Jeff Tatarchuk of TensorWave appears on theCUBE and NYSE Wired at the New York Stock Exchange, NYSE to discuss TensorWave's AMD-first strategy for scaling graphics processing unit, GPU infrastructure for artificial intelligence, AI factories. Tatarchuk draws on experience building GPU-first infrastructure to explain the strategic bet on AMD, NeoCloud economics and open-source tooling. They discuss supply constraints, software ecosystems and enterprise adoption and they outline the company's approach to securing data center power and capacity while scaling AI compute across cloud and edge environments.
Key takeaways include TensorWave's $350 million series B led by AMD and Magnetar, a concentrated effort to secure data center power and capacity, a plan to double capacity this year and pursue gigawatt-scale deployments by 2027, and advancement of ScalarLM, an open GPU-agnostic training and inference stack that simplifies migrations between NVIDIA and AMD.
The conversation covers GPU architecture and performance, ROCm and CUDA software ecosystems, open-source tooling and migration strategies, funding and partnerships and operational factors to consider when deploying large-scale AI infrastructure.
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
theCUBE + NYSE Wired: AI Factories - Data Centers of the Future. If you don’t think you received an email check your
spam folder.
Sign in to AI Factories - Data Centers of the Future.
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 the link to automatically sign into the site.
Register for AI Factories - Data Centers of the Future
Please fill out the information below. You will receive 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 AI Factories - Data Centers of the Future.
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: AI Factories - Data Centers of the Future. If you don’t think you received an email check your
spam folder.
Sign in to AI Factories - Data Centers of the Future.
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 the link to automatically sign into the site.
Sign in to gain access to theCUBE + NYSE Wired: AI Factories - Data Centers of the Future
Please sign in with LinkedIn to continue to theCUBE + NYSE Wired: AI Factories - Data Centers of the Future. 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
Jeff Tatarchuk Tensorwave
Jeff Tatarchuk of TensorWave appears on theCUBE and NYSE Wired at the New York Stock Exchange, NYSE to discuss TensorWave's AMD-first strategy for scaling graphics processing unit, GPU infrastructure for artificial intelligence, AI factories. Tatarchuk draws on experience building GPU-first infrastructure to explain the strategic bet on AMD, NeoCloud economics and open-source tooling. They discuss supply constraints, software ecosystems and enterprise adoption and they outline the company's approach to securing data center power and capacity while scaling AI compute across cloud and edge environments.
Key takeaways include TensorWave's $350 million series B led by AMD and Magnetar, a concentrated effort to secure data center power and capacity, a plan to double capacity this year and pursue gigawatt-scale deployments by 2027, and advancement of ScalarLM, an open GPU-agnostic training and inference stack that simplifies migrations between NVIDIA and AMD.
The conversation covers GPU architecture and performance, ROCm and CUDA software ecosystems, open-source tooling and migration strategies, funding and partnerships and operational factors to consider when deploying large-scale AI infrastructure.
>> Palo Alto Studio connecting Silicon Valley and Wall Street. I'm John Furrier here with Dave Vellante, my co-host.
Gemma Allen
>> I'm Gemma Allen coming to you from theCUBE Studio here at the New York Stock Exchange. This is AI Factories, one of our programs with NYSE Wired. And joining me now is Jeff Tatarchuk, co-founder and chief growth officer, chief GPU officer at TensorWave. Welcome, Jeff.
Jeff Tatarchuk
>> Awesome. Great to be here.
Gemma Allen
>> If you're going to be the chief of anything right now, it might as well be a GPU, right?
Jeff Tatarchuk
>> That's right.
Gemma Allen
>> It feels like that's where the money is. So TensorWave, you guys are a company that have gone all in on AMD. Break it down for me. Why AMD? Why now? Share the story.
Jeff Tatarchuk
>> Yeah, that's a great question, Gemma. It has a lot to do with our origin story. So previously before TensorWave, my other co-founder and I, we had an FPGA cloud and so we were working with Xilinx chips and Altera chips at the time. And Altera was closely connected to Intel and then Xilinx got acquired by AMD about four or five years ago. And then our previous company became AMD's internal FPGA cloud where they would send us all of their latest and greatest FPGAs and we get them debugged and working for them in the cloud. And it's how we forged the relationship that we had with them early on. And we saw the path. We saw the roadmap that AMD had laid out with their foundation being on open source and open standards. And when they announced their GPU offering and there was these supply constraints happening back in 2022, 2023 when AI took off, it made sense for us to go all in on AMD and to be the first and best to deploy their chips at scale. And so at the very beginning, it seemed like a very unintuitive thing to do. People were like, why? Everybody loves NVIDIA. AMD seems to be really far behind, but since we had already cut our teeth on FPGA, solving one of the harder problems first, it made supporting them on scaling out their GPUs a sensible answer for us.
Gemma Allen
>> Well, AMD is certainly having a moment. I think the market loves AMD. People love Lisa.
Jeff Tatarchuk
>> We love Lisa.
Gemma Allen
>> The company on the up, I love to hear that. I love a female CEO overcoming the infra space. Some folks say though the chips are great, the software isn't there. CUDA still has a massive moat opportunity for NVIDIA and ROCm is somewhat of a gaining traction, but not quite where it could be from the perspective of like loyal hearts and minds, right?
Jeff Tatarchuk
>> Yeah.
Gemma Allen
>> What do you think?
Jeff Tatarchuk
>> Well, there's definitely a large gap to begin with. NVIDIA definitely had the advantage. They've been doing it for 20 years, they've been working on this. Jensen did a great job at putting CUDA on all of the consumer hardware, putting it in all the research labs. And so everybody who had been doing high performance compute already had familiarity with CUDA. And so he had built quite a large ecosystem around this space for some time. And so this is where AMD had to spend some time getting caught up. They had built superior hardware with more memory, better across the board. But when it came to the software, that's where a lot of people were skeptical. So in the very beginning when we launched, it was just showcasing that yes, AMD does work. At the very beginning, there was a Databricks article that came out back in 2023 where they showcased, they did a test going from NVIDIA's CUDA porting it over to AMD's ROCm and showing that it does work out of the box.
Gemma Allen
>> Wow.
Jeff Tatarchuk
>> And so that was one of the early signals that we were able to use to say like, yes, this works. Yes, there are some gaps, but AMD has done an amazing job at closing those gaps. And even more exciting thing that's happening now with Codex and Claude and all of these different, with Claude Code, where we're able to use AI agents to help do the code development necessary to close the gap. And so frankly, we see Jensen shaking in his boots when it comes to that CUDA moat that exists, but there is some things that need to be filled when it comes to the ecosystem. Jensen has done a great job at filling out the ecosystem. And this is where we at TensorWave, this is one of my things that I'm focused on is building out the ecosystem necessary to support AI infrastructure at scale on AMD, which is why we're entirely focused on AMD for that reason. If our loyalty is split across multiple things, it's harder for us to be the best at one.
Gemma Allen
>> And I mean, you're clearly leading. TensorWave is definitely the leading neocloud for AMD right now.
Jeff Tatarchuk
>> That's right.
Gemma Allen
>> You guys have some exciting news, I believe. The market sees the opportunity here. Fill us in.
Jeff Tatarchuk
>> Yeah. We just closed our series B. We raised $350 million led by AMD and Magnetar. And so it's just a milestone in the progress that we've made over the last two and a half years, and it's just one small step to even greater things that we see to come.
Gemma Allen
>> Wow. Okay. So let's talk about where that money's going to go.
Jeff Tatarchuk
>> Yeah.
Gemma Allen
>> So clearly there was a big opportunity here. The TAM is huge. Some folks say, you know what? We were at a point in the market where people just needed GPUs. It almost didn't matter where they come from, how they come in. Even spend has been in a state of flux as we well know. But there is definitely kind of segments of opportunities here, I am sure.
Jeff Tatarchuk
>> That's right.
Gemma Allen
>> How are you guys thinking about it? What has been the growth to now? Who has been the typical profile of customer? What does that look like and what is the new TAM that maybe you guys haven't even thought about yet that you will?
Jeff Tatarchuk
>> Yeah, great question. The focus has been on the neo labs that need access to compute for training and for inference. They're all sophisticated enough to jump in, grab the compute, and take advantage of it at its full scale. And so that's been our primary focus, on the top neo labs of the world and getting them access to compute, especially those that are willing and excited about being early adopters. What we heard early on is they were tired of giving all of their margin to Jensen and they needed something to help bring competition into the market, and so they knew that we were the place that can help them deploy it at scale. And since seeing where AMD announced these deals with both OpenAI to do six gigawatts worth of capacity and Meta to do six gigawatts worth of capacity, it really sent a signal to the market where it switched us from headwinds where people were still skeptical of AMD to now tailwinds where people are banging down the door. We can't get enough capacity-
Gemma Allen
>> Wow....
Jeff Tatarchuk
>> to support the demand that's currently out there. And so yeah, to answer your question around what we're doing with the capital that we are getting, it's frankly locking up data center capacity. That's currently one of the biggest bottlenecks, being able to get access to as much power and data center capacity as we can because the GPUs are available. But can you get data centers and powers up fast enough to support and plug those things in to give people access to them? So if we had them up and ready and available today, people would take them to the tune of 5, 10, to hundreds of thousands of GPUs. And so locking up as much data center capacity as we can for 2026 and 2027 and obviously growing out our team to be able to support that infrastructure, and, yeah, to continue to build things out. And so our first initial focus was on the neo labs, and we do see now more and more demand happening on the enterprise side where companies are ... I feel like the enterprise was very AI curious at first, not sure how to engage with whether they're engaging with the top two just APIs or they're going to build out their own infrastructure, and now we're seeing more openness to enterprises getting access to compute, which is one of the things that we're able to help with as well.
Gemma Allen
>> So staying on the topic of enterprise for a second and going back to something you said a few minutes ago. You said sometimes folks are trying to discover what the best solution solve is for their specific need.
Jeff Tatarchuk
>> That's right.
Gemma Allen
>> What workload needs what, and AI can actually help with that. You guys are building, I guess, out your own systems and process to help create an opportunity there. What is the other side of that coin from the perspective of enterprise? Are we seeing a lot of self-managed environments, folks taking a lot of this work back captive?
Jeff Tatarchuk
>> Yeah.
Gemma Allen
>> Where are the kind of substrates happening or consolidating?
Jeff Tatarchuk
>> Well, at first it seemed like they were very skeptical and now that there's options in the market, they're a lot more excited. Before, they didn't want to be tied to just one vendor. And so if they got one vendor, they were stuck with using that one vendor with one stack. Or if they even bought the other vendor, they were stuck with using that one vendor and one stack. But one of the things that we've worked on at TensorWave is we built an open source training and inference stack called ScalarLM with our team, Greg Diamos, who was actually one of the original team at NVIDIA who built CUDA has now created a GPU agnostic training and inference stack that allows you to switch from NVIDIA to AMD at scale at the enterprise level. So we've made it as easy as possible. We've taken the complexity out. GPUs are finicky. GPUs are even more finicky at scale and switching from one SKU to another brand is even more finicky and our job at TensorWave has been to make it as easy as possible. And so we see enterprises now a lot more receptive and interested at deploying their own compute and working with neoclouds to scale out their infrastructure more so than ever before because the tools are now available that weren't there before.
Gemma Allen
>> It's interesting because I think in the neocloud space of the broad conversation, there is definitely somewhat of a narrative around are these real estate plays or are these true technology plays, right?
Jeff Tatarchuk
>> Yeah.
Gemma Allen
>> And I guess the software layer and that can added value for customers of discoverability, of managing solutions, not going to short term solves is huge.
Jeff Tatarchuk
>> That's right.
Gemma Allen
>> People are like, "Are we going to see a mass convergence at the neo cloud level broadly?" We're at GTC this year and I heard someone say, "There's a new neocloud every day." As somebody who's actively competing in that space and really making a significant mark, how do you think about those narratives? How would you stay ahead of them?
Jeff Tatarchuk
>> Yeah, I think about it as well, especially on, we know on the NVIDIA side, there's 150 plus neoclouds. And on the AMD side, those that support AMD, less than 10. And so I can't imagine swimming in that pool trying to get access to allocation when I don't even know if there's 100 plus neo labs that can take that much compute. And so I do think with the competition being as hot as it is, there will be consolidation. I do think there will be clouds that aren't going to make it. They're not able to keep up with their financing terms and pay their bills because of the data center spend that they've been able to spend or keep up with. Maybe if a lab isn't able to pay their bills and they can't find somebody else to take on that bill. There's going to be a lot of labs going to be in trouble. There will be a lot of consolidation. But I do believe that those who will rise to the top will be those that have solved the problem to make it as easy as possible for the end user, and that's what our goal is at TensorWave is to make it as easy for the end user to use and consume compute for whatever use case they might have.
Gemma Allen
>> Well, that's what everyone wants, right?
Jeff Tatarchuk
>> That's right.
Gemma Allen
>> So in terms of the broad economics of this, I presume these are CapEx heavy. There's a lot to figure out here. What is the broader plan? Are you guys planning to grow more, scale more? What does the footprint look like five years from now?
Jeff Tatarchuk
>> Yeah. Well, we definitely, we're not stopping here. We're going to continue to grow scale. We're already working on our next round, which will be significant. And yeah, we need as much capital, we need as much resources to scale out, to support the demand that's currently there. And yeah, our goal is to make access to compute as easy as possible and as simple as like switching on your utility or switching on your lights, paying your utility bill where we can abstract away all the complexity and give people access to this new tool that has never really been available. It's kind of the wild west frontier. And we're excited to be at the forefront of all the different pieces, whether it's policy, working with communities, showing how people are integrating this at the application layer, how is this going to impact jobs, and being at the foundation of it all with building out the infrastructure is something that I'm particularly excited about, especially there's a lot of kind of fear around the data center AI space. I was seeing some studies and people are talking about the perspective of data centers in China where people are thinking about how they're excited about AI. They're excited about the impact that AI can make on their daily lives. But in America, North America, we have this fear of AI is going to take our jobs and all these data centers are going to suck up all of our water and they're going to shoot our power bills through the roof and we're going to hear this buzzing going throughout our small communities. When in reality, I do think there's going to be a huge benefit beyond what we could even understand to what AI is going to bring to the world. And it's exciting to me to be at the foundation of it all.
Gemma Allen
>> I mean, there's definitely some sociopolitical challenge. There's also a big energy challenge though, right?
Jeff Tatarchuk
>> Yeah, that's right.
Gemma Allen
>> Some folks think, well, once the chip shortage ends, which maybe eventually will, things, again, market dynamics going to shift fundamentally all over again, right?
Jeff Tatarchuk
>> That's right.
Gemma Allen
>> What are your thoughts on the energy challenge? What's happening through front of line here? What are you seeing? What's the ultimate bottleneck?
Jeff Tatarchuk
>> It is power. It's getting access to as much power as we possibly can and getting it up and available as fast as possible. Yeah, that's the challenge across the board. And so being able to stand up data centers, get access to power, and plug in these GPUs is really the current biggest bottleneck. And yeah.
Gemma Allen
>> Let's go off script for a second. So space, nuclear, what do you think the reality ... Where do you think pragmatism comes into play here versus kind of almost sci-fi like dreams?
Jeff Tatarchuk
>> Yeah. I mean, it is exciting to hear about data centers in space. I see how finicky GPUs and how hard it is to maintain data centers here on earth, and so being able to increase the reliability of data centers here on earth and service those data centers here on earth without the need of a lot of humans involved in that process I think is going to take more time than people think, but it's exciting. It's exciting to see how robots are getting involved and making it easier. And the idea of doing it makes a lot of sense, being able to get access to the resources and power that you can utilize up in space makes sense, but pragmatically-
Gemma Allen
>> How we cool them.
Jeff Tatarchuk
>> Yeah. How can you take care of them? When a GPU goes down, a server goes down, that's going to be the biggest problem that people are going to have to solve along the way. But I'm optimistic. I do think we dream big. We create solutions to some of these problems that we have here on earth. There's great possibility around the corner. And yes, we do need nuclear. We need as many nuclear power plants up as fast as possible. So I'm excited about the prospects around that as well.
Gemma Allen
>> So you mentioned wild west earlier, you guys are based in Vegas.
Jeff Tatarchuk
>> Yeah.
Gemma Allen
>> Interesting choice. We're always in Vegas. John Furrier's in Vegas right now. In fact, he just flew out this morning.
Jeff Tatarchuk
>> Nice.
Gemma Allen
>> Talk a little bit about the business community there and the opportunity, I guess, because it's somewhat of an untapped tech zone. We just know it from a conference perspective, but I'm sure there's plenty to say about it that maybe doesn't get said enough. So fill us in.
Jeff Tatarchuk
>> That's right. So I moved to Las Vegas in 2020 during COVID. I was living in Southern California at the time and my other company was based out in Wyoming. So it made a lot of sense for me to move to Nevada to still be able to reap the same tax advantages that I had being in Wyoming, and Nevada has been very friendly to us. The state of Nevada is an investor. So there was a fund, FundNV invested in us and the state of Nevada matched the initial investment. So they rolled out the red carpet to make sure we had the resources that we needed to grow and scale. There's a lot of great talent in Las Vegas. There's a lot of people who are willing to move from Silicon Valley or the Bay Area to Las Vegas because of how convenient it is. The same time it takes you to drive from San Jose to San Francisco, you can hop on a flight from Las Vegas to any of those destinations and be there a lot faster and not have to sit in traffic. And so there's a lot of great benefits. Taxes being one of them. But I mean, with the entertainment, with food and the growing tech ecosystem, we see a great advantage for more companies to start moving there. We have a lot of companies that are and other founders in tech that are making the move out there. And so we do a lot to help pour into the ecosystem. And yeah, it's more than just a conference town. And it is convenient. All of our customers will come through town at some point and so it makes sense. But yeah, we're excited about being in Vegas. And we're not just going to only be in Vegas. We're going to expand to other areas as well, but it's great. Vegas and Nevada have been great to us.
Gemma Allen
>> Well, you certainly have a captive audience in terms of the foot traffic, right?
Jeff Tatarchuk
>> That's right.
Gemma Allen
>> So that's wonderful. Okay. So Jeff, last question and I think perhaps you answered this already, but maybe give me a little bit more in terms of what does the next year look like for you and the team? Like what's ahead from here? You've had this big raise. You said you're raising a C. What are the top five or three to five priorities as you look outward?
Jeff Tatarchuk
>> Expand our team, buy more GPUs, lock up more data centers. And we're doubling our capacity throughout the rest of this year and we will be going into gigawatts into 2027, and so the faster we can move to make that happen, the better.
Gemma Allen
>> Wow, we love it. Well, Jeff, thanks so much for joining us on theCUBE.
Jeff Tatarchuk
>> Yeah, thank you, Gemma.
Gemma Allen
>> I'm Gemma Allen here at theCUBE Studio at the NYSE. This is AI Factories, one of our programs with NYSE Wired. Thanks for watching.
>> Palo Alto Studio connecting Silicon Valley and Wall Street. I'm John Furrier here with Dave Vellante, my co-host.
Gemma Allen
>> I'm Gemma Allen coming to you from theCUBE Studio here at the New York Stock Exchange. This is AI Factories, one of our programs with NYSE Wired. And joining me now is Jeff Tatarchuk, co-founder and chief growth officer, chief GPU officer at TensorWave. Welcome, Jeff.
Jeff Tatarchuk
>> Awesome. Great to be here.
Gemma Allen
>> If you're going to be the chief of anything right now, it might as well be a GPU, right?
Jeff Tatarchuk
>> That's right.
Gemma Allen
>> It feels like that's where the money is. So TensorWave, you guys are a company that have gone all in on AMD. Break it down for me. Why AMD? Why now? Share the story.
Jeff Tatarchuk
>> Yeah, that's a great question, Gemma. It has a lot to do with our origin story. So previously before TensorWave, my other co-founder and I, we had an FPGA cloud and so we were working with Xilinx chips and Altera chips at the time. And Altera was closely connected to Intel and then Xilinx got acquired by AMD about four or five years ago. And then our previous company became AMD's internal FPGA cloud where they would send us all of their latest and greatest FPGAs and we get them debugged and working for them in the cloud. And it's how we forged the relationship that we had with them early on. And we saw the path. We saw the roadmap that AMD had laid out with their foundation being on open source and open standards. And when they announced their GPU offering and there was these supply constraints happening back in 2022, 2023 when AI took off, it made sense for us to go all in on AMD and to be the first and best to deploy their chips at scale. And so at the very beginning, it seemed like a very unintuitive thing to do. People were like, why? Everybody loves NVIDIA. AMD seems to be really far behind, but since we had already cut our teeth on FPGA, solving one of the harder problems first, it made supporting them on scaling out their GPUs a sensible answer for us.
Gemma Allen
>> Well, AMD is certainly having a moment. I think the market loves AMD. People love Lisa.
Jeff Tatarchuk
>> We love Lisa.
Gemma Allen
>> The company on the up, I love to hear that. I love a female CEO overcoming the infra space. Some folks say though the chips are great, the software isn't there. CUDA still has a massive moat opportunity for NVIDIA and ROCm is somewhat of a gaining traction, but not quite where it could be from the perspective of like loyal hearts and minds, right?
Jeff Tatarchuk
>> Yeah.
Gemma Allen
>> What do you think?
Jeff Tatarchuk
>> Well, there's definitely a large gap to begin with. NVIDIA definitely had the advantage. They've been doing it for 20 years, they've been working on this. Jensen did a great job at putting CUDA on all of the consumer hardware, putting it in all the research labs. And so everybody who had been doing high performance compute already had familiarity with CUDA. And so he had built quite a large ecosystem around this space for some time. And so this is where AMD had to spend some time getting caught up. They had built superior hardware with more memory, better across the board. But when it came to the software, that's where a lot of people were skeptical. So in the very beginning when we launched, it was just showcasing that yes, AMD does work. At the very beginning, there was a Databricks article that came out back in 2023 where they showcased, they did a test going from NVIDIA's CUDA porting it over to AMD's ROCm and showing that it does work out of the box.
Gemma Allen
>> Wow.
Jeff Tatarchuk
>> And so that was one of the early signals that we were able to use to say like, yes, this works. Yes, there are some gaps, but AMD has done an amazing job at closing those gaps. And even more exciting thing that's happening now with Codex and Claude and all of these different, with Claude Code, where we're able to use AI agents to help do the code development necessary to close the gap. And so frankly, we see Jensen shaking in his boots when it comes to that CUDA moat that exists, but there is some things that need to be filled when it comes to the ecosystem. Jensen has done a great job at filling out the ecosystem. And this is where we at TensorWave, this is one of my things that I'm focused on is building out the ecosystem necessary to support AI infrastructure at scale on AMD, which is why we're entirely focused on AMD for that reason. If our loyalty is split across multiple things, it's harder for us to be the best at one.
Gemma Allen
>> And I mean, you're clearly leading. TensorWave is definitely the leading neocloud for AMD right now.
Jeff Tatarchuk
>> That's right.
Gemma Allen
>> You guys have some exciting news, I believe. The market sees the opportunity here. Fill us in.
Jeff Tatarchuk
>> Yeah. We just closed our series B. We raised $350 million led by AMD and Magnetar. And so it's just a milestone in the progress that we've made over the last two and a half years, and it's just one small step to even greater things that we see to come.
Gemma Allen
>> Wow. Okay. So let's talk about where that money's going to go.
Jeff Tatarchuk
>> Yeah.
Gemma Allen
>> So clearly there was a big opportunity here. The TAM is huge. Some folks say, you know what? We were at a point in the market where people just needed GPUs. It almost didn't matter where they come from, how they come in. Even spend has been in a state of flux as we well know. But there is definitely kind of segments of opportunities here, I am sure.
Jeff Tatarchuk
>> That's right.
Gemma Allen
>> How are you guys thinking about it? What has been the growth to now? Who has been the typical profile of customer? What does that look like and what is the new TAM that maybe you guys haven't even thought about yet that you will?
Jeff Tatarchuk
>> Yeah, great question. The focus has been on the neo labs that need access to compute for training and for inference. They're all sophisticated enough to jump in, grab the compute, and take advantage of it at its full scale. And so that's been our primary focus, on the top neo labs of the world and getting them access to compute, especially those that are willing and excited about being early adopters. What we heard early on is they were tired of giving all of their margin to Jensen and they needed something to help bring competition into the market, and so they knew that we were the place that can help them deploy it at scale. And since seeing where AMD announced these deals with both OpenAI to do six gigawatts worth of capacity and Meta to do six gigawatts worth of capacity, it really sent a signal to the market where it switched us from headwinds where people were still skeptical of AMD to now tailwinds where people are banging down the door. We can't get enough capacity-
Gemma Allen
>> Wow....
Jeff Tatarchuk
>> to support the demand that's currently out there. And so yeah, to answer your question around what we're doing with the capital that we are getting, it's frankly locking up data center capacity. That's currently one of the biggest bottlenecks, being able to get access to as much power and data center capacity as we can because the GPUs are available. But can you get data centers and powers up fast enough to support and plug those things in to give people access to them? So if we had them up and ready and available today, people would take them to the tune of 5, 10, to hundreds of thousands of GPUs. And so locking up as much data center capacity as we can for 2026 and 2027 and obviously growing out our team to be able to support that infrastructure, and, yeah, to continue to build things out. And so our first initial focus was on the neo labs, and we do see now more and more demand happening on the enterprise side where companies are ... I feel like the enterprise was very AI curious at first, not sure how to engage with whether they're engaging with the top two just APIs or they're going to build out their own infrastructure, and now we're seeing more openness to enterprises getting access to compute, which is one of the things that we're able to help with as well.
Gemma Allen
>> So staying on the topic of enterprise for a second and going back to something you said a few minutes ago. You said sometimes folks are trying to discover what the best solution solve is for their specific need.
Jeff Tatarchuk
>> That's right.
Gemma Allen
>> What workload needs what, and AI can actually help with that. You guys are building, I guess, out your own systems and process to help create an opportunity there. What is the other side of that coin from the perspective of enterprise? Are we seeing a lot of self-managed environments, folks taking a lot of this work back captive?
Jeff Tatarchuk
>> Yeah.
Gemma Allen
>> Where are the kind of substrates happening or consolidating?
Jeff Tatarchuk
>> Well, at first it seemed like they were very skeptical and now that there's options in the market, they're a lot more excited. Before, they didn't want to be tied to just one vendor. And so if they got one vendor, they were stuck with using that one vendor with one stack. Or if they even bought the other vendor, they were stuck with using that one vendor and one stack. But one of the things that we've worked on at TensorWave is we built an open source training and inference stack called ScalarLM with our team, Greg Diamos, who was actually one of the original team at NVIDIA who built CUDA has now created a GPU agnostic training and inference stack that allows you to switch from NVIDIA to AMD at scale at the enterprise level. So we've made it as easy as possible. We've taken the complexity out. GPUs are finicky. GPUs are even more finicky at scale and switching from one SKU to another brand is even more finicky and our job at TensorWave has been to make it as easy as possible. And so we see enterprises now a lot more receptive and interested at deploying their own compute and working with neoclouds to scale out their infrastructure more so than ever before because the tools are now available that weren't there before.
Gemma Allen
>> It's interesting because I think in the neocloud space of the broad conversation, there is definitely somewhat of a narrative around are these real estate plays or are these true technology plays, right?
Jeff Tatarchuk
>> Yeah.
Gemma Allen
>> And I guess the software layer and that can added value for customers of discoverability, of managing solutions, not going to short term solves is huge.
Jeff Tatarchuk
>> That's right.
Gemma Allen
>> People are like, "Are we going to see a mass convergence at the neo cloud level broadly?" We're at GTC this year and I heard someone say, "There's a new neocloud every day." As somebody who's actively competing in that space and really making a significant mark, how do you think about those narratives? How would you stay ahead of them?
Jeff Tatarchuk
>> Yeah, I think about it as well, especially on, we know on the NVIDIA side, there's 150 plus neoclouds. And on the AMD side, those that support AMD, less than 10. And so I can't imagine swimming in that pool trying to get access to allocation when I don't even know if there's 100 plus neo labs that can take that much compute. And so I do think with the competition being as hot as it is, there will be consolidation. I do think there will be clouds that aren't going to make it. They're not able to keep up with their financing terms and pay their bills because of the data center spend that they've been able to spend or keep up with. Maybe if a lab isn't able to pay their bills and they can't find somebody else to take on that bill. There's going to be a lot of labs going to be in trouble. There will be a lot of consolidation. But I do believe that those who will rise to the top will be those that have solved the problem to make it as easy as possible for the end user, and that's what our goal is at TensorWave is to make it as easy for the end user to use and consume compute for whatever use case they might have.
Gemma Allen
>> Well, that's what everyone wants, right?
Jeff Tatarchuk
>> That's right.
Gemma Allen
>> So in terms of the broad economics of this, I presume these are CapEx heavy. There's a lot to figure out here. What is the broader plan? Are you guys planning to grow more, scale more? What does the footprint look like five years from now?
Jeff Tatarchuk
>> Yeah. Well, we definitely, we're not stopping here. We're going to continue to grow scale. We're already working on our next round, which will be significant. And yeah, we need as much capital, we need as much resources to scale out, to support the demand that's currently there. And yeah, our goal is to make access to compute as easy as possible and as simple as like switching on your utility or switching on your lights, paying your utility bill where we can abstract away all the complexity and give people access to this new tool that has never really been available. It's kind of the wild west frontier. And we're excited to be at the forefront of all the different pieces, whether it's policy, working with communities, showing how people are integrating this at the application layer, how is this going to impact jobs, and being at the foundation of it all with building out the infrastructure is something that I'm particularly excited about, especially there's a lot of kind of fear around the data center AI space. I was seeing some studies and people are talking about the perspective of data centers in China where people are thinking about how they're excited about AI. They're excited about the impact that AI can make on their daily lives. But in America, North America, we have this fear of AI is going to take our jobs and all these data centers are going to suck up all of our water and they're going to shoot our power bills through the roof and we're going to hear this buzzing going throughout our small communities. When in reality, I do think there's going to be a huge benefit beyond what we could even understand to what AI is going to bring to the world. And it's exciting to me to be at the foundation of it all.
Gemma Allen
>> I mean, there's definitely some sociopolitical challenge. There's also a big energy challenge though, right?
Jeff Tatarchuk
>> Yeah, that's right.
Gemma Allen
>> Some folks think, well, once the chip shortage ends, which maybe eventually will, things, again, market dynamics going to shift fundamentally all over again, right?
Jeff Tatarchuk
>> That's right.
Gemma Allen
>> What are your thoughts on the energy challenge? What's happening through front of line here? What are you seeing? What's the ultimate bottleneck?
Jeff Tatarchuk
>> It is power. It's getting access to as much power as we possibly can and getting it up and available as fast as possible. Yeah, that's the challenge across the board. And so being able to stand up data centers, get access to power, and plug in these GPUs is really the current biggest bottleneck. And yeah.
Gemma Allen
>> Let's go off script for a second. So space, nuclear, what do you think the reality ... Where do you think pragmatism comes into play here versus kind of almost sci-fi like dreams?
Jeff Tatarchuk
>> Yeah. I mean, it is exciting to hear about data centers in space. I see how finicky GPUs and how hard it is to maintain data centers here on earth, and so being able to increase the reliability of data centers here on earth and service those data centers here on earth without the need of a lot of humans involved in that process I think is going to take more time than people think, but it's exciting. It's exciting to see how robots are getting involved and making it easier. And the idea of doing it makes a lot of sense, being able to get access to the resources and power that you can utilize up in space makes sense, but pragmatically-
Gemma Allen
>> How we cool them.
Jeff Tatarchuk
>> Yeah. How can you take care of them? When a GPU goes down, a server goes down, that's going to be the biggest problem that people are going to have to solve along the way. But I'm optimistic. I do think we dream big. We create solutions to some of these problems that we have here on earth. There's great possibility around the corner. And yes, we do need nuclear. We need as many nuclear power plants up as fast as possible. So I'm excited about the prospects around that as well.
Gemma Allen
>> So you mentioned wild west earlier, you guys are based in Vegas.
Jeff Tatarchuk
>> Yeah.
Gemma Allen
>> Interesting choice. We're always in Vegas. John Furrier's in Vegas right now. In fact, he just flew out this morning.
Jeff Tatarchuk
>> Nice.
Gemma Allen
>> Talk a little bit about the business community there and the opportunity, I guess, because it's somewhat of an untapped tech zone. We just know it from a conference perspective, but I'm sure there's plenty to say about it that maybe doesn't get said enough. So fill us in.
Jeff Tatarchuk
>> That's right. So I moved to Las Vegas in 2020 during COVID. I was living in Southern California at the time and my other company was based out in Wyoming. So it made a lot of sense for me to move to Nevada to still be able to reap the same tax advantages that I had being in Wyoming, and Nevada has been very friendly to us. The state of Nevada is an investor. So there was a fund, FundNV invested in us and the state of Nevada matched the initial investment. So they rolled out the red carpet to make sure we had the resources that we needed to grow and scale. There's a lot of great talent in Las Vegas. There's a lot of people who are willing to move from Silicon Valley or the Bay Area to Las Vegas because of how convenient it is. The same time it takes you to drive from San Jose to San Francisco, you can hop on a flight from Las Vegas to any of those destinations and be there a lot faster and not have to sit in traffic. And so there's a lot of great benefits. Taxes being one of them. But I mean, with the entertainment, with food and the growing tech ecosystem, we see a great advantage for more companies to start moving there. We have a lot of companies that are and other founders in tech that are making the move out there. And so we do a lot to help pour into the ecosystem. And yeah, it's more than just a conference town. And it is convenient. All of our customers will come through town at some point and so it makes sense. But yeah, we're excited about being in Vegas. And we're not just going to only be in Vegas. We're going to expand to other areas as well, but it's great. Vegas and Nevada have been great to us.
Gemma Allen
>> Well, you certainly have a captive audience in terms of the foot traffic, right?
Jeff Tatarchuk
>> That's right.
Gemma Allen
>> So that's wonderful. Okay. So Jeff, last question and I think perhaps you answered this already, but maybe give me a little bit more in terms of what does the next year look like for you and the team? Like what's ahead from here? You've had this big raise. You said you're raising a C. What are the top five or three to five priorities as you look outward?
Jeff Tatarchuk
>> Expand our team, buy more GPUs, lock up more data centers. And we're doubling our capacity throughout the rest of this year and we will be going into gigawatts into 2027, and so the faster we can move to make that happen, the better.
Gemma Allen
>> Wow, we love it. Well, Jeff, thanks so much for joining us on theCUBE.
Jeff Tatarchuk
>> Yeah, thank you, Gemma.
Gemma Allen
>> I'm Gemma Allen here at theCUBE Studio at the NYSE. This is AI Factories, one of our programs with NYSE Wired. Thanks for watching.
John Furrier
Gemma Allen