin 51-53%. Tax rate ≈17%. Q1 2026 guided: EUR 8.2-8.9B revenue, 51-53% GM, installed base EUR 2.4B. EUV revenue "up significantly." Non-EUV "similar to 2025." Installed base growing on EUV upgrade appetite.

2030 targets unchanged: EUR 44-60B revenue, 56-60% gross margin.

What's Driving the -12.9% Drawdown

Three converging headwinds, none ASML-specific:

Tariff uncertainty (primary). The Section 232 Proclamation (January 14, 2026) imposed 25% tariffs on advanced AI chips. Semiconductor manufacturing equipment is NOT tariffed in Phase 1 but is in the investigation scope. Commerce provides a July 1 review that could trigger Phase 2. The US-EU trade framework (August 2025) provides partial relief: caps EU semiconductor tariffs at 15% and exempts manufacturing equipment. ASML's CEO flagged on the Q2 2025 call that customers were "keeping cards closer to the chest" on capex due to tariffs.

Employee walkout (noise). 1,000+ employees walked out March 24-25 at Veldhoven protesting 1,700 planned job cuts. This is an organizational restructuring (cutting 3,000 management positions, creating 1,400 engineering roles) designed over 12+ months. Engineers reported 20-30% overhead time. Union response due April 7, works council in May. The restructuring is an agility play during hypergrowth, not cost-cutting on EUR 9.6B net income. But "4,500 workers protest at Europe's most valuable tech company" makes headlines.

Google TurboQuant sympathy (transient). Google announced TurboQuant on March 25 — a KV cache compression algorithm reducing memory usage 6x. Memory stocks sold off (Micron -7%, SK Hynix -6.2%, Samsung -4.7%). ASML dropped in sympathy as a memory-exposed semicap name. Morgan Stanley's counter: TurboQuant doesn't affect model weights or training workloads; it enables 4-8x longer context windows on same hardware, which could increase total AI adoption and therefore wafer demand. Classic efficiency-as-Jevons-Paradox argument. Too early to adjudicate, but the market's same-day reaction was likely oversized.

Semicap Peer Comparison

ASML is the most resilient of the four — lowest beta, lowest short interest, smallest drawdown after KLAC, and smallest decline today. The monopoly provides structural downside protection that competitive semicap names don't have.

ASML also has the smallest QQQ weight (0.67% vs 1.10-1.62%). If the tariff selloff continues and you need to remove a semicap name, LRCX (highest beta at 1.79, most extended at +185% 1-year, highest short interest at 2.4%, and 2.4x the QQQ weight) or AMAT (beta 1.63, -14.3% 1-month, 2.4x the weight) have worse risk profiles and higher filtration impact. Removing ASML before LRCX or AMAT is removing the strongest hand first.

Filtration Economics

At 0.67% weight, the decision math is straightforward:

• If ASML underperforms QQQ by 5% over 15 weeks: removing it captures +3.4 bps
• If ASML outperforms QQQ by 5%: removing it costs -3.4 bps

With P(underperform) approximately 50% (see Bayesian assessment below), the expected value of the remove decision is approximately zero. Compare to MU at 2.52% weight, where a correct remove captures 12.6 bps per 5% underperformance — 3.7x the ASML signal.

Research time has opportunity cost. Analyst bandwidth spent on a 3 bps decision is bandwidth not spent on 10-15 bps decisions. ASML's weight makes it a low-leverage filtration target.

Bayesian Assessment

Prior: P(ASML underperforms QQQ, 15 weeks) = 55%. Semicap names face tariff headwinds, elevated beta amplifies drawdowns, and the sector's 1-year returns (+94-185%) invite reversion. Starting prior is slightly above coin-flip.

Bull product: 1.3 x 1.5 x 1.2 = 2.34 Bear product: 1.3 x 1.2 x 1.1 x 1.1 = 1.89 Net LR: 2.34 / 1.89 = 1.24 (mildly bullish)

Prior odds (underperform): 55/45 = 1.22 Posterior odds: 1.22 / 1.24 = 0.98 P(underperform) = 49.5%

Bull and bear evidence roughly cancel. No directional edge. This is consistent with the "no informational advantage" assessment: $522B market cap, 37 analysts, 84% bullish consensus. Every signal we found — structural and cyclical bull, tariff and valuation bear — is consensus knowledge.

Edge Assessment

There is no informational edge on ASML. At $522B market cap with 37 analysts and 84% bullish consensus, every data point in this memo is priced. The structural monopoly thesis is the least controversial claim in semiconductor equipment. The tariff risk is known and hedged by the US-EU framework. The WFE cycle data is public.

For filtration, the asymmetry matters: we need affirmative evidence of weakness to remove, not evidence of strength to keep. Removing ASML means betting the strongest monopoly in semiconductors will underperform QQQ over 15 weeks. The counterparty on that bet is every semicap analyst, TSMC and Samsung capex guidance, and the physical reality that no alternative to EUV lithography exists.

Export Control Status (from 20-F, February 25, 2026)

• Netherlands expanded controls (January 2025): additional metrology/inspection systems require export licenses
• US Affiliates Rule: suspended until November 10, 2026 (trade deal with China)
• China rare earth controls: suspended until November 10, 2026
• EU controls (November 2025): formalized Dutch national DUV controls into EU-wide list

Net assessment: stable-to-easing through our 15-week window. Both major risk items (Affiliates Rule and rare earth controls) are suspended until November, well past our exit date. The July 1 Commerce review is the only regulatory catalyst within the window.

Customer concentration: China 29.1% of 2025 sales, Taiwan 25.5%, South Korea 25.0%. This is geopolitically concentrated but reflects the entire advanced semiconductor manufacturing ecosystem. The concentration is an industry fact, not an ASML-specific risk.

Risk Calendar

What Would Change Verdict to Remove

1. Q1 orders significantly below Q4 2025 (demand inflection, not tariff pause)
2. TSMC cutting 2026 capex guidance below $52B
3. Phase 2 tariffs explicitly including semiconductor equipment at rates exceeding the 15% US-EU framework cap
4. ASML cutting FY2026 revenue guidance or widening the EUR 34-39B range downward
5. Compounding AI efficiency breakthroughs beyond TurboQuant that structurally reduce memory wafer demand

None are in the current data. Monitor at Q1 earnings (April 15) and TSMC Q1 results (mid-April).

Sources: ASML 20-F (filed February 25, 2026), ASML 6-K (March 11, 2026 — AGM logistics, non-material), ASML Q4 2025 earnings call (January 28, 2026), ASML Q2 2025 earnings call (July 16, 2025), LRCX/KLAC/AMAT earnings calls, EY Tax News Section 232 analysis (January 15, 2026), CSIS Chinese lithography assessment, TrendForce China EUV and TurboQuant analysis, CNBC, yfinance market data (March 27, 2026).