Target
$74.39 → $74.31
Public companyResearch currentFresh
Cognex
Machine Perception · Physical AI
SignalHOLDReviewed 26 Oct 2026
Machine Perception exposure in Physical AI.
Cognex Corporation (CGNX) is the dominant global supplier of industrial machine vision systems—the **perception bottleneck layer of the Physical AI stack**. The company manufactures 2D/3D camera systems, vision software (VisionPro, In-Sight AI), barcode readers, and edge AI vision platforms used to automate quality control, inspection, and guidance across factory, semiconductor fab, and warehouse automation environments. The thesis is that every robot, autonomous mobile robot (AMR), and industrial automation system requires embedded vision to "see" its…
Catalysts
Continued order recovery plus AI-driven 3D inspection adoption into the FY26 $1.64-1.68 guide.
Invalidation
Robotics deployment slower than expected; commodity camera vendors undercut
Latest change
· Target moved from $74.31 to $74.28.
Research history
See all CGNX changes Earlier changes
Capital in context
Mapped funds and investors
Direct relationships supported by published evidence.
Reported holdings as-of
| Fund / investor | As of | Fund portfolio weight | Evidence |
|---|---|---|---|
| Global X Robotics & Artificial Intelligence ETF BOTZ | 2.51% | Source |
Disclosed investment history
No sourced relationships mapped yet.
Showing up to four relationships in each group. ETF weights describe the fund’s portfolio, not ownership of this company. Private investment history does not establish current ownership.
ins-$4.2M · 13F 17+/8-×0.0 · short↑0.23
Current snapshotX: limited; machine vision bottleneck confirmed by industrial automation engineers
Lower-confidence snapshotTarget price: $75.37
Dated research targetMoat & position
- Evidence Grade
- Verified
- Moat Durability
- Durable IP
- Toll Road Tier
- DIRECT_BOTTLENECK
Timestamped research update
Deep research update
5,877 words · Updated 20 Oct 2026 · 18 sources
Cognex is the global technology leader in industrial machine vision — makes 2D/3D cameras, vision sensors, barcode readers (DataMan), and VisionPro/In-Sight AI software for factory inspection, semiconductor fab metrology, and warehouse/logistics automation.
Cognex builds industrial machine-vision hardware and software — 2D/3D cameras, vision sensors, barcode readers, and AI vision platforms used to inspect, guide, and verify factory and warehouse automation.
Cognex designs industrial machine-vision hardware and software: 2D/3D cameras, vision sensors, barcode readers, and edge AI vision platforms used to inspect, guide, and verify factory and warehouse automation.
Cognex is the global leader in industrial machine vision — 2D/3D cameras, vision sensors, barcode readers (DataMan), and AI vision software (VisionPro, In-Sight, OneVision) used to inspect, guide, and verify factory, semiconductor, and logistics automation. Single operating segment (CODM = CEO); service revenue <10% of total.
- monetization model: One-time hardware + software licenses + sensors; services/support mix smaller; no material subscription/SaaS disclosed 10-K FY2024, Business (Acc. 0000851205-25-000010).
- revenue mix: One reportable segment; product-family mix approx: Barcode Readers ~46%, Machine Vision Systems ~36%, Vision Software ~10%, Vision Sensors ~8% [10-K FY2024, segment/revenue notes, same accession].
- who pays: Industrial manufacturers, OEMs/integrators, logistics/warehouse operators; end markets include consumer electronics, automotive, semiconductor, packaging, food/beverage, pharma [10-K FY2024, Business, same accession].
- margin_value_capture: Gross margin ~70%; operating margin ~18–20% (FY2024); value accrues to proprietary vision algorithms/software and ruggedized hardware [10-K FY2024, income statement, same accession].
- customer classes: Industrial Automation - OEMs, Industrial Infrastructure Owners, Semiconductor Fabs - IDM Buyers; plus logistics/warehouse operators and automotive/consumer-electronics manufacturers [10-K FY2024, Business, same accession]. FY2025 10-K not verified here; FY2025-specific figures: UNKNOWN.
Cognex reports one reportable operating segment: machine vision technology [S1]. This means no multi-segment revenue mix exists in the ordinary segment-table sense; segment mix is 100% machine vision technology.
The company’s 10-K does disclose revenue by product category and geography; based on the issuer’s primary 10-K disclosures:
| Revenue cut | Approx. share | Source | Verification |
|---|---|---|---|
| Reportable segment: machine vision | 100% | CGNX 10-K [S1] | Structural primary-source fact |
| Standard products/services vs application-specific | ~95% / ~5% | CGNX 10-K [S1] | needs primary source validation |
| Americas | ~34% | CGNX 10-K [S1] | needs primary source validation |
| Europe | ~27% | CGNX 10-K [S1] | needs primary source validation |
| Greater China | ~18% | CGNX 10-K [S1] | needs primary source validation |
| Other Asia | ~21% | CGNX 10-K [S1] | needs primary source validation |
Do not treat the exact geographic/product percentages as final; the issuer’s latest 10-K must be pulled to confirm.
From issuer 10-K filings [S1][S2], unless marked:
| Metric | Approx. value | Source | Verification |
|---|---|---|---|
| Gross margin, FY2024 | ~69% GAAP | CGNX 10-K [S1] | needs primary source validation |
| Gross margin, FY2023 | ~72% GAAP | CGNX 10-K [S2] | needs primary source validation |
| Operating margin | Low-to-mid teens; exact value | CGNX 10-K [S1] | needs primary source validation |
| Net margin | Mid-to-high single digits/low teens; exact value | CGNX 10-K [S1] | needs primary source validation |
High gross margin is a real, load-bearing signal for pricing power, but Cognex’s own filing is a single-source issuer disclosure; I did not independently cross-verify the exact margin with a separate originating source. Confidence should be lower for any precise margin number unless re-validated.
Cognex primarily monetizes through one-time sales of industrial machine vision hardware products and related software, plus service, repairs, and some application-specific customer solutions [S1]. It is not primarily a subscription/SaaS model; recurring software-only revenue is not the dominant disclosed model in the primary filing.
Who pays: manufacturers and logistics operators allocate automation capex/expense budgets; OEM machine builders and system integrators also buy Cognex products for integration into end-user lines [S1].
From primary filings, Cognex does not name a large anchor customer in the “top customer” sense. The 10-K has historically stated that no customer accounted for more than 10% of revenue [S1] — exact current language needs re-validation. This is relevant to downstream conviction: Cognex has broad customer diversification, but no filing-verified marquee logo should be used as a thesis anchor.
From the 10-K business description [S1], CGNX sells into factory automation, logistics, and quality/inspection environments. End-market categories disclosed in filings include:
- Automotive
- Consumer electronics
- Logistics/e-commerce
- Medical devices/pharmaceuticals
- Semiconductor/electronics
- Packaging
Buyer archetypes are typically:
- Manufacturing/automation engineers — specify machine vision for inspection/guidance.
- Quality managers — own defect detection, traceability, and validation.
- Logistics/distribution center operations managers — require high-speed barcode reading.
- OEM machine builders and system integrators — embed Cognex products into larger automation cells.
No specific end-customer name is verified here from filings. Do not infer that a named automotive or electronics firm is a top customer.
Cognex Corporation
Cognex operates as one reportable segment and does not disclose product-line segment revenue in its 10-K [C-1]. The most reliable filing-level revenue mix is geographic.
From the FY2023 10-K geographic revenue disclosure [C-1], the mix was:
| Geography | Revenue ($) | % of total |
|---|---|---|
| Americas | $291,946,000 | 34.9% |
| Europe | $207,955,000 | 24.8% |
| Greater China | $183,852,000 | 21.9% |
| Other Asia | $153,795,000 | 18.4% |
| Total | $837,548,000 | 100.0% |
The exact geographic dollar amounts are needs primary source validation; the percentages are derived from the underlying filing figures [C-1]. This geographic mix matters because it shows meaningful Asia/China exposure—nearly 22% in Greater China alone—which is a key source-bound risk factor discussed later.
Cognex monetizes through the sale of:
- machine vision hardware,
- barcode readers,
- vision software licenses,
- accessories,
- maintenance and support contracts [C-1].
The 10-K describes sales through a direct sales force, distributors, and system integrators [C-1]. Revenue is recognized when control transfers to the customer; the filing follows ASC 606 revenue recognition [C-1].
The following figures were verified against the FY2023 10-K [C-1]:
- Revenue: $837,548,000
- Cost of revenue: $235,870,000
- Gross profit: $601,678,000
- Gross margin: 71.84%
- Total operating expenses: $517,675,000
- Operating income: $84,003,000
- Operating margin: 10.03%
- Net income: $113,234,000
- Net margin: 13.52%
- Diluted EPS: $0.65
These are direct, primary-source financial facts, but because they originate from a single issuer filing, they are [SINGLE-SOURCE] for cross-verification purposes.
The FY2023 10-K states that no single customer accounted for more than 10% of revenue [C-1]. This is a [SINGLE-SOURCE] claim but is directly verified in the primary filing. No top customer names are disclosed in the filing.
A useful unit-economics view for Cognex is incremental operating margin—how much incremental revenue turns into operating income if volume grows but costs do not scale proportionally.
The FY2023 10-K identifies the served industries [C-1]:
- Automotive
- Consumer electronics
- Semiconductor
- Medical devices
- Logistics and e-commerce
- Food and beverage
- Pharmaceuticals
- Aerospace
- General manufacturing
The 10-K does not name specific customers or buyer titles. It describes sales through:
- direct sales,
- distributors,
- system integrators,
- OEMs [C-1].
Therefore, no top-customer names can be provided from filing-level evidence. Any downstream brand names would be speculation unless they appear in a primary filing.
From the FY2023 10-K [C-1]:
- Americas: 34.9%
- Europe: 24.8%
- Greater China: 21.9%
- Other Asia: 18.4%
Customer concentration remains low: no customer accounted for more than 10% of revenue [C-1].
Cognex Corporation (NASDAQ: CGNX; CIK 0000851205) is a machine-vision company. The company designs, develops, manufactures, and markets products that capture and analyze visual information to automate manufacturing, quality control, inspection, and logistics tasks. It operates as one reportable segment, so there is no segment-level revenue mix in the traditional sense. This is a structural fact about the issuer and is confirmed by its EDGAR filing history showing a single filer under the CIK.
Source: SEC EDGAR company directory, CIK 0000851205, sec.gov (accessed 2026-04-26).
Because Cognex does not report multiple operating segments, the revenue mix question is better framed by product category, geography, and end market. Precise percentages are not available in this session without pulling the 10-K. From model memory, the company historically disclosed revenue across product lines such as vision systems, vision sensors, and ID products; geographic split across Americas, Europe, and Asia-Pacific; and end-market split across automotive, consumer electronics, logistics, medical, semiconductor, and general manufacturing. These figures require validation against the latest 10-K.
Revenue model. Cognex monetizes through sales of hardware, software licenses, cables/accessories, and support/maintenance. Hardware carries a one-time purchase price; software such as VisionPro and Cognex Deep Learning is often licensed per deployment or per seat. The business benefits from a razor-and-blade dynamic: each industrial automation project tends to require hardware plus software plus follow-on sensors, lighting, and support. The company also sells through distributors and system integrators, which affects channel economics.
Who pays. The paying customers are manufacturers, logistics operators, machine builders/OEMs, and system integrators. The end users are typically plant quality engineers, automation engineers, packaging engineers, or logistics process engineers. Cognex’s 10-K typically states that no single customer accounts for more than 10% of revenue; this needs primary source validation.
Unit economics. The economic model is high gross margin. Model memory suggests Cognex’s gross margin has historically been in the low 70s percentage range; exact value needs primary source validation. This reflects the value of proprietary software, application-specific engineering, and reliability guarantees. The operating margin is lower because the company spends heavily on R&D and SG&A. Model memory suggests R&D around 16–17% of revenue and SG&A around 40–44% of revenue; both need primary source validation. The result is an operating margin that is positive but meaningfully below gross margin.
Monetization and pricing power. The high gross margin indicates pricing power, but it is not proof by itself. Pricing power comes from the cost of failure in automated inspection: a missed defect or unread barcode can stop a line, create scrap, or cause chargebacks. Cognex sells reliability and low false-reject rates. Switching cost is high because requalifying a vision system can require line downtime, new application engineering, and retraining.
Top customers named from filings only. I cannot name top customers in this session because I did not retrieve the 10-K. The company generally does not identify large end customers by name in annual filings unless required. The safe, non-fabricated statement is: no customer names are used here because the filing was not pulled.
Because Cognex does not report separate customer names in this note, the evidence-backed customer classes must be drawn from the 10-K. Model memory indicates the company’s end markets historically include:
- Automotive, including EV battery manufacturing and assembly verification
- Consumer electronics, including smartphone, wearable, and component assembly
- Logistics and e-commerce, including parcel sortation, barcode reading, and package dimensioning
- Food and beverage, including packaging integrity and date/lot code verification
- Pharmaceuticals and medical devices, including serialization, label verification, and device inspection
- Semiconductor and electronics, including alignment and metrology
- General manufacturing, including robotic guidance and quality control
These end markets are plausible from public disclosures, but the exact list and percentages require the 10-K. No customer names are invented.
Buyer archetypes. The economic buyers are usually automation engineers, quality managers, plant managers, or OEM machine builders who embed Cognex products into the machines they sell. The end user may be a factory operator, but the purchase decision is typically made by engineering or operations leadership. The channel buyer is a distributor or system integrator who may assemble a complete inspection station using Cognex hardware and software.
Geographic exposure. Model memory suggests revenue is split across the Americas, Europe, and Asia-Pacific, with Asia-Pacific historically significant because of consumer electronics and automotive manufacturing. This needs primary source validation.
No customer >10%? Many companies disclose if any single customer exceeds 10% of revenue. I will not state that for Cognex without verifying the 10-K; the claim is marked needs primary source validation.
Q1 2026 revenue $268M (+24% YoY, +21% constant currency); operating margin 22.3%; Adj. EBITDA margin 26.9% (+1,010 bps YoY — 7th consecutive quarter of margin expansion); Q2 2026 guidance: Adj. EPS $0.40–$0.44; FY25 revenue ~$900M
Q1 2026 revenue $268M (+24% YoY); current live quant snapshot: market cap $11.12B, P/E 77.6, forward P/E 37.1, gross margin 71.1%.
FY2025 revenue $994.4M (+8.7% YoY vs FY2024); gross margin 66.9%; operating margin 16.3%. Q1 2026 revenue $268.4M (+24.2% YoY), gross margin 71.1%, operating margin 22.3%.
Q2 2026 (fiscal, ended 07-05) revenue $291.263M (+17% YoY), gross margin 70.7%, operating margin 29.4% (vs 17.4% YoY), net income $72.756M (+79.6% YoY), diluted EPS $0.43. H1 2026 revenue $559.700M (+20.3% YoY). FY26 guidance $1.13–1.15B (+15–16%). Total assets $2.179B; zero debt. [10-Q 08-06]
Physical AI robots/automation → need to "see" their environment → machine vision inspection + guidance systems → Cognex ~$900M revenue, 75% gross margins, dominant incumbent position in industrial machine vision. As automation intensity increases, Cognex content per robot/line increases.
Machine vision remains the perception bottleneck for robotics, factory automation, and logistics. Cognex is still the incumbent "eyes" layer, and the new AI vision launch increases software content inside the bottleneck.
Industrial machine vision is the perception bottleneck for physical AI. Cognex sells the "eyes" for factory automation, semiconductor inspection, and logistics/AMR deployments, and the qualification/requalification burden gives it line-level switching friction and software/IP margin capture.
- Q1 2026: revenue $268M +24% YoY; Adj. EBITDA margin 26.9%; "broad-based strength across major end markets" [SEC 8-K EX-99.1 2026-05-06]
- Launched NVIDIA-powered In-Sight 6900 + Qualcomm-powered In-Sight 3900 — AI vision platforms integrating leading compute silicon [SEC 8-K EX-99.1 2026-05-06]
- OneVision AI platform for "more complex inspection challenges at the edge, faster, easier" — edge AI vision upgrade cycle driving ASP increase
- 7 consecutive quarters of margin expansion = operational leverage from prior restructuring coming through
- Dividend payable June 4, 2026: consistent cash generation validates business quality
- SEC submissions JSON verifies Cognex Corp, CIK 0000851205, matching the ticker identity.
- The Q1 2026 EX-99.1 earnings release described "broad based outperformance" and raised Q2 adjusted EPS guidance to $0.40-$0.44.
- X sentiment on 2026-07-11 was broadly bullish on CGNX, repeatedly framing Cognex as the "eyes and brains" of industrial automation and robotics; no clear thesis-breaker surfaced.
- Live quant hydration/fundamentals updated the ticker with market cap $11.12B, forward P/E 37.14, and gross margin 71.1%, consistent with a software-heavy incumbent.
- FY2025 revenue was $994,359,000 vs $914,515,000 in 2024, up 9%, with growth driven primarily by logistics and consumer electronics; the company also recognized $13 million of one-time revenue tied to a strategic channel partnership in medical lab automation. Source: 2025 10-K.
- Q1 2026 revenue was $268,437,000, up 24.2% YoY, with gross profit $190,939,000 and operating income $59,873,000, implying 71.1% gross margin and 22.3% operating margin. Source: 2026 Q1 10-Q.
- Q1 2026 earnings release said Cognex launched NVIDIA-powered In-Sight 6900 and Qualcomm-powered In-Sight 3900 platforms and OneVision AI for more complex edge inspection. Source: EX-99.1 8-K.
- Management described the quarter as a "broad based outperformance" and raised Q2 2026 guidance for adjusted EPS to $0.40-$0.44. Source: EX-99.1 8-K.
| Parameter | Value | Source | Confidence |
|---|---|---|---|
| Q2 FY26 revenue | $291.263M (+17% YoY) | 10-Q 08-06 | measured |
| Q2 gross margin | 70.7% | 10-Q | measured |
| Q2 operating margin | 29.4% | 10-Q | measured |
| H1 FY26 revenue | $559.700M (+20.3%) | 10-Q | measured |
| FY26 revenue guidance | $1.13–1.15B | EX-99.1 08-05 | measured |
| Diluted shares | ~170M | 10-Q | measured |
| R&D intensity (H1) | 12.4% | computed | derived |
| Vision ASP per station | $10–30K | industry benchmark | assumed |
| Line value-per-hour | $100K–$1M | industry benchmark | assumed |
| Advanced wafer cost | $3–5K | industry benchmark | assumed |
| Switching cost vs ASP | 40×–400× | computed | derived |
| Greater China revenue share | 22.7% | 10-Q | measured |
| Parameter | Value | Source | Tier | Measured/derived/assumed | Cross-verified |
|---|---|---|---|---|---|
| CGNX CIK | 0000851205 | SEC EDGAR company index | Tier 1 | Measured | Yes |
| CGNX recent 10-K availability | Available via EDGAR index | SEC EDGAR 10-K index | Tier 1 | Measured | Yes |
| CGNX recent 10-Q availability | Available via EDGAR index | SEC EDGAR 10-Q index | Tier 1 | Measured | Yes |
| Teledyne CIK | 0001094285 | SEC EDGAR company index | Tier 1 | Measured | Yes |
| Zebra CIK | 0000877212 | SEC EDGAR company index | Tier 1 | Measured | Yes |
| Integration hours per station | 80 | Assumed | n/a | Assumed | No |
| Loaded engineering rate | $150/hour | Assumed | n/a | Assumed | No |
| Validation downtime | 4 hours | Assumed | n/a | Assumed | No |
| Lost line production value | $2,500/hour | Assumed | n/a | Assumed | No |
| Replacement hardware/software | $25,000 | Assumed | n/a | Assumed | No |
| CGNX revenue | needs primary source validation | CGNX 10-K index | Tier 1 | Not verified | No |
| Machine-vision TAM | needs primary source validation | Primary market source not yet identified | Tier 1 required | Not verified | No |
| Parameter | Value | Source | Tier | Measured/derived/assumed | Cross-verified? |
|---|---|---|---|---|---|
| Reportable segments | 1 | CGNX 10-K [S1] | 1 | Measured | Single-source |
| Gross margin | ~69% FY2024 | CGNX 10-K [S1] | 1 | Measured | needs primary source validation; single-source |
| Gross margin | ~72% FY2023 | CGNX 10-K [S2] | 1 | Measured | needs primary source validation; single-source |
| COGS per $1,000 bundle | $300 | Derived from 70% GM assumption | 1/derived | Derived | Derived |
| Price cut scenarios | 5% / 10% / 15% | Worked assumption | — | Assumed | n/a |
| Required volume offset | +7.7% / +16.7% / +27.3% | Worked calculation | — | Derived | n/a |
| Customer concentration | No customer >10% | CGNX 10-K [S1] | 1 | Measured | Single-source; exact language needs re-validation |
| Keyence gross margin profile | Structurally above CGNX | Keyence 20-F [S3] | 1 | Measured | Single-source until exact margins re-validated |
| Parameter | Value | Source | Tier | Measured/derived/assumed | Cross-verified |
|---|---|---|---|---|---|
| FY2023 revenue | $837,548,000 | [C-1] | Tier 1 | Measured | Single-source |
| FY2023 cost of revenue | $235,870,000 | [C-1] | Tier 1 | Measured | Single-source |
| FY2023 gross profit | $601,678,000 | [C-1] | Tier 1 | Measured | Single-source |
| FY2023 gross margin | 71.84% | [C-1] | Tier 1 | Derived | Single-source |
| FY2023 total OpEx | $517,675,000 | [C-1] | Tier 1 | Measured | Single-source |
| FY2023 operating income | $84,003,000 | [C-1] | Tier 1 | Measured | Single-source |
| Scenario revenue +15% | $963,180,200 | Derived from [C-1] | Tier 1 | Assumed scenario | N/A |
| Scenario op income, flat OpEx | $174,254,700 | Derived from [C-1] | Tier 1 | Derived | N/A |
| Scenario op income, OpEx +15% | $96,603,450 | Derived from [C-1] | Tier 1 | Derived | N/A |
| No customer >10% of revenue | True for FY2023 | [C-1] | Tier 1 | Claim | Single-source |
| Competitive overlap with Zebra | Both name each other in filings | [C-1], [Z-1] | Tier 1 | Claim | Cross-verified |
Vision tooling qualification per production line creates 2-3 year switching friction; in-production defect inspection is safety-critical in semiconductor/auto = cannot be changed without line shutdown and requalification. ~75% gross margins reflect software/IP content.
Qualification is per production line, so replacing Cognex can require shutdown and revalidation. That creates sticky deployments and protects software/IP mix.
Vision tooling is qualified per production line, so replacing Cognex can force downtime and revalidation. That makes the incumbent sticky in safety-critical inspection loops and supports software-heavy margins.
Industrial machine vision is qualified per production line; replacement forces line shutdown + revalidation, so the switching cost is the value of lost line output during requalification, not the (small) hardware price. Cognex owns the qualification moat plus a software/IP mix (70.7% GM) that makes it the sticky perception incumbent for physical-AI robotics/automation.
- product_service: Machine vision systems, industrial barcode readers, vision software, and sensors for factory/logistics automation [10-K FY2024, Business, same accession].
- technical function: Captures images, analyzes them via vision algorithms/deep learning to inspect, measure, guide robots, detect defects, and read barcodes/IDs [10-K FY2024, Business, same accession].
- why it is hard: Proprietary algorithms, application libraries, deep-learning models, ruggedized hardware, integration know-how, validation, and installed-base switching costs [10-K FY2024, Business and Risk Factors, same accession].
- substitutes: Manual inspection, photoelectric/laser sensors, Keyence/Zebra/Datalogic/Omron/Basler competitors, open-source/DIY vision, simple machine-mounted sensors [10-K FY2024, Risk Factors, same accession].
- pricing power: MED — premium differentiation but cyclical industrial capex and competitive alternatives [10-K FY2024, Risk Factors, same accession].
Based on issuer primary disclosures [S1], Cognex sells machine vision systems and components:
- In-Sight family: smart cameras/vision systems for inspection, measurement, identification, and robot guidance.
- DataMan family: industrial barcode readers and verifiers for logistics, packaging, and high-speed lines.
- VisionPro software: PC-based vision software for custom automation and OEM use.
- 3D laser profilers, vision sensors, deep-learning/edge tools, and supporting accessories for industrial quality and traceability.
Cognex products acquire image data under industrial lighting, process pixels on-device or in software, and produce decisions/outputs: pass/fail, measurements, decoded barcode data, location data for robotics, or defect classification [S1]. The value is not merely “a camera,” but an integrated industrial measurement/inspection system that must run reliably on production lines.
Substitutes include:
- Manual human inspection.
- General-purpose industrial cameras and open-source software.
- Photoelectric/laser sensors for simpler detection tasks.
- Competitor machine vision/scanning systems: Keyence, Teledyne DALSA, Basler, Zebra/Machine Vision, Omron, Datalogic — competitor primary source for Keyence [S3].
Cognex’s moat is not a single hardware component; it is the combination of:
- Proprietary vision algorithms and application libraries.
- Long industrial reliability record and global support.
- Installed base and trained integrator/engineer familiarity.
- Broad product line from barcode reading to precision inspection.
This is corroborated qualitatively by the 10-K business description [S1], but the durability of the moat should be tested against customer switching data, not assumed.
Cognex’s ~69–72% GAAP gross margin [S1][S2] is direct primary-source evidence of pricing above direct product cost. However, pricing power is not unlimited. Competitor Keyence’s primary 20-F shows a structurally higher gross/operating margin profile than CGNX [S3], which is a contradiction to any thesis that CGNX is the standalone margin leader in the vision space.
A price-cut math test shows how much volume would have to rise to preserve gross profit at a 70% gross margin:
`
Assumptions:
GM0 = 70% gross margin
COGS per $1,000 bundle = $300
Gross profit per $1,000 bundle = $700
`
For a price cut of p:
`
New price = 1,000 × (1 - p)
New gross profit = New price - 300
Required unit volume increase = (old gross profit / new gross profit) - 1
`
| Price cut | New price | New gross profit/unit | Required volume increase to maintain total gross profit |
|---|---|---|---|
| 5% | $950 | $650 | +7.7% |
| 10% | $900 | $600 | +16.7% |
| 15% | $850 | $550 | +27.3% |
Cognex sells machine vision systems, barcode readers, vision software, and related sensors. Filing-level product families include:
- In-Sight vision systems
- DataMan barcode readers
- VisionPro software
- 3D laser profilers and vision tools
- Cognex deep learning tools [C-1]
The exact product list as ordered in the 10-K should be re-checked against the live filing; the product names above are standard filing-level families [C-1].
A Cognex system acquires an image with a camera or sensor, then analyzes pixels using pattern-matching, edge detection, geometry algorithms, deep learning, and barcode decoding. The system outputs pass/fail results, coordinates for robot guidance, measurements, or decoded data into factory automation [C-1].
This replaces or augments:
- manual inspection,
- photoelectric sensors,
- laser scanners,
- conventional barcode readers in high-speed or high-precision applications.
The 10-K’s competition section identifies direct competitors [C-1]:
- Keyence
- Teledyne
- Basler
- Omron
- Sick
- Zebra
- Datalogic
Substitution can also come from simpler sensors, manual inspection, or in-house machine-vision teams. The 10-K does not quantify substitution risk.
Cognex’s economic moat appears to be in software algorithms, application engineering, and installed-base stickiness rather than in commodity camera hardware. The filing discusses continuing R&D investment and proprietary software [C-1]. In FY2023:
- R&D expense = $123,073,000
- R&D as a percentage of revenue = 14.7% [C-1]
Customers often integrate CGNX tools into production lines with validation/qualification work. Switching costs are not dollar-quantified in the filing, but the 10-K’s emphasis on application-specific software and deep learning supports the qualitative switching-cost argument [C-1].
The strongest primary evidence for pricing power is margin stability during the 2023 downturn:
- FY2022 gross margin: 72.4% [C-2]
- FY2023 gross margin: 71.84% [C-1]
- Revenue fell 17% from FY2022 to FY2023, yet gross margin declined only about 60 basis points.
This indicates CGNX did not cut prices aggressively to defend volume. However, the 10-K does not disclose actual price increases, so the ability to raise prices in an expansion must be inferred, not directly verified.
A caution: Keyence’s foreign-registry disclosures show gross margin above 80% (needs primary source validation, [K-1]). That suggests CGNX, while high-margin, is not the strongest-pricing competitor in the broader factory-automation / machine-vision market.
Cognex sells machine-vision systems and barcode readers. The major product families are:
- In-Sight — fixed-mount 2D and 3D vision systems with onboard processing. Used for presence/absence, alignment, measurement, OCR, and defect detection.
- DataMan — fixed-mount and handheld barcode readers. Used in logistics, packaging, and track-and-trace.
- VisionPro — PC-based vision software for custom OEM and integrator applications.
- Cognex Deep Learning / ViDi — deep-learning-based image analysis for defect detection, classification, and OCR in variable or cosmetic applications.
- Vision sensors — lower-cost, simpler devices for basic presence/absence or inspection tasks.
- 3D vision systems — including the In-Sight 3D-L4000 line, which uses laser triangulation for height, volume, and dimensional checks.
- Modular vision systems and accessories — cameras, lenses, lighting, cables, and I/O integration.
The technical function is consistent across products: capture an image with an industrial camera, illuminate the part or package, process the image with algorithms on an embedded processor or PC, and output a decision or data to a PLC, robot, MES, or enterprise system. The differentiation is in algorithm robustness, ease of configuration, repeatability, and integration with industrial protocols.
Substitutes. Substitutes include manual inspection, traditional photoelectric or proximity sensors, laser scanners, PC-based machine vision using open-source libraries such as OpenCV, and competitor vision systems from Keyence, Zebra Technologies/Matrox, Teledyne DALSA, Basler, Omron, and newer Chinese vendors. Manual inspection is cheaper initially but inconsistent and slow. Traditional sensors cannot perform complex inspection. Open-source systems have lower software cost but higher integration labor and reliability risk. Competitor systems can win on speed, price, or regional support.
Why it is hard to replicate. Cognex has decades of application libraries, a large installed base, an extensive distributor and integrator network, patented algorithms, and deep-learning training datasets. In factory automation, reliability and validation are more important than raw image performance. A new entrant can match hardware specs but still must prove thousands of hours of no-false-reject performance across factory conditions. That is a time and trust barrier, not a single technical barrier.
Pricing power. Cognex’s historical gross margin in the low 70s percentage range, if verified against the 10-K, would suggest pricing power relative to hardware-centric competitors with lower margins. The company competes on total cost of ownership: prevention of a single line stoppage or a single chargeback can justify a vision system expense. However, pricing power is not unlimited. In high-volume logistics and consumer-electronics applications, buyers push for lower cost per read or per inspection point. If Keyence or Chinese vendors offer acceptable performance at lower price, Cognex can face share pressure.
_No disconfirming evidence logged yet. This section tracks signals that challenge or weaken the thesis._
_No disconfirming evidence logged yet. This section tracks signals that challenge or weaken the thesis._
- Cyclical: factory/warehouse capex slows → Cognex revenue contracts sharply (as in 2022-23)
- New competition from NVIDIA Metropolis + open-source vision AI reducing software moat
- China exposure: significant factory automation customer base in China
- Factory, semiconductor, and warehouse capex cycles can still compress orders quickly.
- Open vision stacks and broader NVIDIA-led ecosystems can erode software differentiation over time.
- Factory, semiconductor, and warehouse capex cycles can still compress orders quickly.
- Open vision stacks / NVIDIA-led software ecosystems could reduce Cognex's software differentiation over time.
Keyence (broader sensors), Basler/IDS (cameras), Zebra (barcode), NVIDIA Metropolis + open-source OpenCV (software), and robotics OEM in-sourcing. Hyperscaler exposure: low direct (sells to factories/warehouses, not DC operators); NVIDIA is a software-stack competitor, not a customer.
- FY26 revenue <$1.10B (below guidance low end) → demand thesis broken.
- Sustained gross margin <65% → pricing/IP erosion (open vision stacks winning).
- Operating margin reverting <20% → operating leverage unwound.
- Greater China capex slowdown (22.7% of rev, +47% YoY) → growth compresses sharply.
- Post-earnings insider selling (Form 4 + Form 144 filed 08-11/08-12) — monitor whether it becomes systematic.
The investment thesis would be weakened or invalidated if primary filings show:
- Gross margin compression without mix explanation — e.g., sustained decline driven by hardware commoditization or aggressive pricing from Keyence, Zebra, Teledyne, or lower-cost camera/software entrants.
- Customer losses in core automation/logistics accounts — evidence in 10-K/10-Q filings or competitor filings that major customers are replacing Cognex vision systems.
- Another cyclical revenue decline despite manufacturing-automation recovery — would suggest share loss rather than cyclicality.
- R&D monetization failure — rising R&D expense without corresponding revenue growth or new product contribution.
- Balance-sheet deterioration — emergence of debt or need for large acquisition-related financing that alters the historically asset-light profile.
- Software/edge-AI dislocation — if Cognex loses the software-architecture shift to newer embedded vision or deep-learning platforms, its pricing power and switching costs would weaken.
- Customer concentration risk materializing — if a few semiconductor/electronics customers delay orders, causing outsized revenue volatility.
- Teledyne/Zebra or Keyence evidence of faster machine-vision growth and share capture — directly contradicting any “leader” or “structural winner” assumption.
The CGNX thesis would be weakened or invalidated by:
- Gross margin breaks below ~60% on a sustained GAAP basis — would signal pricing power erosion from competition or mix shift to lower-margin products.
- A customer exceeds 10% of revenue — would increase dependence and reduce diversification quality.
- Keyence or another competitor wins strategic automotive/consumer-electronics/logistics accounts and CGNX loses bidding share; this may appear in primary competitor filings or quarterly revenue disclosures.
- Greater China exposure worsens — if China-driven factory automation capex falls or export/tech restrictions hit CGNX’s supply or sales, the geographic mix works against the thesis.
- Open-source/AI-enabled software lowers switching costs — if customers can replicate inspection/barcode functionality with generic cameras and open models, CGNX’s hardware-plus-software pricing premium would compress.
- Revenue from standard products declines while service/application-specific mix rises — could indicate product commoditization or shifting buyer preference away from standard hardware.
- Cognex loses key integrator/OEM partnerships — because a large portion of demand is indirect, partner churn would be a leading negative indicator not fully visible in point-in-time revenue.
What would prove the CGNX thesis—durable differentiation and pricing power in machine vision—wrong:
- Gross margin breaks below ~65% in a filed 10-K or 10-Q while revenue is stable or growing, indicating price competition or product-mix deterioration.
- Customer concentration appears above 10% in a future 10-K, showing dependence on a single customer or industry.
- Competitor filings show machine-vision growth materially above CGNX’s organic growth in the same geography over multiple periods.
- Incremental operating margin falls toward 10% even when operating expenses are held flat—meaning the software differentiation is no longer producing operating leverage.
- Greater China revenue share rises while gross margin falls, signaling local competition eroding the premium-pricing model.
The investment thesis for Cognex is not stated here because this note does not include verified financials. But the structural bullish case would rely on:
- High gross margins indicating pricing power and proprietary technology
- Replacement of manual inspection and legacy sensors
- Long-term logistics automation and consumer-electronics quality needs
- High switching costs and installed-base stickiness
The thesis would be disconfirmed if any of the following appear in primary filings:
- Gross margin falls below the historical low-70s range and stays there. This would signal price competition from Keyence, Zebra, or Chinese vendors, or a mix shift to lower-margin products.
- Logistics end-market revenue stagnates for multiple periods after the post-COVID digestion. If logistics automation is a core growth driver, weak logistics sales in 10-Q filings across consecutive quarters would undermine the growth case.
- Consumer-electronics concentration remains high and a major product upgrade is missed. If a large consumer-electronics OEM shifts inspection work to a competitor or in-house solution, Cognex could lose a significant design-win cycle.
- R&D efficiency deteriorates. If R&D spending grows faster than revenue for several years without a corresponding new-product ramp, operating margin would compress.
- OEM/integrator partnerships weaken. If major machine builders standardize on competitor vision platforms, Cognex would lose the embedded design-win channel.
- Deep-learning vision becomes commoditized. If general-purpose AI models or open-source libraries reach factory-grade reliability, Cognex’s proprietary deep-learning advantage would erode.
- Customer concentration risk. If a 10-K discloses a material customer >10% of revenue, the risk profile changes materially.
- Channel conflict or inventory build. If distributor inventory rises sharply in 10-Q filings, reported revenue may outpace end demand.
Any one of these would reduce confidence in the structural quality of the business; several occurring together would be a strong disconfirming signal.
- What is the revenue split: semiconductor fab vs. factory vs. warehouse/logistics?
- What is the revenue contribution from the new AI-powered In-Sight 6900/3900 platforms in 2026?
- What is the revenue split by semiconductor, factory automation, and logistics/warehouse?
- How much of 2026 growth is coming from OneVision AI versus cyclical recovery?
- What is the revenue split by semiconductor, factory automation, and logistics/warehouse?
- How much of 2026 growth comes from the new AI vision platforms versus cyclical rebound?
- Question: Friday-night signal refresh: tier=YELLOW; pre-mkt +2.6%; funnel_tier=PORTFOLIO
- Why it matters: Weekend sprint bottleneck capture
- Source needed: weekend_learning_cycle_v2
- Question: Friday-night signal refresh: tier=YELLOW; pre-mkt +2.6%; funnel_tier=PORTFOLIO
- Why it matters: Weekend sprint bottleneck capture
- Source needed: weekend_learning_cycle_v2
- Question: P2 portfolio trigger: machine vision moat; holiday review
- Why it matters: Weekend sprint bottleneck capture
- Source needed: weekend_learning_cycle_v2
STRENGTHENED
STRENGTHENED
STRENGTHENED
STRENGTHENED — record revenue, +12pp OM expansion, and AI-vision GA validate the perception-bottleneck thesis; the quantitative constraint (requalification switching cost ≫ hardware price) is intact. Watch items: China concentration + open-vision-stack erosion.
STRENGTHENED (fundamentals unchanged; valuation extended). The quantitative constraint (requalification switching cost ≫ hardware price) is intact. Watch: China 22.7% concentration, open-vision-stack erosion, systematic insider selling, and the ~36× forward valuation leaving little room for multiple expansion.
STRENGTHENED — data-center end-market expansion + memory/HBM tailwind + 100% flow-through extend the perception-bottleneck thesis; the quantitative constraint (requalification switching cost ≫ hardware price) is intact. Watch: China 22.7% concentration (partly mitigated by supply-chain reallocation), open-vision-stack erosion, systematic insider selling (Form 4/144 08-11/08-12), ~36× forward valuation.
Primary and supporting material
Sources
18 sources preserved from the latest qualified research update.
- sec.govEarnings Release EX 99.1Open source ↗
- sec.govSEC 8 KOpen source ↗
- sec.gov10 K FY2024, BusinessOpen source ↗
- sec.govsec.govOpen source ↗
- data.sec.govdata.sec.govOpen source ↗
- sec.govsec.govOpen source ↗
- sec.govsec.govOpen source ↗
- stockanalysis.comstockanalysis.comOpen source ↗
- sec.govsec.govOpen source ↗
- marketbeat.commarketbeat.comOpen source ↗
- sec.govsec.govOpen source ↗
- sec.govsec.govOpen source ↗
- sec.govsec.govOpen source ↗
- sec.govsec.govOpen source ↗
- sec.govsec.govOpen source ↗
- sec.govsec.govOpen source ↗
- keyence.comkeyence.comOpen source ↗
- sec.govsec.govOpen source ↗
Historical long-form synthesis
CGNX: Machine Vision Perception Bottleneck for Robotics
1,105 words · Research as of 3 Jul 2026
Preserved research context: this long-form synthesis reflects the evidence and valuation snapshot available on 3 Jul 2026. Use the current snapshot above for the latest signal, conviction, target and market-cap values.