brain/
conceptstock-market

Picks-and-shovels: leading-edge fab buildout

Notes

Picks-and-shovels: leading-edge fab buildout

One-line summary: Whichever foundry wins (Intel, TSMC, Samsung), every leading-edge wafer needs the same toolset — ASML is now #1 WFE by revenue (took the spot from Applied Materials in early 2026), with AMAT, LRCX, KLA exposed to the same capex flywheel regardless of foundry winner.

The insight

The us-fab-capacity-bottleneck thesis depends on Intel converting anchor partnerships into shipped silicon — that's medium-conviction with execution risk. The picks-and-shovels trade is structurally cleaner: any meaningful expansion of leading-edge capacity (Intel ramping 14A, TSMC pouring $56B into capex, Samsung's recovery if it materializes, Terafab's prototype) must flow through ASML/AMAT/LRCX/KLA. There is no scenario where leading-edge capacity expands and these four don't benefit.

The chain

TSMC saturation + Intel build-out + Samsung Taylor = capex flows to all three fabs. Semicap leaders (ASML #1 WFE post High-NA cycle, AMAT/LRCX/KLA) capture demand regardless of which foundry wins anchor orders.

Canonical: tsmc-saturation-to-intel-anchor-stack.

Evidence

TSMC Q2 FY2026 — the buildout gets bigger and the tools get more expensive (2026-07-17 ingest)

  • Capex raised to $60–64B from $52–56B (January) / ~$56B (April) — wendell-huang in 2026-07-16-earnings-tsm-q2-fy2026. Allocation: 70–80% advanced process, ~10% specialty, 10–20% advanced packaging, testing, mask making and others. Plus an additional $100B for Arizona (2nm-and-below fabs and advanced packaging fabs) and 13 leading-edge/advanced-packaging fabs in Taiwan — cc-wei in 2026-07-16-earnings-tsm-q2-fy2026. This is the single largest upward revision to the buildout the thesis has recorded.
  • ★ The buyer confirms vendor pricing power — cc-wei in 2026-07-16-earnings-tsm-q2-fy2026, naming the two drivers of the capex raise: "The second reason is inflation. Now we buy the tools with inflation price." The most powerful customer in the industry publicly conceding it pays inflated tool prices is the strongest statement of semicap pricing power this thesis has. See semicap-tool-price-inflation — and note the double edge: it also means TSMC's capex dollars buy less capacity than the number implies.
  • No tool-supply bottleneck — wendell-huang in 2026-07-16-earnings-tsm-q2-fy2026: "We always collaborate closely with the tool suppliers well in advance to prepare the capacity, whether it is a strong upcycle or downcycle... We do not foresee any bottlenecks to our capacity expansion plans." Vendors can fill the orders; the constraint is TSMC's own ramp, not tool availability.
  • Capex is deliberately fungible between front end and back end — cc-wei in 2026-07-16-earnings-tsm-q2-fy2026: "Sometimes we have a bottleneck, we put more money to buy the bottleneck tools, sometimes it's in the front end, sometimes it's in the back end." TSMC will not break out packaging capex because it manages the two as one flexible pool — which is why a clean advanced-packaging capex figure does not exist for modelling.
  • A14/A13/A12 extend the tool-demand runway — A14 pre-production 2027, volume 2028; A13 and A12 volume 2029 (cc-wei in 2026-07-16-earnings-tsm-q2-fy2026). Each generation is a fresh equipment cycle.
  • A named limit on vendor pricing power — cc-wei in 2026-07-16-earnings-tsm-q2-fy2026 on High-NA: "We always consider that technology maturity and the cost, and whether we use it or not." ASML can price High-NA; it cannot make TSMC buy it.

Names and exposures

TickerExposureConviction
ASMLSole EUV/High-NA supplier; #1 WFE; multi-year High-NA cycleHigh — monopoly + secular driver
AMAT18.4% WFE share; 31% DRAM/HBM exposure (highest in peer group); EPIC Center multi-foundry R&D; Q3 guide $8.95BHigh (raised from medium-high post Q2 beat-and-raise) — China 30% revenue overhang remains
LRCXEtch + deposition for TSMC, Samsung, Micron, IntelMedium-high
KLAC52% process control; service revenue moatHigh — process complexity makes inspection essential
LINLinde — controls global helium distribution; pricing power vs fab customers during Ras Laffan-driven shortageMedium-high — helium is one leg of diversified industrial gas business; see helium-supply-crisis-semicap
APDAir Products — co-controls helium distribution; Q1 2026 EPS beat + FY2026 guidance raised citing helium tailwindMedium-high — already confirmed in guidance; see helium-supply-crisis-semicap

Why this is higher conviction than INTC re-rate

  • Diversified win: works whether Intel succeeds, TSMC stays dominant, or Samsung resurfaces. The thesis only fails if leading-edge capacity globally contracts — which is implausible given AI demand.
  • No customer-acquisition risk: ASML/AMAT/LRCX/KLA aren't depending on Apple converting a preliminary deal; their customers are already buying.
  • Stable revenue cadence: tool shipments are scheduled years in advance.

Contradictions / tensions

  • ASML is already richly valued and widely owned — re-rating headroom is narrower than INTC.
  • China export restrictions are a real, ongoing drag on AMAT and LRCX specifically.
  • If Intel 18A/14A stumbles AND TSMC's Arizona ramp slips, near-term order pace could disappoint vs consensus.
  • TSMC skipping High-NA at 2nm (using Low-NA + multi-patterning) reduces ASML's 2026 High-NA buyer pool; partially offset by SK Hynix and Samsung memory adoption. ASML's 2026 High-NA revenue is now Intel/Samsung/SK Hynix-driven — TSMC-independent for this cycle.
  • AMAT beat Q2 guide ($7.91B vs $7.65B) and raised Q3 to $8.95B. Citi raised PT to $520 (30x→33x fwd P/E). Valuation risk now is Q4/2027 guide relative to expanding embedded expectations.
  • WFE 2027 estimate range: EE Times $156B vs Morgan Stanley $185B+. AMAT's $8.95B Q3 guide implies ~$35B AMAT annualized — already past EE Times total if AMAT holds share. Morgan Stanley's $185B+ appears more accurate on current trajectory.

What would weaken this thesis

  • A breakthrough that bypasses High-NA EUV (no credible candidate today)
  • Coordinated foundry capex pause (no precedent at this point in an AI buildout)
  • AI silicon demand normalizing dramatically (unlikely on current evidence)
  • China export restrictions expanding to Low-NA EUV (currently only advanced DUV is restricted)

Valuation snapshot

Last refreshed 2026-07-20 (pre-open; marks are the Friday 2026-07-17 close, markets closed over the weekend). Every price fill tagged twelvedata. Mkt cap / Fwd P/E are not in the Twelve Data free tier and were not re-sourced for these names this run — no figure is shown rather than a stale one carried. The prior snapshot's "Near 52w high" framing on the semicap names is now wrong and has been corrected: 2026-07-16 was a sharp AI-infra risk-off (SOXX −4.5%, SMH −3.7%, XLK −2.2%) and the WFE complex has de-rated hard from its highs, while the industrial-gas leg went the other way.

TickerPrice52w rangeMkt capFwd P/EDay / vs 52w hiWhat's priced in (one line)
ASML$1,747.58$683.48–$1,999.96——−2.09% day; −12.6% from hiThe most resilient name in the complex on a −4.5% SOXX day — monopoly is the defensive leg. EUV monopoly priced; TSMC's raise to $60–64B capex + $100B Arizona (2026-07-16-earnings-tsm-q2-fy2026) is fresh order-book support not yet in the tape
AMAT$529.66$154.47–$739.67——−5.57% day; −28.4% from hiNo longer "near 52w high" — round-tripped from $739.67 to $560.93. Still above the Citi PT $520; Q2 beat+raise priced. China ~30% revenue overhang is the peer-worst exposure and is what the de-rate is expressing
LRCX$313.30$90.94–$438.50——−2.39% day; −28.6% from hiNo longer "near 52w high." All-foundry breadth priced; advanced-packaging TAM upside still deferred to 2027
KLAC$212.75$83.22–$307.37——−3.02% day; −30.8% from hiNo longer "near 52w high" — most de-rated of the four despite the 52% process-control moat. Q4 FY2026 earnings (Aug) = guide for tool intensity
LIN$513.22$387.78–$548.20——−1.44% day; −6.4% from hiGreen on a −4.5% SOXX day — the defensive/consumer rotation rewarded industrial gas while WFE sold off. Helium still excluded from FY2026 guidance = unpriced optionality; JP Morgan OW $525 PT now ~1% above spot. See helium-supply-crisis-semicap
APD$295.62$229.11–$314.87——−0.56% day; −6.1% from hiAlso green in the rotation. FY2026 guidance raise (helium tailwind) partially priced; Q3 FY2026 earnings (late July) = the first post-strike contract-repricing tell — the nearest catalyst on this page

Forward-looking outcomes (12-month)

Bull case — the buyer's own capex raise flows through, and the buyer says the price is inflating: TSMC lifted 2026 capex to $60–64B from $52–56B (January) and committed an additional $100B to Arizona plus 13 leading-edge/advanced-packaging fabs in Taiwan (wendell-huang and cc-wei in 2026-07-16-earnings-tsm-q2-fy2026) — the single largest upward revision to the buildout this thesis has recorded, with 70–80% going to advanced process. And the pricing power is now disclosed by the payer: cc-wei, same source, names inflation as one of only two reasons for the raise — "Now we buy the tools with inflation price." Supply-side confirmation from wendell-huang: "We do not foresee any bottlenecks to our capacity expansion plans" — vendors can fill the orders, so the raise converts to revenue rather than backlog. Implied price: AMAT back toward the $739.67 high (~+30%); ASML tracks prior ATH trajectory. Cited: 2026-07-16-earnings-tsm-q2-fy2026, 2026-05-15-autoresearch-amat-q2-fy2026-earnings-results-guidance.

Base case — WFE grows toward $160–185B by 2027; AMAT sustains an $8.5–9B run-rate; ASML ships 60+ EUV as planned; Intel 18A/14A and Samsung Taylor absorb High-NA at current pace: thesis intact and materially reinforced by the TSMC raise, but the multiples stay compressed because the market is discounting the AI capex cycle wholesale rather than pricing each vendor's book. Implied price: AMAT $600–660; ASML tracks earnings growth. Cited: 2026-05-15-autoresearch-amat-q2-fy2026-earnings-results-guidance, 2026-05-11-autoresearch-picks-shovels-semicap-update-may-2026.

Bear case — China export restrictions expand to Low-NA EUV; AI capex consolidation slows foundry ordering; Intel 18A/14A stumbles AND TSMC's Arizona ramp slips: WFE misses consensus; AMAT misses its guide; multiples compress further from an already-de-rated base. AMAT most exposed (~30% China revenue). A named limit on the vendors' pricing power, first-party: cc-wei in 2026-07-16-earnings-tsm-q2-fy2026 on High-NA — "We always consider that technology maturity and the cost, and whether we use it or not." ASML can price High-NA; it cannot make TSMC buy it, and TSMC is taking 0 High-NA units in 2026 (2026-05-11-autoresearch-picks-shovels-semicap-update-may-2026). Implied price: AMAT to $420–460; ASML most resilient (monopoly + multi-year backlog). Cited: 2026-05-15-autoresearch-amat-q2-fy2026-earnings-results-guidance.

Currently undervalued vs base case? AMAT: changed from No to Marginal. The prior "No" rested on AMAT trading at the Citi PT $520 with 33× forward requiring continued beat-and-raise. It has since round-tripped from $739.67 to $560.93 (−24.2% from the high) while the demand signal underneath got larger, not smaller — TSMC's capex raise to $60–64B landed yesterday. The de-rate is the AI-capex complex being sold as a block (SOXX −4.5%), not a vendor-specific deterioration. Still only Marginal, because $560.93 remains above Citi's $520 and the ~30% China exposure is the live, unresolved drag. ASML: Marginal → the most defensible of the four — it fell just 1.67% on a −4.5% SOXX day and sits only 10.8% off its high; the monopoly is doing exactly what a monopoly should in a de-rate, but that resilience is itself why the re-rate headroom is narrowest here. LRCX / KLAC: Marginal, newly so — both are now 27–29% off their highs on no thesis-level news; the honest statement is that the wiki has never sourced specific forward multiples for either, so this is a distance-from-high observation, not a valuation call. LIN / APD: see helium-supply-crisis-semicap — both green on the day and the only names here with an unpriced, dated catalyst (APD Q3, late July).

  • From 2026-06-01-autoresearch-gtc-taipei-vera-rubin-macro-buckets: Vera Rubin supply chain 2× Blackwell, confirmed June 1 — Jensen Huang GTC Taipei: "The supply chain for Vera Rubin is twice the size of what Blackwell required." Full production fall 2026. $1T cumulative Blackwell + Rubin orders by 2027. TSMC, ASML, AMAT, KLAC are the three-layer beneficiary (logic fab → packaging → metrology/inspection) of the largest Nvidia generation ever. Vera Rubin sole-sourced to TSMC + 150 Taiwan factories. The WFE capex cycle from this demand signal extends through 2027+ — confirming the picks-and-shovels thesis duration.

Catalyst path:

  • AMAT Q3 FY2026 earnings (August): guide for Q4; $9.5B+ confirms continued step-up to bull trajectory
  • ASML Q2 2026 earnings (July): EUV order book update; High-NA pull-forward from TSMC signal
  • Intel 14A tool orders announced: first hard signal of the next WFE cycle leg

Related

Referenced by
Mechanisms
TSMC saturation → 4-year price hikes → marginal buyers shift → Intel anchor stack → re-rate
Concepts
CoWoS packaging capacity crunchHBM supply bottleneckSemicap tool-price inflation: capex dollars ≠ capacityUS fab capacity bottleneck → Intel as the alternative foundry
Questions
Hypothesis: AI power delivery ICs — M&A consolidation premium in MPWR/VICR/TXNHypothesis: CHIPS Act 48D December 31, 2026 construction deadline → H2 2026 fab groundbreaking rush → PWR/FIX near-term revenue accelerationHypothesis: Lattice FPGA + AMI → AI Cluster Management Layer as Standalone CategoryHypothesis: Silicon Photonics → AI Cluster Backend Networking as Undercovered Picks-and-Shovels CategoryHypothesis: SoftBank Roze + ABB Robotics → Robotics as Data Center Construction PlayIs the $830B CSP CapEx cycle a one-year spike or a sustained multi-year structural level?
Entities
Applied MaterialsASMLBE Semiconductor Industries (BESI)Dwarkesh PatelGavin BakerIntelLattice Semiconductor (LSCC)Micron TechnologyReiner PopeWendell Huang