KLA Corporation — Bull & Bear Memo
Prepared 11 September 2026 · Thinking anchor, not a valuation or recommendation
1. Business in one line
KLA sells the inspection and metrology tools that find and measure defects between chipmaking steps, then earns a service stream on every tool it has installed. FY2026 revenue was $13.58bn: 77% product, 23% service. Semiconductor Process Control made up 90% of revenue and 97% of segment profit (FY26 10-K, Note 17).
Cyclical, with a stalwart's service base.Revenue follows the capex budgets of a handful of chipmakers (TSMC alone is ~19%), and quarterly revenue fell 21% in six months in the last down-cycle (Dec-22 → Jun-23 releases). So the bear case lives in the cycle and in China. To win, the bull case has to show that each trough ends higher than the last.
2. Bull case — Peter Lynch pitch
The simple reason this stock could work is that every new generation of chips needs more measurement per wafer. KLA already takes the majority of that spend, and each tool it ships adds to a mostly contract-based service stream. That should make earnings step up from one cycle to the next instead of rising and falling back to the same level.
A1Process-control intensity keeps rising, and KLA keeps its share of it
- What must happen
- KLA's share of total wafer-equipment spend rises from ~7.4% today toward the ~9% (±25bp) assumed in its 2030 model.
- Why KLA specifically
- A defect costs more as dies get larger and stacks get taller (AI dies, HBM, hybrid bonding). KLA is the incumbent across inspection, review and metrology and says it is 6.5× its nearest process-control rival (Investor Day 2026). Lock-in is per production line, so a line KLA wins stays with KLA for that line's life.
- Evidence
- Wafer Inspection grew 43% in FY25 and 7% in FY26. Patterning grew 23% in FY26 (FY26 10-K). KLA guides advanced-packaging process-control revenue to ~$1.1bn in CY26, up about 70%. Its share of that market went from "a couple percent" in 2023 to 7–8% in 2026 (Q4 FY26 call; Citi, Sep 2026).
- What to monitor
- KLA-stated share of wafer-equipment spend, and Inspection and Patterning growth compared with KLA's own market estimate.
A2AI capex stays high through 2027, with no 2023-style air pocket
- What must happen
- Wafer-equipment spend holds at or above the "low $150bn" KLA sees for CY26, and backlog turns into revenue instead of being rescheduled.
- Why KLA specifically
- Orders cover 12–18-month delivery windows, and optics lead times run 18–24 months or more (Citi, Sep 2026). That gives KLA unusual visibility. More foundry/logic players are also investing at the leading edge (Intel 14A, Samsung, Rapidus), which is the most process-control-intensive segment.
- Evidence
- Backlog rose to $12.57bn from $7.86bn (FY26 10-K), about 0.93× FY26 revenue. The Q1 FY27 guide of $4.0bn ±$0.2bn is about 25% above Q1 FY26's $3.21bn. Management expects 2H CY26 revenue to be ~20% above 1H and "significant growth" in 2027 (Q4 FY26 call).
- What to monitor
- Year-end backlog, quarterly revenue against the guided midpoint, and KLA's wafer-equipment outlook for 2027.
A3Service compounds at 13–15% a year and cushions the trough
- What must happen
- Service keeps growing low-to-mid teens, and keeps growing when product revenue falls.
- Why KLA specifically
- About 80% of service revenue is contract-based (Q4 FY26 call). Service revenue per tool is up about 3× from the "class of 2000" to the "class of 2020" (Investor Day). The record shipments of FY25–FY26 enlarge the base of tools that will come off warranty.
- Evidence
- Service revenue was $3.13bn in FY26, up 16%. In FY24 it rose 10% while product revenue fell 11% (10-Ks). From Dec-22 to Jun-23, product revenue fell 26% while service rose 3.5% (releases, computed).
- What to monitor
- Year-on-year service growth, above all in the first quarter when product revenue declines.
A4Scale lifts margins to the 2030 model
- What must happen
- Gross margin rises from 61.3% (GAAP FY26) toward the 63.5% target, and operating margin from 41.7% toward 45–47%.
- Why KLA specifically
- R&D of $1.53bn (11.3% of revenue) is a fixed cost: every competitor has to fund the same roadmap, and KLA spreads it over the largest revenue base. Management incentives are tied to relative free cash flow (Investor Day).
- Evidence
- Operating margin was ~37% at the FY24 trough (34.1% including a $289m impairment) and 41.7% in FY26 (computed from the 10-Ks). FY26 free cash flow was $3.77bn, and 89% of it ($3.35bn) went back to shareholders.
- What to monitor
- Non-GAAP gross margin against the guide, Semiconductor Process Control segment margin, and R&D as a percentage of revenue.
3. Bear case — Munger invert
The most likely way I lose money is paying for peak margins that depend on a China base KLA can't defend and an AI capex wave it can't control, and then finding the next trough is lower than the last, not higher.
Ordered by how permanent and value-destructive the damage is.
B1China shrinks for good while the R&D bill keeps growingPermanent
- How it could fail
- Export rules keep KLA out of fabs that non-US and Chinese competitors can still supply. Domestic suppliers win those lines, and line-level lock-in then works against KLA for the life of each fab.
- What would confirm failure
- China revenue falls in dollars, not just as a share. The Affiliates Rule is reimposed when its one-year suspension ends in November 2026. KLA discloses more returned deposits.
- Damage to economics
- China was $4.05bn, or 29.8% of FY26 revenue (10-K). KLA already admits competitors "have been able to ship into fabs in China that we haven't" (Q4 FY26 call) and that it has had to "return substantial deposits" (10-K). Lost China revenue comes out at a ~61% gross margin while R&D stays, so roughly $0.6bn of operating profit goes per $1bn lost (interpretation).
- Permanent or fixable
- Permanent. A production line won by a competitor does not come back.
B2The AI capex wave peaks and the backlog turns out to be softFixable, but deep
- How it could fail
- Customers pause after pulling 2026–27 capacity forward. KLA itself says the timing of backlog depends on fab readiness and variables "beyond our control" (10-K). One customer is ~19% of revenue, so one budget change moves the whole company.
- What would confirm failure
- Backlog falls year on year, the first quarter lands below the guided midpoint, or the 2027 wafer-equipment outlook is cut.
- Damage to economics
- Last cycle, quarterly revenue fell 21% and product revenue 26% in six months. Operating margin went from 39.7% (FY22) to ~37% ex-impairment (FY24). Service softens the fall (it rose 3.5% in that slide) but did not stop it.
- Permanent or fixable
- Fixable over a cycle, but anyone who bought on peak earnings waits years to get back to them.
B3Intensity is a budget line, not a lawPartly permanent
Attacks A1
- How it could fail
- Process control samples wafers; it is not a production step. The 10-K says costly process changes can lead customers to "reduce inspection and metrology sampling rates." A capacity-driven cycle (greenfield fabs, conventional memory) buys fewer process-control dollars per equipment dollar than a technology-driven one, and larger rivals may build metrology into their process tools (10-K).
- What would confirm failure
- KLA's share of wafer-equipment spend stays flat or falls while the market grows, or memory's share of systems revenue rises with no share gain.
- Damage to economics
- This cuts the number of tools sold, not their price, so it also shrinks the installed base that feeds service (A3).
- Permanent or fixable
- Partly permanent. Sampling can be turned back up, but a line that goes to integrated metrology is gone.
B4Operating leverage is weaker than the model assumesFixable
Attacks A4
- How it could fail
- A shortage of the DRAM chips used inside KLA's tools, plus tariffs, absorbs the gains from higher volume.
- What would confirm failure
- Non-GAAP gross margin falls below 62%, or Semiconductor Process Control segment margin declines a second year.
- Damage to economics
- The memory-cost headwind is now "slightly ahead of 100 bps" of gross margin, up from ~75bp expected (Citi, Sep 2026). The 2030 model already absorbs ~50bp for tariffs. Process Control segment margin fell to 44.8% from 45.7% in FY26 even though revenue grew 12% (10-K Note 17, computed).
- Permanent or fixable
- Fixable. Management plans to pass costs through in new-product pricing.
Each bull assumption beside the attack on it
Must happen: KLA's share of total wafer-equipment spend rises from ~7.4% today toward the ~9% (±25bp) assumed in its 2030 model.
Evidence: Wafer Inspection grew 43% in FY25 and 7% in FY26. Patterning grew 23% in FY26 (FY26 10-K). KLA guides advanced-packaging process-control revenue to ~$1.1bn in CY26, up about 70%. Its share of that market went from "a couple percent" in 2023 to 7–8% in 2026 (Q4 FY26 call; Citi, Sep 2026).
How it fails: Export rules keep KLA out of fabs that non-US and Chinese competitors can still supply. Domestic suppliers win those lines, and line-level lock-in then works against KLA for the life of each fab.
Confirms it: China revenue falls in dollars, not just as a share. The Affiliates Rule is reimposed when its one-year suspension ends in November 2026. KLA discloses more returned deposits.
How it fails: Process control samples wafers; it is not a production step. The 10-K says costly process changes can lead customers to "reduce inspection and metrology sampling rates." A capacity-driven cycle (greenfield fabs, conventional memory) buys fewer process-control dollars per equipment dollar than a technology-driven one, and larger rivals may build metrology into their process tools (10-K).
Confirms it: KLA's share of wafer-equipment spend stays flat or falls while the market grows, or memory's share of systems revenue rises with no share gain.
Must happen: Wafer-equipment spend holds at or above the "low $150bn" KLA sees for CY26, and backlog turns into revenue instead of being rescheduled.
Evidence: Backlog rose to $12.57bn from $7.86bn (FY26 10-K), about 0.93× FY26 revenue. The Q1 FY27 guide of $4.0bn ±$0.2bn is about 25% above Q1 FY26's $3.21bn. Management expects 2H CY26 revenue to be ~20% above 1H and "significant growth" in 2027 (Q4 FY26 call).
How it fails: Customers pause after pulling 2026–27 capacity forward. KLA itself says the timing of backlog depends on fab readiness and variables "beyond our control" (10-K). One customer is ~19% of revenue, so one budget change moves the whole company.
Confirms it: Backlog falls year on year, the first quarter lands below the guided midpoint, or the 2027 wafer-equipment outlook is cut.
Must happen: Service keeps growing low-to-mid teens, and keeps growing when product revenue falls.
Evidence: Service revenue was $3.13bn in FY26, up 16%. In FY24 it rose 10% while product revenue fell 11% (10-Ks). From Dec-22 to Jun-23, product revenue fell 26% while service rose 3.5% (releases, computed).
How it fails: Customers pause after pulling 2026–27 capacity forward. KLA itself says the timing of backlog depends on fab readiness and variables "beyond our control" (10-K). One customer is ~19% of revenue, so one budget change moves the whole company.
Confirms it: Backlog falls year on year, the first quarter lands below the guided midpoint, or the 2027 wafer-equipment outlook is cut.
Must happen: Gross margin rises from 61.3% (GAAP FY26) toward the 63.5% target, and operating margin from 41.7% toward 45–47%.
Evidence: Operating margin was ~37% at the FY24 trough (34.1% including a $289m impairment) and 41.7% in FY26 (computed from the 10-Ks). FY26 free cash flow was $3.77bn, and 89% of it ($3.35bn) went back to shareholders.
How it fails: Export rules keep KLA out of fabs that non-US and Chinese competitors can still supply. Domestic suppliers win those lines, and line-level lock-in then works against KLA for the life of each fab.
Confirms it: China revenue falls in dollars, not just as a share. The Affiliates Rule is reimposed when its one-year suspension ends in November 2026. KLA discloses more returned deposits.
How it fails: A shortage of the DRAM chips used inside KLA's tools, plus tariffs, absorbs the gains from higher volume.
Confirms it: Non-GAAP gross margin falls below 62%, or Semiconductor Process Control segment margin declines a second year.
4. Signals to monitor
BIS Affiliates Rule status
B15. External challenge notes
Not run. Write "external challenge" to add this section. It searches outside the company's filings (competitors, regulation, technology shifts, short-seller arguments) and fills in only this section.
6. Bottom line
- Why it worksKLA could work because more complex chips keep raising the measurement spend per wafer, KLA owns the majority of it, and every tool shipped adds to a contract-based service stream that lifts each trough.
- What must go rightFor that to happen, KLA's share of wafer-equipment spend has to climb from ~7.4% toward ~9% while AI capex holds through 2027, and gross margin has to recover toward 63.5% despite memory-chip costs.
- How it breaksThe most permanent way the thesis breaks is export rules and Chinese substitution taking the ~$4bn China base for good while the $1.5bn R&D bill keeps rising.
- What changes my mindChina revenue falling in dollars, backlog falling year on year, or the first quarter below the guided midpoint would change my mind, while service growth above 13% through a product downturn would strengthen it.
Sources
Annual reports (sources/Annual_Reports) 2
- FY2026_10-K_2026-08-06 — segments, geography, backlog, customer concentration, risk factors, cash flow
- FY2024_10-K_2024-08-05 — trough-year product/service split and opex
Earnings releases (sources/Earnings_Releases) 3
- 2026-07-28 — FQ4 FY26 results, balance sheet, Q1 FY27 guidance
- 2025-10-29 — FQ1 FY26 revenue (base for guide growth)
- 2023-01-26 and 2023-07-27 — Dec-22 peak and Jun-23 trough quarters
Transcripts (sources/Transcripts) 3
- 2026-07-28 Q4 FY2026 earnings call — WFE view, service 80% contract-based, China share commentary, advanced packaging
- 2026-03-12 Investor Day 2026 — 2030 model, share of WFE 7.4% → ~9%, 6.5× nearest competitor, incentive on relative FCF
- 2026-09-09 Citi Global TMT conference — lead times, memory-cost GM headwind, advanced-packaging share
Other filings (sources/Other_Key_Documents) 1
- 2026-03-12 8-K — dividend raise and $7bn repurchase authorization