Okay. G ood morning, everyone. My name is Chad Dillard. I am the Lead Analyst here at Bernstein, covering the machinery and the electrical equipment sector. And today, I'm really pleased to have Cognex on the stage. And joining me for Cognex is Dennis Fehr, who is the CFO of the company. W e're going to have a fireside chat over these next 40 or 45 or so minutes. If you want to participate and have any questions that you have for Dennis, there's a QR code on your table there. Just click on that. That'll take you to Pigeonhole. Write your question up, and I've got my iPad, and I'll do my best to convey the question. B efore we jump into Q&A, I'd love for you, Dennis, just to take us through Cognex for some of the folks that may not be as familiar with the company.
Just give a really broad overview, and then we'll jump into questions.
Thanks, Chad. G ood morning, everyone, from my side. Thanks for your interest in Cognex. And yeah, a few words to Cognex, a few words to myself. So who is Cognex? What do we do? We are a pure-play machine vision company. That means think about software embedded on device, high-speed cameras, machine vision systems, helping to identify, to inspect, to gauge in factory automation or factory environments and logistics and warehouse automation as well. We're about a 40-year-old company, headquartered just outside of Boston in the town of Natick. We have a global footprint. We do about 40% of our business in the Americas, 20% in Europe, and 40% in Asia. We have been and think about ourselves as the technology leader in the space, basically leading the solving the most difficult tasks to solve in terms of inspection, the most hardest code to read, fastest speed. That's kind of our legacy and how we are positioned.
We're operating in a market of about a $7 billion TAM, which has been growing for most of the decades and kind of a low- to mid-single teens as a growth rate there, so as the company. We traditionally have been working with the most sophisticated, the largest companies in the space, naming some of the most iconic companies, our customers like Amazon, for example, in the logistics space, and some big names in the consumer electronics space. Have been over the last two or three years further kind of expanded our strategy to go after a broader market. That more and more through a direct sales force, which we see as a competitive differentiation and also a way how to competitively win.
We traditionally are a high-margin business, so kind of more software-type of gross margins and bottom-line margins, so 10 years historic average, 28% Adjusted EBITDA. And we run a capital-light business model. That means we are highly cash generative and have low CapEx, less than 2% of revenue per year. Myself, as introduced, Chief Financial Officer of Cognex. I'm with the company since about 20 months. So it has been a good journey. Has been driving a couple of priorities around profitability, capital efficiency, and investor relations. I would say I'm part of a new management team with Matt Moschner, the CEO, basically taking up that role mid this year after being eight years with the company. W e're driving a lot of change and a lot of P&L improvements with the company, and we'll be excited to talk more about that.
Okay. Let's dive right in. So yeah, let's talk a little bit more about that change since it's a relatively new management team. You guys came out with some recent strategic priorities. I'd love for you to just lay those out for all of us, and then we can dive a little bit further.
Sure. Absolutely. T he three strategic objectives we talked about at our Investor Day mid this year. So first is to be the number one leader with AI technology and machine vision. It's kind of extending our legacy as the technology leader in the space. Let's just think about that for 40 years or so. What we call rules-based software coding, so to say, was kind of the state of the art. And we have been and still are the leader in that type of approach. And that's more and more complemented by AI-based vision tools and vision models. W e embraced AI very early on with an acquisition late 2017, early 2018, and bought a company called ViDi in Switzerland, and that built our basis for AI. So that's the first strategic objective. The second is about doubling the customer count.
That ties back to I introduced before. We traditionally have been very strong with large global customers, very sophisticated customers, and very sophisticated problems to solve. And we think that there's a broader market we can go after, and that's kind of tied to that strategic objective. And then the third strategic objective is that one thing is to win new customers. The other thing is to keep them. And that's where basically to be the number one in customer experience in this industry plays a role, right? So think about it like a flywheel. You add new customers, and then you keep them, and you increase the share of wallet over time. You penetrate them further. And that's kind of where being the number one in customer experience plays a role.
Great. Okay. So maybe diving a little bit more into that doubling of your customer base. Can you talk about just what changes in the organization do you need to make to realize that goal?
Sure. Absolutely. See, I think that's a journey which we have really embarked on somewhere like in 2023, where if you look back, I talked about it, how we have been kind of leading and being very good with large-scale customers. That, on the one side, certainly brings a big advantage. If you win these customers, you gain big volume. You can grow with these customers. They have the most challenging topics to solve, so that means it kind of really positions you and keeps on challenging you that you are staying the technology leader, but it certainly has also some downsides, right, so one downside is clearly that, is there a ceiling? How far can you grow with just large customers, and then it drives more volatility and more cyclicality into the business. So that means you see strong ramp-ups in growth cycle, but you also see larger compression in down cycles.
In that regard, there was this kind of the strategic narrative. If we want to go beyond, then you want to do more. You want to serve more customers, and that basically created the idea of what was called the Emerging Customer Initiative. That was an initiative launched and announced in 2023, and it has two or three key components to it. It was A, let's put more salespeople, direct sales force people, more boots on the ground. That means go out, have more sales capacity, reach more customers with a bit of a different approach. That means thinking about like, okay, you need to have easier-to-use products to serve broader customers, and therefore kind of combined with a product strategy.
And as the thinking was like, okay, this is a very different approach. Maybe you sell slightly different products. So it was really made this Emerging Customer Initiative a standalone initiative, a standalone sales organization. And that had great benefits. So really acquiring new customers, really that target kind of we saw a nice ramp-up. We also saw that kind of this new type of sales engineers, which were mostly higher ed graduates out of college, they started to show productivity signs. T here were good things about it. But at the same time, we also realized that this may not be, A, the most efficient way to do it. And then you get kind of dyssynergies with your existing sales force because now you have this kind of completely other part of the sales force, and you have two different leaders.
You have one global leader for that part and another global leader for the other part, so that means we pivoted and slightly adjusted at what we call the sales force transformation, so the objective by itself is still similar. That means we want to increase the customer count, and instead, we want to double the customer count, but there is a second piece to it, which is all about sales force efficiency, so that means what we did is we took these two different sales organizations, we merged them into one, and we are much more focused now about, like to say, let's really drive sales efficiency. That means we are thinking a lot about how can we optimize and drive efficiency all the way from lead generation, how we are converting leads into opportunities, and then opportunities into orders. And that's a lot of data-driven, and there's a lot of adjustment into how we manage the sales force.
We still have different selling profiles, so that means we still have people that are more focused on acquiring new customers, and they're selling kind of an entry-level type of product basket, but then you have also two other types of seller profiles in our sales organization who are then catering to other forms of customers, or if you move these customers on in their journey, they eventually might be served by other parts of the sales organization, and then we broadened a bit the approach of the easy-to-use product, so that's still there. We're still trying to first sell easy-to-use products to new customers, and then at the same time, we're working on this more holistic customer experience, kind of the end-to-end journey.
Like, how easy is it to find Cognex? It plays in order with Cognex, how easy it is to get support from Cognex. W e're trying everything easier, and we call that advanced machine vision made easy.
Got it. Okay. So let's actually start at the base. Let's talk about the product cycle, right? So that you're going to need to imagine you're going to need to change that to some extent to reach the customers that you want. I guess, how do you approach developing products differently? Because I have to imagine you have the white glove service with your large clients. The smaller guys, to profitably reach them, you probably need to have a little bit more DIY. So let's have a conversation about that.
Yeah. Right. I think that really touches a bit to this point of customer experience and easy-to-use products, right? As you rightly say, so the more sophisticated customers who have maybe full teams of factory automation engineers, and in these teams, they're likely specialists around machine vision, machine vision engineers. You don't need to do this type of make it easy for them. They will figure it out. They're the experts in the area. But then it really comes to about building guided workflows for these folks who are much less familiar with machine vision systems. And that's where kind of this theme of customer experience plays a big role. And in our sense, also kind of a unified software ecosystem plays a big role. I f you think back five years or so, right, we had different kind of software stacks for each of the product lines.
Now that's all on one software stack, basically, and that means you start to access this easy-to-use product on the same software environment. If you go to the next level of sophistication, you're still in the same software environment. That means it's so much easier for customers to first enter the ecosystem and then kind of move into the ecosystem and kind of add sophistication level, and in that regard, clearly a very distinct, different approach, so it was a very strategic move probably like three to four years back to go to this ecosystem and then to start to build out these kind of guided workflows and kind of self-tuning auto setups, so that means if we sell one easy-to-use product, think about more like an experience like you buy a smartphone, you take it out of the box, and it basically guides you through a setup process. And a lot of the setup is done by the device itself.
We demoed at our Investor Day such kind of a self-setup. So it's basically we built an AI-enabled auto-tuning and self-setup. So basically, if you buy one of the machine vision systems, basically it goes through four steps, get a few questions, answer these questions, and then the device basically configures itself. And that makes it so much easier for our customers to use it. And that's where product development plays a big role to bring that type of mindset into that development.
Okay. So you get the product development setup. So now let's talk about the sales angle. I guess, how do you change who you recruit differently now versus under your prior model? And then more broadly, can you talk about what sort of incentive structure that you're putting in place so you achieve the outcome that you want?
Yeah. No, it's a great point. And I think one thing is how do we hire, right? I talked a bit about Emerging Customer Initiative where we hired college graduates, and we wanted to build a kind of a seller profile, a sales engineer profile who sells much more transactional and less consultative. And now as we merged into this one sales organization, we still have these three distinct seller types, and we would hire differently for each of them, right? We have one more selling type selling kind of this easy-to-use product, very transactional. So you're looking for either graduates or maybe people with one to two years' professional experience. And basically, you feed them through your lead generation, you feed them these leads, and they go out and they run.
And then you have a second seller type, which is more like working with some of the mid to more sophisticated type of customers where it's really consultative selling. That means you walk the line, the production line with your customer, and you basically really talk it through where your problems, where you have cost and quality issues, and you bring them ideas, and you basically start to create your funnel and your leads with this kind of consultative selling approach. And then there is the third one, which is very technical. It's almost like, right, we're working with machine builders, and they are basically speccing in our machine vision devices into their machines. So that's a very technical selling. It's really on an almost application engineer to a design engineer on the customer side type of conversation.
It's really about like, how do we want to solve this technical challenge we have, and these are kind of the three different buckets how we have structured the sales organization, and that in our mind drives a winning more customers and then keeping, retaining them, but again, I think this portion of driving efficiency into the sales force through this automatic lead generation and to basically streamlining the way how we manage the sales force, right? Start to think about sales dashboards. Each sales engineer has their KPIs. How they're performing this week against this KPI? How are they stack ranking against their peers in their districts or in their regions? It's a very different type of approach of how the salesforce is managed today than it was, I would say, 18 months ago.
Okay. H ere's the next question. So the umbrella of the question is like, how do you manage your brand identity? And you've got, I guess, the legacy of Cognex has been more focused on kind of like the high-end bespoke. And you're going more towards, I don't want to use the word simple, simpler. So obviously not simpler, but less complex product types. I guess, what's the most important factor in executing what seems like maybe a bit of a tightrope between those three?
I think it's clearly, see, on the one side, I think we are clearly positioning ourselves, and it's our core and our legacy, so to say, as being the technology leader. I think it's very core to the brand identity. But then I think where we maybe expand the brand identity is that now, all of a sudden, you're also becoming an easier-to-use and easy-to-interact with company, right? I would really emphasize the both aspects to it, right? One thing is like you have a product which is easy to use, easy to set up the guided workflows, but the other piece is like the easy-to-interact with. And I think that's where we probably is the most transformative for the company itself, right?
Because this very sophisticated company is where you're very entrenched with their engineering organization, where it's a good personal relationship that plays a very different role than if you go to a new customer and you want them that their end-to-end journey is really kind of seamless, right? Oh, okay, I got a demo of your product. Now, how do I order it? How easy is it for me to order? How easy is it for me to get it, to set it up, to install it? And if I need support, how easy it is for me to get support? And that's really kind of an expansion of the brand identity. And that's kind of what the strategic objective of being the number one in customer experience plays a big role. And it's really an expansion of the brand identity.
Got it. Okay. So in your sales, are you using any third-party distributors, or is that fully in-house?
Right. W e are today probably like in a 70%-80% of what we call direct sales approach. But then the other 20%-30% don't think about it like distributors. Think more about integrators. So that means they're providing a value-added service to the customer. So they might be pulled in, for example, to help with a broader system integration connection, maybe to their shop floor management systems or to their all the way to pulling cables and mounting brackets and things like that. So in that regard, that is a piece which we do not necessarily want to do, right? I think that would be dilutive to our margin profile. And at the same time, it would be also a kind of a fixed cost block in certain regions, which you may or may not always be able to fully utilize.
In that regard, it's really a very conscious decision for us to say like, "That's we don't want to have in our P&L. Let's use third-party partners to that." And it's clearly also an area where we still have opportunity for us. Like, how do we manage the channel? I think we'll definitely have also some good ideas there.
Got it. Okay. So is there any way to kind of benchmark the progress that you're making on these targets? Maybe what share of your revenues come from new customers? Just kind of get a sense for where you are today versus doubling that base. And then I guess, secondly, you've given out these 10%-11% top-line kickers. How much of that depends on this customer plan?
Right. B asically, maybe on the customer accounts, we have basically announced it on an annualized basis. W e're just approaching kind of the next cycle to announce that. I'm not yet talking about it. But the end of last year, we were about 10% up in customer count, and so we expect a certain acceleration in that regard. But in general, I think we feel like we're making good progress. We like what we see, especially on acquiring new customers, finding new accounts. But then I think I feel like even more relevant is like, as a CFO, what do I see in the P&L? And maybe first I talk a bit about the objectives, and then maybe where do you really see it to show up, right?
We put out a financial framework, which included or includes a growth algorithm where we basically say a 10%-11% organic growth and then + 3% M&A growth through the cycle. So now if you look at this 10%-11% revenue growth through the cycle, we further split that into two core components. There is a 4% underlying market growth. So that means how much are our customers growing? So we track that about 4% across our verticals. And then we add a 6%-7% from additional penetration. And that additional penetration comes from two angles. So one thing is bringing AI into machine vision that enables more use cases, which you could not solve with traditional ways of rules-based software algorithm. So that means we can sell now more applications to customers, and therefore you get more penetration.
And then the second piece is to go to more customers. So in that regard, it's a core piece in our growth algorithm, but then it more applies to some verticals more than others. So key focus areas are the packaging vertical. So think about fast-moving consumer goods and healthcare, for example, pharmaceuticals, food and beverage, and then also in the automotive industry. So that's kind of two of these key verticals, and they're probably a bit more than a third of our total business. So in that regard, in these verticals, that aspect plays a big role.
Okay. L ast question on the customer part. I know we've spent a lot of time on it, but as you think about the customer acquisition costs for these new customers, how does that compare versus your legacy customers? I guess what I'm trying to get at is to what extent is this move margin accretive?
Yeah. I think there maybe I go through on a P&L structure, right? So first, starting on the gross margin side. So it's gross margin accretive, right? You typically have selling to smaller customers at lower quantities, you'll have better prices, right? So ASPs are higher, also the type of products which we're selling. So in that regard, gross margin accretive. And then it comes back to the dollar of your sales organization, the cost of sales organization per dollar booking, right? So in that regard, it's comparatively more expensive to acquire a new customer because very likely you're getting smaller order sizes, and then these are smaller customers. So that means inherently, your dollar per booking is higher.
At the same time, to some extent, the cost of this type of sales engineer is a bit lower because they have a less of an experience profile that helps a little bit offsetting that. But at the end, really, the business case works.
Okay.
The business case really works when you acquire the customer and keep the customer, and that's why the second piece of the customer experience is so important, right? It doesn't help us if we just do a one-time sell to a customer. It's very likely where you can't make a lot of money, right?
Where you start. [audio distortion]
Where you keep that customer and where you then do further penetration, have repetitive selling, getting larger orders. That's where the business case really starts to become attractive and where it becomes to a margin accretive, basically.
Got it. Okay. So let's switch gears. Let's talk about 2026 to the extent that you can. Yeah, I guess, what are some early thoughts on the evolution of your business into the coming year? If you can provide some color on the top line, maybe talk about the profitability as well.
Sure. Thanks, so maybe before I talk about 2026, I'll take a step back, and at the end, we think about we're a cyclical type of business. We showed on our Investor Day, and it's basically rooted in almost like 50 years of industrial manufacturing data that you have in factory automation and machine vision and Cognex-specific. You have cycles. You have an average length of five to seven years of the cycle, and that means if you think back, the last peak, the peak of the last cycle was 2021, driven by logistics and consumer electronics, and then we saw basically a downturn of a cycle, especially 2022, 2023, and it basically started to flatten out into 2024. We think that it's very likely. I mean, sometimes you only notice in retrospect, but it's very likely that we are beginning of a new cycle.
2025, I would say, start of a new cycle. And then we basically laid out a framework of that cycle and said, "Typically, we have seen that a cycle unfolds in three phases. An initial phase where you see moderate growth, a second phase of the cycle where you see strong outsized growth, and then kind of the tail end of the cycle, which is either flat or even down." So that means early stage of the cycle. We see this year, a growth rate in the mid-single digits, which we are using as we work very hard on P&L optimization, right? So we had a pretty low profitability last year for our standard. It was just 17%. And we basically, our implied guidance we gave is the Q4 guidance for this year was greater than 20%.
We're working on P&L optimization, and we are basically showing this at a mid-single-digit growth. We can drive outsized P&L performance. So in that regard, that's kind of how we start to think about 2026. We are a short-cycled business with limited visibility, I should add that. So that means we typically have in factory automation, we may have just three months visibility. And then in warehouse and logistics automation, we have maybe six months, six to nine months visibility. So in that regard, when we think about 2026, we look at where we today, we see mid-single-digit growth, and then we use an additional macroeconomic indicators like the PMI. And PMI is still in most geographies below 50. So that means it doesn't show kind of an expansion.
That means our baseline as a management team at the moment is to say like, "Okay, it looks like the next six months or perhaps nine months will be mid-single digit growth." But of course, we are a short-cycled business. That means there can always be inflection, and it could definitely go further up. We clearly also talked about some of our end markets, and we see that end markets are generally getting better. So in that regard, I would really emphasize that's a baseline case. W e clearly also said, and I felt like that maybe got a bit lost, that we said like, "We can drive outsized EPS growth in a mid-single digit growth environment." Right? W e're growing our EPS this year greater than 20% on a mid-single digit growth. And we think we can repeat that or exceed it in 2026.
And that's what we're really working very hard on. And the key to that is really tight cost management, OpEx efficiency. Right? Sales force transformation plays into that. But in general, it's really across the whole organization that we are looking for efficiencies and driving OpEx efficiency. I think we have executed well in 2025, but it's also clear that we have more work to do, and we are poised to deliver on that in 2026.
Okay. Great. So actually, let's dive a little bit more into end-market by end-market. How are you thinking about what seems like the trough and the shape of the recovery? So maybe just beginning with consumer. Can you talk about that?
Sure. Absolutely. Consumer electronics is our third largest end market. It's also, to some extent, the most interesting one. We sometimes say packaging is our most boring end market, and then consumer electronics is almost the most interesting one because it really has typically shown strong inflection points in both directions. And consumer electronics last peaked in 2021 and then really contracted a lot in 2022, and since then basically hasn't grown until we saw first signs of growth this year. So in that regard, it's really the most exciting also because it's the one which kind of shows the first signs of changing from flat to a growth in the broader factory automation outside of logistics, which has been growing for the last two years. But we'll get to that. C onsumer electronics, what makes us excited about this vertical?
It's really that in a broader sense, if you think about it, change is good for Cognex. So change in the sense of our customers launching new products, they're changing their supply chain, or changes in regulatory requirements in certain industry. So that means whenever change occurs, it drives good business for Cognex. W hat are we seeing this year? W e definitely see a change in terms of the alignment of supply chain. So that means a lot of our customers in consumer electronics looking to diversify their supply chain from China to other places. First of all, starting on the device assembly. That means thinking about where our device is assembled. And that's maybe places like India, could be places like Vietnam. And that's kind of driving some part of that initial growth, which we're seeing this year, and we think this can extend into 2023.
Beyond that, we think components will be the next item to look at. Eventually, customers really want to diversify the entire supply chain and not only the assembly. So that means component manufacturing would be the next item, potentially 2027, 2028, to make moves. And that's good for Cognex. And then the other thing is that after a long time, we also start to see product lineups, new product lineups coming out, which are rooted in changes to hardware form factors, right? If they're just changes on software, that's not really driving business for Cognex. But the moment where there are changes to the physical manufacturing or additions, that is good for Cognex, and that's what we are seeing as well.
I n that regard, we think we could really see a couple of years of nice growth in consumer electronics. Again, it's sometimes hard for us to exactly call what that will be and when it will be because of the nature of the business, but in general, it's definitely the market we are right now the most positive in the sense of how the growth dynamic could change in that market.
Okay. Any product cycles to get excited about over the next 12-18 months?
Yeah. I think, see, they're clearly so first of all, maybe I should say a step back. What we see this year is a broad-based growth in consumer electronics. So it's not about just phones or it's not about laptops or TVs. It's broader-based. And that is, first of all, nice because obviously, it drives diversification and makes the cycle potentially stronger. But then clearly, in the smartphone side, we have seen two things, right? We have seen new form factors coming out, and there's been buzz about additional form factors. Can't really call that, but the buzz is there. And then certainly, what we also see is that there was a lot of kind of consumer spending on devices in the COVID era. W e're seeing kind of a natural refresh, which kind of drives good business for our customers, and that's good for Cognex.
Let's broaden that conversation. Let's talk through logistics. Let's talk through packaging.
Great. So let's start with logistics.
I'll talk about it if you can too.
Yeah, sure. We'll take them one by one, but I'll go to logistics first. Logistics is our largest end market now. We have seen it growing very strongly since basically the beginning of 2024. So the first vertical market to recover from the down cycle and have seen strong growth last year, see strong growth this year. I would say until mid of this year, it was a broad-based growth. I would say in the second half of this year, the growth was more geared towards larger customers. And in general, we've been very [audio distortion] positively about the long-term outlook for logistics because it is an industry which today has a low penetration of automation. And that basically gives a long run rate. And we have seen in the 2021 peak that was a lot about building out additional square footage, building out steel, concrete, but it was not fully optimized.
It's a low automation rate. And that means there's really this multi-year opportunity by automation into this existing square footage. And therefore, we are very positive, actually, that we would see a multi-year growth cycle in logistics. However, we also know that growth is not linear. So in that regard, we could imagine that 2026 might be a year with a lower growth rate. Could it be flattish? Perhaps. Could it be down? Probably not. But that's not changed our view on the long-term outlook for this market. So in that regard, probably more a sense that after two years of outsized growth, maybe take a breather and then kind of start to climb the next mountain.
Okay. Great. Oh, yeah. Go through.
Which one do you want?
Yeah. Let's go through packaging and then then autos.
Okay. L et's go to packaging next. So packaging, I mentioned before, I would describe our most boring end market, right? P ackaging is like fast-moving consumer. Go think about shampoo bottles or cosmetics in a bottle. Think about food packaging, like cookies in a plastic box with a transparent packaging and then in a carton. Pharmaceuticals, right? So why do we call it packaging? Because it's typically all about package inspection, like is the cookie sealed or not? Does the shampoo bottle have a scratch? Does the glass bottle have some glass chipped away? So that type of kind of inspection tasks. And by itself, the underlying verticals are very stable, right? They don't go through big cycles. They typically have low single-digit growth. We think we can grow there with penetration, bringing sophisticated AI tools into these inspection tasks.
I visited a cosmetic manufacturer in France earlier this year, and they fill their cosmetic into glass bottles, and they want to make sure nothing falls into it, and especially little glass pieces from these bottles. And that's very hard to detect, right? It's transparent, something little chipped away, moves very fast, right? Could be thousands of these little bottles per minute. So in that regard, you need to be extremely good. And that's where really the latest AI vision tools help to solve these problems. And that's how we can drive penetration in this industry. So in that regard, we expect this to be a steady grower. And at the same time, also an area where we can reach more customers and therefore feel like it's a market where we would expect steady growth over these years.
That makes it a bit boring, but exciting at the same time.
Okay. Great.
And then lastly, or there might be semi as well as a market, but maybe I'll talk to automotive first. So automotive clearly was the biggest challenging market in the last two years, you can say, right? So we saw a very good year in 2023 from EV batteries, right? So a lot of was the hype moving towards the EV transition, building out more battery plants, and battery plants need a lot of inspection, complicated, sophisticated inspection. So in short, good for Cognex. Good year in 2023. That transition kind of got a little bit slowed down, and therefore a lot of investments didn't move forward in 2024. And basically, that business on the EV side almost disappeared. But then shortly after, also the traditional ICE side of the vehicles also started to show signs of weakness.
Then the tariffs came into play. A lot of uncertainty for car manufacturers. And that means auto declined 14% in 2024, and it's going to decline in the single-digit percentage this year as well. We'll disclose more about that at the next earnings call. But positively speaking, if you can find something positive in there, we start to see signs that it's stabilizing. W e still see year-over-year declines, but if you start looking sequentially, it looks more stable. W e would like to have probably one or two more quarters to say it has found its bottom, but it feels like we are close to that. And so in that regard, it takes away a headwind if we think about 2026. So that market definitely has a better outlook.
While it may not be a big growth driver, at least the headwind which we have experienced in 2024 and 2025 is expected to go away into 2026.
Okay. Take us home.
Okay. Let's do the last one. Semiconductor. What is that? So basically, think about we are selling machine vision systems to the semi-cap manufacturers. They're building machines to produce chips, all sorts of chips, right? From very simple for standard controllers going into a certain cost for high-bandwidth chips. And we are basically working with these machine builders that are very much specced in there. And we have seen nice growth last year. 2024 semi was our largest growth market, very strong growth. And then we saw a pattern which is kind of what we think may happen in logistics in 2026, is that in 2025, kind of that market takes a bit of a breather. And we originally coming into this year thought it would be flat, no growth. N ow we see a little growth, which is better than what we thought it would be.
We think that we could see further acceleration in this market into the second half of 2026, if not earlier. So in that regard, also a market where we have a rather positive outlook.
Okay. That was a great around the world. Okay, so in the last couple of minutes that we have, I want to spend a couple of moments on AI, and I want to take this from the opportunities perspective first, and then we can talk about the threats, so on the opportunity side, can you talk about how you're integrating AI into your product cycle?
Absolutely. I talked a little bit about it before already, how we made that acquisition of ViDi, that company in 2017, 2018. We launched our first AI-embedded product on our devices in 2022. And basically, since then, each new product which we have launched basically has some AI features and AI components to it. And again, what does it mean? It really means you can do additional applications which you couldn't before. I gave this example with the glass bottles, these little glass chips as one example. Another one could be and I went to a consumer goods manufacturer. They do dust sheets. And in the past, we could inspect the dimensions of the dust sheet. Was it cut properly?
But we couldn't inspect white on white, a white structure on a white background, very hard, a rules-based code to say what of the structure is good, what is bad. But with AI, you can train it, and you can teach it and say, "This is good." Or, "No, this is folded. This is a normal structure, but here it's folded, and that's not good." So in that regard, basically enabling them to sell more machine vision systems to such a customer. So that's really what AI does, and I would say over the last two to three years, AI basically progressed on the product side, on the feature side, maybe in two angles. One is deep learning, and the other one we call edge learning, so think about edge learning. You basically, you're on device. You're not connected to a cloud.
You used existing compute power of the device, and you can train it. We have our easy-to-use products. You maybe train 5 to 10 pictures, right? It has a pre-built model in there, so 5 to 10 pictures [audio distortion] and good quality of the model, and off you go. You can inspect doing pretty good tasks with that for that inspection, but then there are these very hard type of inspections like the glass bottles. You may need deep learning, and that means for deep learning in the past, you needed to have the power of the cloud. That means you couldn't do it on the device, and basically, what you did is you did what we call a PC-based vision. You had a laptop standing somewhere on your line, and then you train in the cloud, and you're always connected to the cloud, and customers don't like that, frankly speaking.
They have cybersecurity issues. They may not have internet. They may have only local area networks in their factories. They have latency issues. So that means going through the cloud actually is a problem for them. So in that regard, there was kind of a missing link between edge and the cloud. And this year, we launched OneVision, and OneVision solved that for our customers. Basically, what it means is they can use the device, take the pictures on the device, upload it to the cloud, train in the cloud, and bring an improved model back to the device. And at that moment, you don't need the cloud anymore, right? So that means you're going from an always-on to an I only need to be online to train, like a virtual training room. And once I'm trained, I can just exist back on the device.
That means you're bridging the edge with the cloud. And that, again, drives tremendous opportunity for additional applications. But at the same time, it makes it so much easier for our customers to go after these applications.
Got it. Okay. So we've got one more question left. So in the age of AI, how does Cognex differentiate itself so it doesn't become disintermediated by the next native AI competitor?
Sure. No, absolutely. It's a great question. And we talked a lot about that at our Investor Day. So in that regard, so first of all, what are we solving for our customers, right? We need high accuracy. We need high speed because production lines can move pretty fast. We also need scalability because very often, think about you produce maybe today you produce a shampoo, and maybe tomorrow you produce the shower gel, and maybe you have a different version of shampoo. So you have variability. So that means you need to have scalability. You cannot just have a solution which solves. You calibrate it for one thing, and then that works, but for the next thing, it doesn't. And then certainly, you need customer experience, and it needs to be easy to be used, and it needs to be cost-effective.
In that regard, sometimes getting asked, some of the large language models out there, are they going to disrupt what we do? And large language models, as the name says, they're trained on a lot of generic pictures. And that means they maybe can distinguish a zebra from a lion, right? But we are talking about very specific factory automation-related identifications and high accuracy. So that means we have our proprietary vision model, which we are training with very proprietary information that's highly specific to this area. And these are pre-deployed models to our devices, and that enables us to train them just with 5 to 10 pictures and still have a much better accuracy than these large language models. So that means we beat on accuracy. We beat on speed because you can't go through the cloud. You can't use three seconds to make a compute, right?
You need basically think about 50 milliseconds kind of response, which you would not achieve with a large language model. They're too big, too generic, too general. And then certainly, the question of the cost, right? Do you need to have clusters of GPUs, or do you just use kind of that one chip processor on a single device? So it's much cheaper in that sense. And then this topic of variability and scalability, where again, kind of our proprietary machine vision model plays a big role. So in that regard, we feel pretty confident that AI is an opportunity, a great opportunity, and much less a threat.
Great. Okay. Actually, I've got one more question for you.
Right.
Okay. So Amazon has a plan to have 1 million robots. What is Cognex's role in Amazon's robot project? What is a business opportunity for Amazon?
Right. So probably a bit hard to talk about very specific customers, but I'll try to give maybe a general sense. I mentioned before that logistics and warehouse automation is logistics vertical is low in automation, right? So that means you see still a lot of people working around moving things. You see people maybe manually scanning thing and manually identifying things [audio distortion] over the last 12-18 months, more automation [audio distortion] to take out cost. How much does it cost?
Then also clearly, we're working with all major robotics manufacturers, some of them who have this as their business to sell externally. Some of them might be just in-house. And they basically use us to be the eyes of the robotic arm or of whatever form the robot will have. So in that regard, you will find Cognex machine vision systems very often where you have maybe a kind of a picking application, like a robotic arm takes something and moves it into a certain position where it's then being processed. So you will very likely find a Cognex machine vision system either directly mounted on that arm or somewhere overhead so that it basically tells the robotic arm where does it need to go, right? Because it's not enough to just think like, "Oh, this piece is here." You may need to hold that piece very specifically.
We're talking really about very precise guidance of these robots. So in that regard, when you think about the opportunity for Cognex, think about very much about this guidance type of systems very likely in kind of a fixed robotic environment, probably less so in a mobile robot environment.
Okay. Is there a question over here?
Sure. Yeah. M arket share overall, somewhere in the mid-teens. So, it's like a fragmented market. So, think about the two largest players together clearly have maybe 30%-35% together, and then you have a lot of highly fragmented market. Part of our growth algorithm, that 10%-11% growth, which I mentioned before, would indicate some share gain in the market, but I would say it's not the key driver of the growth, right? So, that means there is some portion that the market by itself is already growing itself. That could change certainly with M&A, right? W e said also that we may look at 3% or so growth from M&A, and that could be certainly also part that we would acquire some companies playing in that space. So, that means M&A might change that.
Excellent. That's a wrap. Thank you, Dennis. Appreciate it.
Cool, Chad. Thanks a lot. Thanks a lot, everyone, for your interest.
That one is great, right?
That's good.