Cognex：视觉探寻——[商业拆解，第207期]

All right, Brett, it is great to have you back. It was about a year ago that we covered Trane Technologies, maybe a little more than a year ago, but it’s an episode that I’ve referenced quite a bit over the past few months and one that I consistently go back to. I’m hoping that we get the same thing today with Cognex. It was a name that I wasn’t familiar with, and it’s a very interesting name that seems to not get a ton of attention or conversation as I was doing my research. The easiest place to start is with an introduction to who Cognex is and what they do.

Thanks for having me, Matt. It was fun, so I was excited to be able to join you again. Cognex stands for “cognition experts,” and they are leaders in machine vision. What that means specifically is that they sell ruggedized cameras with embedded processing and software, which is the real value-add. The software captures images and analyzes them in order to automate decisions at high speeds in manufacturing and logistics environments.

The product might look like 2 cell phones stacked together, in Cognex yellow, sitting right out there on the factory floor. The applications generally fall into 4 categories: guide, gauge, inspect, and ID. ID is the DataMan family of products, which uses vision for reading barcodes and optical character recognition—that is, reading letters and numbers that are printed or etched on something. The easiest example to picture would be an Amazon sorting facility that’s moving tens or hundreds of thousands of packages per day at extremely high speeds. Cognex’s vision products scan all those barcodes, and the automation then performs the required actions.

In-Sight is the other main product category, and that covers guide, gauge, and inspect. Inspect would be quality inspection—2D and 3D vision systems that are used to inspect something. A common example would be a printed circuit board. The system would make sure that it has the correct number, size, and placement of the different components, as well as quality soldering. Another classic example would be making sure bottles in a bottling machine have lids and labels.

Guide would be vision systems used to guide robotic arms, and gauge would be measuring the dimensions of something. Ultimately, you can think about the value proposition they’re trying to deliver as improving quality and throughput, reducing costs and waste, and increasingly addressing labor constraints. The payback period is usually quite quick for these types of products.

You mentioned there’s the hardware, which I can imagine means you have these cameras on a conveyor belt. The software has to link into whatever you’re working on. In some ways, that software is machine-to-machine execution. Is there a lot that’s also going back to humans, or human interaction with the software?

Ideally, the only time humans are really interacting with the software would be upfront, when they’re programming the application. After that, the camera is sitting out on the factory floor, capturing the data, analyzing it, and then communicating that data to something more like the programmable logic controller that Rockwell might sell, which is then performing whatever action needs to be done. Ideally, once it’s out on the factory floor, the human is out of the loop other than overseeing it.

I think you brought up a great visual in terms of an Amazon sorting facility and how much of that is now automated. You don’t have humans involved moving the big packages here and the smaller packages there. It’s gotten more advanced, but how big is this machine vision market today? I’d also be curious how much it has grown recently—whether there’s been a step change in the growth of the market.

There are a lot of different total addressable market estimates out there for machine vision, but Cognex doesn’t play in all the different niches that get captured in those estimates. Cognex last estimated its serviceable addressable market at $6.5 billion, which was up from $2.9 billion in 2017. They’ll probably update their SAM again at their investor day, and it’ll probably be something like $8 billion to $9 billion based on the new product categories they’ve entered. The industry has probably grown at a 10% type of CAGR over the past decade. Obviously, there are cycles in there, and there’s dispersion between some companies that are growing above that range, like Cognex, and some that have been below it.

I like the SAM terminology. I might have to mix that in with TAM in the future. Where does Cognex rank in terms of market share within the market? What are some of the characteristics of their positioning, and then just the market overall?

Cognex is number 2 behind Keyence, a company based in Japan. The 2 companies obviously compete, but they’ve also historically focused in slightly different areas, which is interesting. Cognex has historically focused on the top of the pyramid when you think about the sophistication of the customer.

Typically, they’re hiring very well-trained engineers who are working with customers to spec systems for specific tasks. It’s a more technical sale with very sophisticated customers that are automating very complex tasks. 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 and the best technology, which they can differentiate at the spec level.

For example, in a logistics facility where you’re scanning 100,000 packages a day, Cognex can usually deliver read rates that are 100 basis points, or even 200 or 300 basis points, better than a peer. Over even just a day, that’s thousands of packages that don’t need to have a human there to look at something. That’s usually how they go about competing, and there’s a really strong brand reputation as well.

If a new COO of a company walks into a factory for the first time and sees the yellow Cognex cameras, that says something to them. The installed base is also very sticky for all the vendors. I’ll foreshadow a bit, but they’re currently focused on the top of the pyramid. They’re also broadening out a little bit lower, which is interesting.

And then Keyence, as I mentioned, is number one. They're based in Japan. It's been a really successful company and it's also very much an enigma just in terms of it's a public company but there's just not a lot of intel out there on it. My best guess, from what I've gathered, is maybe 20% or so of their sales are comparable to Cognex. They also do things like scientific microscopes, PLCs, direct part marking, and stuff like that. But what's also quite interesting is that they spend 2% of sales, or low single digits, on R&D and have mid-80s gross margins.

That compares to Cognex, which spends more like the mid-teens on R&D and has more like 70% gross margins. The other machine-vision vendor peers spend a little less and have even slightly lower gross margins than Cognex. It really comes down to the way Keyence goes to market and how they focus. They generally focus just on the middle to lower tiers of customers. They're also trying to develop more standardized products that go after very high-frequency applications.

It's a very process-oriented sales organization. They'll hire more college graduates compared to Cognex, train them on the products, put them out in the field, and use more of a scripted sales process. They're keeping track of more of those activity KPIs in the CRM. How many calls are you making? How many shop visits? How many demos? It's really good coverage and very relentless.

If you get down the rabbit hole, it's become a meme in the community. There are some pretty funny memes out there. If you're going to download a product spec sheet from the website, you should use your buddy's email and phone number, not your own. You'll never hear the end of it.

Those are the 2 main players. Maybe I'll briefly mention a couple of other competitors while we're here. The other bucket, I would say, is China. These are primarily for manufacturers in China.

Hikrobot, which is a division of Hikvision, is probably the third-largest player in the industry overall. They've grown quite quickly, driven by that domestic market and the domestic manufacturers, and they're probably about half the size of Cognex in terms of sales. There are a number of other smaller Chinese players as well. Outside of China, there are more legacy players.

Within Teledyne, there's DALSA and Point Grey. There are SICK and Basler in Germany, Datalogic, and Matrox, which was acquired by Zebra. There are even companies like MVTec. So that's kind of a lay of the land in terms of the industry structure.

I did read quite a bit about Keyence and their sales process and how effective of a sales organization they've built. But in the sales process for someone like Cognex, where it sounds like it's much more technical, built to spec, perhaps advanced, and not so much off the shelf, are they selling to Amazon, which then works to integrate it into the other hardware and equipment that they're using? Or are they selling to the equipment manufacturers—whoever builds the conveyor belt? Who is the customer for Cognex?

The answer is yes. If you think about how they go to market, 70% of sales are direct. That would be direct either to the factory floor, where Cognex and the automation engineers at the customer are working to implement the system. But direct also includes selling to a machine builder or an OEM, who would then integrate it into the machine, and then they take that machine to the factory floor.

The remaining 30%—about half of that is going through systems integrators, and that's primarily logistics at this point. The other half of that 30% would be distribution, and that's primarily just for markets like Cambodia or something, where they don't have any presence. That's kind of how I would separate the customer.

That makes sense. It's an interesting dynamic within the overall industry-structure conversation. Before we move on too far, I do want to get into a bit of the history. What came across through the work that I was doing is that there is technical expertise here and a focus on that technical expertise. Can you bring us back to the beginning and the origin story here, when they got into the market, and some of the evolution over time?

It's really important to understand, for Cognex, the DNA of stacking S-curves, essentially, over 40 to 50 years. The company was founded in 1981 and came to market with what was the world's first industrial optical character recognition system. So, again, reading numbers and letters, and at the time, the system looked more like a big video camera with external processing. The first application was actually to read serial numbers on semiconductor wafers for IBM. That was the DataMan product family, which still exists.

From there, they got a call from Johnson & Johnson to do optical character recognition for labels. They did that application, but then J&J asked them, "Hey, can you do some of these other, at the time, novel applications, like verifying the caps were on the bottles and the labels were present and the like?" That's when Cognex added more of these inspection-type applications.

The first 20 years or so of the business was pioneering machine vision for all these different applications, but it was very much for semiconductors and the electronics capital equipment industries, primarily. As high as 80% of sales going into the dot-com bubble were to those industries, and that went to 54% following the hardware bust.

Around 2000 was when we saw smart cameras really come onto the scene. That's where you embedded the processing directly into the camera: a smaller footprint and a more ruggedized product. That enabled the product to be ready for the factory floor as we more often think about it. We saw machine vision get adopted in more heavy manufacturing industries like automotive, consumer electronics, food and beverage, and packaging. By the end of 2010, that semiconductor and electronics capital equipment business was down to 15% of sales as they brought in the end markets that they could serve through all those new S-curves.

Around 2010 was when Cognex came out with their first ID product for reading barcodes. It was primarily aimed at displacing laser-based scanners in logistics facilities because it could do it faster and with much higher read rates. If you go back to 2010, they were talking about how they'd be happy if one day, as a long-term target, they got to $75 million of sales from this line of business. They worked very closely with Amazon to develop the technology. In 2021, at the peak, it was about $300 million in sales and 30% of Cognex's overall business. So it turned out to be just an enormous new S-curve for them.

At present, we're really in the early innings of the next technology evolution for the industry. It began in 2017, when Cognex acquired ViDi, and then in 2019 they acquired a company called SUALAB. Those were both essentially IP and acqui-hires in the field of deep learning, or the application of AI to machine vision. ViDi's co-founder and CTO is Cognex's current VP of AI technology.

After those acquisitions, there was a big development effort, and they came to market with these new deep-learning and edge-learning products. The way I conceptualize this is moving from rules-based programming of vision systems to teaching by example, which has large implications. I think we're probably in the first inning of this next chapter, but behind the scenes of a long cyclical downcycle.

Could you paint that picture in terms of an example of rules-based vision? Let's say a package is over 100 lb with these dimensions: send it in the left lane. If it's under that, send it through the middle. If it's super small, send it to the right. It's a pretty standard tree of logic and decisioning. Where would this new example, or this new training, come into play? Do you have any real-world applications that would paint the picture?

There are deep learning and edge learning. As I mentioned, deep learning first. Traditional rules-based vision is great for a lot of things, but not for very subtle or nuanced tasks with a great deal of variation. Those sorts of tasks still usually have a human performing them because, frankly, it's just hard to quantify them and then program them.

Maybe an example of that would be inspecting a phone case for very subtle defects. There's obviously a wide variety of colors and an endless list of potential defects: scratches, dents, and blemishes in the paint. They're very small and very nuanced. There's also usually a scale of what's acceptable and what's not, where a human would be able to look at it and pretty quickly discern that. But it's very hard to program a machine to be able to do that, especially at high speeds.

With deep learning, instead of programming it, you teach it with very large data sets and very large image sets. These are good phone images. These are defects. These are acceptable defects. The machine learns what is acceptable and what is not and is able to accomplish that task.

Just to quantify this a little bit, there are currently still 30 million people in the world doing visual inspection, and that's something that humans actually aren't great at. They can do it quickly, but they get fatigued and miss things. That's a big opportunity.

Another example would be something like deboning chickens, where every chicken's different. It comes down the line in a different orientation. The wings are somewhere different. 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 implication of deep learning would be potentially new applications for machine vision. Edge learning is the other end of the spectrum, and it's probably more financially tangible for the investor community right now. These products come pre-programmed for specific applications, but then they're trained with as few as 5 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. The major implication here is that Cognex can now sell to less sophisticated customers.

Again, thinking about where they've historically focused, at the top of the pyramid, they can do it with a much less technical sales force. So, they're broadening their customer base below where they've typically served.

Now they have the product, and Cognex is currently making a sales force investment called the Emerging Customer Initiative. It's really taking Keyence's playbook. These emerging-customer salespeople—they call their employees Cognoids—are hiring young salespeople instead of technical applications engineers. They're arming them with these new edge-learning products that are really easy to implement, and they're going after these less sophisticated customers, more SMB types, and even less sophisticated tasks within existing customers where those exist.

Then it's a much more formalized selling playbook that tracks the KPIs just like Keyence. They're getting compensated on both their selling activity as well as their commissions. Last year, cohort 1 of these sales Cognoids hit the field, and they did 80,000 customer visits, added 3,000 new customers to a base that was previously 30,000, had accretive gross margins, and exited the year at about $1 million a week in sales from that first cohort. The second cohort entered this year, in 2025, and hopefully will do a bit better just from lessons learned.

Obviously, they're now competing in Keyence's bread-and-butter target customer set, so they are running into them sometimes with these customers being served. But the majority of the time, these customers have never used machine vision at all. I think it's more like 60% to 70% of their sales have been to people that have never had a camera in their factory.

Obviously, the financial implications here are that they're trying to grow their customer base from 30,000 to hundreds of thousands over time. That could be an obvious revenue opportunity, add accretive gross margins, and reduce customer concentration as well over time in the customer base.

I think you painted a picture just in terms of the various applications and who might use them, but what does it look like in terms of a customer—whether it's a contract, whether it's a sale, the ramping of that contract or sale, and then what the stickiness is once you build something out?

The stickiness of a customer is quite high once you're on the factory floor and the people managing the floor are familiar with the software used to program the cameras. It's usually more or less standardized, especially for some customers. For the sake of simplification, they don't want a bunch of different machine vision vendors out there and the associated software.

It's a capex sale, and they have, as I mentioned, historically been focused on the top-of-the-pyramid customers. So, they've been tied to the large capex buildouts of companies like Apple and Amazon. That obviously comes in a wave, and then it's more a question of stacking new S-curves and when the next waves in capex are going to come. But the customers themselves are generally quite sticky.

Maybe this is a good segue into the different end markets, and we can start with consumer electronics.

It's interesting because that end market is tied to the capex spending around new consumer electronics or features. There was a tailwind from the adoption of cell phones, and then there was the tailwind from Apple building out iPhone manufacturing. At one point, Apple was as high as 20% of Cognex's sales, which is pretty incredible. It's now probably down to the mid- to high-single-digit range.

There was another big tailwind when OLED came on the scene, and that was with Samsung. Looking forward, it's interesting to think about if there will be new form factors or changes to the existing ones for consumer electronics tied to things like LLMs, if AR or VR is ever going to take off, or if humanoid robots are going to take off. Basically, if any of those things were to be manufactured at high volumes, that would fall right into Cognex's sweet spot: sophisticated customers and big manufacturing capability.

Consumer electronics in 2024 was 17% of sales, and longer term they expect that market to grow at mid-teens; that's the target.

Logistics is the largest end market. It's 23% of sales. At its peak, it was 30% of sales, and Amazon was up to 17% of Cognex's sales. Obviously, there was the well-publicized down cycle in that end market. They did return to growth in 2024—they grew 20%—and the long-term target for that end market is 20% growth. Amazon is probably back around the high-single-digit to 10% range in terms of a customer.

In Cognex fashion, they're stacking new S-curves in logistics, which is really interesting. They're taking the vision tunnels, which they've really honed with Amazon, to new customers and new geographies, and those non-Amazon customers are growing very quickly. They're also identifying other applications outside of barcode reading, things like vision inspection to see damaged boxes coming down the line or damaged labels, and to make sure there are labels. Even things like dimensioning can help estimate shipping costs and things like that. So, they get their foot in the door, and then they find additional applications within the end market. That's the story of logistics at the moment.

Automotive is the second-largest market, at 22% of sales. In theory, EV batteries should be a good tailwind for them. Cognex has solutions, and the transition frankly necessitates a bunch of capex, whether it be new automotive lines or battery facilities, which otherwise would not have been necessary if it were just new internal-combustion-engine models.

But frankly, the growth just didn't come through last year as people had expected it to. 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, at least. Long term, the target for this end market is 10% growth for Cognex.

Two other quick ones: semiconductors are probably about 10% to 15% of sales now. They're selling primarily to the semiconductor capital equipment manufacturers, and the growth outlook there is positive at this point in time. The other 20% to 25% is general factory automation, with PMI being the primary thing to keep an eye on there. It's been a historically long down cycle for industrial manufacturing, and that's been felt by Cognex in that part of their customer base.

Is there an average useful life in terms of the equipment that they're installing? Are there replacement cycles for the equipment as well?

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. It's more about getting the customer into your installed base, and then they'll be doing brownfield capex on existing lines, which will create revenue opportunities for Cognex.

Every now and then, there will be these greenfield opportunities with those customers, or just new end markets and new customers, and that's what drives the growth of the industry over time.

On the software side of the business, is that an actual revenue driver for them? When I think about how much they will make from the sale of equipment, is it all recognized in year 1, or is it spread out with some type of software component?

The joke I say is that Cognex doesn't adjust out their stock-based compensation, 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. The software is tied to the hardware capex.

It sounds like you get the S-curves when there is a paradigm shift or a new form factor introduced into an industry that would require new capex spend on a factory or a production line, where new Cognex equipment would have to come in. But if it's just producing the same phones or the same cars without added features, you're not going to see a big step-up in revenue because you don't have to make adjustments to the factory.

I think that's right. You can think of the core business as supporting customers as they're doing their big capex spends. They're tied to that. So, at the moment in consumer electronics, for example, we're keeping an eye out for what it could be, but there's not something that's driving a big capex spend at Apple, for example. That end market is at more of a steady-state maintenance level. You characterized it correctly.

Within that context, on semiconductors, it sounds like it's a meaningful chunk of the business, but not something that maybe has captured the same tailwind of the actual market itself with semiconductors. What drives the disconnect there? What would you tie that to, where it's not just this massive chunk of the business right now as we're seeing this cycle play out?

They actually did a bolt-on acquisition of a company called Moritex in late 2023. Moritex sells optics and lighting, and a lot of that goes into the semiconductor end market. Prior to that, Cognex's semiconductor exposure was quite small, probably sub-5%. After that acquisition, it's bigger. It grew quite quickly in 2024, for example, and the expectation is that it's going to grow very strongly again in 2025 at this point in time. So, I think that kind of explains why it's the size that it is.

Yeah. Everything you mentioned, just in terms of the various big customers and the auto market being tepid or modest at best at the moment, all aligns. One thing we didn't get into when we talked about the history was the culture, more so the leadership. What would you point to just in terms of the evolution of the culture and then the management team?

I think it's rare that you get to this point in the discussion before you talk about the culture at Cognex. It really is very unique. As I mentioned, their employees are called Cognoids, and they've had 2 CEOs in the past 4.5 decades.

The founder, Dr. Robert Shillman, is one of a kind. He goes by Dr. Bob. I was thinking about how to describe him quickly for a podcast.

There’s one theory that he shares tongue-in-cheek that I think captures him quite well. He says that he doesn’t believe in exercise because, basically, your body is a bunch of mechanical joints, and anything mechanical has a finite amount of use. Just from that, you can see he’s clearly an engineer. He’s really smart, very skeptical, and a bit rebellious, and he’s got a really good sense of humor.

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. I believe they have the largest collection of PhDs in machine vision in the world working on just advancing the field. They hire really smart people and give them the room to make autonomous decisions, and they say, “Be right most of the time.”

They’re willing to fail and try new things, but they also try to have a lot of fun when they’re doing it. Some examples of that: on leap years, a few employees are selected to go jump out of a plane with Dr. Bob, and I think he dresses up like a frog or something ridiculous. One year, they rolled in an armored vehicle to deliver the cash bonuses, and Halloween every year is a huge event. They have, without question, the best annual reports in the business. They’re themed, so you should take a look at those. They’re really fun.

But what’s interesting is they’ve really managed to maintain the culture, from what I can tell. I think there are 2 reasons they’ve been able to do this after Dr. Bob. First, they have what are called ministers of culture in every office around the world who are responsible for maintaining the culture. It’s actually an incremental job in addition to their existing job. They have meetings, and they actually get a separate check in terms of compensation for this job, and it’s viewed as very important to the company.

So, 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. He really ensured the culture endured through the transition, and Rob’s really embraced it. That long handoff as well, I think, is really interesting and unique.

The fruits of the culture are obviously things like voluntary attrition at half of their industry peers, but I also think it’s just critical in terms of being nimble and adopting new technology to stack all these new S-curves over time. It ultimately comes back to the culture. There’s no culture like a funky or fun engineering group. I think it’s one of the more unique things out there in the market when you find one.

Quite interesting to hear about. It did bring up one point, just given the technical background. Is there a big patent portfolio or IP focus for the business?

You can go on their website, and they list hundreds of patents.

I wanted to get to just some of the numbers on the business as well. I think we’ve danced around them a bit, but if I’m thinking about what the cost of their equipment is, to visualize a factory purchasing these, is there a general sense of what the average selling price would be?

These new edge-learning products might only be $1,000 or $2,000 per system, and a customer might need a couple of those. At the low end of the spectrum, you can think of an order being maybe $10,000 or something like that. Then, obviously, at the high end of the spectrum, the individual sensor or vision system can cost over $10,000, and there could be many of them on a single implementation. Those orders would be running in the hundreds of thousands of dollars.

What about the turnaround time for something being contracted to actually being delivered? I would imagine there’s a lot of lead time in terms of when they’re building out a facility. What does that look like? Is it something where they have to be building this to spec when the customer gives them the order, or do they have off-the-shelf stuff that they can deliver?

The smaller purchase orders, I think, are more book-and-ship type of business—a very short cycle. Then, for their largest, most strategic customers with complex capex plans and the like, there’s more of a lead time. You could think of it as they’re going to build the shell of the factory first, then the lines, and then, after that, the machine-vision cameras go on. Or they’re selling to the machine builders, which obviously have their own lead times. For those larger customers, I think they generally have—not crazy visibility, but a little bit of visibility.

Taking it down to gross margins and operating margins, where do those typically hover, and how much cyclicality is there?

Longer term, they target 15% top-line growth in constant currency and 40% incrementals, and then they layer on capital allocation from there, starting with revenue. In the 10-year period ending in either 2020, 2021, or 2022—so, just before this down cycle—and adjusted for the divestiture that they did, they’ve grown about 13% excluding M&A, and then mid-teens in constant currency. So, they’ve been right there.

I think the right long-term bogey, and what most people have in mind, understanding we’re at a cyclical place, so a few years above this, is more like low-double-digit top-line growth longer term. That’s how most people think about it. But again, we’re at the end of a very long down cycle, and there’s this new customer base being added to Cognex that hopefully should drive some revenue, at least in the near term, that’s above that.

On margins, similarly, they’re depressed cyclically right now. They also have the headwind from the emerging-customer initiative investment and the sales force investment, and they actually expect the emerging-customer initiative to be operating-margin accretive. It already is gross-margin accretive, but operating margins last year were 13%, and that’s down from a peak above 30%. There are a couple hundred basis points of emerging-customer-initiative headwind in there.

But longer term, they target regaining the 30% level as the leverage returns to the business. They get back to growing, and then, obviously, the sales force investment flips from a headwind to a tailwind to margins.

That swing is pretty meaningful between the 40% incrementals, the 30% target, and the 13% last year. I know we’re in a down cycle right now, but when you go back over time, is that the type of swing that you tend to see?

Not to that magnitude. I think part of it is that they’ve invested very heavily in this current strategy, which makes it deeper than it otherwise would have been. For example, 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%.

The depth and duration of this downturn, along with investing through it, is what makes the operating margins what they are. The target, again, is to regain the 30% level and then grow incrementally at 40% from there.

In terms of the cycle, are there signposts that you monitor, the industry monitors, or the management team monitors to suggest that there are inflections? I know there are several end markets here, so it’s going to vary, but what could you point to just in terms of the timing of that and what you typically look for?

For logistics specifically, that market’s returned to growth now. That’s their largest end market. It grew quickly last year and should probably grow quickly again this year.

Similarly, for semiconductors, you watch the semicap companies—the companies selling equipment to TSMC and the like. That’s primarily who they’re selling to. Again, the outlook there is relatively positive right now.

For consumer electronics, like we mentioned, we’re at a steady place as we wait to see if there’s going to be a next consumer-electronics type of capex cycle that’s needed. That could be new features or new form factors tied to LLMs, AR and VR, or whatever it might be, but that’s usually what drives the growth there. There’s nothing tangible on the horizon right now for that.

Then automotive: we’ve found a place of stabilization now, or we’re closer to that than the down cycle, but just watching the automotive OEMs and their capex is the signpost that you watch for that one. Lastly, PMI is the catchall for the rest of the business. General industrial activity, industrial sentiment, and the like are what drive capex for those types of customers.

Would you classify their performance as somewhat of a leading indicator on cycles? I’m just trying to visualize. You have the capex announcements, which will ultimately drive the business for Cognex, but it does feel like they’re still at the front end of the actual cycle, where you will eventually see that come on board and whatever they’re producing be released. Is that a fair way to categorize it relative to just general macro trends?

I think you’ve captured it right. 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, just given the nature of that. So, it’s pretty interesting in that regard.

Definitely very interesting. You see that a little bit in the transports, but this is a completely different play on something similar. Going back to the business, you mentioned cyclicality and a little bit about investing through the cycle. Just on the balance sheet and capital allocation more broadly, how would you categorize them in terms of capital allocators and the risk of managing the balance sheet versus conservatism?

First, on free-cash-flow generation for them: they’ve generally converted 100% of net income to free cash flow over time. They produce good free cash flow, and when they’ve returned it, they’ve returned over 100% of that to shareholders—one-third through dividends and two-thirds through share repurchases.

They do the occasional M&A, and that’s primarily, to this point at least, been more of those IP-acquire-type deals, so ViDi and SUALAB. They have a fortress balance sheet, I would say. They have net cash on the balance sheet. Cash and investments, actually, right now are 10% of the market cap.

They’ve always run that cash on the balance sheet. When you piece it all together—a cyclical business, which they won’t argue with—how do you approach valuation when you can see sharp changes in cycles and different dynamics like that? How do you go about approaching it?

I think there are 2 ways that I approach valuation on this one. The first is that we look at what’s implied in terms of future free cash flow growth to justify the current valuation. We do that by assigning an exit free cash flow yield out in the future and then applying our internal hurdle rate, which is at least 10%. Then we place that within the range of outcomes for the business to assess the attractiveness.

At present, we think, frankly, that Cognex needs to compound free cash flow at more of a low-double-digit rate over the long term. That’s below the long-term model and also coming from a place that’s at the low point of a cyclical downcycle, with a lot of hopefully margin recovery ahead. So that’s the bull case.

The other way we look at it is the historical multiples. Given the margin dynamics, we’re specifically looking at EV to next-12-month sales as the relevant one for us. Currently, it’s within reach of its 10-year low, at 5.5 times next-12-month sales. It traded as high as 16 times sales during ZIRP, which, like most things, is just hilarious in hindsight. Otherwise, in a more normal range, it’s been more like 6 to 10 times sales.

My takeaway from that would be that the market might be suggesting there’s some more meaningful compression in the margin profile of the business. We’re certainly seeing it beaten up through the cycle, but is there any reason to believe that margin compression—or the margin recovery—won’t be as strong as the cycle recovers?

I think that’s certainly a discussion. Historically, we have seen margins decline in magnitude with cycles, and they’ve always recovered back to 30%. We just saw, even in the last quarter, where some growth returned to the business and they grew in the double digits organically, that you can see the leverage that fell through in the quarter. There are signs like that that we’re watching.

That’s certainly a debate on the name: what normalized margins will look like. It’s always the debate on anything in this space.

I can appreciate that. What are the other risks you would point to? I think we probably outlined a lot of fairly obvious risks, but what really stands out to you?

There are a couple that are top of mind. The first one is obviously just the cyclicality. Early is the same thing as wrong. Basically, every incremental investor in Cognex over the last year or 2 is probably feeling very early right now. With regard to that one, frankly, it’s just a matter of whether the cycle will turn, or if we’re sitting here a year from now and still feeling early—hopefully still temporarily.

The other one is China, which we briefly mentioned. It was 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 and Foxconn. For those customers, it seems unlikely they’re going to put Hikrobot in their factories, for obvious reasons.

For the remainder—call it the other mid- to high-single-digit percentage of overall Cognex sales that comes from domestic Chinese manufacturers—I think, on a 5- to 10-year basis, it’s probably going to be an uphill battle to grow those customers, at least. That’s a well-understood risk as well, that people talk about.

The last one would just be the technology transition. Ultimately, I think the technology transition opens a great window of opportunity for Cognex, and there are reasons to believe they’re early and ahead in terms of applying machine learning to machine vision. But with any technology transition, it opens the window for disruption as well. So that’s the third risk out there that’s on my mind.

Yeah, it’s an interesting cyclical name because, when I think of the majority of cyclicals, they’ll trade in terms of the slowdown or the growth within a range. It is not this S-curve-type growth where you do have major pickups. To the extent that they do have new S-curves emerge—which, from a very, very, very high-level view, I would imagine exist out there—it’ll just be a matter of how Cognex aligns. It’s an interesting one, with a slightly different tilt on cyclicality than a normal business.

What’s also interesting is that cyclicals, whether it be analog semis, semicap equipment, or metals and mining, will trade at peak multiples on trough earnings and the like. You look historically at Cognex’s multiple, and it’s basically trended up when sales are growing and down when sales are declining. It’s really interesting. I don’t know exactly why that is, but that’s the reality.

The peak-on-peak phenomenon can be painful on the way down. It’s always interesting to watch.

This has been very informative. It filled in a lot of blanks that I had in terms of the research that I did, and there were many. What would you point to as the key lessons to take away from Cognex and apply elsewhere?

It’s funny. I was thinking about all the episodes of Business Breakdowns I’ve listened to, and I feel like the number-one answer has to be culture. I’d be curious if that’s accurate, but I think it’s clearly culture for Cognex, too. What’s interesting about it, more than just having a unique culture, is how they’ve maintained it through the founder’s departure.

Specifically, like I mentioned, the ministers of culture have the job of maintaining it. Secondarily, there was that long overlap of the founding CEO with the new CEO as the new CEO took over the reins, just ensuring that it lasted through the transition. I think those were unique, and I hadn’t really seen that before elsewhere.

It’s definitely one of the more common answers, often related to businesses that have had these very long-term durations of success. It’s interesting and very hard to measure, which probably makes it even more valuable to study. It’s a very fascinating business. Thank you again for sharing the knowledge, Brett.

It was a pleasure. It was a lot of fun. Thanks for having me back on.