Applied Materials Inspection Tools Explained: Enlight, SEMVision, PROVision and Where AI Fits
Averroes
Nov 07, 2026
Applied Materials’ inspection portfolio covers the optical, e-beam and mask sides of fab process control. Enlight optical inspection finds defect candidates at speed, SEMVision e-beam review resolves and classifies them, PROVision e-beam inspection hunts what optics cannot see, Aera inspects photomasks, and ExtractAI links optical detection to e-beam review in real time. Together they form one of the two major inspection ecosystems in modern fabs, alongside KLA’s.
These tools run across the inspection gates of Stage 03 and Stage 04 of the 12-stage semiconductor manufacturing process. This guide from Averroes.ai, an AI visual inspection platform for manufacturers, maps the portfolio to the line, explains the AI Applied Materials builds into it, and shows where a cross-fleet AI layer fits alongside it.
Key Notes
Enlight finds, SEMVision resolves: Applied Materials’ optical and e-beam tools are built as a linked pair, with ExtractAI deciding what moves between them.
SEMVision H20, introduced in March 2026, is the current flagship review tool, pairing cold field emission e-beam with deep-learning defect extraction.
Applied Materials states that advanced nodes produce about 100X more defect candidates for e-beam review, which is why nuisance filtering is the portfolio’s central theme.
Applied Materials’ AI is embedded in Applied Materials’ tools. Mixed fleets need a layer that works across vendors, which is where a tool-agnostic overlay fits.
The Applied Materials Inspection Portfolio, Mapped to the Line
Applied Materials’ inspection and metrology tools are the process-control arm of the industry’s largest equipment maker, spanning optical inspection, e-beam inspection and review, mask inspection and CD metrology. The portfolio’s organizing idea is a division of labor: inspection tools find defect candidates fast, and review tools resolve and classify what inspection found. Every tool below sits on one side of that split.
Applied Materials inspection tools: what each does and where it runs
Tool
Type
What It Does
Where It Runs
Enlight 3
Optical inspection
Finds defect candidates at speed
Inline gates: litho, etch, CMP, films
SEMVision G10 / H20
E-beam review
Resolves and classifies detections
After every inspection gate
PROVision 3E / PrimeVision 10
E-beam inspection
Finds what optics cannot see
Hotspots, voltage contrast
VeritySEM 10
CD-SEM metrology
Measures printed dimensions
Litho and etch metrology
Aera 6
Mask inspection
Qualifies photomask patterns
Mask shop and requalification
VeritySEM 7AP / SEMVision G7AP
Packaging metrology and review
Bump, TSV and bond control
Advanced packaging lines
The map makes one point: the portfolio is not a list of rivals but a relay, and each handoff in it happens at a gate this hub covers.
The optical line. Enlight 3 is the current optical inspection platform, and Applied Materials’ product materials describe it as covering brightfield and greyfield imaging in one system, with throughput reaching 30 wafers per hour in darkfield applications. It is the candidate-finding end of the relay, running the inline gates where bright field inspection and its darkfield counterpart earn their keep.
The review line. SEMVision is the resolving end. Applied Materials describes the family as the most widely deployed e-beam review line in the industry, and its newest member, the SEMVision H20 introduced in March 2026, pairs the company’s second-generation cold field emission source with deep-learning image models, delivering review results the company states are up to 3X faster than prior techniques.
The e-beam inspection line. PROVision 3E and PrimeVision 10 do something different from review: they are inspection tools that use e-beam inspection physics to find defects optical tools cannot see at all, including buried and electrically significant defects surfaced through voltage contrast.
Mask and metrology. Aera 6 carries mask inspection, the qualification gate the photomask inspection guide covers end to end, and VeritySEM 10 carries CD-SEM metrology. For advanced packaging, the VeritySEM 7AP and SEMVision G7AP pair extends the same find-and-resolve logic to bumps, TSVs and bonding, where Applied Materials’ own materials describe optical tools flagging defects that the e-beam system then resolves and classifies.
The AI Inside the Tools: ExtractAI and the H20
ExtractAI is Applied Materials’ machine learning technology that links Enlight optical inspection to SEMVision e-beam review, using review results to classify the optical defect map in real time. The company’s Enlight 3 materials describe it, now in its third generation, as the intelligent link between the two platforms.
The problem it exists to solve is the industry’s oldest trade restated at modern density. Raising optical sensitivity raises nuisance detections with it, and Applied Materials states that today’s advanced nodes produce about 100X more defect candidates for e-beam review than earlier generations. Review capacity cannot grow 100X, so something has to decide which candidates deserve it. ExtractAI makes that call on the optical side; the SEMVision H20 adds deep-learning extraction of true defects from nuisance on the review side, which is the capability its March 2026 launch coverage led with, quoting the company on helping chipmakers distinguish signal from noise.
One sentence of context belongs here. The industry’s largest equipment maker now builds AI into both ends of its inspection pair, which confirms that separating real defects from nuisance is the central problem of modern inspection, the same problem every gate page in this hub traces from coat to final test.
Architecturally, the design is a closed loop by intent: ExtractAI runs between Applied Materials’ optical tool and Applied Materials’ e-beam reviewer, trained and tuned within that stack. That is a strength inside the pair, and it defines the boundary of what the embedded AI sees, which matters in the last section.
Applied Materials vs KLA in Inspection
Applied Materials and KLA divide fab inspection along historical lines: KLA is the long-standing leader in optical patterned-wafer inspection, while Applied Materials leads in e-beam review and has been expanding into optical. Neither company’s position is in serious dispute, and the 2026 numbers describing the contest come from Applied Materials’ own executives: investor coverage from October 2026 reports the company’s process control business heading toward roughly 3 billion dollars for the year, serving about 2 billion dollars of a 9 billion dollar optical inspection market, with its imaging group head, a 27-year KLA veteran, telling investors that customers want alternatives.
Those are company statements about its own momentum, and this page takes no side. The practical reality for a fab is simpler and more useful than the market-share contest: most fabs run both. Tool selection happens per gate, on sensitivity, throughput and cost of ownership for that layer’s defects, and a line that runs KLA optical inspection at one gate, Applied Materials review behind it, and an Onto tool at a third is not an exception but the norm that wafer inspection programs are actually built from.
That norm is the fact the rest of this page turns on. Each vendor’s embedded AI is excellent inside its own fleet and sees nothing outside it.
Where a Cross-Fleet AI Layer Fits
A cross-fleet AI layer classifies the defect images and maps that inspection tools from every vendor already produce, so one trained judgment applies across an Applied Materials, KLA and Onto fleet at once. The logic follows from the mixed-fleet norm: embedded AI travels with its vendor’s tools, but a fab’s defect classes travel with the fab. The same scratch, residue or opened void crosses every tool in the line, and a classification standard that changes at each vendor boundary means the fab’s false-positive rate is really three different rates stitched together.
The diagram makes one point: the two architectures answer different questions, one optimizing a pair, the other unifying a fleet.
Averroes.ai is that second architecture. It runs as an overlay on the inspection images and defect maps that Applied Materials, KLA and Onto tools already produce, trains AI defect detection and classification models on the fab’s own defect classes from 20 to 40 labeled images per class, with no code, and deploys on premise so image data stays inside the fab. It replaces nothing. The tools, their recipes and their embedded AI keep doing exactly what the fab bought them for, and ExtractAI keeps optimizing its pair; the overlay adds the judgment layer that spans the whole fleet, so one taxonomy and one false-positive standard hold from the first gate to the last, with defect classification sorting confirmed detections into the fab’s own classes wherever they were imaged.
Running a Mixed Inspection Fleet?
See AI defect detection on the images your Enlight and SEMVision tools already produce
Frequently Asked Questions
What is SEMVision?
SEMVision is Applied Materials’ family of e-beam defect review tools, which the company describes as the most widely deployed in the industry. The current flagship, the SEMVision H20 introduced in March 2026, combines second-generation cold field emission with deep-learning image models, and Applied Materials states it reviews defects up to 3X faster.
What is ExtractAI?
ExtractAI is Applied Materials’ machine learning technology connecting Enlight optical inspection to SEMVision e-beam review. It uses the results of e-beam review to classify the optical defect map in real time, so the system directs review capacity at the detections most likely to be real defects. It is in its third generation.
What is the Enlight inspection system?
Enlight is Applied Materials’ optical wafer inspection platform, with Enlight 3 as the current generation. The company’s materials describe it covering brightfield and greyfield imaging in one system, reaching 30 wafers per hour in darkfield applications, and generating the defect candidates that SEMVision review resolves.
What is PROVision?
PROVision is Applied Materials’ e-beam inspection line, distinct from SEMVision’s review role: inspection finds defects, review resolves them. PROVision 3E and PrimeVision 10 use e-beam imaging to find defects optical tools cannot see, including buried defects and electrical faults surfaced through voltage contrast on test and product structures.
Does Applied Materials make mask inspection tools?
Yes. The Aera family, currently Aera 6, inspects photomasks, covering the qualification gates mask shops run before a reticle prints and during requalification. Mask inspection has its own methods, comparison modes and defect classes, which are covered in depth in this hub’s photomask inspection guide.
How do Applied Materials and KLA compare in inspection?
KLA is the long-standing leader in optical patterned-wafer inspection, and Applied Materials leads in e-beam review while expanding into optical, a push its executives described to investors in 2026. Most fabs run tools from both, selected per gate on sensitivity, throughput and cost of ownership rather than by vendor.
Applied Materials’ inspection portfolio covers the optical, e-beam and mask sides of fab process control. Enlight optical inspection finds defect candidates at speed, SEMVision e-beam review resolves and classifies them, PROVision e-beam inspection hunts what optics cannot see, Aera inspects photomasks, and ExtractAI links optical detection to e-beam review in real time. Together they form one of the two major inspection ecosystems in modern fabs, alongside KLA’s.
These tools run across the inspection gates of Stage 03 and Stage 04 of the 12-stage semiconductor manufacturing process. This guide from Averroes.ai, an AI visual inspection platform for manufacturers, maps the portfolio to the line, explains the AI Applied Materials builds into it, and shows where a cross-fleet AI layer fits alongside it.
Key Notes
The Applied Materials Inspection Portfolio, Mapped to the Line
Applied Materials’ inspection and metrology tools are the process-control arm of the industry’s largest equipment maker, spanning optical inspection, e-beam inspection and review, mask inspection and CD metrology. The portfolio’s organizing idea is a division of labor: inspection tools find defect candidates fast, and review tools resolve and classify what inspection found. Every tool below sits on one side of that split.
Applied Materials inspection tools: what each does and where it runs
The map makes one point: the portfolio is not a list of rivals but a relay, and each handoff in it happens at a gate this hub covers.
The optical line. Enlight 3 is the current optical inspection platform, and Applied Materials’ product materials describe it as covering brightfield and greyfield imaging in one system, with throughput reaching 30 wafers per hour in darkfield applications. It is the candidate-finding end of the relay, running the inline gates where bright field inspection and its darkfield counterpart earn their keep.
The review line. SEMVision is the resolving end. Applied Materials describes the family as the most widely deployed e-beam review line in the industry, and its newest member, the SEMVision H20 introduced in March 2026, pairs the company’s second-generation cold field emission source with deep-learning image models, delivering review results the company states are up to 3X faster than prior techniques.
The e-beam inspection line. PROVision 3E and PrimeVision 10 do something different from review: they are inspection tools that use e-beam inspection physics to find defects optical tools cannot see at all, including buried and electrically significant defects surfaced through voltage contrast.
Mask and metrology. Aera 6 carries mask inspection, the qualification gate the photomask inspection guide covers end to end, and VeritySEM 10 carries CD-SEM metrology. For advanced packaging, the VeritySEM 7AP and SEMVision G7AP pair extends the same find-and-resolve logic to bumps, TSVs and bonding, where Applied Materials’ own materials describe optical tools flagging defects that the e-beam system then resolves and classifies.
The AI Inside the Tools: ExtractAI and the H20
ExtractAI is Applied Materials’ machine learning technology that links Enlight optical inspection to SEMVision e-beam review, using review results to classify the optical defect map in real time. The company’s Enlight 3 materials describe it, now in its third generation, as the intelligent link between the two platforms.
The problem it exists to solve is the industry’s oldest trade restated at modern density. Raising optical sensitivity raises nuisance detections with it, and Applied Materials states that today’s advanced nodes produce about 100X more defect candidates for e-beam review than earlier generations. Review capacity cannot grow 100X, so something has to decide which candidates deserve it. ExtractAI makes that call on the optical side; the SEMVision H20 adds deep-learning extraction of true defects from nuisance on the review side, which is the capability its March 2026 launch coverage led with, quoting the company on helping chipmakers distinguish signal from noise.
One sentence of context belongs here. The industry’s largest equipment maker now builds AI into both ends of its inspection pair, which confirms that separating real defects from nuisance is the central problem of modern inspection, the same problem every gate page in this hub traces from coat to final test.
Architecturally, the design is a closed loop by intent: ExtractAI runs between Applied Materials’ optical tool and Applied Materials’ e-beam reviewer, trained and tuned within that stack. That is a strength inside the pair, and it defines the boundary of what the embedded AI sees, which matters in the last section.
Applied Materials vs KLA in Inspection
Applied Materials and KLA divide fab inspection along historical lines: KLA is the long-standing leader in optical patterned-wafer inspection, while Applied Materials leads in e-beam review and has been expanding into optical. Neither company’s position is in serious dispute, and the 2026 numbers describing the contest come from Applied Materials’ own executives: investor coverage from October 2026 reports the company’s process control business heading toward roughly 3 billion dollars for the year, serving about 2 billion dollars of a 9 billion dollar optical inspection market, with its imaging group head, a 27-year KLA veteran, telling investors that customers want alternatives.
Those are company statements about its own momentum, and this page takes no side. The practical reality for a fab is simpler and more useful than the market-share contest: most fabs run both. Tool selection happens per gate, on sensitivity, throughput and cost of ownership for that layer’s defects, and a line that runs KLA optical inspection at one gate, Applied Materials review behind it, and an Onto tool at a third is not an exception but the norm that wafer inspection programs are actually built from.
That norm is the fact the rest of this page turns on. Each vendor’s embedded AI is excellent inside its own fleet and sees nothing outside it.
Where a Cross-Fleet AI Layer Fits
A cross-fleet AI layer classifies the defect images and maps that inspection tools from every vendor already produce, so one trained judgment applies across an Applied Materials, KLA and Onto fleet at once. The logic follows from the mixed-fleet norm: embedded AI travels with its vendor’s tools, but a fab’s defect classes travel with the fab. The same scratch, residue or opened void crosses every tool in the line, and a classification standard that changes at each vendor boundary means the fab’s false-positive rate is really three different rates stitched together.
The diagram makes one point: the two architectures answer different questions, one optimizing a pair, the other unifying a fleet.
Averroes.ai is that second architecture. It runs as an overlay on the inspection images and defect maps that Applied Materials, KLA and Onto tools already produce, trains AI defect detection and classification models on the fab’s own defect classes from 20 to 40 labeled images per class, with no code, and deploys on premise so image data stays inside the fab. It replaces nothing. The tools, their recipes and their embedded AI keep doing exactly what the fab bought them for, and ExtractAI keeps optimizing its pair; the overlay adds the judgment layer that spans the whole fleet, so one taxonomy and one false-positive standard hold from the first gate to the last, with defect classification sorting confirmed detections into the fab’s own classes wherever they were imaged.
Running a Mixed Inspection Fleet?
See AI defect detection on the images your Enlight and SEMVision tools already produce
Frequently Asked Questions
What is SEMVision?
SEMVision is Applied Materials’ family of e-beam defect review tools, which the company describes as the most widely deployed in the industry. The current flagship, the SEMVision H20 introduced in March 2026, combines second-generation cold field emission with deep-learning image models, and Applied Materials states it reviews defects up to 3X faster.
What is ExtractAI?
ExtractAI is Applied Materials’ machine learning technology connecting Enlight optical inspection to SEMVision e-beam review. It uses the results of e-beam review to classify the optical defect map in real time, so the system directs review capacity at the detections most likely to be real defects. It is in its third generation.
What is the Enlight inspection system?
Enlight is Applied Materials’ optical wafer inspection platform, with Enlight 3 as the current generation. The company’s materials describe it covering brightfield and greyfield imaging in one system, reaching 30 wafers per hour in darkfield applications, and generating the defect candidates that SEMVision review resolves.
What is PROVision?
PROVision is Applied Materials’ e-beam inspection line, distinct from SEMVision’s review role: inspection finds defects, review resolves them. PROVision 3E and PrimeVision 10 use e-beam imaging to find defects optical tools cannot see, including buried defects and electrical faults surfaced through voltage contrast on test and product structures.
Does Applied Materials make mask inspection tools?
Yes. The Aera family, currently Aera 6, inspects photomasks, covering the qualification gates mask shops run before a reticle prints and during requalification. Mask inspection has its own methods, comparison modes and defect classes, which are covered in depth in this hub’s photomask inspection guide.
How do Applied Materials and KLA compare in inspection?
KLA is the long-standing leader in optical patterned-wafer inspection, and Applied Materials leads in e-beam review while expanding into optical, a push its executives described to investors in 2026. Most fabs run tools from both, selected per gate on sensitivity, throughput and cost of ownership rather than by vendor.