Cognex

Phase A — Understand the business

Lens 1 · Company Overview

Cognex makes industrial machine vision — hardware + software systems that "see": they locate, identify, inspect, measure, and read barcodes on discrete items moving down an assembly line or through a warehouse. Founded 1981; HQ Natick, MA; 2,745 employees ("Cognoids") at 2025-12-31, of whom 1,876 are outside the US. It is one of the two global leaders in machine vision alongside Japan's Keyence.

What it sells. Three product families: (1) vision systems & sensors (In-Sight 2D/3D smart cameras, deep-learning and "edge-learning" AI tools); (2) vision software (VisionPro, the OneVision cloud platform launched 2025); (3) barcode readers (DataMan fixed-mount + handheld). Moritex (acquired 2023) added optical components — lenses and lighting. Services (maintenance, consulting, training) are <10% of revenue.

Customers / end markets. Sells into nearly every discrete-manufacturing and distribution industry. The four largest end markets — logistics, packaging, consumer electronics, automotive — are ~85% of revenue; semiconductor is the notable fifth. Cognex does not disaggregate revenue by end market in dollars — it operates as one reportable segment ("machine vision technology") and only breaks revenue out by geography. 67% of 2025 revenue was non-US.

Contract structure. Largely transactional product sales recognized at a point in time; a sub-set ("application-specific customer solutions") recognized only on solution validation (risks/rewards transfer on acceptance, not shipment). In Q1'26 ~95% of revenue was "standard products and services" ($254.4M) vs. $14.0M application-specific. No long-term take-or-pay backlog — it is a book-and-ship, cyclical capex-exposed model, which is the single most important fact about the business.

Customer concentration (rising — flag). One customer was 15% of total revenue in 2025, up from 10% in 2024 and <10% in 2023. Unnamed, but the disclosure pattern + the logistics-recovery commentary points overwhelmingly to a large e-commerce/parcel buyer (consensus reads this as Amazon; ``, not stated in the filing).

Lens 2 · Supply Chain

Upstream → Cognex → end customer, every named link from the filings:

• Inputs: semiconductors/ICs (incl. memory chips, called out as a supply risk ) sourced partly from Taiwan-based vendors — a China-Taiwan flashpoint dependency the company explicitly names; optical glass + components.
• Contract manufacturing: the majority of hardware (vision systems, sensors, barcode readers) is built by third-party electronics-manufacturing-services (EMS) suppliers in Indonesia and Malaysia — Cognex specifies components and controls test processes but owns no fab. (A June 2022 fire at its primary Indonesian contract manufacturer destroyed inventory and cost the company through 2023 — a concrete single-point-of-failure realization.)
• In-house manufacturing: optical components (lenses/lighting) made at Cognex plants in China and Vietnam.
• Distribution: assembled goods routed to Cognex distribution centers (US, Europe, Asia) where Cognex personnel load software, align imagers, QC, and ship.
• Go-to-market: direct salesforce (three teams — Market Creation/Expansion, Market Penetration, Partner Enablement) plus a partner network of OEMs, machine builders, and system integrators.
• End buyers: e-commerce/parcel (logistics), CE OEMs, automakers, packaging/FMCG, semiconductor fabs.

Chokepoints / single-source dependencies: (1) the primary Indonesian EMS contractor — concentration proven costly by the 2022 fire; (2) Taiwan ICs — geopolitical; (3) China + Vietnam in-house optical plants — tariff/sanction exposure; (4) memory chip availability. Asset-light is a margin/FCF positive but pushes resilience risk onto a thin supplier base.

Lens 3 · Competitive Advantages (moats)

The moat is real but narrow — "best-in-class point technology + brand," not a system lock-in like Keyence's.

• Brand + performance leadership in barcode reading and high-end deep-learning vision — Cognex self-identifies its e-commerce logistics differentiation as "the high performance of our barcode reading". Decades of installed base and application know-how at the most sophisticated customers.
• Software/AI estate: ~a decade in industrial AI; VisionPro + edge-learning + the new OneVision cloud-training platform; AI-assisted internal coding tools to cut RD&E intensity.
• Switching costs: moderate — once a vision recipe is validated on a production line, re-qualifying a competitor is costly and risky. But this is per-line, not enterprise-wide; it does not produce Keyence-level pricing power.
• Bargaining power: strong vs. fragmented SIs/OEM partners and its EMS contractors; weak-and-weakening vs. its largest customer (one buyer now 15% of revenue) and increasingly vs. buyers who can reach for free/open-source AI vision tools — a threat Cognex itself flags as lowering entry barriers.
• Bottlenecks (commercial layer, kb/robotics/wiki/bottlenecks.md): machine vision is the perception bottleneck for physical automation — favorable secular position, but the value increasingly migrates toward software/AI where moats are shallower than in proprietary optics/sensors.

Verdict on moat: durable enough to defend ~21% global share and 67-71% gross margins, not durable enough to justify a Keyence multiple. The structural question (Lens 13) is whether commoditizing AI compresses the software premium faster than Cognex broadens its base.

Lens 4 · Segments

One reportable segment (machine vision) — so the only `` disaggregation is by geography. End-market splits are qualitative only.

Revenue by geography (FY, $000s):

Trend + cause: Americas (logistics/e-commerce led, +1 one-time channel deal) and Europe (CE procurement shifting out of China to Europe; packaging recovery; FX) accelerating; Greater China and Other Asia declining on CE regional purchasing shifts, an extra month of Moritex revenue booked in 2024 (lag elimination), and automotive weakness.

Q1'26 geography ($000s) shows the recovery broadening everywhere:

Greater China inflecting from −3% (FY25) to +40% (Q1'26) is the standout — semiconductor + CE base growth.

---

Phase B — Measure performance

Lens 5 · Earnings Result (latest print: Q1 FY2026, quarter ended 2026-04-05)

A clean, large beat that the stock rewarded with a ~15% one-day pop.

P&L, Q1'26 vs Q1'25 ($000s):

• vs. consensus / own guidance: revenue $268M beat the $235-255M guide and consensus; adjusted diluted EPS $0.34 (+113% YoY) beat the $0.22-0.26 guide — 7th straight quarter of growth.
• Drivers: broad-based — CE base growth, higher semiconductor + packaging, continued large-logistics growth; FX a ~3pt tailwind.
• Margin moves: GM to 71% on favorable end-market mix + volume leverage; opex +11% (incentive comp on stronger performance, FX, reorg charges, partly offset by cost management) → operating leverage drove OM to 22%.
• One-time tailwinds to flag: a $5.0M SBC credit from setting the former CEO's award forfeiture to 100% on his March 2026 board retirement, plus forfeiture-rate true-ups; and a small $1.5M pre-tax loss on the Moritex-Japan divestiture (in Other income). Net, a couple cents of EPS was non-operational.
• Balance-sheet flags: healthy. Inventory down to $135.5M (from $137.9M; raw materials drawn down) — no channel stuffing signal. Deferred revenue up to $34.1M (from $21.1M) — a positive forward tell. Receivables roughly flat. Zero debt; ~$642M cash+investments.
• Guidance (Q2'26): revenue $280-300M (+16.5% midpoint YoY) and adjusted EPS $0.40-0.44 (+68% midpoint) — implies the re-acceleration continues.

FY2025 context (10-K):

FY2025 GM (67%) was dented by a $13M Q4 excess-&-obsolete inventory charge from the new leadership team's strategic portfolio review (de-emphasizing legacy products) + mix + tariffs. The Q1'26 snap-back to 71% suggests that was a clean-up, not a structural margin break. FY2025 GAAP EPS was distorted down by a $33.2M discrete OBBBA tax charge (raised the effective rate to 37%; normalized rate ~17%). FY2025 adjusted EPS was $1.02.

Lens 6 · Earnings Calls (sentiment trend)

No transcripts on disk (transcripts=0). From web coverage of the last several calls:

• Tone has shifted decisively from defensive to offensive over ~4 quarters. The 2023/early-2024 narrative was "painful reset of the logistics growth market" and digestion of e-commerce capacity. By Q4'25 → Q1'26 the language is "broad-based strength," "AI push," "margin surge," "seventh straight quarter of growth."
• Recurring phrases under CEO Moschner: "#1 in AI for industrial machine vision," "best customer experience," "double the customer base in five years," "salesforce transformation," "cost discipline / operating leverage."
• Things they stopped saying: the apologetic "digestion"/"time-out" framing of the e-commerce customers; emphasis on the emerging-customer headcount build (replaced by a data/KPI-driven "salesforce transformation" that is explicitly less headcount-additive).
• Net read: management sentiment is the most bullish it has been since the 2021 peak — appropriate given the tape, but it is now selling a growth story, which raises the bar for disappointment.

Lens 7 · Comps

Peer set: the global machine-vision / industrial-automation leaders. Multiples are ``, dated; where a clean figure isn't sourced, n/a. USD.

Read: Cognex trades at a Keyence-like multiple (~34x fwd P/E) on Zebra-like margins (~16% OM vs Keyence's ~51%). The market is paying for (a) the cyclical EPS snap-back (NI forecast +77% in 2026 ) and (b) an AI-vision optionality narrative. Against Zebra at ~12x forward — a direct vision competitor consolidating the space — Cognex is ~3x the forward multiple. That gap is the crux of the valuation debate. Market context: machine-vision TAM ~$21B (2025) → ~$23.5B (2026), ~11% CAGR; Cognex ~21% / Keyence ~19% share.

Lens 8 · Stock-Price Catalysts (last ~5 years)

The pattern is unambiguous: CGNX is a logistics/automation-capex cycle trade with one giant e-commerce customer as the swing factor.

• 2021: record revenue $1.037B on the e-commerce/logistics buildout boom — the cycle high.
• 2022: revenue declined from the record as Amazon shifted from "investment" to "digestion" of distribution capacity; compounded by the June 2022 Indonesian contract-manufacturer fire and elevated component costs. Stock de-rated hard.
• 2023: revenue −17% to $837.5M — the trough; logistics + CE + semiconductor all down as large customers paused capex.
• 2024-2025: recovery, +9% / +9%; salesforce transformation + AI product cycle; CEO transition announced Q1'25.
• Q1'26 (May 2026): ~15% one-day surge on the +24% revenue / +113% adj-EPS beat and raised outlook.
• Analyst re-rating (2026): Goldman → $80, KeyBanc → $80, JPMorgan upgrade to Overweight, PT $75, Barclays → $75.

What the market actually reacts to: (1) the direction of large-logistics/e-commerce capex (Amazon), (2) beats/raises vs. guidance on the recovery slope, (3) gross-margin direction (mix-driven), and (4) management-credibility events (CEO change, board refresh). It does not react much to incremental product launches absent revenue proof.

---

Phase C — Judge people & books

Lens 9 · Management

A fresh, young, internally-promoted CEO with a board deliberately re-tooled toward automation + sales execution.

• CEO: Matt Moschner (38) — became CEO June 27, 2025, succeeding Robert Willett (CEO 2011-2025). Joined Cognex 2017; ran ID Products/barcode platform engineering (VP, 2022); as President/COO oversaw global engineering, products, sales, operations. Track record: built/scaled the DataMan barcode franchise — the very product line Cognex calls its logistics differentiator. Founder vs professional manager: professional manager, but a homegrown operator, not an outside parachute — continuity of culture with a sharper cost/KPI edge.
• Founder still present: co-founder Dr. Robert Shillman ("Dr. Bob") remains Chairman Emeritus/large holder historically — founder DNA intact (verify current role/ownership; not in the on-disk filings).
• Willett fully exited the board March 2, 2026 (triggering the $5.0M SBC forfeiture credit) — a clean break enabling Moschner's reset.
• Board refresh (the tell): six independent directors added since 2021; the Feb 2026 additions are heavyweight — Dr. Sami Atiya (62), ex-ABB President of Robotics & Discrete Automation and Group Executive Committee member (30 yrs automation/robotics/AI), and Chris Donato (60), President & CRO of Zendesk (enterprise go-to-market). This is a board being rebuilt around deep industrial-automation domain + sales execution — precisely Moschner's agenda. Reads as governance-driven (possibly investor-nudged) modernization, not a hostile 13D campaign (no named activist surfaced in search; "activist shareholder activities" appears only as a boilerplate risk factor ).
• Capital-allocation history (disciplined, shareholder-friendly): zero debt; consistent dividend (raised to $0.085/q in Q4'25, ~$54.6M paid in 2025); aggressive buyback — $151.2M repurchased in 2025 (4.234M shares), a fresh $500M authorization Feb 2026 on top of the nearly-exhausted prior $500M program. M&A is bolt-on and rational: Moritex ($270M EV, 2023) for optics — and notably, the new team divested the Moritex Japan trading business (April 2026, ~$12M, was ~2% of revenue), i.e. it is pruning the prior regime's deal to keep the portfolio clean. That is a constructive signal about capital-allocation honesty.
• ROE/ROIC: FY2025 net income $114.4M on equity $1.49B → ROE ~7.7% — optically low because the balance sheet carries ~$642M idle cash + $383M deferred-tax assets + $467M goodwill/intangibles. On the cyclical EPS recovery, returns improve mechanically.
• Red flags: none material on comp/related-party in the on-disk filings; the main governance watch-item is simply execution risk on an unproven CEO selling an ambitious "double the customer base" story into a cyclically favorable tape.

Lens 10 · Forensic Red Flags

Acting as a forensic analyst: the books are clean. This is a conservative, cash-rich balance sheet with high earnings quality — the risks are cyclical and tax, not accounting.

• Earnings quality — strong. FY2025 OCF $245.5M vs net income $114.4M (OCF/NI ≈ 2.1x); FY2024 $149.1M, FY2023 $112.9M. Cash conversion exceeds earnings — the opposite of a manufactured-earnings profile. (Part of FY25's OCF lift is a +$40.1M non-cash deferred-tax add-back from the OBBBA charge — strip it and OCF/NI is still ~1.8x.) FCF ≈ $236.8M FY2025.
• Revenue recognition. Point-in-time for standard products; validation-gated for application-specific solutions — conservative (revenue deferred until customer acceptance). Unbilled revenue rose to $17.0M (from $3.1M) — small and consistent with more solution work; not a concern at this scale.
• Receivables / inventory vs. revenue. Receivables ~flat ($146.7M) on +9% revenue — clean. Inventory deliberately reduced ($157.5M → $137.9M → $135.5M in Q1'26) with a transparent $15.6M FY25 E&O charge disclosed and explained. Management is writing inventory down, not building it — no stuffing.
• SBC. $48.5M FY2025 (declining from $52.4M/$54.8M) ≈ 42% of net income — material, the standard tech caveat. It flatters non-GAAP (adj EPS $1.02 vs GAAP $0.68) and the buyback partly exists to offset its dilution. Watch, but trending the right way; $68.9M unrecognized, ~1.8yr.
• Goodwill / intangibles. $382.8M goodwill + $67.1M intangibles post-Q1'26, all from Moritex; no impairment, and the Japan-trading divestiture test left "substantial excess of fair value over carrying value". Customer-relationship intangible amortizes over 15yr — long, but disclosed.
• Tax (the one genuinely messy line). FY2025 effective rate 37% (vs 19% FY24) purely from the $33.2M OBBBA discrete charge raising a deferred-tax liability; normalized ~17%. Pillar Two top-up tax effective Jan 1, 2026 (US safe-harbor expired end-2025) — immaterial in Q1'26 but a structural creep risk to the rate. ~$383M deferred-tax assets on the balance sheet is large and worth monitoring, but valuation allowance is trivial ($2.7M).
• Controls / auditor. ICFR effective; no disagreements with accountants; auditor Grant Thornton LLP. No going-concern or material-weakness language.

Regulatory findings (required):

• SEC Litigation Releases: none naming Cognex (2021-06-22 → 2026-06-22).
• SEC AAERs: none.
• Item 3 Legal Proceedings (10-K) & Q1'26 10-Q Note 7: only "various claims and legal proceedings generally incidental to the normal course of business… any liability… will not have a material adverse effect" — i.e. no material litigation disclosed.
• Non-SEC (FTC/DOJ/FDA/etc.): web search surfaced no material enforcement action, consent decree, or fine against Cognex.
• Conclusion: No material regulatory or legal findings — verified via SEC EDGAR EFTS (LR, AAER), web search, and 10-K Item 3 / 10-Q Note 7 as of 2026-06-22.

---

Phase D — Project & stress-test

Lens 11 · Forward Projection (FY2026 / FY2027 / FY2028 EPS)

Built bottom-up from FY2025 actuals + Q1'26 print + Q2'26 guide. All outputs with arithmetic; consensus is.

Anchor facts: FY2025 revenue $994.4M, adj EPS $1.02. Q1'26 revenue $268.4M (+24% rep / +21% cc); Q2'26 guide $280-300M (+16.5% mid) & adj EPS $0.40-0.44. H1'26 run-rate implies ~$560M; seasonality typically H2 ≥ H1 for Cognex. Diluted shares shrinking ~1.5%/yr on buyback. Normalized tax ~17-19% (Pillar Two creep).

Base case (FY2026): revenue ~$1.13B; GM ~70%; OM ~21%; adj EPS ~$1.45 — essentially in line with the ~$1.43 Street consensus.

• Bull (FY2026): revenue ~$1.18B (cycle re-accelerates, China/semi sustain +30%+), GM ~71%, adj EPS ~$1.60.
• Bear (FY2026): large-logistics customer pauses again / CE air-pocket → revenue ~$1.05B (+6%), GM ~68% (mix), adj EPS ~$1.20.

FY2027 base: revenue ~$1.26B (+11%), adj EPS ~$1.75. FY2028 base: revenue ~$1.38B (+10%), adj EPS ~$2.05. These assume the cycle does not roll over and the salesforce transformation adds a few points of share — an above-trend path the ~34x multiple already embeds.

Brier forecast (NOT logged — --watchlist rule skips forecast.ts create): the base call I would track is "CGNX FY2026 non-GAAP diluted EPS ≥ $1.45, p≈0.55" — coin-flip-plus, because consensus is ~$1.43 and Cognex has beaten 7 straight quarters, but H2 is cycle-dependent and the comps stiffen.

Lens 12 · Bull vs Bear

Bull case. Cognex is the #1/#2 global franchise in a secularly growing (~11% CAGR), structurally under-penetrated category (logistics automation is "early stage," mostly manual today ). The 2023 trough is behind it; Q1'26 (+21% cc, GM 71%, 7th straight up-quarter) proves operating leverage is intact on the way up. A credible young CEO + automation-stacked board are running a real cost/KPI transformation while guiding for continued double-digit growth. Fortress balance sheet (zero debt, ~$642M cash) funds a $500M buyback that shrinks the count every quarter. AI vision (OneVision, deep/edge learning) is a genuine product cycle that expands the addressable base ("double the customer base"). If the logistics + semiconductor + CE cycles all run together, FY26-28 EPS compounds ~20%+ and the multiple is defensible.

Bear case (permanent-impairment candidates). (1) Customer concentration + capex cyclicality: one customer is now 15% of revenue and the whole model is book-and-ship into discretionary automation capex — the 2021→2023 −17% revenue collapse can recur if Amazon/e-commerce digests again. (2) AI commoditizes the software premium: Cognex itself warns that "free or low-cost" open-source AI vision and well-resourced tech entrants lower barriers — the moat is shallower in software than in Keyence's proprietary sensor/measurement stack, and that is precisely where value is migrating. (3) The multiple is the risk: ~34x forward EPS / ~32-48x EV/EBITDA prices a flawless multi-year re-acceleration with ~16% operating margins; any cycle wobble re-rates it toward Zebra's ~12x.

Pre-mortem (18 months out, thesis broke): H2'26 large-logistics orders flatten as the customer "digests" again; China/CE give back the Q1 spike (FX reverses too); a couple of beats become a miss; the AI narrative gets repriced as open-source tools erode low-end pricing; the stock round-trips from ~$66 to the high-$40s as the ~34x multiple compresses to ~22x on flat-to-down EPS revisions.

Are multiples too high? Yes, on an absolute and relative basis for a ~16%-operating-margin, customer-concentrated, cyclical hardware-led name — the valuation is a quality + cycle-recovery + AI-optionality multiple, and at least one of those three has to keep delivering every quarter to hold it.

Contrarian view (what the market is refusing to see): The Street is extrapolating the Q1'26 +21% cc and "7 straight quarters" into a smooth secular ramp, but this is still fundamentally the same logistics-capex cycle stock that fell 17% in 2023 — the customer concentration is higher now, not lower, and the very AI wave being celebrated is the thing most likely to erode the software pricing power over a 3-5yr horizon. The bull and bear both depend on Amazon's warehouse capex plans more than on anything Cognex controls.

Lens 13 · Devil's Advocate (short-seller)

Dismantling the bull case:

• Where revenue is concentrated: ~85% in four cyclical end markets; one customer = 15% (rising); ~59% of revenue outside the Americas with heavy FX and China/Taiwan exposure. A single large-customer capex pause is a double-digit revenue event — and it has happened twice this decade.
• Why the moat is weaker than bulls think: the durable moat is in optics/sensors/brand; the growth story is in AI software, where Cognex explicitly concedes open-source and free tools are lowering entry barriers. Margins are ~16% operating vs Keyence's ~51% — Cognex does not have Keyence's pricing power, yet trades near Keyence's earnings multiple.
• Most dangerous competitor bulls underestimate: Zebra Technologies — explicitly rolling up machine vision (Photoneo, Matrox Imaging, Aurora) to offer a one-stop logistics/AIDC portfolio to the exact same warehouse customers, at ~12x forward vs Cognex's ~34x. Plus Keyence's relentless direct-sales machine and commoditizing open-source/foundation-model vision from hyperscalers.
• Worst capital-allocation history: mostly disciplined — but Moritex required a partial divestiture three years later, a tacit admission the original deal over-reached; and the heavy buyback at a high multiple to "offset dilution" from a 42%-of-NI SBC program is shareholders funding employee comp.
• Assumptions that must hold for ~$66: double-digit revenue growth for 3+ years, GM holding ~70%, operating margin expanding toward the high-teens/low-20s, the large customer keeping/raising spend, and AI not compressing software pricing. Miss two and the multiple halves.
• −20-30% growth-disappointment scenario: if FY26 growth comes in at +6% (bear) not +14% (base), adj EPS ~$1.20 not ~$1.45; at a de-rated ~20x that's ~$24 of value vs ~$66 today — i.e. the downside on a cycle stumble is on the order of −50-60%, far larger than the ~+15% to the $76 average target.
• Single permanent-impairment scenario (and plausibility): durable commoditization of industrial AI vision by open-source/foundation models structurally caps Cognex's software premium and resets it to a mid-teens-multiple hardware grower. Plausibility: moderate over 3-5yr, low over 12-18mo — Cognex's brand/installed-base buys time, but the direction of travel is real.

Lens 14 · Management Questions (ordered by information value)

1. Customer concentration: the 15%-of-revenue customer — is that one logistics buyer, what is its FY26 capex commitment vs FY25, and how exposed is the recovery if it digests again?
2. AI commoditization: as foundation-model and open-source vision improve, what specifically defends Cognex's software pricing in the next 3 years — and which revenue tiers are most at risk?
3. Margin durability: is the Q1'26 71% gross margin the new normal or a mix peak? What's the sustainable operating-margin target once the salesforce-transformation investment fully runs through?
4. "Double the customer base in five years": what's the implied new-customer count and revenue per new customer, and what are the leading-indicator metrics you'll report so we can track it?
5. Zebra & one-stop competitors: how do you compete with Zebra's bundled vision+AIDC portfolio for warehouse share without sacrificing price?
6. Salesforce transformation ROI: quantify the revenue uplift and CAC improvement vs the old emerging-customer model — what proof do you have it's working beyond the macro cycle?
7. Capital allocation: with ~$642M cash, zero debt, and a fresh $500M buyback — at ~34x forward earnings, why is buyback the best use of capital vs larger, transformative M&A?
8. Moritex: after divesting the Japan trading arm, is the rest of Moritex meeting the original deal thesis, and what did you learn about your M&A diligence?
9. China/Taiwan: what is the contingency if Taiwan IC supply is disrupted or US-China export controls tighten on your China optical plants and customers?
10. Contract-manufacturing concentration: post-2022 fire, how diversified is EMS now — is there still a single primary contractor that is a single point of failure?
11. Pillar Two / tax: what is the steady-state effective tax rate once Pillar Two top-up fully applies, and how much of the ~$383M deferred-tax asset is at risk?
12. Semiconductor + China inflection: Q1'26 China +40% and semiconductor strength — how much is durable AI-chip-driven demand vs a restocking bounce?
13. SBC: SBC is ~42% of net income; what's the multi-year plan to bring it down as a share of earnings?
14. New board: what specifically will Sami Atiya (ex-ABB robotics) and Chris Donato (Zendesk GTM) change about strategy and sales execution?
15. Cyclical guardrails: what early-warning signals would make you cut opex fast, and how do you avoid the over-build that preceded the 2022-23 reset?

---