Real Time with... productronica 2025: Exploring AI Native Systems With Delvitech

Estimated reading time: 8 minutes

At productronica 2025, Delvitech CEO Roberto Gatti discussed the company’s explosive growth and disruptive vision behind its AI-native inspection platform. Its technology—capable of processing massive data volumes in milliseconds—is reshaping what’s possible in electronics manufacturing and beyond. From ultra-high-speed optical systems to emerging predictive capabilities and microelectronics-ready resolutions, Delvitech is pushing past the limits of conventional AOI. Roberto says what’s coming next could redefine machine vision across entire industries.

For our full coverage of productronica 2025, visit the Real Time with... productronica event site, located here.

Video Transcript:

Marcy LaRont: This is Marcy LaRont with I-Connect007 here at productronica. Here with me is Roberto Gatti of Delvitech in Switzerland. Delvitech is a very exciting company that I met two years ago at productronica, and Roberto Gatti is the man with the idea, the vision, the intelligence. We're going to talk about AOI and AI and all that Delvitech has been up to in the last couple of years.

Roberto, you started in about 2018. You're a Swiss company and you have had tremendous growth since then. So, why don't we talk a little bit about the development of the organization, and then we'll get into the development of your amazing technology.

Roberto Gatti: Thank you for this interview, and yes, the company is a Swiss-based company. We started our operation in 2018. We are in the southern part of Switzerland. And let's say that our operation officially started in 2019. We have grown up quite significantly in the last year. Yes, this is true.

The organization was initially few people. We dedicated our efforts the first five years on research and development on this new technology that is quite disruptive compared to what is available in the market. It's a technology that is AI native and being AI native and being based on a microservice, this is a technology that can be used in electronics today, but in the future, it can be used for whatever other systems. It can be for food. It can be for a farmer, it can be for agriculture, it can be for solar panels, whatever.

At the end of 2023, we really started our sales. We expanded our operation and right now we have a facility for citizen support in Germany, in Norberg. We have one in Dallas in the United States for sales support. We have one that is quite large right now in Bangalore in India, where we have more than 25 people throughout India with the main base of business in Bangalore; and in Bangalore, very soon—I would say more or less, at the end of Q1—we will establish our production for mass volume. Right now, the organization consists of more than 100 people.

This year our turnover will be roughly 400% compared to year 2024, and we predict that 2026 will be, again, a growth of roughly 400% compared to 2025. In this moment of time, it’s a very good moment for Delvitech. Important customers are understanding our technologies. Important customers are understanding our path of growth, and now, I would say we are shaping the organization to fulfill orders, to fulfill support, and to fulfill whatever is requested from the market.

LaRont: That is tremendous. Congratulations to you and to Delvitech for all of that tremendous success. So, you mentioned that your customers are starting to really understand the technology, which is so important. It's always the first part of this when you're doing something disruptive. You also mentioned “native AI.” So, you know the term AI is thrown around a lot these days, and I remember this from the last time we spoke, there's a lot of misunderstanding around it. So, can you explain “native AI” and what that means specifically, and for Delvitech, because I know we're talking about proprietary stuff here?

Gatti: The word “native AI” is very important. Now, if we go back four years ago in the market when we decided to develop a system that is native AI, the word was easy because, as you say, the word AI is on the mouth of everybody, but the problem is how to make an AI native system.

When we wanted to enter the market with this technology, we immediately understood that if you talk about AI, typically the first thing that comes to you is amount of data. Normally, if you go with the classic 3D system, the amount of data is significantly low. If you go on an AI native solution, like the one we are having right now, we normally manage between 96 96-gigabit of information in approximately 0.3 seconds, up to 200 gigabytes of information in the same time. So, how can we match this data?

For this reason, we have done a huge investment in technology because everything that resides in the platform is totally customized by us and totally designed by us. We have cameras dedicated to be so fast. We have a projectors with exceptional lighting and exceptional high speed. We have a communication system between the optical head, the GPUs, the PC, and so on, that are customized on purpose to get the speed. Out of this, you get a system that is capable to run on an AI native platform.

The point that comes immediately after is the software, which is our richest part. We have developed our neural network, and the neural network we are using is not a neural network that you find in the market. It is a dedicated neural network that we grew up, acquiring more than 12 million data sets to enrich it—these neural nets. And right now, we can finally show the market what it means to have an AI native machine.

So, in simple words, it means capability to detect whatever you want, capability to learn whatever you want, capability of flexibility for the customer. The customer can analyze a component, learn a component, and deploy the inspection of the component without calling the supplier, in this case, Delvitech. A lot of our competitors are complaining that AI technology is not possible on an optical inspection system, and are complaining about the fact that Delvitech does not have real AI. No, this is not true.

You can visibly see that if you come and see us at the booth or if you come and see us in our different branches, because you come with boards, you come with the data, and we show you what it means to use AI. So, the whole technology offers the possibility for the customer to analyze on a board, everything that is electronic, everything that is mechanical, everything that could be plastic, and other materials. We are capable of analyzing glass, to analyze scratches on glass, for example. We are capable of analyzing glue. We are capable of detecting the thickness of glue. We are capable of analyzing metallic parts and any type of components you can have on a board.

To me, AI native means flexibility, the capability of full detection. And it is what our customers are understanding. Because when you get a salesman or when you get an application engineer explaining the Delvitech machines, it looks like a dream. So, most of the customers, they say, “Ah, it cannot be true. We do not believe this. We want to touch it. We would like to get a machine as a demo in house. We want to run a benchmark.” True. We have run a lot of benchmarks during the last year. We won most of them. In some cases, we are still running these benchmarks, and I have to be honest, there is not a single customer who, after having seen the machine, is complaining about the technology.

They are all saying, “Okay, now we really understand, and now we really appreciate it. Now, we are really capable to understand what it means to have a platform and not a machine. A platform is capable of running from SPI to pre-reflow to post-reflow and to other areas, as a single system. A machine (by contrast) is something that, inside the production environment, you (would) need to buy for every single phase. So, this is our strength.

LaRont: That's amazing. It's also beneficial to invite customers to come in, get in touch, and experience the technology firsthand by running their boards. Because I know this industry and I know the customers, and they really do have to get comfortable, which is the education piece of this, which it sounds like you've been doing so well.

So, originally when we spoke, you had your first machine (Aton). Not too long after that, you had your first Horus machine. Let's talk about that tech development and where you are with Horus today and what that does, what the other machines don't do, and the edge over your competition.

Gatti: We are developing our platform. As you correctly say, we have the Aton. We have the Horus. The Horus has got some new features like the internal tilt unit, for example. It means you can get the board, you can inspect the board, you can tilt the board inside the machine, and you can run inspection of the other side.

We have the dual lane conveyor for high speed lines. We have a new high performance optical head that can be used for very fast lines. This is a unique technology. It’s a technology that is patented and it basically decreases the acquisition time for roughly 190 gigabit of information from 240 milliseconds down to 60 milliseconds. So, it's something amazing in terms of speed, and this technology really needs to be seen by our customers.

On top of this, we have launched—on the market right now—the version that is called HD, where the Horus with a particular modification to the optical head is reaching a resolution of three microns. This is the perfect system to be used for microelectronics, for, let's say the die bonding, ball bonding, wire bonding inspections.

Of course, Horus and Aon are enhanced every day. We keep on adding new features, self-programming features, and special features to handle with artificial intelligence. The library is a special feature to handle multiple plants connected to a single server where on this single server, you get the neural network that can be used in all the manufacturing plants and the special features, all in the direction of speed, in the direction of battery error classification, and in the direction of new technology to give the possibility to the customer to really program machines in just a few minutes. That is the point where we're willing to go.

Down the road with these systems, we will be capable of implementing technologies that will become more predictive. In fact, in less than one month, we will install a very particular line where there will be three Horus machines, one for post-solder paste, one for pre-reflow, and one for post-reflow. At this particular location, we will test the predictivity software, which is a patented feature from Delvitech. And after this test period, we will make a white paper that we'll make public to the whole world. And that would be—if we get the correct numbers—a solution that can really touch the pockets of our customers because it can really save (reduce) errors, so it can save a lot of money inside the production environment.

LaRont: We look forward to hearing about that and the release of that because of course, it's all about ROI for the customers, always. Roberto, do you think the way that this industry approaches test and inspection—do you think that we've limited ourselves? Because I feel like you're a technologist and that you have a very big vision for the technologies that you're working with, and it almost feels like maybe we're just scratching the surface a little bit. We're doing it this way right now, but there's lots of places for us to go that we may not even realize yet and that we're probably going to need with the way technology is going.

Gatti: Yes, this is totally true. Let’s say, you can see from what happened two weeks ago that Delvitech raised more than $40 million in our increase of capital, and we had to limit the amount of money injected in the company because we were oversubscribed. And the reason is that these investors really understood that what we are doing right now is almost nothing compared to what can be done.

These technologies are scratching the surface of the problem. But down the road, this technology will be, in my opinion, the technology to follow independent of it is Delvitech or other suppliers. But this AI applied to machine vision is for electronics, but it’s not only for electronics. It is the technology to be used to make life simpler to the different customers (markets), but especially to get the best quality with the lowest cost of manufacturing. And this can only be done if we use AI totally integrated. This is what it is, absolutely.

LaRont: Roberto Gatti, it's always a pleasure, always so interesting talking to you. I feel like I get a good education when I do. Congratulations on $40 million and it's great to see you here at productronica. Thanks for spending the time.

Gatti: Thank you.

LaRont: This is Marcy LaRont with Real Time with… productronica. Thanks for tuning in.