What this covers
The episode is a company breakdown of Cognex, the machine vision maker, led by Brett Larson. It starts with the business's history of end markets. Semiconductor and electronics equipment was as much as 80 percent of sales going into the dot-com bubble and about 15 percent by the end of 2010, as smart camera adoption reached heavier manufacturing. A barcode reader launched around 2010 displaced laser scanners in logistics, worked closely with Amazon, and peaked near 30 percent of the business. Apple once reached roughly 20 percent of sales. Larson stresses that Cognex looks like software but recognizes hardware CapEx revenue upfront, so it behaves like an industrial.
The middle of the discussion covers the technology turn. Rules-based vision struggles with subtle judgments such as blemishes on phone cases, which deep learning can learn from labeled examples. Acquisitions in 2017 and 2019 brought in that expertise. Edge learning goes further, with preprogrammed products trained on five to ten images, which supports a Keyence-style sales push at smaller, less sophisticated customers. Larson contrasts Keyence's low R&D spend and mid-80s gross margins with Cognex's mid-teens R&D and roughly 70 percent.
The contested parts are investing questions: whether margins recover to about 30 percent, why the multiple tends to rise with sales, and his view that being early is the same as being wrong for cyclical investors. He also lists the capex signposts he watches, and he sees consumer electronics, 17 percent of 2024 sales, growing only with new device form factors.
He closes on culture. Founder Dr. Bob stayed as Chief Culture Officer until 2021, after Rob Willett became CEO, and local "ministers of culture" are paid to maintain it.
Brett Larson argues Cognex is a cyclical industrial company whose growth comes in S-curves triggered by new factory capital spending, and that its shift from rules-based to deep and edge learning vision, backed by a culture that has survived a founder transition, can open a new growth wave.
- Cognex revenue follows upfront factory CapEx waves, so growth arrives in S-curves rather than steady recurring streams
- Deep and edge learning let Cognex sell to less sophisticated customers with a Keyence-style sales force, opening a new S-curve
- Culture, maintained through a long CEO handoff, is a central reason the company can adopt new technology and stack S-curves
Cyclical revenue from factory capex and concentrated end markets amplifies swings and uncertain recovery timing.
- Smaller purchase orders are book-and-ship with very short cycles, while large strategic customers with complex capex plans have longer lead times because Cognex's systems follow the factory build sequence of shell, lines, and then cameras, and machine builders have their own lead times.
“The smaller purchase orders I think it's more book and ship type of business, very short cycle”
- Machine vision is moving from rules-based programming, where engineers hand-code decision logic, to teaching by example with large labeled data sets, which has large implications, and the industry is probably in the first inning of a long cyclical down cycle for this next chapter.
“the way I conceptualize this is moving from rules based programming of vision systems to teaching by example, which has large implications.”
- Customers hope a Cognex camera lasts 10 to 20 years, so there is no regular replacement cadence; instead revenue comes from brownfield capex on existing lines and occasional greenfield or new end-market opportunities that drive industry growth.
“Ideally when a customer puts the camera in the factory, they're hoping it's going to last for 10 to 20 years. So there's not really a regular replacement cadence.”
- Because Cognex's business cycle is short, its performance leads the broader cycle, and consumer electronics orders will inflect before anyone knows what the spending is for.
“Given their short cycle and then especially with something like consumer electronics, that business will inflect before we even know what it is they're spending on”
- Early is the same as wrong for cyclical investors, because recent incremental investors in Cognex are feeling early and the risk is that the cycle may not turn within a year.
“Early is the same thing as wrong. Basically every incremental investor in Cognex the last year or two is probably feeling very early right now”
- Cognex's cycle signposts are end-market capex indicators: semi cap equipment makers selling to TSMC for semiconductors, automotive OEM capex for automotive, consumer electronics product-cycle capex, and PMI and industrial sentiment for the rest of the business.
“just watching the automotive OEMs and their CapEx is the signpost that you watch for that one. And then lastly PMI is the catch all”
- Most of Cognex's recent sales, around 60 to 70 percent, have gone to customers who had never had a camera in their factory, and the company aims to grow its base from about 30,000 customers toward hundreds of thousands and reduce customer concentration.
“I think it's more like 60 to 70% of their sales have been people that have never had a camera in their factory.”
- Apple was as high as 20 percent of Cognex's sales during iPhone manufacturing build-outs and is now probably in the mid to high single digits, showing how concentrated its revenue has been around a few large capex waves.
“At one point Apple was as high as 20% of Cognex's sales which is pretty incredible. It's now probably down to more mid to high single digits range.”
- Consumer electronics is 17 percent of Cognex sales in 2024 and only grows when new form factors or features, possibly tied to LLMs, AR, VR or humanoid robots, trigger high-volume factory capex; without them, the business sits at steady-state maintenance.
“if any of those things were to be manufactured at high volumes that would fall right into Cognex's sweet spot.”
- Cognex's semiconductor exposure was sub 5 percent before its late 2023 bolt-on acquisition of Moritex, which sells optics and lighting mainly into semiconductors, and it grew quickly in 2024 with strong growth expected in 2025.
“They actually did a bolt-on acquisition of a company called Moritex in late 2023 and Moritex sells optics and lighting.”
- China was 18 percent of Cognex's 2024 sales, two-thirds of which went to Western multinationals, about half of that or more to Apple and Foxconn, and the remaining mid to high single digits went to domestic Chinese manufacturers, which will be an uphill battle to grow over five to ten years.
“So it's 18% of sales in 2024. Two thirds of that is to western multinationals and a good chunk of that, probably half of that or more is Apple Foxconn.”
A trough valuation, net-cash balance sheet, and cash returns frame the investment case.
- Cognex does not adjust out stock-based comp, so it looks like a software company, but it carries the cyclicality of an industrial company because all revenue is recognized upfront as hardware capex, with software tied to the hardware.
“Cognex doesn't adjust out their stock based comp which is great for a company that's essentially a software company. But you get stuck with the cyclicality of an industrial company as a trade off. So it's all an upfront sale.”
- Cognex runs a fortress balance sheet with net cash at all times, and cash and investments are about 10 percent of market capitalization.
“So they have net cash on the balance sheet. Cash and investments actually right now is 10% of the market cap.”
- Cognex trades at about 5.5 times next-twelve-month sales, near its 10-year low, having traded as high as 16 times sales during the zero-interest-rate period and typically 6 to 10 times sales historically.
“Currently it's within reach of its 10 year low at 5 and a half times next 12 months sales. It traded as high as 16 times sales during ZIRP”
- Cyclical stocks such as analog semis, semi caps, and metals and mining trade at peak multiples on trough earnings, and Cognex's multiple has historically risen when sales grow and fallen when sales decline, which Brett cannot fully explain.
“they'll trade peak multiple on trough earnings and the like. And you look historically at Cognex's multiple and it's basically trended up when sales are growing and down when sales are declining.”
- Cognex has converted about 100 percent of net income to free cash flow over time and returns over 100 percent of that to shareholders, split about one-third dividends and two-thirds share repurchases, with occasional IP-focused acquihire M&A.
“they've generally converted 100% of net income to free cash flow over time. So they produce good free cash flow when they've returned over 100% of that to shareholders. So a third dividend and two thirds share repurchases.”
Long-run returns depend on margin recovery toward 30 percent and low double-digit growth.
- Using an exit free cash flow yield and an internal hurdle rate of at least 10 percent, Brett's firm concludes Cognex needs to compound free cash flow at a low double digit rate over the long term, which is below the long-term model and assumes margin recovery from a cyclical trough.
“we look at what's implied in terms of the future free cash flow growth to justify the current valuation. So we do that by assigning an exit free cash flow yield out in the future and then have our internal hurdle rate which is at least 10%.”
- Historically Cognex's margins have declined with cycles but always recovered back to about 30 percent, and early signs of leverage in a recent quarter with double-digit organic growth are being watched, though normalized margin levels remain the central debate.
“Historically, we have seen margins decline in magnitude with cycles and they've always recovered back to 30%.”
- In 2019, before the current down cycle, Cognex's margins fell from 27 percent to 20 percent, and the current downturn is deeper because management invested heavily through it.
“back in 2019, which was prior to this down cycle, the last time that they had a down cycle, margins went from 27% down to 20%.”
- Cognex's operating margin was 13 percent last year, down from a peak above 30 percent, with a couple hundred basis points of that drag from the Emerging Customer Initiative sales investment, and management targets regaining the 30 percent level as leverage returns.
“operating margins last year were 13% and that's down from a peak above 30%. And there's a couple hundred basis points of the emerging customer initiative headwind in there.”
- Cognex's long-term target is 15 percent top-line growth in constant currency and 40 percent incremental margins, though it grew about 13 percent excluding M&A over the ten years ending around 2020 to 2022, so the realistic long-run bogey is low double digit top-line growth.
“Longer term they target 15% top line and constant currency and 40% incrementals.”
Cognex wins sophisticated engineering buyers through spec-level application engineering and premium pricing against Keyence.
- Cognex focuses on the top of the customer pyramid with sophisticated engineering buyers, is a more expensive vendor, and wins by spec-level application engineering that delivers read rates 100 to 300 basis points better than peers, which at 100,000 packages a day means thousands of packages needing no human check.
“Basically they're not winning on price, they're a more expensive vendor of machine vision. But they have a reputation for having really good application engineers”
- Keyence spends about 2 percent of sales or low single digits on R&D and earns mid-80s gross margins, while Cognex spends mid-teens percent on R&D with roughly 70 percent gross margins, reflecting Keyence's go-to-market approach.
“they spend 2% of sales or low single digits on R&D and have mid-80s gross margins. And that compares to Cognex which spends more like mid teens on R&D and has more like 70% gross margins.”
- Keyence sells to middle and lower customer tiers with standardized products, hires college graduates, trains them on products, and runs a scripted sales process tracking KPIs such as calls, shop visits and demos in the CRM.
“So they'll hire more college graduates compared to Cognex and then they'll train them with the products, put them out in the field. And it's more of a scripted sales process”
- Cognex is the number two machine vision company behind the Japanese firm Keyence, and the two compete while historically focusing on different segments of the customer base.
“Cognex is number two behind Keyence, which is a company based in Japan.”
- Hik Robotics, a division of Hikvision, is probably the third largest machine vision player overall, has grown quickly on domestic Chinese demand, and is about half Cognex's size in sales.
“Hik Robotics, which is division of Hikvision, is probably the third largest player in the industry overall and they've grown quite quickly driven by that domestic market”
Factory-floor stickiness and software, not hardware, are Cognex's real value add.
- Once a Cognex camera is programmed, humans are ideally out of the loop on the factory floor, with the camera capturing and analyzing data and passing results to a programmable logic controller that performs the action.
“Ideally the only time the human's really interacting with the software primarily would be upfront when you're programming the application.”
- Cognex's applications fall into four categories, Guide, Gauge, Inspect, and ID, with ID covering barcode and optical character reading and Insight covering Guide, Gauge and Inspect.
“And the applications generally fall into four categories. So there's Guide, Gauge, Inspect, and ID.”
- Cognex is a machine vision leader that sells ruggedized cameras with embedded processing, and the software that captures and analyzes images to automate decisions at high speed in manufacturing and logistics is the real value add.
“they sell ruggedized cameras with embedded processing and then software which is the real value add which captures images and analyzes them in order to automate decisions at high speeds in manufacturing and logistics environments”
- Customer stickiness is high once cameras are on the factory floor because plant staff know the programming software, which is usually standardized, and customers want to avoid managing multiple machine vision vendors and their software.
“The stickiness of a customer is quite high once you're on the factory floor and the people that are managing the floor are familiar with the software used to program the cameras”
Machine learning's shift to teaching by example opens an opportunity Cognex is well positioned to exploit.
- About 30 million people in the world still do visual inspection, a task humans perform quickly but poorly over time because they fatigue and miss defects, which represents a large opportunity for machine vision.
“So there's currently still 30 million people in the world doing visual inspection. And that's something that humans actually aren't great at.”
- Deep learning enables robots to handle tasks like deboning chickens, where each bird differs in orientation and wing position, making it impossible to program explicit grasp rules but feasible to learn them.
“It'd be impossible to quantify and program a robot to be able to grab the right places. But with deep learning, that's an application that they can now do”
- The technology transition to machine learning opens a great window of opportunity for Cognex, where it is early and ahead in applying machine learning to machine vision, but any technology transition also opens a window for disruption.
“But with any tech transition it opens the window for disruption as well.”
- Cognex lists hundreds of patents on its website, reflecting a strong IP focus.
“You can go on their website and they list hundreds of patents.”
Cognex's technology evolved from OCR and smart cameras into deep learning through acquisitions.
- After the OCR work for Johnson and Johnson labels, J&J asked Cognex to verify bottle caps and label presence, which led Cognex to add inspection applications and begin its first 20 years pioneering machine vision mainly for semiconductors and electronics.
“So that's when Cognex added more of these inspection type of applications.”
- Cognex was founded in 1981 and came to market with the world's first industrial optical character recognition system, whose first application was reading serial numbers on semiconductor wafers for IBM, the origin of the DataMan product family.
“The company was founded in 1981 and they came to market with what was the world's first industrial optical character recognition system.”
- Traditional rules-based vision cannot handle subtle, nuanced tasks with great variation, such as spotting small scratches, dents and paint blemishes on phone cases where acceptability is hard to quantify, but deep learning can learn these judgments from large labeled image sets of good and defective examples.
“traditional rules based vision is great for a lot of things, but not for things very subtle or nuanced tasks with a great deal of variation.”
- Cognex began its current technology evolution in 2017 with the ViDi acquisition and in 2019 with SUALAB, both primarily IP and acqui-hire deals in deep learning, and ViDi's co-founder and CTO now serves as Cognex's VP of AI technology.
“It began in 2017 when Cognex acquired ViDi. And then in 2019 they acquired a company called SUALAB.”
- Around 2000, smart cameras with processing embedded directly in the camera, which were smaller and more ruggedized, made machine vision ready for the factory floor and drove adoption in automotive, consumer electronics, and food and beverage packaging.
“Around 2000 was when we saw smart cameras really come onto the scene. So where you embedded the processing directly into the camera.”
Cognex's culture is sustained by founder-built nimbleness and long leadership overlap through leadership changes.
- Cognex has ministers of culture in every office around the world who maintain its culture as an incremental job, meet regularly, and receive a separate check in compensation, which signals the company views culture maintenance as important.
“they have what are called ministers of culture in every office around the world who are responsible for maintaining the culture. And it's actually an incremental job”
- Rob Willett joined Cognex from Danaher in 2008, became CEO in 2011, and founder Dr. Bob stayed on as Chief Culture Officer until 2021, a long overlap that ensured the culture endured the transition and that Cognex has had only two CEOs in about 45 years.
“Rob Willett is the current CEO. He joined from Danaher in 2008 and became CEO in 2011. Dr. Bob stayed on as Chief Culture Officer until 2021.”
- Cognex's culture produces voluntary attrition about half that of industry peers, which Brett argues supports nimbleness in adopting new technology and stacking S-curves.
“The fruits of the culture is obviously things like voluntary attrition is half of their industry peers.”
- Cognex's founder Dr. Robert Shillman, known as Dr. Bob, is a skeptical, rebellious, engineer-minded leader who built the company's culture from the beginning around the motto work hard, play hard, move fast, emphasizing autonomous decisions and a large PhD-heavy engineering base.
“he was also very intentional with culture from the very beginning. Cognex's motto is work hard, play hard, move fast. It's very engineering centric.”
- The most important lesson from Cognex is culture, specifically how the company maintained it through the founder's departure via ministers of culture and a long CEO overlap, which Brett says he had not seen elsewhere.
“I feel like it's got to be the number one answer is culture.”
Cognex is copying Keyence's scripted, volume-based sales model to reach smaller customers.
- About 70 percent of Cognex sales are direct, including sales to factory floors and to OEMs that integrate the vision system into their machines, while the remaining 30 percent splits roughly evenly between systems integrators, mainly in logistics, and distribution in markets where Cognex lacks presence.
“70% of sales are direct... And then the remaining 30%, about half of that is going to be through systems integrators and that's primarily logistics at this point.”
- Cognex's Emerging Customer Initiative is a salesforce investment that takes the Keyence playbook, hiring young salespeople instead of technical applications engineers, arming them with edge learning products, and targeting SMB and less sophisticated tasks with a formal KPI-tracked, commission-based selling process.
“And it's really taking the Keyence playbook. These emerging customers, salesnoids, they call their employees cognoids, their salespeople, salesnoids, they're hiring young salespeople instead of technical applications engineers.”
- The first cohort of young salespeople made 80,000 customer visits, added 3,000 new customers to a base of about 30,000, delivered accretive gross margins, and exited the year at about a million dollars a week in sales.
“Last year Cohort One of these salesnoids hit the field and they did 80,000 customer visits, added 3,000 new customers to a base of what was previously 30,000, had accretive gross margins”
- Edge learning products cost about one to two thousand dollars per system, so a low-end order might be around 10,000 dollars, while high-end vision systems can exceed 10,000 dollars each and orders can reach hundreds of thousands of dollars.
“These new like Edge Learning products might only be one or $2,000 per system and a customer might need a couple of those.”
- Edge learning products come pre-programmed for specific applications and need as few as five to ten images to train, so an application can be running in a few hours, which lets Cognex sell to less sophisticated customers with a less technical sales force.
“they're trained with just as few as five to 10 images. So it's essentially very easy to deploy relative to traditional rules-based vision. You can usually get an application up and running in a few hours.”
Cognex's end markets have shifted from semiconductors toward logistics and automotive.
- Semiconductor and electronics capital equipment was as high as 80 percent of Cognex sales going into the dot-com bubble, fell to 54 percent after the bust, and was down to 15 percent by the end of 2010 as smart camera adoption brought in heavy manufacturing end markets.
“It was as high as 80% of sales going into the dot com bubble were to those industries. And that went to 54% following the hard way.”
- Cognex's first ID barcode product around 2010 displaced laser scanners in logistics, was originally targeted at a 75 million dollar long-term sales line, worked closely with Amazon, and peaked around 2021 at about 300 million dollars or 30 percent of the business.
“they were talking about they'd be happy if one day, long term target $75 million of sales from this line of business.”
- Logistics is Cognex's largest end market at 23 percent of sales, down from a peak of 30 percent, with Amazon once at 17 percent of Cognex sales and now around high single digits to 10 percent as non-Amazon customers grow quickly.
“Logistics is the largest end market. It's 23% of sales. At peak it was 30% of sales and Amazon was up to 17% of Cognex's sales.”
- Automotive is 22 percent of Cognex sales and was expected to benefit from EV battery transition capex, but automotive sales fell in the mid-teens percent in 2024 because the growth did not come through as expected, with 2025 expected to decline less sharply.
“Automotive sales were down mid teens in 2024. It's not expected to be a good year in 2025, but it shouldn't be a deep decline like 2024”
Machine vision is growing steadily, and Cognex's addressable market keeps expanding with new products.
- Cognex estimated its serviceable addressable market at 6.5 billion dollars, up from 2.9 billion in 2017, and expects an update at its investor day to around 8 to 9 billion based on new product categories.
“Cognex last estimated their serviceable addressable market at 6.5 billion, which was up from 2.9 billion in 2017.”
- The value proposition of machine vision is improving quality, throughput, and reducing costs and waste, increasingly addressing labor constraints, and the payback period for these products is usually quite quick.
“ultimately you can think of the value prop that they're trying to deliver as improving quality, throughput, reducing costs and waste. And then a more and more common one's just addressing labor constraints. And the payback period is usually quite quick”
- The machine vision industry has grown at roughly a 10 percent CAGR over the past decade, with cycles and dispersion, as some firms including Cognex grow above that rate and others below.
“Industry's probably grown at a 10% type of CAGR the past decade and obviously there's cycles in there”