Cognex Zen Horizon Framework Deep Dive: The Eyes of Machine Vision, a Great Business at an Expensive Price

2. Vertical History: From a Tufts Professor's Lab to a "Standard-Component Supplier" for Machine Vision

2.1 Origin: A University Professor and "Machines That Can See"

Cognex was founded in 1981 by Tufts University electrical engineering faculty member Dr. Robert Shillman, known as "Dr. Bob". The name came from "Cognition Experts." The company was born in the early days of industrial automation, when machines could move and compute, but could not "see." Shillman's bet was that enabling machines to understand images would become a manufacturing necessity. That judgment has been validated repeatedly over more than four decades: from early character recognition for semiconductor and electronics plants to today's parcel-barcode reading in e-commerce warehouses, the central line has never changed: turn vision into an industrial component that can be embedded in production lines.

The company listed on NASDAQ in 1989. Shillman was more than the founder; he shaped a distinctive corporate culture, with employees calling themselves "Cognoids," Shillman serving as chief culture officer, and the "Dr. Bob" image deeply embedded. This will return in the governance section, because the people running Cognex today are no longer Shillman.

2.2-2.4 Three Stages: Hidden Champion of the Electronics Era -> Logistics Boom -> Cycle and M&A

The clearest way to compress Cognex's history is into three stages:

Stage 1: hidden champion in electronics and semiconductors, from IPO to around 2017. This was Cognex's "smile-curve" period. It captured inspection steps in consumer-electronics and semiconductor manufacturing through machine vision. The largest single customer was Apple for many years, with machine vision used in iPhone screen and OLED assembly inspection, and financial media named it directly more than once. This created its high-margin, high-share leadership, while also planting the fragility of customer concentration plus the consumer-electronics cycle. In 2018, large-customer order cuts plus a China slowdown drove the stock down 37% in a single year. 2019 revenue of $726 million was even below 2017's $766 million. This is the first key to understanding Cognex: it has already proven that one cycle can interrupt the "growth stock" narrative.

Stage 2: logistics boom, 2018-2021. E-commerce warehouse automation became the new engine. Amazon-like customers built sorting centers aggressively, and every line needed Cognex barcode readers and vision systems. Together with the post-pandemic automation frenzy, 2021 revenue reached an all-time peak of $1.037 billion, up 28%, operating margin hit 30.4%, and the stock topped out around $102. The market then believed Cognex was a pure beneficiary of an "automation supercycle."

Stage 3: cyclical downturn plus M&A integration, 2022 to present. Logistics-customer investment peaked and rolled over, consumer-electronics replacement demand weakened, and EV-battery investment softened. Under this triple hit, revenue retreated over three years to a trough of $838 million in 2023, down 19% from the peak, while operating margin was cut roughly in half to 15.6%. Management did two major things in response. First, in August 2023, Cognex acquired Japanese optics company Moritex for about ¥40 billion / $275 million in all cash, the largest acquisition in its 42-year history. Then-CEO Willett described it as "Cognex's vision system is the brain, and Moritex's optics are the eyes". Second, in 2024 the company completed a CEO transition and began portfolio optimization and cost reduction. The key fact in this stage is that by 2025, revenue was $994 million, still below the 2021 peak after four years. Moritex is a double-edged sword. It supported reported growth in 2024, up 9%, but only +1% excluding Moritex. It also brought in a lower-gross-margin optics business, one reason 2024 gross margin stepped down. In April 2026, Cognex divested the Japan-focused low-margin trading business attached to Moritex, with about $16 million of revenue and a sale price of about $11.9 million. That is one source of the recent margin improvement, and it is portfolio optimization rather than organic demand.

3. Vertical Financial Review: A High-Quality Cash Machine, Yet "Revenue Grows While Profit Does Not"

Spread out the five-year financials, on a GAAP basis, using data from SEC filings and quarterly 8-K releases:

This table hides three business facts that matter far more than the numbers themselves:

First, growth comes from cycles and M&A, not structural compounding. The five-year revenue CAGR is only about 3.5%, and revenue remains below the 2021 peak after four years. Even more striking, EPS has declined over five years: $1.56 in 2021 to $0.68 in 2025, almost halved. This is not a compounding company. It is a mature company that rises and falls with cycles, with M&A helping hold the line.

Second, gross margin is structurally declining. It fell from 73.3% in 2021 to 66.9% in 2025, down about 6 percentage points over five years. The company acknowledged in its 10-Q that the causes were unfavorable business and regional mix, tariffs, and pricing pressure. This partially falsifies the narrative that a machine-vision leader has strong pricing power. If pricing power were unbreakable, gross margin would not step down in one direction.

Third, cash quality is extremely high, and this is its hardest card. Accounting profit understates Cognex's cash generation. The business is extremely capital-light, with 2025 capex of only $8.7 million, less than 1% of revenue. Free cash flow rose year by year: FY2023 $90 million -> FY2024 $134 million -> FY2025 $237 million, a 23.8% FCF margin, equal to 2.1× that year's net income, helped by working-capital release and large non-cash items.

The balance sheet is textbook clean: at the end of 2025, cash plus investments were $642 million, with zero interest-bearing debt. The roughly $76.6 million of "debt" on the statements is operating leases. By Q1'26, Cognex still had $622 million, or net cash of about $3.7 per share. Goodwill was $386 million, brought in by Moritex, with no impairment over five years. Capital returns have fallen with the cycle: ROE declined from 19.6% in 2021 to 7.67% in 2025. Independent ROIC estimates are about 9% on an all-equity basis and about 16% on an operating basis excluding net cash. The metric is sensitive to methodology, but the direction is consistent: capital returns have compressed materially from the peak. A third party therefore described Cognex as having "ROIC below WACC and destroying value." That figure, 5.52%, uses an opaque and non-replicable methodology, so it should be treated only as directional reference, not as a firm conclusion.

Operating leverage cuts both ways. From peak to trough, revenue fell 19%, while operating income fell 58.5%, an amplification of about 3×. Fixed costs, with R&D routinely around $139 million / 14% of revenue, are rigid, and the high-gross-margin model magnifies profit damage in downturns. The good news is that the reverse also holds: Q1'26 revenue growth of +24% lifted operating margin from 12.1% a year earlier to 22.3%. This is the core of the current bull narrative, and also the fragility in the valuation, because upside leverage depends on the assumption that the cycle is moving up.

4. Share Price and Valuation History: Structurally High-Multiple Stock, Currently Near the Top of Its Historical Range

Cognex has never been a "cheap stock." Looking back at year-end TTM PEs from 2019 to 2025, year-end price divided by that year's GAAP EPS, the range was roughly 37× to 77×. When cyclically depressed earnings reduced the denominator, PE actually surged, reaching 62× at the 2023 trough and 57× in 2024. This is a franchise stock to which the market has long assigned high multiples.

Yet "long-term high multiple" does not mean "not expensive now." Three facts sit side by side:

• The current 76× TTM PE is at the upper end of the range since 2019, close to the 2020 extreme.

• The share price almost doubled in the first half of 2026: $35.4 at the end of 2024, $35.9 at the end of 2025, and $64.67 now. The 52-week range is $29-$71.8, with the low at $27 in April 2025. Current valuation reflects "revenue +24% recovery plus AI-vision rerating," rather than a long-term normal price.

• The current price implies a forward PE of about 44×, based on FY2026 adjusted EPS consensus of about $1.47, pricing a company with a five-year revenue CAGR of 3.5% like a high-growth stock.

The market has given Cognex various labels: automation-supercycle growth stock in 2021, cyclical casualty returned to reality in 2024, and AI-vision reacceleration reversal stock in 2026. The repeated changes in valuation center reflect shifts in market preference more than changes in business quality. That is precisely the warning: the multiple paid today is a bet that market sentiment can continue, not just that fundamentals are good.

5. Business Model and Moat: The Real Moat Is "Software + Embedding," but Its Edges Are Being Worn Down

5.1-5.2 Revenue Mix and Operating Leverage

Cognex has a single reportable segment and discloses by end market and geography. The FY2024 end-market mix was roughly logistics 23%, automotive 22%, consumer electronics 17%, and semiconductors 11%, with the remaining 27% in diversified markets such as packaging, food and beverage, and medical. Geographically, about 67% of revenue comes from outside the United States: Europe about 24%, Greater China about 18%, and the Americas about 33%, with the Americas rising to 45% in Q1'26 because of strength there.

This structure ties Cognex's fate to three legs: e-commerce capital expenditure in logistics, the consumer-electronics replacement cycle, and automotive / EV production-line investment. All are highly capex-cyclical end markets. Combined with the high operating leverage described above, Cognex is essentially a high-beta automation-capex stock.

5.3 Moat: Which Parts Really Hold

Using the standard of "still effective over 3-5 years in an unfavorable environment," Cognex has three real moats:

1. Vision software plus decades of imaging know-how. This is the strongest one. The difficulty in machine vision is not camera hardware; it is enabling machines to recognize correctly and consistently amid production-line noise. Cognex's In-Sight, VisionPro, DataMan, and deep-learning edge learning, which was already commercialized in 2022 and is not a new 2026 development, reflect four decades of algorithms and application libraries. A 68% gross margin is itself financial evidence of this moat.

2. Switching costs after embedding into production lines. Once a vision system has been integrated into a high-speed production line, tuned, and proven on yield, changing suppliers risks downtime and revalidation. This "embed and lock in" dynamic is Cognex's main weapon against low-priced Chinese competition. Foreign suppliers commonly bind customers through end-to-end sales, installation, and integration services.

3. Brand and global service network. In high-mix, high-reliability settings such as semiconductors and automotive safety components, Cognex remains a default option.

But the moat's edges are being worn down. Three pieces of evidence matter. First, gross margin has fallen 6 percentage points in five years, and the company acknowledged "pricing pressure." Second, Cognex itself lists "large technology companies plus open-source tools" as threats in its 10-K risk factors and admits that AI and open-source platforms are commoditizing machine vision, lowering barriers to entry and intensifying pricing pressure. This is the company's own admission that general AI is eating into the software barrier. Third, Chinese local vendors, including Hikrobot with a global share of 6.2% and OPT at 2.1%, continue to replace foreign brands on price in mid-end applications. The moat remains, but it is no longer the "wide and deep" moat of 2021.

5.4 Management and Governance: Smooth Handoff, Clean Governance, but Light Insider Commitment

On governance, Cognex has several notable features:

• Smooth CEO transition. Matt Moschner became CEO on 2025-06-27, succeeding retiring Robert Willett. Willett stayed on as director and adviser, and the transition agreement had restrained economic terms and no rich severance, a governance positive. Moschner joined in 2017 and was promoted internally from product-line roles. He has no external public-company CEO experience, so execution continuity is strong, while strategic force still needs proof. CFO Dennis Fehr joined from the Siemens system in 2024.

• The founder has exited operations. Shillman, about 80, stepped down as chairman and left the board in 2021. He is now only chairman emeritus / adviser. He still owns about 3.41%, around 5.7 million shares worth $190 million, making him the largest individual shareholder.

• The share structure is clean: one class of common stock, one share one vote, no dual-class structure, with institutional ownership around 95.5%, including BlackRock at 9.76%. There is no VIE, family control, or super-voting right. This is a governance advantage relative to many technology stocks.

• But current management's skin in the game is thin. Excluding the founder, who is no longer operating the company, current executives plus directors together own only about 0.4-0.5%. Insiders have been net sellers over the past year: former CEO Willett sold $10.3 million in February 2026 and another $3.3 million in May, with an additional Form 144 planned sale of 420,000 shares; chief accounting officer MacDonald sold $9.3 million in May. The only meaningful buying signal was CFO Fehr adding 6,570 shares after joining in 2024. Much of the selling was tied to option exercise and may include tax or expiration factors, making intent hard to judge. Taken together, the degree to which operators are betting real money is not high.

The capital-allocation track record is broadly rational: five-year buybacks of $664 million plus dividends of $245 million, with buybacks 2.7× dividends and opportunistic timing, light at the 2021 high and accelerated after the pullback. Share count declined by a net 5.9% over five years. Moritex is the only large acquisition. It is still being integrated, with the low-margin tail already divested, and has not yet proven clear value creation.

6. Industry and Cycle: A Good 8% Growth Track, but Cognex Is the High-Beta Name Within It

Machine vision is a structurally growing track, although growth is less dramatic than the narrative suggests. Several institutions give different estimates: Grand View Research puts the total 2024 market at $20.4 billion with a 13% CAGR; MarketsandMarkets puts the 2025 industrial machine-vision market at $15.8 billion with an 8.3% CAGR; BCC gives 9.1%. Using the midpoint, this is a mature growth track with high-single-digit to low-teens growth, good enough to be attractive, but insufficient by itself to support high-growth pricing at 44× forward PE.

Drivers include warehouse automation, with warehouse-automation TAM growing 17%+ as the fastest subsegment, semiconductor inspection, and consumer-electronics replacement cycles. Cyclically, Cognex is exposed at the same time to capital-expenditure cycles, consumer-electronics cycles, and semiconductor cycles, which explains the severe volatility in its history. Current position: early-to-mid recovery from the 2023 trough.

Geopolitics and tariffs are real variables. Revenue outside the United States is 67%, and the company acknowledged tariff pressure on gross margin, while saying it has "substantially mitigated" it and that there is no material EPS impact. Greater China is about 18% of revenue, with weak growth in recent years, flat in FY2024 and under pressure in single digits in FY2025. It also faces local substitution, procurement migration, and Taiwan Strait supply-chain risk, all listed in the 10-K. These are not one-off shocks; they are long-term structural constraints.

7. Horizontal Peers: One of the Leaders, but Far from the Most Profitable Player in the Industry

Machine vision is a competitive track with many peers, a Scenario C setting. The most representative comparison is Japan's Keyence (6861.TSE), together with barcode / vision player Zebra (ZBRA.US), Teledyne (TDY.US), Italy's Datalogic (DAL.MI), and China's Hikrobot, under Hikvision 002415.SHE, plus Omron (6645.TSE).

The horizontal comparison clarifies Cognex's real ecological niche: it is one of the leaders, but it is certainly not the most profitable company in the industry. Keyence's FY2024 revenue was about ¥1.06 trillion, or $7 billion+, with an operating margin around 50% and a net margin of 37.6%. Its scale is 7-8× Cognex, and its operating margin is roughly 3× Cognex's 16%. Keyence relies on direct sales, fabless operations, and a consultative sales model with very high markups. The phrase "Cognex is a machine-vision leader" needs a discount. It ranks second in global share, but Keyence overwhelms it in profitability and scale, while continuing to pressure its share.

The valuation contrast is even more cautionary, based on same-day prices and standard metrics:

The conclusion is obvious: Cognex is the most expensive name in this group. Its PE premium is 2-3× versus Zebra, Teledyne, and Datalogic, and about 1.7× even versus the strongest peer, Keyence, at 76× versus 43.6×. The only peer in the same P/S neighborhood is Keyence, and what supports that multiple is an operating margin of about 50%, while Cognex has only 16-22%. Put differently, the market is giving Cognex something more than "peer average": it is pricing it as the most expensive peer plus another step higher. For this premium to hold, investors must bet that growth and margins will keep beating expectations.

8. Current Fundamentals and Bull-Bear Split: The Cycle Is Truly Turning, but the Debate Is Durability

8.1-8.2 Recent Quarters: Clear Reacceleration

Quarterly revenue growth has clearly formed an accelerating curve: Q1'25 +2% -> Q2 +4% -> Q3 +18% -> Q4 +10% -> Q1'26 +24% (constant currency +21%). Q1'26 revenue was $268.4 million, GAAP gross margin was 71.1%, operating margin was 22.3%, adjusted EPS was $0.34, up 113%, and adjusted EBITDA margin was 26.9%. Q2'26 guidance is revenue of $280-300 million and adjusted EPS of $0.40-0.44. Profit-side elasticity is much greater than revenue-side elasticity, driven by portfolio optimization through the divestiture of low-margin business, cost reduction with a target of another $35-40 million of annualized cost cuts by year-end 2026 and a 25% adjusted EBITDA margin, and operating leverage.

The drivers are logistics, with double-digit growth for 9 consecutive quarters, as the base, plus semiconductor strength from the second half of 2025, Q1'26 up 20%+, and improving consumer electronics. Q1'26 was already broad-based. The market is now trading this line of "cyclical reversal plus structural margin lift."

8.3 Bull-Bear Split: Everything Depends on Whether This Acceleration Can Last

Specific bull evidence: 1. the cycle has just turned from a three-year downturn, with logistics, semiconductors, and consumer electronics improving at the same time; 2. margins have improved sequentially for 7 straight quarters and have a clear cost-reduction path, so this is not a one-quarter one-off; 3. FY2025 FCF of $237 million set a record high, the balance sheet has zero debt, and buybacks continue to reduce share count; 4. AI-vision products, including In-Sight 3900 and the OneVision cloud platform, plus the long-term physical-AI option.

Specific bear evidence: 1. revenue is still below the 2021 peak after four years, the five-year revenue CAGR is only 3.5%, and EPS has declined. The +24% recovery is built on an extremely low base after three down years; 2. gross margin has fallen 6 percentage points over five years, and the company acknowledges pricing pressure plus the commoditization threat from open-source AI; 3. 44× forward PE / 76× TTM PE is fragile pricing for a stagnant-growth company, and multiple independent sources, including StockStory's "better stocks to buy" and GuruFocus's "significantly overvalued by about 35%", have issued overvaluation warnings; 4. even the CEO leaves room on durability: demand looks strong, but how long can it last? We will have to see, and market uncertainty is high.

The fact-level verdict: cyclical recovery from a low base and structural improvement from cost reduction plus portfolio optimization both truly exist in this acceleration, but a "structural growth inflection point" has not yet been proven. Management itself guided logistics growth to "mid-to-high single digits" and automotive to "flat to low single digits," which shows they do not see +24% as the new normal.

9. Valuation Analysis: Current Price Has Priced In the Recovery, Leaving Zero Margin of Safety

9.1-9.2 History and Peers

Historical percentile: the current 76× TTM PE is at the top of the range since 2019; the 44× forward PE is also far above its historical forward midpoint. Peers: the premium is 2-3× versus Zebra, Teledyne, and Datalogic, and about 1.7× even versus the strongest peer, Keyence, as shown in Section 7. Valuation discipline warning: the fact that machine-vision peers are all not cheap does not make Cognex cheap. When the entire track carries a high narrative premium, absolute valuation must still be tested independently.

9.3 Absolute Valuation: Cash-Flow Look-Through Plus Three Scenarios

Cash-flow look-through, required before any method: Cognex's accounting earnings seriously understate cash generation. FY2025 OCF / net income was 2.15×, and the FCF margin of 23.8% was far above the net margin of 11.5%. Maintenance capex is extremely low, less than 1% of revenue, and owner earnings are close to or even above net income. Therefore valuation should use normalized earnings / FCF, rather than reported GAAP EPS of $0.68, which is depressed by the cycle and restructuring.

Normalized earnings anchors: GAAP TTM $0.85, depressed by cycle and restructuring, < adjusted FY2025 $1.02 < consensus FY2026 adjusted $1.47 / FY2027 $1.76; TTM FCF per share is about $1.45; the 2021 peak already proved earnings power of about $1.58 EPS. Three scenarios, with endpoints matching the front-matter valuation ranges:

The current $64.67 sits between the upper end of the base range, $63, and the lower end of the bull range, $76. The market has already priced in a move from "base" toward "bull." One tension must be made explicit: the multiples method and absolute DCF diverge sharply. Running a DCF on the same normalized FCF, with FCF around $215 million, a 9% growth rate fading to 3%, and WACC of 8.5%, yields only about $35 per share, far below the $54-63 base range from the multiples method. This is typical for high-multiple franchise stocks: the market pays a premium far above DCF present value for "franchise plus growth option." The two are presented side by side and are not forced into reconciliation. The multiples method represents the "relative price the market is willing to pay"; DCF represents the "absolute price of cash."

9.4-9.5 Expectations Gap and Margin-of-Safety Recheck, Independent Discipline

The market's current implied expectations are that this +24% growth is not a low-base flash, margins can exceed 25%+, and the machine-vision leader premium can persist. If any one of these is falsified, such as revenue growth slowing over the next one or two quarters or margins missing guidance, valuation compression could be triggered.

Margin-of-safety recheck, independent conclusion:

• The current price is at a large premium of 60%+ to the bear scenario of $34-40. There is zero margin of safety.

• The most fragile assumption is that "44× forward PE can be maintained." If the multiple is cut by 30% back to about 30×, still higher than most peers, base value falls to about $45.

• If earnings do not grow over the next 3 years, annualized return at the current price would be far below the 10-year Treasury yield. This purchase price has no margin of safety.

• This is a typical "great company but expensive price" case: it is worth waiting for a better price rather than buying now.

• Conclusion on margin-of-safety sufficiency: none.

10. Risk Analysis

Business risks: 1. cycle plus customer concentration, high / high. Consumer electronics, including the Apple chain, and logistics, including large e-commerce customers, swing with capex. In 2021, the largest customer accounted for 17% and the second largest for 15%. A single 2018 cycle already produced a -37% stock decline. Observable indicators are quarterly revenue growth and large-customer end-market capex. 2. moat erosion, medium / high. Open-source AI vision plus Chinese local price competition; gross margin below 65% would be an alarm.

Financial risks: overall low, given zero debt and strong FCF. The main risks are potential impairment of Moritex goodwill of $386 million if the optics business underperforms, and FX risk from 67% overseas revenue.

Valuation risk, the highest: with 44× forward PE / 45× EV-EBITDA, valuation compression could be large if rates rise or risk appetite shifts. A theme fade, with physical-AI narrative cooling, would directly compress the multiple.

Governance and external risks: continuing insider net selling, medium / medium; strategic uncertainty after CEO transition, medium / medium; Greater China at 18% of revenue, exposed to geopolitics / tariffs / decoupling, medium / high.

11. Catalysts and Tracking Indicators

Positive catalysts: earnings continue to beat expectations, margin reaches the 25% EBITDA target, logistics / semiconductor orders accelerate, buybacks increase, AI-vision products scale. Negative catalysts: guidance cut, revenue growth slows, falsifying the "structural" story, gross margin declines again, large customers cut orders, valuation sells off, insiders continue large sales.

Tracking dashboard, with 6 priority items: 1. quarterly revenue growth, where the normal range is single digits; sustained double digits signal structural improvement, while negative growth signals a cyclical peak; 2. adjusted EBITDA margin, especially whether it can hold above 25%+, while a fall below 20% falsifies the margin story; 3. gross margin, with 68%+ healthy and two consecutive quarters below 65% indicating moat breakage; 4. logistics end-market growth, with management guiding mid-to-high single digits and actual results needing to match; 5. global share versus Keyence / Hikrobot, where share loss means intensifying competition; 6. insider transactions, Form 4, sustained net selling versus insider buying. For valuation, track forward PE, with >40× overdrawn, <30× approaching reasonable, and a return to the $40-49 range creating margin of safety.

12. Zen Horizon Synthesis: Good Business, Expensive Price, Wait for a Better Price

Vertically, Cognex has truly proven its ability to turn machine vision into an industrial standard component. Over four decades, it built high-margin, high-share leadership in inspection steps across electronics, semiconductors, and logistics, based on software know-how plus the real moat of embedded switching costs. Its cash-generation ability, with a 24% FCF margin, zero debt, and $3.7 per share in net cash, is real proof of asset quality. But it has also proven something else: it does not qualify as a "high-quality compounder." Revenue remains below the peak after four years, EPS has declined over five years, and gross margin has fallen in one direction. It is a mature franchise that moves up and down with cycles and uses M&A and cost cuts to support the picture, rather than a stable compounding machine.

Horizontally, its real advantage is being the global number two in machine vision behind Keyence, while its weaknesses are structural: Keyence overwhelms it in profitability and scale, Chinese vendors are eating at the bottom of its share, and general AI commoditization is wearing down the software barrier. Its strengths are real, but the word "leader" carries less weight than the market narrative implies.

Is the current valuation rewarding past success or pulling forward the future? The answer leans toward the latter. With 76× TTM / 44× forward PE and a 2-3× premium to peers, the market is pricing a low-base cyclical recovery after a three-year downturn as structural high growth. The market's most likely mistake is treating the slope of a cyclical rebound as the durability of structural growth, while management itself remains cautious about durability.

Key future variables: over 1 year, whether this acceleration continues, meaning whether revenue growth slows and whether margin can reach 25%; over 3 years, whether the moat can withstand open-source AI plus Chinese price competition, and whether gross margin can hold; over 5 years, whether the physical-AI / embodied-robotics option can actually translate into revenue, since it is currently pure narrative.

12.1 Bull and Bear Cases, Both Traceable to Prior Sections

Bull case: 1. extremely capital-light cash machine, with FY2025 FCF of $237 million / 23.8% margin, zero debt, $3.7 per share in net cash, and a five-year buyback-driven share-count reduction of 5.9%, Section 3; 2. clear cyclical reversal, with quarterly revenue growth accelerating from +2% to +24% and margins improving for 7 consecutive quarters, Section 8; 3. real moat remains, with 68% gross margin, machine-vision software plus embedded switching costs, and number two global share, Sections 5 and 7; 4. long-term physical-AI option, as machine vision is the "eyes" of automation / embodied intelligence, backed by real industrial revenue today, Section 1.

Bear case: 1. structurally stagnant growth, with revenue still below the 2021 peak after four years, a five-year revenue CAGR of 3.5%, and negative EPS growth, Section 3; 2. valuation extremely overdrawn, with 76× TTM / 44× forward PE, a premium to every comparable company, mostly 2-3×, near the top of the historical range, and a doubled share price in half a year, Sections 4, 7, and 9; 3. moat edges are wearing down, with gross margin down 6pct over five years, management acknowledging pricing pressure plus open-source AI commoditization threat, and Keyence's 50% margin overwhelming it, Sections 5 and 7; 4. historical evidence of cycle plus customer concentration, including a single-cycle -37% drop in 2018, 2019 revenue below 2017, and the largest customer once at 17%, Sections 2 and 10; 5. thin management skin in the game plus insider net selling, with current management owning about 0.5% and insiders net selling over the past year, Section 5.

12.2 Pre-mortem: A 50% Loss Scenario Three Years Later

Scenario 1, cycle falsified plus valuation compression double hit: In the second half of 2026 to 2027, e-commerce and consumer-electronics capex retreats after restocking ends, and Cognex quarterly revenue growth slides from +24% back to single digits or even turns negative. The market realizes that "+24% was only a low-base rebound" and cuts forward PE from 44× back to a historical low of about 25-28×, still not cheap. EPS stays around $1.5, PE falls to 26×, and the stock price is about $39, down 40% from current. This would be a repeat of the 2018 playbook, with a higher starting valuation and a harder fall.

Scenario 2, moat and margin both falsified: In 2027, Chinese vendors such as Hikrobot accelerate replacement in mid-end applications, and open-source AI vision tools mature, pushing Cognex gross margin below 64%. At the same time, after the $35-40 million cost-reduction benefit is exhausted, the 25% EBITDA margin target is missed and margin falls back to 19-20%. The market re-rates Cognex as a mature industrial company with stagnant growth and peaked margins, assigning 22× PE. EPS is revised down to $1.3, implying a stock price of about $29, down 55% from current.

If the two scenarios overlap, with growth falsified plus margin falsified, a -50% outcome is mathematically quite possible from a starting point of 44× forward PE.

12.3 Final Research Conclusion

【Company Profile Score】 Fundamental quality: high | Growth: medium, cyclical with structural stagnation | Moat: medium, real but wearing at the edges | Financial stability: strong, zero debt and strong FCF | Management credibility: medium, smooth transition and clean governance, but thin skin in the game and insider net selling | Valuation attractiveness: low | Risk level: medium-high, mainly valuation risk | Suitable investor type: long-term value investors waiting for a pullback; the current price is unsuitable for chasing.

【Investment Rating】

• Rating: Watch

• One-sentence investment thesis: The world's second-largest machine-vision cash machine has high-quality fundamentals, but the current 44× forward PE has already pulled forward the cyclical reversal and leaves zero margin of safety; wait for a better price of $40-49.

• Three price signals, endpoints from the Section 9.3 valuation scenarios:

Ideal Buy Price: <= $49 (below the upper end of the bear scenario, with about 15-25% margin of safety; the truly attractive range is $40-49)

• Holdable Price: $54-63 (implied value range in the base scenario)

• Clearly Overvalued Price: >= $91 (above the upper end of the bull scenario)

• Current Price Classification: upper end of holdable range / slightly expensive ($64.67 is above the $63 upper end of the base range, leaving no margin of safety, but has not exceeded the bull range enough to be "clearly overvalued").

• Worth waiting for a better price: yes. Specific buy triggers: share price falls below $49, ideally $40-49, or fundamentals further validate "structural growth," with 2-3 consecutive quarters of double-digit revenue growth plus EBITDA margin holding at 25%+, lifting base value. The opportunity cost of waiting is possibly missing the bull-case +18-41% upside, but the permanent-loss risk from chasing at zero margin of safety, -40-55%, is asymmetrically larger.

• Target holding period: 3-5 years, if entered at a reasonable price.

• Expected annualized return, about 3-year holding period, price only, with dividend yield of about 0.5% counted separately: bear about -15%/year, falling toward the bear range; base about 0-3%/year, with price near the base range and mild earnings growth; bull about +6-12%/year, moving toward the bull range plus multiple persistence.

• Maximum loss risk: based on the pre-mortem, worst case -50% to -55%, from growth plus margin both being falsified and valuation compression, triggered by revenue growth turning negative and the EBITDA margin target being missed.

• Signals that trigger reassessment: 1. two consecutive quarters of double-digit revenue growth and EBITDA margin holding at 25%+ (upgrade); 2. two consecutive quarters of gross margin below 65% (downgrade); 3. quarterly revenue growth turns negative (downgrade); 4. share price falls below $49 (rating can be upgraded to "Cautious Buy"); 5. physical AI / embodied robotics begins to contribute quantifiable revenue (revalue the option).

Again: this report is based on public information and does not constitute investment advice.

13. Key Data Table

Research Uncertainties and Blind Spots

• The GAAP / adjusted basis for FY2026 full-year consensus EPS is not fully pinned down: this report's forward PE and valuation ranges use the adjusted basis, FY26 about $1.47. If the market uses GAAP instead, valuation at the same price is more expensive. The conclusion direction is unchanged, but the degree is stronger.

• The latest single-customer concentration lacks first-hand figures: the 2021 figures of largest customer 17% and second-largest 15% are the latest available point values. The 2024/2025 10-K did not disclose specific percentages, so the actual current concentration of the Apple chain / large logistics customers remains to be confirmed.

• This round's +24% split between "low-base cyclical recovery" and "structural improvement" cannot be precisely quantified: management has not disclosed the respective contributions from cost reduction / portfolio optimization and organic demand. This report's judgment that cyclical factors dominate is an inference.

• The third-party claim that ROIC < WACC and "destroys value" uses opaque methodology: independent ROIC estimates are in the 9%-16% range, depending on basis. This report treats it only as directional reference and not as a firm conclusion.

• There are differences in how Greater China FY2025 growth is framed, with about -3% and Q3 adjusted +9% both appearing. Geographic segment numbers depend on secondary citations of primary sources, and the 10-K original text was not checked word for word.

This report is based on public information and does not constitute investment advice. Markets carry risk; invest with caution.