This episode explores edge artificial intelligence architectures and field programmable gate array solutions at the New York Stock Exchange as part of AI Factories presented by theCUBE Research and NYSE Wired. The discussion examines how programmable hardware enables real-time edge intelligence and accelerates deployment across industries.
Sammy Cheung of Efinix and Bob Beachler of Efinix join theCUBE Research hosts Gemma Allen and John Furrier to discuss small, power-efficient FPGA devices, the Efinity software development kit, and applications that require multimodal sensor fusion and low-latency inference.
Cheung explains that edge artificial intelligence demands deterministic low-power compute in milliwatt ranges and that Efinix roadmap targets high performance in compact form factors. They highlight field programmable gate arrays and the Efinity software development kit tool flow as enabling technologies for humanoid robotics, autonomous machines, medical imaging and industrial machine vision.
Beachler emphasizes that the Efinity SDK and programmable tool flow create a competitive moat by simplifying hardware-software integration. They note a market shift from data center GPU scale toward a "do more with less" edge architecture and accelerating real-world deployments.
Topics covered include edge AI FPGA architectures, Efinity SDK tool flow, sensor fusion, real-time inference, and applications in robotics and machine vision.
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
theCUBE + NYSE Wired: Physical AI & Robotics Leaders. If you don’t think you received an email check your
spam folder.
Sign in to theCUBE + NYSE Wired: Physical AI & Robotics Leaders.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open this link to automatically sign into the site.
Register For theCUBE + NYSE Wired: Physical AI & Robotics Leaders
Please fill out the information below. You will recieve an email with a verification link confirming your registration. Click the link to automatically sign into the site.
You’re almost there!
We just sent you a verification email. Please click the verification button in the email. Once your email address is verified, you will have full access to all event content for theCUBE + NYSE Wired: Physical AI & Robotics Leaders.
I want my badge and interests to be visible to all attendees.
Checking this box will display your presense on the attendees list, view your profile and allow other attendees to contact you via 1-1 chat. Read the Privacy Policy. At any time, you can choose to disable this preference.
Select your Interests!
add
Upload your photo
Uploading..
OR
Connect via Twitter
Connect via Linkedin
EDIT PASSWORD
Share
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
theCUBE + NYSE Wired: Physical AI & Robotics Leaders. If you don’t think you received an email check your
spam folder.
Sign in to theCUBE + NYSE Wired: Physical AI & Robotics Leaders.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open this link to automatically sign into the site.
Sign in to gain access to theCUBE + NYSE Wired: Physical AI & Robotics Leaders
Please sign in with LinkedIn to continue to theCUBE + NYSE Wired: Physical AI & Robotics Leaders. Signing in with LinkedIn ensures a professional environment.
Are you sure you want to remove access rights for this user?
Details
Manage Access
email address
Community Invitation
Sammy Cheung & Bob Beachler, Efinix
This episode explores edge artificial intelligence architectures and field programmable gate array solutions at the New York Stock Exchange as part of AI Factories presented by theCUBE Research and NYSE Wired. The discussion examines how programmable hardware enables real-time edge intelligence and accelerates deployment across industries.
Sammy Cheung of Efinix and Bob Beachler of Efinix join theCUBE Research hosts Gemma Allen and John Furrier to discuss small, power-efficient FPGA devices, the Efinity software development kit, and applications that require multimodal sensor fusion and low-latency inference.
Cheung explains that edge artificial intelligence demands deterministic low-power compute in milliwatt ranges and that Efinix roadmap targets high performance in compact form factors. They highlight field programmable gate arrays and the Efinity software development kit tool flow as enabling technologies for humanoid robotics, autonomous machines, medical imaging and industrial machine vision.
Beachler emphasizes that the Efinity SDK and programmable tool flow create a competitive moat by simplifying hardware-software integration. They note a market shift from data center GPU scale toward a "do more with less" edge architecture and accelerating real-world deployments.
Topics covered include edge AI FPGA architectures, Efinity SDK tool flow, sensor fusion, real-time inference, and applications in robotics and machine vision.
>> ...Palo Alto studio connecting Silicon Valley and Wall Street. I'm John Furrier. I'm Dave Vellante, my co-host.
Gemma Allen
>> Welcome back to theCUBE Studio here at the New York Stock Exchange. This is AI Factories, one of our programs with NYSE Wired. And on AI Factories, we talk a lot about maximizing GPUs in data centers. But joining me now are two gentlemen who want to talk about bringing AI to the edge, Sammy Cheung, CEO of Efinix, joined by Bob Beachler, head of marketing and corporate development. Welcome, folks.
Bob Beachler
>> Thanks for having us.
Gemma Allen
>> Delighted.
Sammy Cheung
>> Thank you.
Gemma Allen
>> Delighted to have you. So folks, you were here at the NYSE in January, a little over five, six months ago. Feels like a long time though in the world of AI Factories and this space. Talk us through some developments at Efinix since you've last been on the show.
Bob Beachler
>> Yeah, sure. So I'll turn it over to Sammy to talk a little bit about the business. It's, as you said, five, six months, a lot can change. We entered the year feeling pretty positive, but so far it's really exceeded our expectations. So, Sammy?
Sammy Cheung
>> Yeah. Since January, January was first time after we closed 2025, and we got an exceptional year in '25. We grew over 115% and thought, okay, was it like a one-time phenomenon for big growth because of market demand? But it turns out that momentum is actually continues. And as we have validated that, I mean, most of our design wins and also the growth is supported and propelled by the demand in the edge AI space. And this year, we are looking at another record year. So we are very convinced that, of course, not only AI is real, everyone understand that. Essentially what you said, the data center AI is fantastic. I mean, it set up the foundations for a lot of compute capacity. But now we personally get involved in driving the growth on the edge AI side and witnessing in real time the explosive growth in the business.
Gemma Allen
>> I mean, I think it's been an interesting year for the world of the edge and what it means, what physical AI on the edge means. It was a topic that really wasn't a headline, front-and-center topic just even 18 months ago, but it's certainly having a moment because I think we're increasingly seeing the importance of having intelligence on physical devices in all sorts of proximities. What is it exactly that you folks, and the team at Efinix, are hoping to achieve? Where are you really doubling down on opportunity?
Bob Beachler
>> And so for us, the difference between the data center and the edge is that, first of all, the data is much different. In the data center with generative AI, it's primarily text-based, whether you're coding, whether you're using large-language models. But at the edge, we're in the physical world. So the data that's coming in is multimodal. There's imaging, there's thermal, there's pressure sensors, LiDAR, radar, ultrasound, you name it. We're trying to understand what's going on in our physical world and then being able to process that in real time in a deterministic manner so that whether you're working on a factory floor, a vision-guided robotics, autonomous vehicles. The problem is actually much more difficult when you talk about AI on the edge. And that's why it needs a hardware solution, and in particular, a programmable hardware solution like FPGA that Efinix makes because without that you just don't have the compute power, you don't have the speed and the latency necessary to deal with what's happening in the real world. And so that's what we're seeing. And that's why we're seeing such great growth is that all of these companies, whether you're talking about humanoid robotics or you're talking about machine vision applications, medical equipment, they're all using AI, they're all changing their sensors, the data sets that they need to train for and therefore the problems that they're trying to solve in the inference, and that's why we're seeing such a really great opportunity.
Gemma Allen
>> When we think about the world of data centers, it seems as though the narrative is do more, more and more, and more with more, more with more. When we think about the edge, it's really almost like a counter narrative. It's like do more with less, lower power, lower, I guess, demands or requirements from the perspective of what you can achieve with less. Bring that to life for us. Give us some examples of some customer use cases or some prototypes that you folks are both working on that really, I guess, exemplify how the edge differs to the other race that's happening in industry right now.
Bob Beachler
>> Certainly. So the space constraints and the power constraints that you mentioned are critical. So if you think about even an application like a humanoid robotics in the hand, we're seeing people looking at solving the robot hand problem by putting multiple sensors in every finger. And then inside of the finger of the robot, you have the processing. So putting in an FPGA that's only four millimeters by four millimeters in size, but it's doing the sensor aggregation, processing, fusion and quick low-latency response. And then it's also communicating back to the main processing unit in the head of the robot. Those are the types of things that are very different than a data center.
Gemma Allen
>> And this world moving increasingly away from what was a fixed function world to a world where devices and sensors and different types of machinery, cameras, robots, vehicles, whatever it might be, are actually almost updating instantaneously is obviously a huge technical shift. It involves a whole different level of architecture. What does that look like? Break that down for the audience.
Bob Beachler
>> Sammy, you want to take that one?
Sammy Cheung
>> Sure. I think you're right on. I think this is what we talk about at the edge. It is the real world. In data center, it's a machine world, it's a virtual world. But in the real world, everything is different and they happen at different time, very fragmented. But what's interfaced with real world is what Bob mentioned just now. It's about all sort of sensors, and all these sensors and what we need in today is not just a sensor passing, better data centers coming, but everything need in real time and real time and low latency. So in order to do that, the FPGA is the best. But what make Efinix stand out in this area is truly the efficiency of our innovations and the real-time programmability adding on top or be able to deliver highest performance, lowest power and smaller size is where we see we are shining in any space, be it the sensor or data aggregation or after aggregation helping the processing. And on top of it, I mean, you can view what the world really need is like another platform. GPU is clearly a standard platform for data center. However, in terms of looking at real handle very fragmented different applications in real time near the edge, FPGA is the best. And FPGA provide the efficiency in turnout and standing out in the industry in the marketplace.
Gemma Allen
>> We had an event here at the NYSE last night with the neocloud player, TensorWave, and we talked about the competing world of neoclouds right now. They're all in an AMD. We talked about CUDA and that moat and the fact that the software layer of the data center divide has been pretty effective in basically building a defensible moat. When it comes to the edge, what does that software strategy, that software layer look like and what are, I guess, the competitive forces at play there?
Bob Beachler
>> So in our case, because our devices, when we ship, them are blank, the user designs the hardware that gets programmed into the FPGA. So unlike a processor or a GPU where you have sequential C code or Python, we're working in what's called a hardware description language. So the tool flow that we provide to allow our customers to get their hardware design into our FPGA, along with their C code because we have embedded RISC-V processors, along with getting their AI model into our FPGA because we have an entire AI design flow to lower the model and accelerate it on our FPGAs, that's a tremendous amount of software. And that development kit, in our case, we call it the Efinity SDK, is a big amount of our engineering resources, and we provide that free of charge to our customers. But it is certainly a moat in that regards that once they become familiar with it and use it, they don't want to change. And so we spend a lot of time engineering that and making it as easy to use as possible.
Gemma Allen
>> Well, that's a great position to be in, Bob. In terms of these customers, these use cases, can you talk us through some examples, bring this to life for us? You mentioned humanoid robots with sensors in every finger, fascinating stuff. How has this been actually used effectively in the field? And I guess what's this maturity curve look like as well? I'm interested to understand is a lot of this stuff still prototype or is it actually live active devices we may not know are running on Efinix?
Bob Beachler
>> Sure. It absolutely is live active devices. So what AI has done in the edge is that it's made existing applications better. So think of everything from a portable ultrasound system that now has AI or an endoscopy system that can now recognize tumors in real time as opposed to going back to a radiologist. Think about your autonomous vehicles, processing LiDAR, radar information, machine inspection in terms of finding defects on the factory floor. It's really broad, and it's all over the place. So it's making those existing applications better. They're faster. They have more features. But it's also enabling these new applications that you couldn't have had without AI, like humanoid robotics. I mean, 10 years ago, it was an intractable problem. Now people are really in the nuts and bolts solving these issues in terms of robot hands, in terms of guidance, in terms of what's called transferability. Can my robot go from a bedroom to a laundry room to outside walking the dog without totally losing its mind because each of those is a different data set and a different set of tasks that it needs to do. So it's a mix of both. It's the new applications also making existing stuff better.
Gemma Allen
>> We hear a lot in the generic AI Factories conversation around bottlenecks, and there's, of late, somewhat of a divide as to what the key bottleneck is. Is it energy? Is it power? Again, it's a moving problem, if you like. In this world, in the space you guys are in, what are the true bottlenecks right now? What is the one thing that if you could solve tomorrow you would?
Bob Beachler
>> Sammy, you want to take that?
Sammy Cheung
>> Sure. It's going back, when you mentioned energy, the tolerance in the edge is very small. I mean, basically it's all limited by the thermal and size. Imagine, we do a lot of sensors. Sensors is very sensitive to heat. So in this case that in the past, I mean, when there's no real-time requirement or no compute required near a sensor, that's perfectly fine. You can put a very passive processing device next to the sensor. But nowadays, when we start talking about real AI at the edge and everything else is real time, need to process into it and then has to change as well, so the thermal, which mostly is driven by the power consumption itself, it's very, very critical. So as I mentioned previously, the requirement, it has to be all three factors. It has to be in a real-time processing, has to be lowest power that enable the thermal, has to be high performance and then small. Everything has to come together. It turns out we have that solution, and we have the technology that can enable it, and also we are looking at the roadmap that can even make it better and better. And that is actually why we see explosive growth in the marketplace. And just a reminder, FPGA is not a single custom chip. The same silicon in the same chip can be used for what so many things like LIDAR in the autonomous machines and also industrial cameras. And the exciting part what Bob mentions is the physical AI, robotics, human non-deployments and they have the same equations, thermal and energy, and it is super critical. It's not like you build more, electricity can solve it. It just cannot deliver if you cannot deliver lowest power, highest performance today. And I'm very grateful that we have that technology, we have that innovation that can support all the physical AI innovations.
Bob Beachler
>> So the data center, you're talking to the TensorWave folks, they think in kilowatts and gigawatts. We think in milliwatts. We're orders of magnitude divorced from the data center in terms of the constraints and our biggest problems that we're solving for our customers, which is that energy efficiency and power consumption.
Gemma Allen
>> And it also seems like a space that had some long-tail opportunities that have become very real, very fast. Where are the highest demands that you guys are seeing? Where are those customer demands coming from? What sorts of industries and verticals, I guess, are quickest to the edge to land and expand on the edge?
Bob Beachler
>> Boy, it's really hard to nail down just one. We've mentioned some during this interview. Certainly anything autonomous, whether it's a robot on a factory floor in terms of vision-guided robotics, whether it's an autonomous vehicle on the road, a humanoid robot, anything that needs to be able to have self-actualization and be able to interpret the world and be able to use that, that's certainly a super-fast growing market. But also, as I mentioned, some of the which you consider older, more industrial applications, like factory floors and machine vision, things like medical equipment, AI is just pervasive at this point, and so it's changing the way people are designing these systems. And whenever there's change, they need a programmable solution like what we offer.
Gemma Allen
>> Well, folks, what's ahead? I mean, a lot has happened since you were here in January. It sounds as though you guys have certainly really found an incredibly competitive market, and you're leaning in, and now you're a market leader in this space. What does the next 12 months look like? Heavy is the crown, and all that. How do you stay ahead?
Bob Beachler
>> Well, I'll talk a little bit about the product and let Sammy talk about the business. So on the product side, we're going to be announcing some new products here in just about another week or so that are really focused again to continue to improve our solution for edge AI applications. So we're continuing to focus on that in terms of making better products and delivering them in the shortest amount of time that we can to our customers because there is this insatiable demand. But, Sammy, perhaps you can comment on the business side.
Sammy Cheung
>> I think on the business side, there are two ways. The market is actually really pulling us into that because of our technologies, and the adoption rates is huge. I got to mention just now a specific market. You talk about industrial, automotive, communications are growing crazy, now adding on the physical AI on the robotics side. So that is good. But on the other hand, we are also pushing our growth as well. I think we are investing in all fronts, scaling up pretty fast. It's not just adding people, but we have, Bob mentioned we have a new product announcement. We also have a very interesting roadmap, continue our commitment to support the physical AI area on the silicon side, on the system integration side, also on the two flow integrations. And then on the marketplace, one biggest things that we are investing in helping our growth is the ecosystem. Imagine, we are not just trying to sell a silicon, a chip. We, actually, it's a platform. It's enabler for a lot of innovator. And well edge AI, or even, I mean, totally with a data center as well, I mean, a data center needs a lot more quality data. Where it comes from? It come from physical AI. It come from edge AI. So we are the enabler, and we want to scale ourselves to embrace the ecosystem. On and off, I would expect the market will see us working closely with different partners, whether it's bigger companies, smaller company, hardware, software, IP on an AI company. And that's where we are, and we are committed to be that centerpiece in helping the innovations to come to the market as quick as possible.
Gemma Allen
>> Well, I love that one word you mentioned Sammy, ecosystem. Because if tech is about anything, it's about the power of ecosystem. And a rising tide lifts all ships, as they say. So Sammy and Bob, thank you so much for joining us on theCUBE.
Bob Beachler
>> Hey, Gemma, thanks for having us. It's been wonderful.
Sammy Cheung
>> .
Bob Beachler
>> We look forward to working with you in the future.
Gemma Allen
>> We hope so.
Sammy Cheung
>> Thank you very much.
Gemma Allen
>> I'm Gemma Allen here at theCUBE's NYSE studio. This is AI Factories, one of our programs for NYSC Wired. Thanks for watching.
>> ...Palo Alto studio connecting Silicon Valley and Wall Street. I'm John Furrier. I'm Dave Vellante, my co-host.
Gemma Allen
>> Welcome back to theCUBE Studio here at the New York Stock Exchange. This is AI Factories, one of our programs with NYSE Wired. And on AI Factories, we talk a lot about maximizing GPUs in data centers. But joining me now are two gentlemen who want to talk about bringing AI to the edge, Sammy Cheung, CEO of Efinix, joined by Bob Beachler, head of marketing and corporate development. Welcome, folks.
Bob Beachler
>> Thanks for having us.
Gemma Allen
>> Delighted.
Sammy Cheung
>> Thank you.
Gemma Allen
>> Delighted to have you. So folks, you were here at the NYSE in January, a little over five, six months ago. Feels like a long time though in the world of AI Factories and this space. Talk us through some developments at Efinix since you've last been on the show.
Bob Beachler
>> Yeah, sure. So I'll turn it over to Sammy to talk a little bit about the business. It's, as you said, five, six months, a lot can change. We entered the year feeling pretty positive, but so far it's really exceeded our expectations. So, Sammy?
Sammy Cheung
>> Yeah. Since January, January was first time after we closed 2025, and we got an exceptional year in '25. We grew over 115% and thought, okay, was it like a one-time phenomenon for big growth because of market demand? But it turns out that momentum is actually continues. And as we have validated that, I mean, most of our design wins and also the growth is supported and propelled by the demand in the edge AI space. And this year, we are looking at another record year. So we are very convinced that, of course, not only AI is real, everyone understand that. Essentially what you said, the data center AI is fantastic. I mean, it set up the foundations for a lot of compute capacity. But now we personally get involved in driving the growth on the edge AI side and witnessing in real time the explosive growth in the business.
Gemma Allen
>> I mean, I think it's been an interesting year for the world of the edge and what it means, what physical AI on the edge means. It was a topic that really wasn't a headline, front-and-center topic just even 18 months ago, but it's certainly having a moment because I think we're increasingly seeing the importance of having intelligence on physical devices in all sorts of proximities. What is it exactly that you folks, and the team at Efinix, are hoping to achieve? Where are you really doubling down on opportunity?
Bob Beachler
>> And so for us, the difference between the data center and the edge is that, first of all, the data is much different. In the data center with generative AI, it's primarily text-based, whether you're coding, whether you're using large-language models. But at the edge, we're in the physical world. So the data that's coming in is multimodal. There's imaging, there's thermal, there's pressure sensors, LiDAR, radar, ultrasound, you name it. We're trying to understand what's going on in our physical world and then being able to process that in real time in a deterministic manner so that whether you're working on a factory floor, a vision-guided robotics, autonomous vehicles. The problem is actually much more difficult when you talk about AI on the edge. And that's why it needs a hardware solution, and in particular, a programmable hardware solution like FPGA that Efinix makes because without that you just don't have the compute power, you don't have the speed and the latency necessary to deal with what's happening in the real world. And so that's what we're seeing. And that's why we're seeing such great growth is that all of these companies, whether you're talking about humanoid robotics or you're talking about machine vision applications, medical equipment, they're all using AI, they're all changing their sensors, the data sets that they need to train for and therefore the problems that they're trying to solve in the inference, and that's why we're seeing such a really great opportunity.
Gemma Allen
>> When we think about the world of data centers, it seems as though the narrative is do more, more and more, and more with more, more with more. When we think about the edge, it's really almost like a counter narrative. It's like do more with less, lower power, lower, I guess, demands or requirements from the perspective of what you can achieve with less. Bring that to life for us. Give us some examples of some customer use cases or some prototypes that you folks are both working on that really, I guess, exemplify how the edge differs to the other race that's happening in industry right now.
Bob Beachler
>> Certainly. So the space constraints and the power constraints that you mentioned are critical. So if you think about even an application like a humanoid robotics in the hand, we're seeing people looking at solving the robot hand problem by putting multiple sensors in every finger. And then inside of the finger of the robot, you have the processing. So putting in an FPGA that's only four millimeters by four millimeters in size, but it's doing the sensor aggregation, processing, fusion and quick low-latency response. And then it's also communicating back to the main processing unit in the head of the robot. Those are the types of things that are very different than a data center.
Gemma Allen
>> And this world moving increasingly away from what was a fixed function world to a world where devices and sensors and different types of machinery, cameras, robots, vehicles, whatever it might be, are actually almost updating instantaneously is obviously a huge technical shift. It involves a whole different level of architecture. What does that look like? Break that down for the audience.
Bob Beachler
>> Sammy, you want to take that one?
Sammy Cheung
>> Sure. I think you're right on. I think this is what we talk about at the edge. It is the real world. In data center, it's a machine world, it's a virtual world. But in the real world, everything is different and they happen at different time, very fragmented. But what's interfaced with real world is what Bob mentioned just now. It's about all sort of sensors, and all these sensors and what we need in today is not just a sensor passing, better data centers coming, but everything need in real time and real time and low latency. So in order to do that, the FPGA is the best. But what make Efinix stand out in this area is truly the efficiency of our innovations and the real-time programmability adding on top or be able to deliver highest performance, lowest power and smaller size is where we see we are shining in any space, be it the sensor or data aggregation or after aggregation helping the processing. And on top of it, I mean, you can view what the world really need is like another platform. GPU is clearly a standard platform for data center. However, in terms of looking at real handle very fragmented different applications in real time near the edge, FPGA is the best. And FPGA provide the efficiency in turnout and standing out in the industry in the marketplace.
Gemma Allen
>> We had an event here at the NYSE last night with the neocloud player, TensorWave, and we talked about the competing world of neoclouds right now. They're all in an AMD. We talked about CUDA and that moat and the fact that the software layer of the data center divide has been pretty effective in basically building a defensible moat. When it comes to the edge, what does that software strategy, that software layer look like and what are, I guess, the competitive forces at play there?
Bob Beachler
>> So in our case, because our devices, when we ship, them are blank, the user designs the hardware that gets programmed into the FPGA. So unlike a processor or a GPU where you have sequential C code or Python, we're working in what's called a hardware description language. So the tool flow that we provide to allow our customers to get their hardware design into our FPGA, along with their C code because we have embedded RISC-V processors, along with getting their AI model into our FPGA because we have an entire AI design flow to lower the model and accelerate it on our FPGAs, that's a tremendous amount of software. And that development kit, in our case, we call it the Efinity SDK, is a big amount of our engineering resources, and we provide that free of charge to our customers. But it is certainly a moat in that regards that once they become familiar with it and use it, they don't want to change. And so we spend a lot of time engineering that and making it as easy to use as possible.
Gemma Allen
>> Well, that's a great position to be in, Bob. In terms of these customers, these use cases, can you talk us through some examples, bring this to life for us? You mentioned humanoid robots with sensors in every finger, fascinating stuff. How has this been actually used effectively in the field? And I guess what's this maturity curve look like as well? I'm interested to understand is a lot of this stuff still prototype or is it actually live active devices we may not know are running on Efinix?
Bob Beachler
>> Sure. It absolutely is live active devices. So what AI has done in the edge is that it's made existing applications better. So think of everything from a portable ultrasound system that now has AI or an endoscopy system that can now recognize tumors in real time as opposed to going back to a radiologist. Think about your autonomous vehicles, processing LiDAR, radar information, machine inspection in terms of finding defects on the factory floor. It's really broad, and it's all over the place. So it's making those existing applications better. They're faster. They have more features. But it's also enabling these new applications that you couldn't have had without AI, like humanoid robotics. I mean, 10 years ago, it was an intractable problem. Now people are really in the nuts and bolts solving these issues in terms of robot hands, in terms of guidance, in terms of what's called transferability. Can my robot go from a bedroom to a laundry room to outside walking the dog without totally losing its mind because each of those is a different data set and a different set of tasks that it needs to do. So it's a mix of both. It's the new applications also making existing stuff better.
Gemma Allen
>> We hear a lot in the generic AI Factories conversation around bottlenecks, and there's, of late, somewhat of a divide as to what the key bottleneck is. Is it energy? Is it power? Again, it's a moving problem, if you like. In this world, in the space you guys are in, what are the true bottlenecks right now? What is the one thing that if you could solve tomorrow you would?
Bob Beachler
>> Sammy, you want to take that?
Sammy Cheung
>> Sure. It's going back, when you mentioned energy, the tolerance in the edge is very small. I mean, basically it's all limited by the thermal and size. Imagine, we do a lot of sensors. Sensors is very sensitive to heat. So in this case that in the past, I mean, when there's no real-time requirement or no compute required near a sensor, that's perfectly fine. You can put a very passive processing device next to the sensor. But nowadays, when we start talking about real AI at the edge and everything else is real time, need to process into it and then has to change as well, so the thermal, which mostly is driven by the power consumption itself, it's very, very critical. So as I mentioned previously, the requirement, it has to be all three factors. It has to be in a real-time processing, has to be lowest power that enable the thermal, has to be high performance and then small. Everything has to come together. It turns out we have that solution, and we have the technology that can enable it, and also we are looking at the roadmap that can even make it better and better. And that is actually why we see explosive growth in the marketplace. And just a reminder, FPGA is not a single custom chip. The same silicon in the same chip can be used for what so many things like LIDAR in the autonomous machines and also industrial cameras. And the exciting part what Bob mentions is the physical AI, robotics, human non-deployments and they have the same equations, thermal and energy, and it is super critical. It's not like you build more, electricity can solve it. It just cannot deliver if you cannot deliver lowest power, highest performance today. And I'm very grateful that we have that technology, we have that innovation that can support all the physical AI innovations.
Bob Beachler
>> So the data center, you're talking to the TensorWave folks, they think in kilowatts and gigawatts. We think in milliwatts. We're orders of magnitude divorced from the data center in terms of the constraints and our biggest problems that we're solving for our customers, which is that energy efficiency and power consumption.
Gemma Allen
>> And it also seems like a space that had some long-tail opportunities that have become very real, very fast. Where are the highest demands that you guys are seeing? Where are those customer demands coming from? What sorts of industries and verticals, I guess, are quickest to the edge to land and expand on the edge?
Bob Beachler
>> Boy, it's really hard to nail down just one. We've mentioned some during this interview. Certainly anything autonomous, whether it's a robot on a factory floor in terms of vision-guided robotics, whether it's an autonomous vehicle on the road, a humanoid robot, anything that needs to be able to have self-actualization and be able to interpret the world and be able to use that, that's certainly a super-fast growing market. But also, as I mentioned, some of the which you consider older, more industrial applications, like factory floors and machine vision, things like medical equipment, AI is just pervasive at this point, and so it's changing the way people are designing these systems. And whenever there's change, they need a programmable solution like what we offer.
Gemma Allen
>> Well, folks, what's ahead? I mean, a lot has happened since you were here in January. It sounds as though you guys have certainly really found an incredibly competitive market, and you're leaning in, and now you're a market leader in this space. What does the next 12 months look like? Heavy is the crown, and all that. How do you stay ahead?
Bob Beachler
>> Well, I'll talk a little bit about the product and let Sammy talk about the business. So on the product side, we're going to be announcing some new products here in just about another week or so that are really focused again to continue to improve our solution for edge AI applications. So we're continuing to focus on that in terms of making better products and delivering them in the shortest amount of time that we can to our customers because there is this insatiable demand. But, Sammy, perhaps you can comment on the business side.
Sammy Cheung
>> I think on the business side, there are two ways. The market is actually really pulling us into that because of our technologies, and the adoption rates is huge. I got to mention just now a specific market. You talk about industrial, automotive, communications are growing crazy, now adding on the physical AI on the robotics side. So that is good. But on the other hand, we are also pushing our growth as well. I think we are investing in all fronts, scaling up pretty fast. It's not just adding people, but we have, Bob mentioned we have a new product announcement. We also have a very interesting roadmap, continue our commitment to support the physical AI area on the silicon side, on the system integration side, also on the two flow integrations. And then on the marketplace, one biggest things that we are investing in helping our growth is the ecosystem. Imagine, we are not just trying to sell a silicon, a chip. We, actually, it's a platform. It's enabler for a lot of innovator. And well edge AI, or even, I mean, totally with a data center as well, I mean, a data center needs a lot more quality data. Where it comes from? It come from physical AI. It come from edge AI. So we are the enabler, and we want to scale ourselves to embrace the ecosystem. On and off, I would expect the market will see us working closely with different partners, whether it's bigger companies, smaller company, hardware, software, IP on an AI company. And that's where we are, and we are committed to be that centerpiece in helping the innovations to come to the market as quick as possible.
Gemma Allen
>> Well, I love that one word you mentioned Sammy, ecosystem. Because if tech is about anything, it's about the power of ecosystem. And a rising tide lifts all ships, as they say. So Sammy and Bob, thank you so much for joining us on theCUBE.
Bob Beachler
>> Hey, Gemma, thanks for having us. It's been wonderful.
Sammy Cheung
>> .
Bob Beachler
>> We look forward to working with you in the future.
Gemma Allen
>> We hope so.
Sammy Cheung
>> Thank you very much.
Gemma Allen
>> I'm Gemma Allen here at theCUBE's NYSE studio. This is AI Factories, one of our programs for NYSC Wired. Thanks for watching.
John Furrier
Gemma Allen