speaker-0: hi, Michael. Nice to meet you Maybe we can start with the first set of questions. So actually, for our listeners, tell us please a little bit more about our region history and story of how our HETONICS was set up. And maybe separate interesting point is association with universities. if you have it or if you had it.
speaker-1: Yeah. So yeah, great to meet you again, and thank you very much for having me. And the story of Iketonics actually begins around the COVID time, where we all had quite a bit of time on our hands. And at the time I was really into chip security and making chips more secure against any sort of side channel attacks or real hardware attacks instead of just trying to exploit software. was trying to exploit hardware. And I was looking into, how can you make a very secure chip? And this led me down this huge rabbit hole of, okay, a secure chip is fairly slow. So how do you make a fast chip that can be secure? And slowly got to learn about optical computing. And optical computing at that time, there was already some attempts to it, but was still very new and was still academic driven. And that's how I met my co-founder, Leonardo, who's been working on optical computing for years before. Even I thought of moving into that direction. And right at the end of COVID, we decided, okay, let's meet up. Let's have a chat in a German beer garden as you have it here in Germany. And we decided, hey, optical computing, the time is right that we can make an optical computer. And so we decided to go for it and we created Akatronics in 2021.
speaker-0: good to know. And maybe you can tell us a little bit more about your team, how you formed the team during this already four plus years.
speaker-1: since we are, I like to say a greenfield startup, so we are not a spinoff from university. We created everything from scratch. We also, of course, meant that we had to hire a fully new team. So we couldn't just take our university teams like some spinoffs, right? So we decided, okay, let's look into the academic field. Who are the top talents in Optical Computing? We reached out to everyone, had a chat with everyone. And then decided to hire our initial group of team members who designed our optical chips, who created the hardware setups, who created the lab setups and got everything done. And so, our initial journey was really like a regular startup trying to find great people. And we had a great initial bunch who are still with us until today. So we're very happy about that.
speaker-0: Great. maybe you can also explain, because I know that you have now like half-half software engineers and half physicists or something like this. So maybe you can explain how you came to this split and at which point you came actually to the split of half-half of the team.
speaker-1: Yeah, in our company, we like to split the technology stack as a top down and a bottom up approach. So for example, since I come from a computer science from a, I do have a physics background, but I do come more from the chip design and computer science part. So I do the top down approach and Leo, comes from the experimental physics from the nonlinear optics side. So he does the top up approach and that really showed us two things. You need both approaches and you have to think. like a business if you want to create an all-optic computer. Because even if you have some great technology, if the user can't use it, you have nothing that you can actually sell and make available to your customers. So we looked at, what is actually Nvidia's biggest selling point? Is it the hardware or is it actually the software stack on top of it? And we realized early on, it's a mixture of both, course, but... The lock-in into CUDA is a huge selling point for Nvidia hardware because there are companies out there that make better hardware than Nvidia. But if you're locked into the CUDA ecosystem, if you're reliant on all those SDKs, et cetera, that come with the CUDA ecosystem, it's very hard for you to switch up. So we decide, okay, we need the best driver. We need the best compiler. We need the best software development kit. So our version of CUDA, so to speak, in order for us to really bring our product to market. And that's why our company split half software developers and half hardware developers.
speaker-0: Yeah, quite clear how you formed the team. Thank you. And maybe because we don't have all of our listeners who are very deep in optical computing topic, it would be great to know why do in general we need an all-optical general purpose computer. So why we need it in terms of like speed, energy, efficiency, or any other points we can mention here.
speaker-1: Yeah, and the answer is actually quite simple almost. If you look at the electronic world, right, you have processors that got faster until 2006 approximately. Around 2006, we started, instead of making processes faster, we started parallelizing them, adding more cores instead of more speed, more performance, right? And adding more parallelization can only help you so much. There's something called Amdahl's law. If you look at it, It's basically diminishing returns, the more parallelization you have. And we've already reached that long times ago, this kind of diminishing return. You might not see it in the AI world at the moment, mainly because we still have a huge batch sizes that we can deal with. But if you look at edge AI, right, even your most complex, your biggest cores, can't handle. edge AI in a valuable speed for you to do, for example, real-time computer vision or anything like that. The problem here comes down to the sequential speed of your process. And if you want to increase that electronics comes to a limit because in electronics, right, even if you clock at a few gigahertz, your processor, your actual switching speed on chip is a few hundred gigahertz or about a hundred to 200 gigahertz is how fast your transistors can switch. And that means your Let's say the electromagnetic wave that travels through your system is about a millimeter to almost a centimeter in size. That is fine. If you want to clock faster, if you want to go into the terahertz domain, suddenly your electromagnetic wave that travels through your processor becomes smaller. And now you have to guide your wave. You have to guide the electromagnetic wave on your chip. And that's where you start coming into this photonic area, right? Where you have to think about, okay, I have to think about how to compute, but I also have to think about how to guide my electromagnetic wave through the processor. What I actually want to say with all of this is that if you want to do terahertz computing, naturally your computer will be an optical computer. It is just by the nature of the science of the electromagnetic wave traveling through your computer. so, yeah, we have to have an optical computer to go into the terahertz domain. And we have certain opportunities that come with it, like the energy efficiency that you mentioned, right? optical waveguides are much more efficient than electronic wires and that allows us to make more energy efficient and a faster computer.
speaker-0: Maybe also interesting question because you know, like in this quantum optical photonics area for us as venture capitalists, it's very important to meet some timeline because we have usually seven, 10, 12 years old funds. And we need to understand how we could return the money during this timeline. And this is why the next question, where do you see first applications and which applications are the biggest in terms of market size?
speaker-1: Yeah, the cool thing about photonic computing that some people see as a downside, we see as a huge advantage is that it is so-called time of flight, meaning your optical wave traveling your computer, can't really stop it. You can't really stop light. So you always have to have it perfectly synchronized. Everything in your process has to be perfectly synchronized. This might seem like a huge, huge pain to deal with. But it brings an opportunity that everything in your processor, you know exactly how long everything takes. And this weird quirk is actually something that allows you to make a processor that is extremely fast and real time. So you exactly know, or you have a very deterministic way of calculating how long a certain operation takes, how long a certain program will execute. And this deterministic character is something you critically need for anything safety related. We all know airplanes, right? They need to run on deterministic hardware. Otherwise you have the problem of, okay, my calculation of the angle my flaps have to be can either take a millisecond or it might take five seconds. I don't know. Okay. My plane crashes. That's why you need this deterministic compute in your airplane. And when you think about optical computing, you have that by default. And so for us, interesting first market applications are safety critical environments. where you need that. There are other things that contribute that in the optical domain, right? Because we are less prone to electronic interference. We are less prone to electronic warfare and all these problems that you might find in a more constrained or safety critical environment. Our process ideal for these edge safety critical appliances.
speaker-0: Got it. Maybe here one more additional question. What sets are HETONICs apart from other optical computing companies? If we compare more general purpose and accelerators here.
speaker-1: When people hear of optical computing, immediately either think of optical quantum computing or optical analog computing. Both have a huge problem when it comes to noise. If you have an analog computer, you're very prone to noise because you're calculating numbers in a range of zero to some number, and you need accurate values in between. And if you have some noise, that value changes. So it's not something you want to run your bank account on, for example, right? If your bank account fluctuates just by looking at it, that's a problem. And you have, of course, have a similar problem in the quantum domain where you need extremely low losses to not lose your coherent state. And so these types of optical computing are very prone to these problems. That's why we at Akatronics, we are at the moment, the only company that works on optical digital computing, meaning instead of having like a continuous value or very prone to noise value, we have just a zero and one. So it's light on or light off. And that's what we compute on. And we don't care if the light on state, so to speak, is a one or if it's a 0.9, we just care that the light is on or the light is off. And that allows us to recreate digital architectures. So yes, we're creating an optical digital computer. and digital architecture. So we can actually create a CPU, we can actually create a GPU. And that's actually what we're doing. We're creating a general purpose computer. So a CPU, and that allows us to do any type of compute, which is also different to the analog domain, where you're mostly focused on just arithmetics like addition and multiplication. Ours is a general purpose computer.
speaker-0: Yeah, and this actually, as far as I understand, this helps us to understand how you are able to overcome the issues that have stalled other companies with their photonic integrated circuits approach. Am I right? Or something else we can add here?
speaker-1: Exactly. That's exactly right. And a lot of photonic integrated circuit companies that are working on the analog computing side, at some point they face this noise issue or the accuracy issue. And for us, we never have this problem because yeah, if our on state, so to speak, is a bit noisy, we can clean that up. That's not an issue, something you can't do in analog space. And so for us, iterating on this technology is much simpler. And for example, we are now on our 25th tape out, which is also insane to think about, right? Whereas a normal electronic startups did one tape out and then it's go or it's bus for them. We are now 25th and we can iterate very fast simply because our system has a very, very simple protocol in that sense.
speaker-0: Okay, got you. maybe the following question about your milestones is that you already reached by the end of 2025, maybe some K milestones you reached. In your business, in technology, product, IP, customers,
speaker-1: Yeah, believe our biggest milestone is that we've actually completed our first customer project. And that project actually involves building a processor for the avionics domain. Of course, it's not our full CPU just yet. It's a slice of our main technology, meaning it has some optical digital logic. It has some optical memory, also something very important to us. And that allows us to showcase, yes, our technology, if we add more of these devices, we can build any type of circuitry. And this milestone also showcased that yes, we can also fulfill customer projects at this early stage, which is also quite insane to think about that we creating a new computing paradigm, but we already working with customers to provide them value. And we are going to enhance our speed in the coming year where we are going to build our first optical computer, a full scale computer with enough memory, enough compute power to actually run any type of use case And one showcase we are very excited to show is of course running Doom on our computer.
speaker-0: Yeah, this is already big milestones. We are happy as an investor that you know you're progressing so actively and for us for sure it's very important that you have also customer traction and customers pipeline. So thank you very much for this news for us and for you and for the market. Regarding maybe people are comparing what hyperscales are doing. And in this regard, the following questions. So hyperscales as AWS, Google, Microsoft could build all optical processors. Why haven't they? And what is their incentive to use a ketone X instead of building internally?
speaker-1: Yeah, that's a very good question because our idea of optical computing is actually very old. The first ideas of this type of digital optical computer go back all the way to von Neumann himself, who holds the first patent in this space. So we're talking almost 50 years or 60 years even of history in this regard. And we aren't even the first to build an all optical computer that is digital. It was actually Bell Labs who created the first version of it. Sure, there were... electro-optical, but we are now creating all optical version. The reason why the hyperscalers are afraid of building an all optical computer is, I would say twofold. On the one hand, of course, they have this, let's call it the sunk cost fallacy, right? They already are invested so deeply in the new nodes, right? With the two nanometer, three nanometer nodes and If you want to switch a hardware stack to a completely different technology that is reliant on a completely different way of how you make the chips, then you have a problem, right? Because you need to establish new nodes in an older technology. Yeah, they won't want to do that. And on the other hand, they of course have the problem that if you look at it from just a broad picture, if you just think about an Nvidia who has their new tends to hundreds billion transistor chips and they want to transport that architecture to their, to an all optical version, they will see, wait, optics is much larger. We can actually fit a hundred billion transistors on our chip. How would we do that? Would it be a chip that is the size of two football fields or how would we even manage that? And there lies a huge misconception of how an optical computer works. because we don't need billions of transistors, we just need a few thousand of transistors to make them equivalent processor to an H200, for example. And this insight is something that is lacking in these larger companies. yes, of course, Nvidia, if they decided to go all in on all of the computer, they have, of course, already half of the stack ready, right? They already have their software stack ready, but the hardware is a new paradigm. It's a new way of thinking that is lacking in a company like Nvidia.
speaker-0: Yeah, this is, know, like, because market likes to compare actually where such products as Ahetonex would be potentially sold and maybe also used and being some kind of collaboration services is big hyperscale. So it seems that your competitive advantage relies partly on European tech sovereignty. Does a shift in geopolitics reduce that advantage?
speaker-1: So for us, the main advantage is of course the technical aspects, right? The performance, the energy efficiency, the deterministic part, but you're perfectly right. At the moment, our biggest advantage is European serenity because everything we do is done here in Europe. So our chips are designed here in Europe. The software is written here in Europe. And most importantly, our chips are also made here in Europe. They're packaged here in Europe. And that is something very unique. We are actually currently the only company that can make a full computer, a high-performance computer entirely in Europe. That is not something anyone else can claim at this point, which is very, very special to us. And that means yes, at the current geopolitical situation, right? We have a very, call it far right government in the U S we have tensions with China, with tensions with Russia. We have the main hubs where you currently produce chips are currently problematic, so to speak, Taiwanese chips can become more troublesome if we think about, for example, defense products. That's why our current market is actually very defense focused, that we have chips that we can guarantee a certain level of security in. And that is something that is absolutely unique throughout Europe.
speaker-0: a close question also to competitive advantage regarding the fact that you use European manufacturing on major nodes. And how do you turn this into strengths compared to cutting edge advanced nodes?
speaker-1: Yeah, and to me personally, that is actually our biggest strength that we can make this type of process in all of these legacy nodes. use instead of like these three nanometer nodes, right? We use 180 nanometers, just as an example. And when you think about 180 nanometer nodes, a lot of things become simpler. And I'm not just talking about, okay, we have bigger lithography, meaning we don't need expensive machinery, right? We don't need. who spend billions of dollars for a single lithography machine. But the really cool thing is you actually save on water, you save on power on making these chips, right? You save on the key economical issues that are facing, for example, Taiwanese chip production. You have droughts in Taiwan because of chip production. You have material shortages because of chip production. With our technology stack, you don't have that. I like to say that in the future, I can imagine a world where anyone can make chips in their backyard, high-performance computers in their backyard, like we used to do in the 60s and 70s, right? Where you could still make your own little processors at home.
speaker-0: The next question maybe to finalize this part about the biggest risk I ketonics faces technical market regulatory or something else at this stage.
speaker-1: Yeah, so we have one big technological challenge, of course, that revolves around optical memory. Why we, of course, have optical memory, right? We have all different types of optical memory. It's the quantity that can really hamper what you can do with the technology. So for example, our current design for our next processor, we are talking about a processor that has around the range sub one gigabyte of memory. It still has a lot of memory, right? You can still do a lot of things. But when we're talking about volatile memory, we're not there yet for electronics. And that means if you want to run, for example, AI training on your processor, that will not yet work. If you want to run AI inference, of course, that is not a problem because read-only memory is not an issue. It's just this volatile memory that is an issue. And this can of course mean that when we want to enter the market in a certain field, like I mentioned AI training, we will face a lot of new challenges to go with. Another big challenge is of course, when we're talking about regulatory issues, right? For example, if you want to certify our processor for avionics use cases, or if you want to certify it for security related use cases, the certifying bodies, they have never heard of an optical computer. So they will look at our technology stack and they will first need to understand what we're actually doing before they can certify it. Meaning our certification. journey is going to be very interesting to say the least, and it will take some time for us to actually have the stamp of approval for us to be used in a safety critical environment. It's something we have to do. Sadly, we of course would have wished that someone else would do for us, but once that is done, sky's the limit for our technology.
speaker-0: And maybe you can mention now a little bit more about your milestones for next one to years, for sure, which you could announce some of them according to your roadmap. So which milestones do you have for next period of time?
speaker-1: We actually have three big milestones that we are now focused on. The first one is that we want to open source our software stack. This might seem strange, right? We are building this high tech new paradigm and we want to open source it, but it's critical for us that we open source our instruction set architecture, our driver, our compiler and our software development kit so that people can actually start using the device without needing a physical version of it. And that will happen sometime Q3 of next year. Then Q4, so end of next year in pretty much 12 months, we will build our first all optical high-performance computer. And I really mean fully all optical in the sense of that you have one full rack of compute power that allows you to do any sort of tasks from running Doom to running your newest AI models all optically. And that really showcases, yes, this technology is ready for the prime time. We can take that rack. We can sell it to... Amazon to a Microsoft, they can put it in the data center if they wanted. But that is not our end goal. Our end goal is miniaturizing the technology, which is our third milestone for 2027, where we want to take that entire rack unit, shrink it down, that you can actually put it in edge device. And that is our final milestone before we go to the big market and sell our product at scale.
speaker-0: Yeah, so quite, you know, a huge roadmap and milestones and we will be happy when you will reach them. So imagine five years from now, what does the road look like if Ahetronics is successful?
speaker-1: So of course, our hope in five years is that many people use optical computers. And we envision a roadmap similar to the electronic world, right? Where we used to have computers that were based on relays and mechanical devices. And then we face those out in favor of the transistor. And we want to replay that history that we will slowly face out the electronic transistor for the optical. right, moving from the gigahertz domain to the terahertz domain. That is what we are working towards. And in five years, of course, it will take time for us to scale that up. But we believe that in five years, the first data centers will turn all optical using Akatronics devices. This will be isolated data centers, but it's the first step for us going this direction. And then in 10 years, hopefully the entire cloud computing environment is already optical. And that means that we are in every data center and are now focused in 10 years on moving towards the edge devices on moving towards the computers at home. And at some point, hopefully everything will be an optical computer.
speaker-0: Good. We also hope that in five, I don't know, seven years we will see it. And you know, because they are quite deep investing in quantum space, they are very interested about the bet on quantum photonics integration. And the question is the following. Is quantum a core part of your long-term vision or not?
speaker-1: Yes, quantum is actually a core part of our vision. We're actually involved in a few quantum projects ourselves. Even though our computer is not quantum in itself, we like to joke that our computer is two-thirds quantum, right? We use principles like interference. We use principles like coherence. What we don't do is of course entanglement. So yeah, we use two of the key three principles that make a quantum computer. But, and the reason I stress this is... It means that our computer, our optic computer is naturally compatible with a quantum computer. And the big problem that the quantum computing world has, or at least in certain areas of quantum computing is something called feed forward, enabling very deep quantum circuits. And a big problem here is that electronics is too slow or too indeterministic for you to be able to do this. Let's say daisy chaining of quantum gates. And this is where optical computing comes in and enables that new technology. So for us, the quantum computer and classical, like our optical computer together would make the next paradigm in computing. And we are mainly focused on this classical world. Let's say the janitor of the future quantum computer.
speaker-0: Yeah, thank you. Thank you very much, Michael.
speaker-1: Yeah. So yeah, great to meet you again, and thank you very much for having me. And the story of Iketonics actually begins around the COVID time, where we all had quite a bit of time on our hands. And at the time I was really into chip security and making chips more secure against any sort of side channel attacks or real hardware attacks instead of just trying to exploit software. was trying to exploit hardware. And I was looking into, how can you make a very secure chip? And this led me down this huge rabbit hole of, okay, a secure chip is fairly slow. So how do you make a fast chip that can be secure? And slowly got to learn about optical computing. And optical computing at that time, there was already some attempts to it, but was still very new and was still academic driven. And that's how I met my co-founder, Leonardo, who's been working on optical computing for years before. Even I thought of moving into that direction. And right at the end of COVID, we decided, okay, let's meet up. Let's have a chat in a German beer garden as you have it here in Germany. And we decided, hey, optical computing, the time is right that we can make an optical computer. And so we decided to go for it and we created Akatronics in 2021.
speaker-0: good to know. And maybe you can tell us a little bit more about your team, how you formed the team during this already four plus years.
speaker-1: since we are, I like to say a greenfield startup, so we are not a spinoff from university. We created everything from scratch. We also, of course, meant that we had to hire a fully new team. So we couldn't just take our university teams like some spinoffs, right? So we decided, okay, let's look into the academic field. Who are the top talents in Optical Computing? We reached out to everyone, had a chat with everyone. And then decided to hire our initial group of team members who designed our optical chips, who created the hardware setups, who created the lab setups and got everything done. And so, our initial journey was really like a regular startup trying to find great people. And we had a great initial bunch who are still with us until today. So we're very happy about that.
speaker-0: Great. maybe you can also explain, because I know that you have now like half-half software engineers and half physicists or something like this. So maybe you can explain how you came to this split and at which point you came actually to the split of half-half of the team.
speaker-1: Yeah, in our company, we like to split the technology stack as a top down and a bottom up approach. So for example, since I come from a computer science from a, I do have a physics background, but I do come more from the chip design and computer science part. So I do the top down approach and Leo, comes from the experimental physics from the nonlinear optics side. So he does the top up approach and that really showed us two things. You need both approaches and you have to think. like a business if you want to create an all-optic computer. Because even if you have some great technology, if the user can't use it, you have nothing that you can actually sell and make available to your customers. So we looked at, what is actually Nvidia's biggest selling point? Is it the hardware or is it actually the software stack on top of it? And we realized early on, it's a mixture of both, course, but... The lock-in into CUDA is a huge selling point for Nvidia hardware because there are companies out there that make better hardware than Nvidia. But if you're locked into the CUDA ecosystem, if you're reliant on all those SDKs, et cetera, that come with the CUDA ecosystem, it's very hard for you to switch up. So we decide, okay, we need the best driver. We need the best compiler. We need the best software development kit. So our version of CUDA, so to speak, in order for us to really bring our product to market. And that's why our company split half software developers and half hardware developers.
speaker-0: Yeah, quite clear how you formed the team. Thank you. And maybe because we don't have all of our listeners who are very deep in optical computing topic, it would be great to know why do in general we need an all-optical general purpose computer. So why we need it in terms of like speed, energy, efficiency, or any other points we can mention here.
speaker-1: Yeah, and the answer is actually quite simple almost. If you look at the electronic world, right, you have processors that got faster until 2006 approximately. Around 2006, we started, instead of making processes faster, we started parallelizing them, adding more cores instead of more speed, more performance, right? And adding more parallelization can only help you so much. There's something called Amdahl's law. If you look at it, It's basically diminishing returns, the more parallelization you have. And we've already reached that long times ago, this kind of diminishing return. You might not see it in the AI world at the moment, mainly because we still have a huge batch sizes that we can deal with. But if you look at edge AI, right, even your most complex, your biggest cores, can't handle. edge AI in a valuable speed for you to do, for example, real-time computer vision or anything like that. The problem here comes down to the sequential speed of your process. And if you want to increase that electronics comes to a limit because in electronics, right, even if you clock at a few gigahertz, your processor, your actual switching speed on chip is a few hundred gigahertz or about a hundred to 200 gigahertz is how fast your transistors can switch. And that means your Let's say the electromagnetic wave that travels through your system is about a millimeter to almost a centimeter in size. That is fine. If you want to clock faster, if you want to go into the terahertz domain, suddenly your electromagnetic wave that travels through your processor becomes smaller. And now you have to guide your wave. You have to guide the electromagnetic wave on your chip. And that's where you start coming into this photonic area, right? Where you have to think about, okay, I have to think about how to compute, but I also have to think about how to guide my electromagnetic wave through the processor. What I actually want to say with all of this is that if you want to do terahertz computing, naturally your computer will be an optical computer. It is just by the nature of the science of the electromagnetic wave traveling through your computer. so, yeah, we have to have an optical computer to go into the terahertz domain. And we have certain opportunities that come with it, like the energy efficiency that you mentioned, right? optical waveguides are much more efficient than electronic wires and that allows us to make more energy efficient and a faster computer.
speaker-0: Maybe also interesting question because you know, like in this quantum optical photonics area for us as venture capitalists, it's very important to meet some timeline because we have usually seven, 10, 12 years old funds. And we need to understand how we could return the money during this timeline. And this is why the next question, where do you see first applications and which applications are the biggest in terms of market size?
speaker-1: Yeah, the cool thing about photonic computing that some people see as a downside, we see as a huge advantage is that it is so-called time of flight, meaning your optical wave traveling your computer, can't really stop it. You can't really stop light. So you always have to have it perfectly synchronized. Everything in your process has to be perfectly synchronized. This might seem like a huge, huge pain to deal with. But it brings an opportunity that everything in your processor, you know exactly how long everything takes. And this weird quirk is actually something that allows you to make a processor that is extremely fast and real time. So you exactly know, or you have a very deterministic way of calculating how long a certain operation takes, how long a certain program will execute. And this deterministic character is something you critically need for anything safety related. We all know airplanes, right? They need to run on deterministic hardware. Otherwise you have the problem of, okay, my calculation of the angle my flaps have to be can either take a millisecond or it might take five seconds. I don't know. Okay. My plane crashes. That's why you need this deterministic compute in your airplane. And when you think about optical computing, you have that by default. And so for us, interesting first market applications are safety critical environments. where you need that. There are other things that contribute that in the optical domain, right? Because we are less prone to electronic interference. We are less prone to electronic warfare and all these problems that you might find in a more constrained or safety critical environment. Our process ideal for these edge safety critical appliances.
speaker-0: Got it. Maybe here one more additional question. What sets are HETONICs apart from other optical computing companies? If we compare more general purpose and accelerators here.
speaker-1: When people hear of optical computing, immediately either think of optical quantum computing or optical analog computing. Both have a huge problem when it comes to noise. If you have an analog computer, you're very prone to noise because you're calculating numbers in a range of zero to some number, and you need accurate values in between. And if you have some noise, that value changes. So it's not something you want to run your bank account on, for example, right? If your bank account fluctuates just by looking at it, that's a problem. And you have, of course, have a similar problem in the quantum domain where you need extremely low losses to not lose your coherent state. And so these types of optical computing are very prone to these problems. That's why we at Akatronics, we are at the moment, the only company that works on optical digital computing, meaning instead of having like a continuous value or very prone to noise value, we have just a zero and one. So it's light on or light off. And that's what we compute on. And we don't care if the light on state, so to speak, is a one or if it's a 0.9, we just care that the light is on or the light is off. And that allows us to recreate digital architectures. So yes, we're creating an optical digital computer. and digital architecture. So we can actually create a CPU, we can actually create a GPU. And that's actually what we're doing. We're creating a general purpose computer. So a CPU, and that allows us to do any type of compute, which is also different to the analog domain, where you're mostly focused on just arithmetics like addition and multiplication. Ours is a general purpose computer.
speaker-0: Yeah, and this actually, as far as I understand, this helps us to understand how you are able to overcome the issues that have stalled other companies with their photonic integrated circuits approach. Am I right? Or something else we can add here?
speaker-1: Exactly. That's exactly right. And a lot of photonic integrated circuit companies that are working on the analog computing side, at some point they face this noise issue or the accuracy issue. And for us, we never have this problem because yeah, if our on state, so to speak, is a bit noisy, we can clean that up. That's not an issue, something you can't do in analog space. And so for us, iterating on this technology is much simpler. And for example, we are now on our 25th tape out, which is also insane to think about, right? Whereas a normal electronic startups did one tape out and then it's go or it's bus for them. We are now 25th and we can iterate very fast simply because our system has a very, very simple protocol in that sense.
speaker-0: Okay, got you. maybe the following question about your milestones is that you already reached by the end of 2025, maybe some K milestones you reached. In your business, in technology, product, IP, customers,
speaker-1: Yeah, believe our biggest milestone is that we've actually completed our first customer project. And that project actually involves building a processor for the avionics domain. Of course, it's not our full CPU just yet. It's a slice of our main technology, meaning it has some optical digital logic. It has some optical memory, also something very important to us. And that allows us to showcase, yes, our technology, if we add more of these devices, we can build any type of circuitry. And this milestone also showcased that yes, we can also fulfill customer projects at this early stage, which is also quite insane to think about that we creating a new computing paradigm, but we already working with customers to provide them value. And we are going to enhance our speed in the coming year where we are going to build our first optical computer, a full scale computer with enough memory, enough compute power to actually run any type of use case And one showcase we are very excited to show is of course running Doom on our computer.
speaker-0: Yeah, this is already big milestones. We are happy as an investor that you know you're progressing so actively and for us for sure it's very important that you have also customer traction and customers pipeline. So thank you very much for this news for us and for you and for the market. Regarding maybe people are comparing what hyperscales are doing. And in this regard, the following questions. So hyperscales as AWS, Google, Microsoft could build all optical processors. Why haven't they? And what is their incentive to use a ketone X instead of building internally?
speaker-1: Yeah, that's a very good question because our idea of optical computing is actually very old. The first ideas of this type of digital optical computer go back all the way to von Neumann himself, who holds the first patent in this space. So we're talking almost 50 years or 60 years even of history in this regard. And we aren't even the first to build an all optical computer that is digital. It was actually Bell Labs who created the first version of it. Sure, there were... electro-optical, but we are now creating all optical version. The reason why the hyperscalers are afraid of building an all optical computer is, I would say twofold. On the one hand, of course, they have this, let's call it the sunk cost fallacy, right? They already are invested so deeply in the new nodes, right? With the two nanometer, three nanometer nodes and If you want to switch a hardware stack to a completely different technology that is reliant on a completely different way of how you make the chips, then you have a problem, right? Because you need to establish new nodes in an older technology. Yeah, they won't want to do that. And on the other hand, they of course have the problem that if you look at it from just a broad picture, if you just think about an Nvidia who has their new tends to hundreds billion transistor chips and they want to transport that architecture to their, to an all optical version, they will see, wait, optics is much larger. We can actually fit a hundred billion transistors on our chip. How would we do that? Would it be a chip that is the size of two football fields or how would we even manage that? And there lies a huge misconception of how an optical computer works. because we don't need billions of transistors, we just need a few thousand of transistors to make them equivalent processor to an H200, for example. And this insight is something that is lacking in these larger companies. yes, of course, Nvidia, if they decided to go all in on all of the computer, they have, of course, already half of the stack ready, right? They already have their software stack ready, but the hardware is a new paradigm. It's a new way of thinking that is lacking in a company like Nvidia.
speaker-0: Yeah, this is, know, like, because market likes to compare actually where such products as Ahetonex would be potentially sold and maybe also used and being some kind of collaboration services is big hyperscale. So it seems that your competitive advantage relies partly on European tech sovereignty. Does a shift in geopolitics reduce that advantage?
speaker-1: So for us, the main advantage is of course the technical aspects, right? The performance, the energy efficiency, the deterministic part, but you're perfectly right. At the moment, our biggest advantage is European serenity because everything we do is done here in Europe. So our chips are designed here in Europe. The software is written here in Europe. And most importantly, our chips are also made here in Europe. They're packaged here in Europe. And that is something very unique. We are actually currently the only company that can make a full computer, a high-performance computer entirely in Europe. That is not something anyone else can claim at this point, which is very, very special to us. And that means yes, at the current geopolitical situation, right? We have a very, call it far right government in the U S we have tensions with China, with tensions with Russia. We have the main hubs where you currently produce chips are currently problematic, so to speak, Taiwanese chips can become more troublesome if we think about, for example, defense products. That's why our current market is actually very defense focused, that we have chips that we can guarantee a certain level of security in. And that is something that is absolutely unique throughout Europe.
speaker-0: a close question also to competitive advantage regarding the fact that you use European manufacturing on major nodes. And how do you turn this into strengths compared to cutting edge advanced nodes?
speaker-1: Yeah, and to me personally, that is actually our biggest strength that we can make this type of process in all of these legacy nodes. use instead of like these three nanometer nodes, right? We use 180 nanometers, just as an example. And when you think about 180 nanometer nodes, a lot of things become simpler. And I'm not just talking about, okay, we have bigger lithography, meaning we don't need expensive machinery, right? We don't need. who spend billions of dollars for a single lithography machine. But the really cool thing is you actually save on water, you save on power on making these chips, right? You save on the key economical issues that are facing, for example, Taiwanese chip production. You have droughts in Taiwan because of chip production. You have material shortages because of chip production. With our technology stack, you don't have that. I like to say that in the future, I can imagine a world where anyone can make chips in their backyard, high-performance computers in their backyard, like we used to do in the 60s and 70s, right? Where you could still make your own little processors at home.
speaker-0: The next question maybe to finalize this part about the biggest risk I ketonics faces technical market regulatory or something else at this stage.
speaker-1: Yeah, so we have one big technological challenge, of course, that revolves around optical memory. Why we, of course, have optical memory, right? We have all different types of optical memory. It's the quantity that can really hamper what you can do with the technology. So for example, our current design for our next processor, we are talking about a processor that has around the range sub one gigabyte of memory. It still has a lot of memory, right? You can still do a lot of things. But when we're talking about volatile memory, we're not there yet for electronics. And that means if you want to run, for example, AI training on your processor, that will not yet work. If you want to run AI inference, of course, that is not a problem because read-only memory is not an issue. It's just this volatile memory that is an issue. And this can of course mean that when we want to enter the market in a certain field, like I mentioned AI training, we will face a lot of new challenges to go with. Another big challenge is of course, when we're talking about regulatory issues, right? For example, if you want to certify our processor for avionics use cases, or if you want to certify it for security related use cases, the certifying bodies, they have never heard of an optical computer. So they will look at our technology stack and they will first need to understand what we're actually doing before they can certify it. Meaning our certification. journey is going to be very interesting to say the least, and it will take some time for us to actually have the stamp of approval for us to be used in a safety critical environment. It's something we have to do. Sadly, we of course would have wished that someone else would do for us, but once that is done, sky's the limit for our technology.
speaker-0: And maybe you can mention now a little bit more about your milestones for next one to years, for sure, which you could announce some of them according to your roadmap. So which milestones do you have for next period of time?
speaker-1: We actually have three big milestones that we are now focused on. The first one is that we want to open source our software stack. This might seem strange, right? We are building this high tech new paradigm and we want to open source it, but it's critical for us that we open source our instruction set architecture, our driver, our compiler and our software development kit so that people can actually start using the device without needing a physical version of it. And that will happen sometime Q3 of next year. Then Q4, so end of next year in pretty much 12 months, we will build our first all optical high-performance computer. And I really mean fully all optical in the sense of that you have one full rack of compute power that allows you to do any sort of tasks from running Doom to running your newest AI models all optically. And that really showcases, yes, this technology is ready for the prime time. We can take that rack. We can sell it to... Amazon to a Microsoft, they can put it in the data center if they wanted. But that is not our end goal. Our end goal is miniaturizing the technology, which is our third milestone for 2027, where we want to take that entire rack unit, shrink it down, that you can actually put it in edge device. And that is our final milestone before we go to the big market and sell our product at scale.
speaker-0: Yeah, so quite, you know, a huge roadmap and milestones and we will be happy when you will reach them. So imagine five years from now, what does the road look like if Ahetronics is successful?
speaker-1: So of course, our hope in five years is that many people use optical computers. And we envision a roadmap similar to the electronic world, right? Where we used to have computers that were based on relays and mechanical devices. And then we face those out in favor of the transistor. And we want to replay that history that we will slowly face out the electronic transistor for the optical. right, moving from the gigahertz domain to the terahertz domain. That is what we are working towards. And in five years, of course, it will take time for us to scale that up. But we believe that in five years, the first data centers will turn all optical using Akatronics devices. This will be isolated data centers, but it's the first step for us going this direction. And then in 10 years, hopefully the entire cloud computing environment is already optical. And that means that we are in every data center and are now focused in 10 years on moving towards the edge devices on moving towards the computers at home. And at some point, hopefully everything will be an optical computer.
speaker-0: Good. We also hope that in five, I don't know, seven years we will see it. And you know, because they are quite deep investing in quantum space, they are very interested about the bet on quantum photonics integration. And the question is the following. Is quantum a core part of your long-term vision or not?
speaker-1: Yes, quantum is actually a core part of our vision. We're actually involved in a few quantum projects ourselves. Even though our computer is not quantum in itself, we like to joke that our computer is two-thirds quantum, right? We use principles like interference. We use principles like coherence. What we don't do is of course entanglement. So yeah, we use two of the key three principles that make a quantum computer. But, and the reason I stress this is... It means that our computer, our optic computer is naturally compatible with a quantum computer. And the big problem that the quantum computing world has, or at least in certain areas of quantum computing is something called feed forward, enabling very deep quantum circuits. And a big problem here is that electronics is too slow or too indeterministic for you to be able to do this. Let's say daisy chaining of quantum gates. And this is where optical computing comes in and enables that new technology. So for us, the quantum computer and classical, like our optical computer together would make the next paradigm in computing. And we are mainly focused on this classical world. Let's say the janitor of the future quantum computer.
speaker-0: Yeah, thank you. Thank you very much, Michael.
