China Chip Tooling Talent Monitor

China · Semiconductor equipment

China Chip Tooling Talent Monitor

What public records reveal about the people and organizations behind China's semiconductor-equipment push.

The monitor tracks firms, city clusters, role families, and source records tied to etch/clean, deposition, metrology/inspection, and lithography-adjacent tooling. It is built for analysts tracking how Chinese toolmakers move from R&D headcount to customer-site capability.

The dataset

What the dataset catalogues today. Counts measure record coverage.

Sources
43

Distinct public records currently catalogued.

Non-taxonomy evidence rows
143

Observations tied to filings, directories, policy, or proxies.

Taxonomy scaffold rows
65

Structural placeholders that organise the monitor, not measurements.

Core tooling segments
4

Etch and clean, deposition, metrology and inspection, lithography sidebar.

Findings01–04

What the public record shows, and where it thins out

Four findings from the current dataset, with the source pattern behind each and how it should change a reader’s priors.

  1. Finding 012 sources cited

    Thirty-eight percent of rows are structural scaffolding

    Source pattern
    Sixty-five of 172 rows map disciplines, roles, and segments. They make the monitor navigable, but they do not show firms hiring, training, or deploying people.
    Analyst read
    Broad STEM counts cannot tell a toolmaker where Chinese competitors are building production support.
  2. Finding 023 sources cited

    Employer disclosures carry the strongest public signal

    Source pattern
    AMEC, ACM Research Shanghai, and NAURA publish workforce categories that expose R&D, technical, and service structure better than most education data.
    Analyst read
    The customer-site support work between R&D and the customer fab deserves more attention than raw graduate output.
  3. Finding 033 sources cited

    Shanghai and Beijing dominate the visible record

    Source pattern
    Shanghai leads the city view, followed by Beijing. The concentration reflects firm filings, park records, and shortage notices. It does not map the full labor market.
    Analyst read
    Shanghai-area filings, Beijing park records, and a small set of coastal shortage notices are the first stops for a primary-source check.
  4. Finding 042 sources cited

    The decisive know-how sits behind the public record

    Source pattern
    Chamber recovery, field calibration, tool matching, and customer-ramp support rarely appear in public records, even when firms publish product lines and R&D counts.
    Analyst read
    Production-support signals, including service hiring, field application language, and customer-ramp staffing, carry more weight than product-line announcements.
Exhibit AGeography of visibility

Visible evidence concentrates in a handful of cities

Public records cluster around Shanghai, Beijing, and a small set of coastal hubs. The cluster shows coverage, not capability depth.

City source coverage · province outline

Source coverage by city

The map shows where this beta dataset has public records. Node size tracks source-backed records, not engineers, workforce size, or city capability.

Node size = source-backed records in this datasetProvince outline for orientation. City placement is approximate.
Exhibit BWhat the work looks like

Tooling talent, by segment

Each tool family pulls on a different mix of roles, bottlenecks, and public signals. STEM totals don’t separate them.

Capability and role matrix

Tooling talent, by segment

Equipment work depends on more than STEM headcount. Each tool family relies on a specific mix of role families, process-window judgement, and customer-ramp know-how. Public records expose each piece unevenly.

Segment

Etch, clean, and strip

Tools and process work that remove material, residues, or photoresist without damaging the wafer pattern.

4 capabilities · 4 role families catalogued

Key role families

  • Plasma and surface scientists
  • Etch and clean process engineers
  • Field application engineers
  • Maintenance and debug technicians

Likely bottlenecks

  • High-aspect-ratio profile control with low damage
  • Post-etch residue removal without pattern loss
  • Chamber matching and recovery after maintenance

Public signals

  • Etch and clean product pages
  • CNINFO annual-report disclosures
  • Process and application job language
  • Patents and technical papers on plasma or wet clean

What to watch

  • Repeat references to advanced memory and high-aspect-ratio use cases
  • Service and application-team expansion
  • Training language tied to customer-site ramps

Caveat

Public sources expose product families and role language better than yield-learning depth.

Segment

Deposition

Tools and process work that grow or place thin films with the right thickness, composition, coverage, and stability.

4 capabilities · 4 role families catalogued

Key role families

  • Thin-film scientists
  • ALD, CVD, and PVD process engineers
  • Vacuum and gas-delivery engineers
  • Customer validation and service engineers

Likely bottlenecks

  • Conformal film growth at production-worthy throughput
  • Precursor and chamber behavior under high-volume use
  • Particle control and chamber matching

Public signals

  • ALD, CVD, PVD, and furnace product coverage
  • Annual-report language on scale and R&D intensity
  • Film-process and service job language
  • Patents on chamber architecture and gas delivery

What to watch

  • Batch or vertical tool claims tied to throughput
  • Customer validation and repeat-order language
  • Hiring or training for field support and application work

Caveat

Repeat orders and product breadth help, but they do not reveal full process-window or maintenance sensitivity.

Segment

Metrology and inspection

Tools and workflows that measure, inspect, classify, and feed defect or process information back into production.

4 capabilities · 4 role families catalogued

Key role families

  • Optical and imaging scientists
  • Algorithm and data engineers
  • Metrology application engineers
  • Calibration and field-support specialists

Likely bottlenecks

  • Optics-plus-algorithms integration
  • False-alarm reduction without missing real defects
  • Fab-specific calibration and defect-library tuning

Public signals

  • Optical, algorithm, and system-engineering job language
  • Metrology product filings and product pages
  • Service-team and support-network disclosures
  • Research-output proxies in optics and instrumentation

What to watch

  • Evidence of application engineers at customer sites
  • Training or support language around calibration
  • Product claims that link measurement to yield workflows

Caveat

Customer datasets, nuisance-signal handling, and calibration routines are usually private.

Segment

Lithography sidebar

A sidebar category for optics, alignment, exposure, focus, and process-window signals adjacent to lithography bottlenecks.

2 capabilities · 4 role families catalogued

Key role families

  • Optical scientists
  • Stage, control, and mechatronics engineers
  • Lithography process engineers
  • Optical assembly and calibration technicians

Likely bottlenecks

  • Opto-mechatronic integration for overlay and focus
  • Contamination discipline and calibration recovery
  • Process-window tuning under customer conditions

Public signals

  • Optics and control job language
  • Lithography product and service pages
  • University optics and instrumentation signals
  • Policy and park references to IC equipment ecosystems

What to watch

  • Specialized optics hiring
  • Service and training infrastructure
  • Connections between lithography, metrology, and precision motion

Caveat

This sidebar provides context. Public evidence in this dataset cannot measure lithography readiness.

Exhibit CEmployer-side evidence

Employer disclosures anchor the public record

Three listed firms publish enough workforce structure to anchor the picture. Categories describe firms, not segments.

Firm workforce snapshots

What AMEC, ACM Research Shanghai, NAURA, and Piotech publish about their workforce

These categories come from each firm's filing as published. Firms use different labels and denominators, and the figures describe whole-firm staffing. Tooling-segment headcount would need separate disclosure.

Shanghai

AMEC

  • Etch, clean, and strip
  • Deposition

AMEC is the clearest public anchor for the etch story, with disclosed R&D staffing that points to a science-heavy equipment organization.

R&D personnel
1,548

Denominator: Firm-level R&D personnel headcount.

The filing does not split this total by etch, deposition, service, or customer-ramp work.

Source checked
R&D personnel, share of total staff
52.24%

Denominator: Total staff, as disclosed in the filing.

Use as an R&D intensity signal. Segment-specific staffing would need a separate disclosure.

Source checked
Doctoral researchers in R&D
280

Denominator: R&D personnel.

Filing-disclosed count of PhD holders inside the R&D layer.

Source checked
Master's researchers in R&D
616

Denominator: R&D personnel.

Filing-disclosed count of master's holders inside the R&D layer; together with the 280 PhDs this gives the 57.88% master's-or-doctoral share within R&D.

Source checked
Doctoral degree holders, company-wide
291

Denominator: Company-wide.

Filing-disclosed absolute count; the checked filing sections do not disclose a company-wide total-employee figure, so a share cannot be computed.

Source checked
Master's degree holders, company-wide
989

Denominator: Company-wide.

Filing-disclosed absolute count; the checked filing sections do not disclose a total-employee denominator.

Source checked
After-sales / field-service headcount
Not disclosed

Denominator: —

The checked filing sections do not break out an after-sales or field-service category.

Not disclosed

Limit · Filing figures are firm-level. They do not split people by etch, deposition, service, or customer-ramp work.

Source

  • AMEC 2025 annual report · CNINFO / AMEC

Shanghai

ACM Research Shanghai

  • Etch, clean, and strip
  • Deposition

ACM Research Shanghai makes the wet-clean and strip workforce layer more visible than most firms, especially through technical and service-heavy disclosures.

Total employees
2,485

Denominator: Whole firm.

Company-level total; it cannot be read as clean/strip-specific headcount.

Source checked
Technical personnel
1,228

Denominator: Technical personnel, as disclosed by the firm.

Technical-staff category as filed; do not relabel as R&D.

Source checked
Technical personnel, share of total employees
49.42%

Denominator: Total employees, as disclosed by ACM Research Shanghai.

Share signal; do not add to R&D or service categories.

Source checked
After-sales service personnel
672

Denominator: Whole firm.

Filing-disclosed after-sales / service category; the only clean post-sale headcount among the four firms in this dossier set.

Source checked
Master's degree or above, company-wide
708

Denominator: Total employees.

Filing-disclosed absolute count of master's-and-above across the firm.

Source checked
Master's or doctoral share within technical personnel
Not disclosed

Denominator: Technical personnel.

The filing reports master's-and-above company-wide but does not break that group down inside the technical-personnel category.

Not disclosed

Limit · Role categories are aggregated and cannot be read as clean/strip-specific headcount.

Source

  • ACM Research Shanghai 2025 annual report · CNINFO / ACM Research Shanghai

Beijing

NAURA Technology Group

  • Deposition
  • Etch, clean, and strip

NAURA gives the strongest scale signal among the listed equipment firms in this layer, but its breadth makes segment attribution especially risky.

Total employees
21,101

Denominator: Whole firm.

Broad equipment group workforce figure.

Source checked
Production personnel
8,065

Denominator: Total employees.

Filing-disclosed production category; spans the full equipment group.

Source checked
Sales and customer-service personnel
3,950

Denominator: Total employees.

Filing-disclosed sales-and-customer-service category; the sales-expense increase was partly attributed to growth in this team.

Source checked
Technical personnel
6,511

Denominator: Total employees.

Filing-disclosed technical-staff category; equals the R&D-personnel figure in this filing.

Source checked
R&D personnel
6,511

Denominator: Total employees.

Firm-level R&D scale signal for the broad equipment group.

Source checked
R&D personnel, share of total employees
30.86%

Denominator: Total employees.

R&D intensity signal. Filing role labels and job-posting role labels do not map cleanly.

Source checked
Doctoral researchers in R&D
268

Denominator: R&D personnel.

Filing-disclosed count of PhD holders inside the R&D layer.

Source checked
Master's researchers in R&D
4,137

Denominator: R&D personnel.

Filing-disclosed count of master's holders inside the R&D layer.

Source checked
Master's degree or above, company-wide
6,271

Denominator: Total employees.

Company-wide count of master's-and-above degree holders.

Source checked
Headcount by tool family
Not disclosed

Denominator: —

The filing does not split production, R&D, or service headcount by tool family within the equipment group.

Not disclosed

Limit · The filing covers a broad equipment group, so the figures should frame scale and R&D intensity. Tooling-segment capacity would need a separate disclosure.

Source

  • NAURA 2025 annual report · CNINFO / NAURA Technology Group

Shenyang

Piotech

  • Deposition

Piotech is the dedicated deposition vendor in this dossier set; the 2025 financing report supplies its first source-checked workforce figures.

R&D personnel (as of 2025-06-30)
638

Denominator: Total staff at the reporting date.

Filing-disclosed R&D personnel; the financing report is dated September 2025 and reports the position as of 30 June 2025.

Source checked
R&D personnel, share of total staff
40.66%

Denominator: Total staff at the reporting date.

R&D intensity signal for a deposition-focused vendor.

Source checked
Doctoral researchers in R&D
53

Denominator: R&D personnel.

Filing-disclosed count of PhD holders inside the R&D layer.

Source checked
Master's researchers in R&D
384

Denominator: R&D personnel.

Filing-disclosed count of master's holders inside the R&D layer.

Source checked
After-sales / installed-base support headcount
Not disclosed

Denominator: —

The financing report does not break out a full after-sales or installed-base support category.

Not disclosed

Limit · Piotech's disclosures come from a financing report. Treat product and demand framing as corporate filing evidence.

Source

  • Piotech 2025 financing report · CNINFO / Piotech

Firms use different workforce categories. R&D share, technical staff, service staff, and advanced-degree counts should stay separate.

  • Source checked
  • Needs check
  • Staging
  • Not disclosed
ImplicationsWho this is for

Who this is for, and where the evidence stops

Equipment firms

Track specific role combinations in public records.

Read where role families, segment bottlenecks, and firm or city evidence overlap. Single-axis counts will mislead.

Use the explorer for

  • Comparing public records by segment
  • Finding where service, application, and production roles become visible
  • Separating R&D visibility from customer-ramp know-how

Caution · Do not read public evidence strength as a hiring pipeline or competitor capability score.

Policymakers

Shortage lists track pressure across local directories.

Local directories and park plans help identify where public institutions see pressure, but they rarely measure tacit production learning.

Use the explorer for

  • Checking which cities disclose shortage or cluster signals
  • Seeing which segments rely on proxy evidence
  • Identifying evidence gaps before funding or training claims harden

Caution · Avoid converting qualitative shortage signals into exact workforce targets without separate validation.

Researchers / analysts

The decisive variable is usually missing from the public record.

Public data is best at showing institutions, filings, and role language; it is weaker at showing chamber recovery, calibration habits, and customer-specific learning.

Use the explorer for

  • Tracing each observation back to source_id
  • Distinguishing direct public records from analytical proxies
  • Building a verification queue for the most visible claims

Caution · Treat observations.csv as staging until the visible claims are checked against primary sources.

Exhibit DEvidence ladder

The 172 rows sit in separate evidence tiers

Three tiers sit under the headline count. Mixing them inflates what the public record shows.

Evidence ladder

Three tiers of evidence

The monitor combines direct public records, analytical proxies, and taxonomy scaffolding. Each tier is reported separately.

  1. Tier 01

    Load-bearing

    Strongest analytical weight

    Direct public record

    Filings, official directories, industry presence, and shortage notices. These are disclosures that exist on the public record and can be re-checked at source.

    • Industry presence77
    • Shortage signal17
    • Institutional capacity14
    • Official policy3

    Rows in tier

    111

  2. Tier 02

    Supporting

    Use as context, not as measurement

    Analytical proxy

    Job postings, research-output indicators, and expert secondary sources. Directional context that helps frame the picture but should not be read as a measurement.

    • Research output proxy16
    • Expert secondary source12
    • Job posting proxy4

    Rows in tier

    32

  3. Tier 03

    Navigational

    Scaffolding only, no real-world signal

    Taxonomy scaffold

    Manual inferences that link MOE disciplines and structural categories to tooling segments. They make the dataset navigable and should not be read as real-world activity.

    • Manual taxonomy scaffold65

    Rows in tier

    65

    Navigational scaffolding. This tier does not show real-world activity on its own.

Counts reflect the current observation set. The tiers stay separate and stay out of composite scoring.

Exhibit EMonitoring brief

Four signals worth monitoring

For corporate and policy readers: what to watch, what the stronger version of each signal looks like, and where the public record runs out.

Corporate and policy briefing

What to watch next

Corporate and policy users need more than product announcements. Stronger signals show whether a firm is building the people and routines that move a tool from lab result to customer-site operation.

Watch question

Firms hiring for fab support

Watch signal

Listed toolmakers disclosing growth in R&D, technical, after-sales, service, or customer-support categories in annual filings.

Stronger signal

Growth concentrated in customer-site categories: field application, service training, and after-sales. Those are the categories where broad engineering supply converts into deployed customer-site teams.

Do not infer

Whole-firm workforce categories are firm-level. They do not show segment-specific headcount.

Compare firm dossiers→

Watch question

Customer-site support

Watch signal

Product announcements naming a new tool family, a customer, or a process win.

Stronger signal

Language about field application engineers, customer validation, calibration, repeat orders, or service training tied to a specific tool family. These are the routines that move a tool from lab result to fab operation.

Do not infer

An announcement does not show ramp success, installed-base quality, or yield. Service language confirms the function shows up in public records; it does not measure customer-site depth.

Open ACM Research Shanghai dossier→

Watch question

Scarce role combinations

Watch signal

Hiring or training in single disciplines: plasma, optics, motion control, ALD chemistry, algorithms.

Stronger signal

The combinations are harder than the parts: plasma plus chamber hardware, ALD chemistry plus vacuum behavior, optics plus algorithms, precision motion plus calibration. STEM totals do not separate these combinations.

Do not infer

Aggregate graduate output does not show whether a firm has assembled the right cross-discipline teams.

See the optics-plus-algorithms case→

Watch question

City clusters worth checking first

Watch signal

Aggregate city evidence counts. Shanghai and Beijing dominate the visible record.

Stronger signal

Inside those clusters, pull listed-firm filings, park disclosures, shortage notices, and policy directives before raw evidence totals.

Do not infer

City evidence concentration shows visibility coverage. Sparsely covered cities may still have material activity.

Open Shanghai rows in explorer→
Continue readingPick a thread

Pick how you want to dig in

Segments covered: Etch, clean, and strip, Deposition, Metrology and inspection, Lithography sidebar.

Interactive

Explore evidence

Filter the observation set by segment, evidence type, and city. Each row links back to its source.

Open explorer →

Reference

Methodology

How observations are collected, classified, and what the evidence categories mean.

Read methodology →

Audit

Source ledger

Every public record behind the monitor, with publisher, type, reuse value, and retrieval dates.

Open ledger →

Editorial evidence product. Counts describe public-record coverage; observation rows remain in the beta dataset until manually verified against the underlying source. Mainland PRC only.