Machine vision is a category of industrial hardware + software systems that use cameras, lighting, and algorithms to inspect products on manufacturing lines, guide robot arms, read barcodes, and detect defects. The unit of demand is a vision system — typically a camera module bundled with processing hardware and software sold to a factory. The global machine vision market was approximately $12.6 billion in 2025, growing at roughly 9% annually toward a projected ~$27 billion by 2034 est. (Fortune Business Insights). The three US-listed tickers with exposure are Cognex (CGNX), the US pure-play leader in factory inspection; Teradyne (TER), which owns the cobot leader Universal Robots and touches vision through its robotics and test divisions; and Keysight Technologies (KEYS), a test-and-measurement company whose Electronic Industrial Solutions segment serves manufacturing inspection. Cognex derives substantially all of its revenue from machine vision. TER and KEYS are diversified — machine vision is a minority of each business.
The business model is selling inspection hardware (cameras, sensors, lighting) plus software licenses to factories. Revenue is driven by new factory construction, automation retrofits, and quality-control upgrades. Deep learning replaces hand-crafted inspection rules, expanding addressable use cases from structured defect detection to universal visual QA — but this is a software upgrade layered onto existing camera hardware, not a step-function surge in camera unit demand.
A machine vision system is a camera (2D or 3D) combined with lighting, optics, and processing — either a self-contained "smart camera" with onboard compute, or a PC-based system where cameras feed images to an external processor running inspection software. The system is installed on a production line and runs continuously, inspecting every item at speeds that often exceed one part per second.
The money comes in two streams. First: hardware sales. A customer buys cameras, lighting modules, and sometimes a controller. A single smart camera from Cognex (the In-Sight series) typically costs $2,000–$10,000 depending on resolution and processing power; a full multi-camera line inspection system can run $50,000–$500,000+ installed est.. Systems are replaced or upgraded every 5–10 years est. as technology improves or production lines change. Second: software and services. Cognex's Edge Intelligence platform and VisionPro deep-learning tools carry license fees. The company is shifting toward a unified software platform to reduce fragmentation and increase stickiness.
Cognex's 67% gross margin (FY2025, 10-K) reflects the blend of hardware (lower margin) and software (higher margin). The premium over a commodity camera maker (30–40% margins est.) comes from proprietary algorithms, application-specific calibration, and switching costs — once a factory has trained a Cognex system to inspect a specific product, replacing it requires re-validating all inspection logic.
For Teradyne, machine vision is embedded within its Robotics segment (Universal Robots cobots) and its test equipment, where vision sensors guide robots and verify product quality. This is not a separable revenue line — it is a feature bundled into a $25,000–$50,000 cobot arm est. or a multi-million-dollar test cell. For Keysight, the Electronic Industrial Solutions Group (EISG) sells test and measurement instruments used in manufacturing — some incorporating visual inspection — but Keysight is fundamentally a test instrument company, not a vision company.
Cognex product info: cognex.com; system pricing: industry estimates; gross margin: CGNX FY2025 10-K; Teradyne cobot pricing: Universal Robots published base prices; Keysight EISG: KEYS FY2025 10-K.
Machine vision systems are not sold on long-term contracts. Customers place orders on a project basis — when building a new production line, retrofitting an existing one, or responding to quality problems. There is no public order backlog for Cognex or the vision portion of TER/KEYS. Cognex management describes "limited visibility" beyond one quarter (Q1 2026 earnings call, May 2026).
Cognex reported Q1 2026 revenue of $268M, up 24% year-over-year, and guided Q2 2026 to $280–$300M (~17% growth at midpoint). End-market trends from that call: logistics recovering, packaging growing mid-to-high single digits, semiconductor growing high single-to-double digits, consumer electronics growing high single-to-double digits, automotive flat-to-low single digits.
Cognex Q1 2026 earnings: MarketBeat, May 6, 2026; Q2 guidance: same source.
The global machine vision market is projected to grow from ~$12.6B (2025) to ~$27B by 2034, a CAGR of roughly 9% est. (Fortune Business Insights). The semiconductor end market is the largest single segment, projected at ~35% of demand by 2026 est., driven by automated wafer and die inspection. Automotive (ADAS cameras, EV battery inspection) is the second-largest at ~24% est..
Deep learning replaces hand-coded inspection rules with learned representations: the system learns what "good" looks like and flags anomalies without explicit programming for each defect type. This expands the addressable market from structured inspection (known defects on known products) to universal visual QA. The dollar uplift is primarily in software (license fees for AI models), not hardware units.
The other demand lever is broader factory automation. Every robot arm needs eyes. As cobot and industrial robot deployments grow — driven by reshoring, labor shortages, and increasingly capable manipulation — each new robot cell typically includes one or more vision systems. This demand grows with robot deployments, which follow industrial cycles (mid-single-digit annual growth est., punctuated by surges when robots become capable of new tasks).
Market size: Fortune Business Insights, Machine Vision Market Report (2025); sector context: 500-stocks sector scan, Robotics & Automation, Section 08.
Machine vision supply is not constrained (sector scan assessment). The hardware components — CMOS image sensors, FPGAs, LEDs, lenses — are commodity electronics available from multiple suppliers (Sony, ON Semiconductor, ams-OSRAM). Assembly is straightforward. There are no multi-year lead times, no scarce raw materials, and no government permitting bottlenecks.
Cognex operates with minimal capex: $8.7M in capital expenditure on $994M of revenue in FY2025 (10-K) — capex intensity under 1%. The company outsources manufacturing and can scale production by placing more orders with contract manufacturers. Keyence (Japan-listed, not one of the three tickers but the #1 or #2 global player) runs a similarly asset-light model with operating margins above 50% est..
Teradyne's Robotics segment (Universal Robots) manufactures cobots in Denmark and more recently in the US. Cobot production is not capacity-constrained — UR can ramp by adding shifts and expanding factory footprint over quarters, not years.
The binding constraint is customer adoption speed, not physical supply. Factories are conservative; production-line changes require validation, integration, and downtime. A secondary constraint is the application engineering talent needed to deploy vision systems — each installation requires calibrating lighting, training models, and integrating with factory control systems.
Chinese competitors (Hikrobot/Hikvision subsidiary, Daheng Imaging, MindVision) are undercutting Western vendors on price in Asia-Pacific, the fastest-growing region (~40% of global demand, ~10% CAGR est.). Premium vendors retain advantages in software, support, and regulatory trust in Western markets.
Supply assessment: 500-stocks scan, Section 08; Cognex capex: FY2025 10-K; Chinese competition: Mordor Intelligence Machine Vision report (2025).
There is no structural demand-supply gap in machine vision hardware. Vision system hardware can be manufactured and shipped within weeks to months. Supply can match demand without a multi-year lag.
The relevant gap is in software and AI capability. The companies that develop the best deep-learning inspection algorithms — generalizing across defect types with minimal training data — capture pricing power. This is a competition-driven gap (who has the best model), not a supply-driven gap (who can make enough units).
Pricing direction: hardware prices face downward pressure from Chinese entrants and general deflation of CMOS sensors and compute. Software and AI-based inspection pricing faces upward pressure — customers pay more for systems requiring less engineering setup and catching more defects. Cognex's seven consecutive quarters of margin expansion (through Q1 2026) are consistent with this mix shift.
| Factor | Demand side | Supply side |
| Current size est. | ~$12.6B global (2025) | Unconstrained; asset-light manufacturing |
| Forward change | ~9% CAGR to ~$27B by 2034 (forecast) | Production can scale in weeks/months |
| Speed of response | Demand limited by factory adoption cycles | Supply responds within a quarter |
| Constraint | Customer capex budgets, integration timelines | Application engineering talent; Chinese price competition |
| Net direction (per scan) | No supply shortage; value accrues to software/AI, not hardware volume | |
Market projections: Fortune Business Insights; gap assessment: 500-stocks scan, Section 08.
The three tickers have very different levels of machine-vision purity.
| Ticker | What it sells | Vision exposure | FY2025 Rev | Mkt Cap (Jun 2026) | Gross Margin | Debt |
| CGNX | Machine vision cameras, sensors, software for factory inspection | ~100% pure play | $994M | $11.0B | 67% | $0 (net cash $642M) |
| TER | Semiconductor test equipment + cobots (Universal Robots) + industrial automation | ~10–12% est. (Robotics segment $365M includes cobots + vision; semi test is 66% of revenue) | $3.19B | $64.1B | ~58% | Low; $294M cash at year-end |
| KEYS | Test & measurement instruments (RF, 5G, AI data center, general electronics) | ~5–10% est. (EISG includes some industrial inspection; CSG dominates at ~70% of revenue) | $5.38B | $59.9B | ~65% | Moderate; ~$2B acquisition spend in FY2025 |
Not listed but relevant: Keyence (Japan: 6861.T) has operating margins above 50% est. and a market cap above $100B. Omron (Japan: 6645.T) is the third member of the ~35% combined share trio. Basler (Germany), Teledyne/FLIR, and Chinese players (Hikrobot, Daheng) compete but are not in this ticker set.
CGNX financials: FY2025 10-K, Q1 2026 earnings; TER: FY2025 press release, Q4 segment data; KEYS: FY2025 10-K, Q2 FY2026 earnings; market caps: StockAnalysis.com, Jun 3 2026; Keyence: public filings (Japan).
Market cap $11.0B on FY2025 revenue of $994M and net income of $114M. That is ~96x trailing earnings and ~11x trailing revenue. Free cash flow (operating cash flow minus capex) was $237M in FY2025, putting the price at ~46x trailing FCF. The company carries zero debt and $642M of net cash. Trailing 12-month FCF as of Q1 2026 was $241M, up ~50% year-over-year.
Revenue is accelerating: Q1 2026 was +24% YoY, Q2 guidance implies ~17% growth. If the full year reaches ~$1.15B est. (annualizing recent quarters), the forward revenue multiple compresses to ~9.5x. Adjusted EPS grew 113% YoY in Q1 2026 on operating leverage.
Market cap $64.1B on FY2025 revenue of $3.19B and GAAP net income of $554M. That is ~116x trailing earnings and ~20x revenue. The stock has risen ~391% in the past year, driven by AI-related semiconductor test demand. The Robotics segment ($365M in FY2024, ~$340M run-rate in FY2025 est.) is roughly 11% of revenue and has been shrinking (down 3% in FY2024, with 400-employee layoffs in FY2025). Machine vision is a subset of the Robotics segment, not separately reported. The company repurchased $702M in stock in FY2025.
Market cap $59.9B on FY2025 revenue of $5.38B and net income of $850M. That is ~70x trailing earnings and ~11x revenue. The stock has risen ~113% in one year, driven by AI data center test demand (Communications Solutions Group, ~70% of revenue). EISG generated ~$486M in Q2 FY2026 (up 24% YoY) but is not broken down further into vision-specific revenue.
All multiples computed from: CGNX FY2025 10-K + Q1 2026 earnings; TER FY2025 press release; KEYS FY2025 10-K + Q2 FY2026 earnings; market caps: StockAnalysis.com, Jun 3 2026.
What was used:
Hard vs. approximate: