theCUBE + NYSE Wired: AI Factories - Data Centers of the Future | Alexis Crowell, Axelera AI

Alexis Crowell, Axelera AI

In this segment from theCUBE + NYSE Wired’s “AI Factories – Data Centers of the Future” series, theCUBE’s Dave Vellante sits down with Rob Biederman, managing partner at Asymmetric Capital, to unpack a disciplined approach to early-stage investing amid AI-scale infrastructure shifts. Biederman explains Asymmetric’s founder-first model: writing $1–$10M checks (often via SAFEs), joining boards as they form and helping operators with go-to-market, operations, finance and strategy (not product/engineering). He shares why the firm avoided 2021’s lofty SaaS multiples in favor of backing proven builders earlier (single-digit pre-money), and highlights portfolio execution such as a cash-efficient LATAM e-commerce company scaling from ~$1-2M to about $50M in revenue. The discussion also explores Asymmetric’s subscale buy-and-build plays (e.g., pool cleaning in San Diego, sleep apnea clinics in Houston), where density, tech-enabled services and platform ops expand margins and enterprise value. Biederman weighs in on AI economics as enterprises race to “AI factories,” cautioning that not every AI workload creates ROI and that overbuilt compute assumptions could face a reckoning. He argues that winners will prove a clear 10× value equation and avoid scaling go-to-market before product-market fit. Additional insights include early liquidity discipline (returning $0.20 on the dollar before the fund’s third anniversary), portfolio survivability (34 of 35 companies still operating; three positive exits), and guidance to founders: make your value proposition relevant, credible and differentiated. Tune in for candid perspective on how capital efficiency, ownership discipline and anti-thematic sourcing intersect with a world where GPU-dense data centers and AI-scale software are reshaping enterprise infrastructure and economics.

Share this session

Clips
More from theCUBE + NYSE Wired: AI Factories - Data Centers of the Future

Alexis Crowell

CMO & President of the Americas

Axelera AI

play_circle_outline Inference-First Edge AI: Axelera’s In-Memory Compute Runs 8B Models at Thousands FPS on 4–8W

play_circle_outline Edge hyper-convergence: bringing compute near data with wireless, wireline, and diverse power profiles.

play_circle_outline Edge-to-Data-Center Ambition: Five-Year Startup Led by NVIDIA/Intel/Google/Qualcomm/AMD Alumni — Four Chips Taped Out, Two in Production

Info
Transcript

Alexis Crowell, Axelera AI

Alexis Crowell

CMO & President of the Americas Axelera AI

In this interview from theCUBE + NYSE Wired: AI Factories - Data Centers of the Future event, Alexis Crowell, CMO and president of Americas at Axelera AI, joins theCUBE’s John Furrier to discuss the evolution of high-performance, low-power AI inference at the edge. Crowell reveals how Axelera AI is tackling infrastructure bottlenecks by uniting digital and memory compute, enabling the deployment of large language models and high-speed computer vision in a remarkably small four-to-eight-watt footprint. She dives into the strategic impact of hyper-converging th... Read more

explore Keep Exploring

What does the company do, and what is its product and technological approach (the "secret sauce") for delivering high-performance, low-power AI inference at the edge and in data centers? add

How do you see AI impacting edge computing, and what would happen if the edge becomes hyper‑converged? add

What has Axelera accomplished in its first five years (team, chip development, and deployments), and how does its technology address edge computing and enable scaling into data center solutions? add

bolt Powered by CUBE AI

Alexis Crowell, Axelera AI

search

John Furrier

>> Welcome back. I'm John Furrier, host of the theCUBE here at the New York Stock Exchange Cube Studios. Of course, we have our studio in Palo Alto, California, connecting Silicon Valley and Wall Street. This is our AI factory series where we talk to the leaders who are making a difference in accelerating the future of AI, AI native applications and setting the foundation for the scalable infrastructure. A lot of build out going on, a lot of activity. Alexis Crowell is here, CMO and President of America's Axelera AI. Alexis, thanks for coming on theCUBE. Really appreciate it.

Alexis Crowell

>> Thank you for having me. I'm excited to be here.

John Furrier

>> So you guys are in the AI infrastructure business, so talk about what you guys do. Then I want to get into some of these conversations around AI factories, AI factories at the edge, how the computing and infrastructure is evolving. So set the table.

Alexis Crowell

>> Sure. So the short version is we do inference specifically. So we started really honing in on how do you do high performance imprints in a very low wattage footprint. So we actually started at the edge. So we do 4 to 8 watts range, but you can run an 8-billion-parameter model. You can do thousands of frames per second when it comes to computer vision, all with the predication of we think inference is going to be the future of what gets run the most. So that's what we do. We do inference specifically, started at edge, but now we're actually scaling into the data centers.

John Furrier

>> Inference is the hottest area. We saw Groq get acqui-hired by NVIDIA. You starting to see, now, developers and hardware and software manufacturers wireless. And even the telco infrastructure and the carriers really looking at how do they deploy AI in their network and data. Obviously, power's a huge issue. You guys nailed that. What specifically are you guys doing? What's the secret sauce? What's the product?

Alexis Crowell

>> Sure. We took a different approach. So our CEO was incredibly intentional when he built the company. And what he did was he looked at all of the different methodologies to build silicon, basically. And honed in on the best research at the time indicated that digital and memory compute was going to be the best way. So what does that mean in layman's terms? Basically, it means the memory and the compute happen in the same location because one of the biggest bottlenecks, typically, for doing this type of computing is how do you move the data around? So we keep all of the data together, which is what makes this different for us. That's how we can get a lot of performance into a small footprint.

John Furrier

>> Alexis, talk about the use cases that you guys are seeing at the edge. What specifically is it? Is it the classic computer vision? Is it specific to industries or is it more horizontal? What are some of the things you guys are seeing?

Alexis Crowell

>> That's a great question, John. So we do horizontal. Our goal is to make hardware plus software that's applicable everywhere. But in terms of the fastest adopters, absolutely, classic computer vision. So think retail and how are companies doing inventory management, supply chain management. We've got one customer that has figured out if they know how many donuts are on their shelves, they can predict what their sales are going to be. So they've actually networked a number of their stores together on a single one of our systems, and then they watch it. Every second, they have a live count of how many donuts are, because that completely changes whether or not they make their sales quotas for the day or not.

John Furrier

>> Yeah, smart retail.

Alexis Crowell

>> In defense, that's a really big, really big space. Surveillance, how are we looking at not surveillance so much of people, but of spacing. What's happening within our environments?

John Furrier

>> How about robotics? In terms of robotics, physical AI is the hottest thing. We'll hear that at GTC. We're going to hear that at MWC.

Alexis Crowell

>> Yeah. Physical AI. For us, that's kind of where we started. When you start at the edge, it really manifests itself in physical AI. So whether it's more of the traditional machine robotics in terms of single arm, how are you connecting machineries together on a factory line to ... We've got one partner, Kudelski. They've built out a robot dog that does perimeter surveillance and security to make sure that the people on the property are exactly who you would want it to be. So absolutely. Robotics is a huge one, but you really need that high-performance, low-power footprint to make that work.

John Furrier

>> Yeah. And the inference is like the brain working. It's like you get trained, then you infer, that's how you apply the training.

Alexis Crowell

>> That's right.

John Furrier

>> We've been seeing the waves of GenAI, the foundation models and specialty models. And then you see the agent wave. Physical AI is like the third wave, but you guys have kind of pulled that forward. How do you see agents? Because agents are just a mechanism for a lot of inference, a lot of delegation, but it also is a path into physical.

Alexis Crowell

>> Agents from our perspective are another way to leverage AI as a tool. It's really a matter of how are you querying your data. Everyone on the planet is creating data at exponential rates. And we look at agents as a mechanism to figure out how can you do something real with that data. So our current product line was really targeting more of that classic CV language model, GenAI type of approach, but our next product line called Europa will be a sweet spot for how do you leverage those agents at the edge. So kind of how are people leveraging their own models within their own environments. If I only want an agent to pay attention to my data but not have access to everything else, then a model like this or a product like this allows you to do that more consistently within your own environment.

John Furrier

>> What I like about what you guys are doing is like NVIDIA and others, you're bringing AI technology to existing things like the edge. MWC's coming up next week. We'll be there, introducing our new narratives around how we see the edge hyper-converging. It's almost funny, the data center converged storage networking can compute, what, 17 years ago. That was the big hot thing. Then it unbundled. But now you have the convergence at the edge because you got wireless, you got wireline, you have a lot of things going on. There's diversity in power, on the energy side, but you're still connected, you got power. How do you guys see this AI factors at the edge? Because our thesis is if you bring a factory to the edge, it could be the size of a brick, a DGX box, or another piece of hardware. It doesn't have to be that big.

Alexis Crowell

>> It doesn't. Absolutely.

John Furrier

>> It could be a small footprint. What's your reaction to that and what happens if the edge hyper-converges?

Alexis Crowell

>> I mean, I've been in this industry for almost 20 years, so I feel like I've seen a lot of these waves. I think this is the right next step. And the reason both Axelera is investing in it and why we're a part of this is we think the closer the compute and the compute moves to where the data is, the faster people are going to be able to make decisions, the faster you're going to be able to react to what's going on around you. So I love this convergence because, to me, what it's doing is it's pushing more and more of that compute to be real time. If I see something happening right in front of me as a human, my brain is instantly computing what I should do about that. And we think that there's this capability of bringing that artificial compute. And instead of your brain computing it, but a computer actually looking at it closer to where that data is, then everybody ends up with a better solution. You still may need people watching it or you may need an alert coming back to a centralized location, but that real-time understanding gets us more and more towards what a human brain would compute if you had a person watching that instance.

John Furrier

>> Alexis, as the training and inference evolves, the edge is different. It's not seeing old data, it's seeing fresh data. And you mentioned real time. Real-time information is fresh data. It's also a lot of opportunities, but it's also technically challenging.

Alexis Crowell

>> You're right. It is technically challenging because just like a human, as we learn in the moment, we constantly basically retrain, reprogram ourselves. Compute is not currently designed to go do that. So you have to do it in software layers. You have to do it with capabilities like RAGs where you think a little bit differently about how you're bringing new information into it. What really I'm excited about, though, is that hardware, software co-design capability that so many more companies are leaning in on, I think is solving that as we look forward. I don't think we're there yet, but I think that there's this opportunity where when the hardware and the software are so tightly interconnected that we can create these better integration points for that real time, what used to be reinforcement learning, but is now just this net new capability to react in the instant.

John Furrier

>> I think there's going to be a lot of opportunities at the edge. Obviously, IoT is converting into full GenAI. Talk about the company momentum. If you can take a minute to explain some of the things you guys are working on, successes. The folks don't know a lot about what you guys are doing. Share some of the stats and momentum points.

Alexis Crowell

>> I appreciate the question, John. So Axelera has been around for a little over five years. And what I love about it is we haven't had a traditional kind of launch for a five-year company, especially in Silicon. Silicon typically takes a long time to really get right. And what Axelera has been able to do is we've brought in experts throughout the industry. If you look at our leadership team, it's ex-NVIDIA, ex-Intel, ex-Google, Qualcomm, AMD, the folks that really know what they're doing. And in four years, we've taped out four chips and we've taken two of those into full production. We're deployed globally. I'm on four different continents in terms of customers. I've got people in 20 countries. So we're in a really good space in terms of being able to fulfill this need of what customers have told us they've wanted. They want compute at the edge. They want to be able to solve for this massive data growth that they've had. And Fabrizio, who's our co-founder but CEO, his premise was, if we could solve for that, we can take that architecture and scale it up. There's not really been a good success story of someone starting in a data center and being able to scale that infrastructure into an edge environment. So we think we've got the starting point right because now we have performance efficiency and power efficiency at high performance that can now scale into data center solutions and server solutions going forward.

John Furrier

>> Yeah. I love the mission and love the success. And I brought it up because I think this is a tipping point for the industry, the true intelligent edge, not just some manufacturing or some use case. We're talking about a generative AI-like experience everywhere. I mean, look at the power.

Alexis Crowell

>> That's right.

John Furrier

>> I mean, power requirements, you don't need to worry about ... I mean, the big data centers do, but if you're at the edge, if you're a telco provider, you've got a base station, you got a cabinet, you can put a device in there with software, connect it to another factory, office space, retail. All have power, and you guys nailed that power envelope. We think this is going to open up massive acceleration in value propositions. And I know you do too. What would you think those would be? Because some of the things that are going to be net new are going to emerge fast. What are some of the things you see evolving that'll be use case that were hard to do pre-AI?

Alexis Crowell

>> Well, I mean, again, I think AI is a tool from an inference perspective that's going to support every use case over time. If you look at medical, I think medical is a great scenario for this because there's a scarcity of clinicians. According to the World Health Organization, there's 10 million fewer clinicians than what we need worldwide. If you bring AI into that, now not only can you read radiology scans faster, you can point doctors towards, hey, maybe you haven't seen this type of disease, but we've seen this in other parts of the world. Maybe you need to think about this as a diagnostic answer to what's going on in that human. That to me is huge, huge untapped ground because we haven't brought it there yet in math. There's pockets, there's amazing pockets, but we haven't brought it there in mass. And I think that what that brings to both humanity, but also from a business perspective is amazing. So I think medical is this absolute boom. It's going to take a little bit with regulation, but to me, there's so much opportunity there and that you do want locally because you don't want that patient's data going everywhere. You want that patient's data staying with their doctor. And then there's the, how do we make inference available everywhere in the retail scenarios, in the smart city, smart parking, right? It's amazing how much gas is wasted in cities these days of people trying to drive around and figure out where to park. We can solve that. If you solve it at the edge, it's kind of done because it's just available. So to me, this is literally, the sky is the limit on what we can go do. It's really limited by imagination. It's no longer limited by compute, which to me is really, really exciting.

John Furrier

>> Well, you guys are doing a great job. I remember the '90s, grid computing was a hot term. Now, we might actually see a lot more connected intelligence. Alexis, thank you so much for coming on our AI factories show. Really appreciate it. Looking forward to meeting you in person on theCUBE. Thanks for coming here.

Alexis Crowell

>> Thank You. Thank you for having me. I appreciate the opportunity.

John Furrier

>> I'm John Furrier with theCUBE. We are here at our NYSE Studios with our AI factory series. AI factories are the centers powering the intelligence, the tokens, the software. It's coming to the edge. We'll see a lot more configurations. Of course, we're bringing you all the action here. Thanks for watching.