Sean Hehir, BrainChip | theCUBE + NYSE Wired: Physical AI & Robotics Series

Palo Alto studio connection Silicon Valley and Wall Street. I'm John F co here with Dave Volante my co-host. Hello, I'm John Furry host of the cube. We are at the cub's NYC studios here in New York City. Of course, we have our Peloto studio connecting Silicon Valley to Wall Street. This is our physical AI robotics series. We talked to the leaders who are making it happen in physical AI and robotics which also includes a little bit of defense tech which we have a whole another series on. But this is part of the AI infrastructure that's booming and we expect to see more action and more growth in 2027 as AI factories come on board that's going to open up the edge that's going to open up robotics, drones, devices. Sean Harris here, CEO of Brainchip. Sean great to see you. Thanks for coming back. It's been a while been on the cube. Good to see you. Good to see you're running Brainchip as CEO. you guys doing great. Uh tell us a little bit about what you guys are doing and where the action is. >> Sure. I could talk a long time, but I'll be brief so we can ask some follow- on questions. So at its simplest level, most people know AI and think of the data center, John, right? These kind of large centralized thing. I know you and I are going to talk about the edge. What we do is enable the edge and everything always starts with silicon, right? So we provide chips. We also IP for those companies that want to build custom AS6 or SOC's. Of course, we have a tool chain to put models on there. comprehensive models, zoo, developers, etc. things like that. >> So, brain chip is there is it a chip for the brain or is that a metaphor because you know Elon loves the the brain chip? >> Explain the the name. >> Sure. Brain chip comes from the idea of being neuromorphic. If those who are not familiar with neuromorphic is it's really event or sparity. If you think about things on the edge, they're typically streaming data. They're typically sparse, not a lot of activity till something happens. So, it's ideal for that. What normorphic means, it's brain-like, brain inspired because your brain is the most efficient computational engine known to mankind. Right? If nothing is happening, your brain's not firing. And I like to use a very simple example, right? Most people we meet in in our daily lives have two eyes, nose, mouth, you know, hair color and things like that. The only thing we notice when we meet new people is what's different. The skin color, facial hair, and things like that. That's what norphic is. Nothing challenging. We don't compute. So on the opportunities you mentioned the edge obviously you're starting to see the formation of what AI factories are doing the big data centers gigawatts there's geothermal going on so energy is being worked on but still a constraint uh power and money is the big constraint in the AI infrastructure but when you have intelligence tokens that that they're doing it has value and everyone's talking about that that's kind of what's happening now right but you connect the dots to the edge those AI factories are just a node in the work. They're smaller. Maybe they're different footprints. >> Y >> that's going to be the next land grab for chips, >> the systems that are around those chips. Uh what's your vision on that? Because it, you know, every enterprise I talk to, the word basically comes out every time like we need a brain for our company. >> Mhm. >> Graph databases are popular, systems are being organized around the resources like a brain. >> Yep. >> The edge is the perfect scenario for that. Well, that's a really comprehensive question. I'll give you a kind of a little bit longer answer, right? So, you talk about data centers and power and all that. You know, most compute models start centralized and then they decentralize. And so, what what the industry is finally realizing is the right tool for the right job. So, you're going to have that data center to do some things, but you're also going to do some edge stuff. And the edge can take form itself as discreet. Maybe it's a standalone medical device, maybe it's something mobile in defense, or maybe it's a wearable device. And I can give some examples a little bit later. Or it can work in that federated way or orchestrated way with the data center. Now our technology we certainly we announced something not that long ago with IBM with Symphfony which is a product they have what goes out is kind of an orchestration layer that looks and uses collects up what we call a key which is our brand name for our chips and uses all that kind of data and brings it into a centralized. So the right compute happens on the edge the right the uh right compute happens in the data center. >> Who are your mention IBM so who are your customers? Who do you work with? What's what's their application? >> No, no. Our our technology is horizontal enabling across all industries. But I'll answer your question directly. The hot markets are ones that you could expect or markets. Mobility that matters. Change matters. Defense right now. If you think about what's going on with defense, it has changed from these kind of large monolithic fixed mobile position systems to mobile, right? Look what's going on in Ukraine. Look what's going on in the Middle East right now. So, you're going to do that with technologies that require mobility. And mobility requires long battery life. You require long battery life. You need good edge technology. The best edge technology is neuromorphic. So defense is really really strong with us right now. People examining cases they need to do. They get decisions quicker. Latency, fear of security, going back to the data center. Defense is ideal. I mentioned wearables a moment ago. You enabling use cases for things that you couldn't possibly do. You can't afford the latency. Maybe you're concerned about security. There's a lot of things. If you're putting a thing on your hand, your wrist, your face, you need long battery life. So those are the kind of hot industries. Of course, IoT and others, but the use cases are limitless. And I I'm going to make one more comment. When I became CEO of this company four and a half years ago, usually the conversations were like this really [snorts] interesting technology, breakthrough performance. Help me understand what I do with it. Now all day I hear, Sean, here's my use case. Here's my performance KPIs. I need you guys to do this. And it's coming from all industries. >> Battery life's huge. You mentioned that. So talk more about that because energy is the bounding function. Whatever conversation you want to have, but on the edge and mobility specifically, drones have distance limitations because of battery. >> That's right. >> As an example, what what is the prerequisite for really strong battery life in your system? >> Well, is it's really the underlying technology how we do it. Because we're neuromorphic or event based, we ignore everything. So basically more conventional computing I've been in this industry for a long time. Most conventional computing it's all around you know brute force matrix multiplication right the von Newman architecture just multiply multiply zeros and ones zeros and ones if it's a zero the way to look at it we do nothing. So just that fact alone if there's nothing happen that saves huge amounts of powers. Secondly we have our our memory right next to the compute elements. So it's called a near-term so you're not moving a lot. So very little energy there and of course the way we quantize models to do that. basically blacklist all the zeros basically you only compute on relevant >> events data that's called event driven or sparity or normorphic exactly right >> where's this going uh connect the dots because you know what's happening we know in AI is state which I I do have a couple questions on that but like it's booming >> we kind of know the impact what it means lower battery more power better better efficiency where does it go what's the next connect the dots >> well it's one of these things that I have the great privilege to lead a company where use cases come up every day. Phone calls come in say, "Sean, can you talk to this customer for me?" I'm like, "Sure." And I'm like, "Hm, I never would have thought of that. Never would have thought of that." And I'll give you a good example. One is a company we're working with out of Oman right now, which has wearable glasses for epileptic seizure prediction, not detection. And it has a little sensor up here on your frame of your glasses and says you're going to have a seizure. 98% accuracy in 1 hour. What that allows you to do is get off the road if you're driving or go seek medical attention, things like that. So, the use cases are limited. They I mean limitless. They can go anywhere. What this can do basically enabling any kind of product, any kind of technology to do more with less. >> These are breakthroughs. >> They're really breakthrough. >> I mean, when you think about that, I was talking to an entrepreneur. They have this, it's in R&D, but it's getting commercialized. Basically, the eye is a the eyes are a lens into the brain. >> That's right. So the neurological pathways from the eyes can predict and without AI they never would have been able to do anything. >> That's right. >> They were using old statistical samples. That's right. >> But they're doing real time eye tracking to look at certain things that AI can predict. >> Sensors on the glasses. What other things are you seeing out there that are kind of I won't say fall out of the chair, but like not obvious. >> Well, we have another lency some a company that licensed us out of Sweden called Front Grade Geyser. They make uh silicon for spaceships. Now you think of space that's the ultimate harsh environment, right? You you cannot rely on a network, right? Because it could go away and you certainly can't the latency could kill you. You need the ability to do the computation, do it quickly and do the inference and that's what's happening in space. So there a whole bunch of missions going on in space right now that are enabled by our kind of technologies. So things like that that just were not possible or conceivable now can be done. So basically the environment of what used to be a data center application could be space edge human wearable >> defense. >> So you're seeing that kind of capability come in. >> Yeah. This is it's basically the right tool for the right job, right? Depending on what the requirements are. The other thing that's driving it is people, you know, the whole world is more and more competitive every single day. So every company's out there looking for more feature functions in their end product, right? And if this enables a feature, function, something allow them to sell more things. It's incredible. >> Talk about the business that you're running because talk about the some of the momentum you have. Where is the hot areas? What's working right now? >> It's literally across the board. The industries I talked about right there. But one of the things that's really important about the edge is anything in AI, the models keep changing. They go quickly. Brainchip is mostly scientists and engineers. We hire from the best schools in the world and we're following very closely model trends and we're building the technology to accelerate that at all times. So we've got some really interesting offers that we have now and we got some incredible offers that are coming even later this year and early into next year. Things that seemed impossible LLMs or SLMs on the edge with incredible performance couple billion parameter models doing you know very rich functionality are very real in today's time. A lot of successful companies that we interview here on the cube and the ones that we see in the industry all the time that that are in AI and winning have deep bench of labs. >> That's right. >> Technical people. Even uh one company I just interviewed this week got a half a billion dollars. A bulk of it going to go to R&D. >> So having a technical team matters. Speak to that nuance because back then oh 10% R&D maybe 20% if you were aggressive. Not anymore. >> No. No. I don't you know I don't have the numbers in front of me but I can simply say the vast majority 70 80% of our people all scientists engineers mostly PhDs from the best universities in the world doing breakthrough research following the model trends looking at the best way to optimize the edge technology. >> It's interesting you went to hire great people you got to have hard problems to solve. That's right. >> What are some of the hard problems that you see out there? I mean money follows constraints. Entrepreneurs follow constraints. businesses enter new markets through constraints that they solve. What are some of those areas that you guys are eyeing down in the market? >> Well, it's just the point I said earlier, the ability to do more complex models in the edge. Most people thought of the edge as relatively simple CNN's and things like that at one particular point. Now, very aggressive breakthrough type of models. So, if you're familiar with a state of models called state space models, we're we have an offering to support them. You know, a lot of uh companies project that it'll surpass transformers as the LLM choice of the future. Uh we of course we will support both, but we put a tremendous amount of energy into state space models, supporting them, developing them because they they are really compelling reasons why there'll be better kind of morphic ways the the benefit. >> Absolutely. The noramorphic way is finally catching on. You know, it's interesting about norphic and it's same thing. you know, you're you're a longtime veteran in Silicon Valley as I am too. Breakthrough technology is wonderful, but if you make it too hard to implement, it doesn't go. So, what we strive to do is make this incredible breakthrough performance, but make it really easy to adopt, meaning the best tool chain to port models on there. Make a model zoo for people to try and buy. We have a whole plethora of kind of uh form factors for people to try and buy. And of course, I put a real customer focus in. I know when customers engage with us, we're enabling their revenue stream. The entire company knows that we're here to make those companies success. >> It's it's all of that, but it starts with the right leadership. And I believe my view of this when I came in was we have got to make this implementable super easy and that means every part of it whether that's documentation customer interactions the tool chains itself make it more intuitive the guies and things like that also to support industry standard frameworks. So as an example, a lot of people think of normorphic as overly complex analog. It's fully digital. You can put this in any fab in the world, right? You know, we have chips that are in global foundaries, TSMC, we're part of the Intel foundry program. So you can move it. So that makes it easy. Models today, a lot of them develop PyTorch, TensorFlow. You don't have to learn a new framework. If you got a bunch of scientists, engineers, you develop those models, use a tool chain and put them on. So it's those kind of thinking. Let the engineers who are familiar with the framework develop it and make it easy to port their models. >> Yeah. And get value quickly. What's the growth strategy? As CEO, you got to look at the northstar. You got to look at the market. What's your execution plan? What's the growth strategy for Brain Chip? >> Yeah. It's it's really on two vectors. We sell both IP and chips. Now IP, if you're familiar with the IP business model, which I know you are, is very very powerful because you go in with a license and a royalty. And so we've seated the market with key licenses and we've got more coming. And as those chips go out, you know, you get you virtually get a revenue stream that's 100% margin at some point. At the same time, we have silicon because silicon matters for companies who cannot afford to build a chip, right? And say, hey, I want to maybe add a co-processor. We have co-processors to add it on and integrate it. Back to the ease. We integrate the runtime very simply. Any engineer in the world can do it. So you grow on both of those and make it easy to adopt. And that's when you look then at the at the industries that are uptaking and you lean harder on those. We mentioned a few earlier. We we enable them all, but when we see one, we lean really hard like defense right now. >> Yeah. And it's it's hot and we think robotics uh and is going to take on many different >> and drones. >> And drones and again hor it's a horizontal play >> as you pointed out. So it's not like the old classic well it's a vertical industry in a way. Robotics is going to be infused everywhere. >> Of course it is. >> And AI will be everywhere and injecting intelligence is the key at a low cost. good power envelope or battery and get that intelligence kicking. Um I have to ask you because you're a Silicon Valley veteran. We both are. Um the narrative right now on AI is half the world oversimplified. Half hate it, think it's dangerous, it's going to kill us. Other half are like so bull and there's also fringe. You get the purists on the left say it's the best thing ever going to save the world. And the other half wants we're going to get killed by it. But a bulk of the 50% that are kind of doomers on AI are really normal people influenced by the media, don't know the Silicon Valley playbook, they don't know what innovation looks like, and they don't really have confidence. So they're confused. They need clarity, right? >> And I want to ask you specifically what your thoughts are on that because >> if you look at all these revolutions, they didn't just start overnight. >> I mean, you had a background in in chips systems. >> There are people that have come before that have built a lot of stuff. >> That's right. maybe oneoffs, maybe for pioneers, first movers. Yeah. But now that we're going to a scale market, what's your view on this whole AI debate? Do you agree that the confusion on the doomer side is lack of clarity or they're being influenced by the politics of it? What's your view on this? Because, you know, normal people are scratching their head asking me all the time privately, what's going on with this AI thing? I'm like, guys, this all overblown, even Jensen, but Jensen's biased because he sells sells GPUs, he sells infrastructure. But of course, he's right. At least my my opinion. >> Yeah. Yeah. I I think I think it's all those things I think is is influenced by politics, lack of knowledge. In the end, this technology is amazing. If you think about what's happened to our world, you the agricultural revolution, industrial, the internet. This is going to make all those look small. I see this as nothing but great for humanity. Clearly, there's always risk with anything. There was risk of when you did, you know, the internet. There was risk when you did, you know, agriculture. But I I I think it's way overblown. I see nothing but goodness for society. The bubble question comes down to like where's the where's the financing coming from. There's real use case unlike the dot bubble or other bubbles. There really wasn't visibility into the unit economics. We're starting to see already productivity numbers. You're starting to see literally transformation projects that have revenue tied to it. So this is like realworld first of course >> horizontal scale. >> Of course I mean the customers I talk to on a daily basis are very you know ROI oriented right? What's the cost? Look at it. that they do their analytics in here, but usually it's because they already have products in the market and they're looking for that next generation to enable. So, it's very very very clear. The only other exception to that is where you got to have it regardless of the cost and that is defense, right? When somebody else makes a first move, you got to make a move to make it better. >> Great. Well, what's what's next for you? What's on your agenda? I know you travel a lot seeing customers. Uh what's your plan for the second half of the year going into 2027? Well, we've got a tremendous roadmap that I'm highly focused with our team on executing on and announcing that later this year. You're absolutely right. I've got a lot of customer meetings I've got to do and I look forward to chatting with everybody around the world here because I'm going to be leaving here. I'll be doing some trips to Australia and Asia in the second half of this year. >> All right, Sean, great to see you. Congratulations on on uh taking the helm at Brain Chip and congratulations. >> Always a pleasure, John. >> I'm John Furrier at the cube. We are here for the robotics AI series. Physical AI is the hottest area. has not yet gone fully mainstream, but you're starting to see the signals. You got robotics, you got drones, you got all kinds of impact across all industries. This is where intelligence meets the physical world, and that's where you're going to start to see new things emerge, new use cases, breakthroughs on the science side, of course, a lot more that affects humanity. I'm John Furrier. Thanks for watching.