Defect inspection & review (process step)

Step 38 of 56 in the semiconductor flow (Control). Prev: Metrology (CD, overlay, film, OCD) (process step) · Next: Parametric / in-line E-test (process step)

What this step does

Defect inspection finds physical anomalies on the wafer (particles, scratches, pattern bridges, opens, residue) by scanning the surface and comparing it to a defect-free reference. It answers “is something there that shouldn’t be?”, which is a different job from metrology’s “is the right thing the right size?”. A flagged defect is then sent to defect review, where a high-resolution image classifies it so engineers can trace the root cause back to a tool or step. Sub-processes inside this step:

• Optical inspection — broadband/laser brightfield and darkfield scanning of patterned and unpatterned (bare) wafers at high throughput to catch yield-killing defects fast.
• E-beam inspection — slower, far higher resolution scanning that resolves defects optics physically cannot, including buried/voltage-contrast (electrical) defects in 3D structures.
• Defect review — a review SEM revisits each flagged coordinate to image and classify it (often AI-assisted).
• Defect classification + sampling — software bins defects, builds wafermaps, and decides which to review; the data feeds yield management.

Where it sits and why it matters

Inspection runs repeatedly through the flow, after litho, etch, deposition and CMP, both inline in production and heavily in process R&D and ramp, where finding a new defect mode first decides who hits yield first. It is the eyes of the fab: a single uncaught systematic defect can scrap thousands of wafers. Value here is overwhelmingly equipment-led (the tools cost millions each and there is little consumable), which makes it one of the highest-margin, most concentrated niches in all of semiconductor equipment.

Equipment market

Wafer inspection equipment was roughly $6.9B in 2025 (some syndicated reads put the broader “defect inspection” line at $7.4-7.8B in 2024), growing ~8-9% CAGR, faster than the equipment average because shrinking nodes and 3D/GAA stacks need ever more sensitive inspection [1][3][7]. This is the most KLA-dominated step in the whole flow:

Concentration note: KLA + AMAT hold >60% of all defect inspection and KLA alone is past 85% in optical, one of the most monopolistic positions in WFE outside EUV litho. E-beam is the contested frontier, growing fastest (~9% CAGR) on multibeam throughput and voltage-contrast needs, and is the one segment where the field is genuinely plural [1][3].

Materials & consumables

Almost none, which is the point. Inspection tools consume electron-source filaments/cathodes (e-beam columns), calibration/reference wafers, optics and vacuum spares, plus the maintenance and service contract. There is no slurry-style razor-and-blades stream as in CMP. The recurring revenue is service + software: maintenance, applications support and yield-management software (e.g. KLA’s Klarity), which is high-margin and sticky but tied to the installed tool base rather than per-wafer consumption.

Volumes, revenue, profitability

Inspection is sampling-based, not every-wafer, so the volume driver is the number of installed tools and inspection points per flow, not wafer-passes. A leading-edge fab runs dozens of inspection steps and buys tools at $2-10M+ each (e-beam at the top). Revenue pool: ~$6.9-7.8B equipment annually, with KLA’s inspection-systems line alone ~$4.3B (51% of KLA’s ~$9.8B FY24) [2][5]. Profitability is the headline: KLA runs ~60%+ corporate gross margin and ~40% operating margin, the richest in large-cap WFE, precisely because process control is concentrated, software-attached and mission-critical [2]. The margin is earned by the toolmaker, overwhelmingly KLA.

Competitive landscape & value capture

The moat is the deepest in equipment: decades of defect-detection physics, the installed base of BBP optical inspectors with thousands of patents, qualification lock-in at every leading-edge fab, and a yield-data flywheel (more installed tools means more defect signatures means better algorithms). Value accrues to the toolmaker, and within that almost entirely to KLA in optical, with the e-beam segment the only place a challenger can win. EU exposure is thin: ASML touches e-beam/multibeam via past acquisitions (HMI) and Carl Zeiss SMT (Germany) supplies the optics/e-beam columns behind several platforms, but there is no EU-headquartered defect-inspection tool champion. Seed-relevant angle is narrow and software/AI-shaped: defect classification, deep-learning review automation, and computational/optical-inspection startups that could be acquired into a platform, rather than a credible new hardware entrant against KLA.

Market drivers, constraints & trends

Net read: a structurally accelerating served market, growing meaningfully faster than WFE through 2030 on AI-die complexity and advanced packaging, with China the one drag on an otherwise compounding base.

• Drivers

• AI-die inspection intensity is rising per wafer: larger AI/HBM dies mean a single flaw scraps proportionally more silicon, so leading-edge logic and DRAM buy more inspection passes per layer. KLA frames a ~15% process-control revenue CAGR off exactly this plus packaging [8]. JPMorgan models a path to ~20% annual process-control growth to 2030 on GAA + backside power delivery [9].

• Advanced packaging is the fastest new pool: KLA’s packaging process-control revenue is on track for ~$1B in CY2026, up from ~$635M (2025) and ~$300M the year before, growing in the high-50s percent; share went from <1% three years ago to >6% now [9]. 2.5D/3D, hybrid bonding and heterogeneous integration each add alignment and bond-void inspection that did not exist at this scale before [10].

• Node transitions force new sensitivity: GAA, 3D stacks and High-NA EUV each create defect modes prior tools cannot resolve, with KLA’s Gen-4 optical and e-beam aimed at GAA structures [9].

• Constraints

• China is the live drag: China fell to ~30% of KLA revenue in Q4 FY25 from ~44% a year earlier, a ~$500M incremental export-control hit, with a further ~20% China decline guided and a 50% domestic-equipment mandate threatening the trailing-edge pool [11].

• E-beam throughput ceiling: single-column e-beam can take hours to days per 300mm wafer, capping how much of the rising defect load can actually be inspected and limiting how fast the highest-margin segment converts demand to revenue [12].

• Cyclicality + concentration: a sampling-driven, equipment-led pool tied to fab capex still rides the WFE cycle, and KLA’s >85% optical share leaves the served market’s growth hostage to one vendor’s roadmap [7].

• Trends & inflections to watch

• Multibeam e-beam scaling: multi-column platforms lifted throughput ~28% with ~53% of logic makers expanding deployment in 2024; tripwire = a >100-beam production-qualified inspection tool shipping at a leading-edge logic/DRAM fab [1][12].

• High-NA stochastic-defect inspection: ppt-level stochastic defects make High-NA layers uninspectable by today’s optics; tripwire = a named foundry qualifying a dedicated High-NA stochastic-defect inspection flow (the planned 2026-27 High-NA era) [12].

• Packaging re-rate: tripwire = KLA’s packaging process-control line printing >$1B in CY2026, confirming the new pool compounds rather than spikes [9][10].

Connections

• Semiconductor Metrology & Inspection — the metrology + inspection equipment cluster page
• Semiconductor Metrology Wedge — thesis on where value accrues across the metrology/inspection layer

Sources

1. Seeking Alpha, Applied Materials e-beam inspection lost share to KLA’s optical (segment dynamics, AMAT vs KLA) — https://seekingalpha.com/article/4565533-applied-materials-ballyhooed-e-beam-inspection-sector-lost-share-to-klas-optical
2. KLA FY2024 results / 10-K (total revenue ~$9.8B, process control 89%, inspection systems 51% of net sales; corporate margins) — https://ir.kla.com/news-events/press-releases/detail/478/kla-corporation-reports-fiscal-2024-fourth-quarter-results
3. GlobalGrowthInsights, Semiconductor Wafer Defect Inspection System Market (~$7.82B 2024; vendor shares; optical vs e-beam) — https://www.globalgrowthinsights.com/market-reports/semiconductor-wafer-defect-inspection-system-market-115186
4. KLA, BBP broadband-plasma patterned wafer inspection 40th anniversary (~3,000 BBP installed, ~1,000 patents) — https://www.kla.com/esl10-ebeam-inspection
5. Statista, KLA revenue by product 2024 (wafer inspection ~$4.33B) — https://www.statista.com/statistics/1358062/kla-revenue-by-product/
6. GMInsights, Semiconductor Metrology & Inspection Market (KLA/AMAT/Onto/Thermo Fisher/Hitachi ~58% combined; AMAT ~9.8%) — https://www.gminsights.com/industry-analysis/semiconductor-metrology-and-inspection-market
7. Maximize Market Research, KLA wafer inspection (market ~$6.92B 2025 → ~$12.03B 2032, 8.22% CAGR; KLA >56% process control, >85% optical) — https://www.maximizemarketresearch.com/news/kla-corporation-semiconductor-wafer-inspection/
8. The Index Times, KLA process control + AI-die inspection intensity (~15% process-control revenue CAGR; larger AI dies, inspection per wafer rising) — https://www.theindextimes.com/post/klac-kla-s-58-process-control-stranglehold-tightens-as-ai
9. TIKR / JPMorgan KLAC note (packaging process control ~$1B CY2026 vs ~$635M 2025 / ~$300M prior, high-50s% growth, share <1%→>6%; ~20% process-control CAGR to 2030 on GAA + backside power; Gen-4 for GAA) — https://www.tikr.com/blog/kla-corporation-jumped-5-after-jpmorgan-conference-heres-what-klac-stock-could-return
10. KLA, the packaging pivot driving AI chip performance (2.5D/3D, hybrid bonding, heterogeneous integration inspection requirements) — https://www.kla.com/advance/innovation/the-packaging-pivot-driving-ai-chip-performance
11. TrendForce, KLA China demand + export curbs (China ~30% of revenue Q4 FY25 from ~44%; ~$500M incremental hit; ~20% China decline guided) + Fintool, China 50% domestic-equipment mandate — https://www.trendforce.com/news/2025/08/01/news-chip-equipment-giants-tel-kla-flag-weak-china-demand-in-2025-amid-trade-tensions/
12. Semiconductor Engineering, e-beam inspection essential for advanced nodes (single-column throughput hours-to-days/wafer; multibeam ~28% throughput gain, ~53% logic adoption 2024) + Averroes, EUV stochastic-defect detection (ppt-level stochastics, 2026-27 High-NA era timeline) — https://semiengineering.com/e-beam-inspection-proves-essential-for-advanced-nodes/