Transcript: Inside the Battle for AI Cloud Dominance — Why Cloud Builders like TensorWave are Rethinking NVIDIA’s Monopoly | Jeff Tatarchuk, Co-Founder of TensorWave

Host: This particular customer said, you only need CUDA if you don't understand, you know, how to really do low level development on the GPU. CUDA is no longer a moat.

It started with, you know, the supply shortage and then the frustration that the customers were having around the pricing and the profit margin that uh, Jensen and Nvidia were demanding.

And we kept having people coming to us saying, hey, we're tired of giving our profit and margin to Jensen. We need another solution. You know, one of my friends at Nvidia said, like, they have 150, you know, neo clouds.

I don't even feel like there are 150 AI labs, you know, in this ecosystem. And so how, how is a ecosystem with 150 neo clouds actually able to to stay alive?

We think like in order to hit the current demand over the next few years, we need like 30 gigawatts.

And if you think like it takes one nuclear power plant to power one gigawatt of of power and how fast are we putting up, you know, nuclear power plants today, it's just not fast enough.

Host: Hello everyone. I welcome to Startup Project. Um, today on the show, we have Jeff, uh, the CEO co-founder of Tensor Wave. Um, with, uh, so much of AI compute spending.

I think we're at an interesting time where uh first time ever we're seeing a bunch of new neo clouds evolve uh around different strategies, uh, in order to provide more AI compute that we need. Uh, and Tensor Wave is one such company.

Um, so today, I'm going to talk about um, you know, how Tensor Wave is solving the AI compute problem. Uh, why they're exclusively working with AMD, uh, and what it takes to build, uh, in the modern day data centers and a lot more.

Uh, with that, uh, Jeff, welcome to the show.

Guest: Fanra, it's good to be here, man. Thanks for having me.

Host: Yeah, so, uh, you know, uh a couple of years back, if you asked me, like, you know, there were smaller cloud companies, you know, CloudFlare was, uh, CloudFlare was a smaller cloud company.

They were trying to be, you know, a larger cloud company. Um, there's Digital Ocean, which is even, you know, much smaller cloud company, but I always wondered, why are more cloud companies not coming out?

Uh, and then once like the whole, uh, AI compute, uh, spending cycle started, then we've seen a lot of neo cloud companies, uh, doing different things.

Some are focused on exclusively bringing more infrastructure online, some are bringing, uh, you know, some companies like Astec are bringing whole new architectures.

Uh, I think every time we sort of have a big wave of new innovation, there's always a new architecture innovation that comes up and then, uh, we see companies built around that. Uh, and I think you are one such company as well.

Uh, so, you know, for those folks who have never heard about Tensor Wave, um, can you give a little bit of introduction of, you know, what is Tensor Wave? What do you guys do? Uh, and how did the whole idea of Tensor Wave come together?

Guest: Yeah. Um, so Tensor Wave is, simply put, a neo cloud that only deploys AMD GPUs. And it started out, um, so my other co-founder, Derek, and I initially had launched a FPGA cloud business about eight years ago.

And, um, we were solving one of the harder problems first. FPGAs are usually more of an edge, you know, chip that's used for really low latency use cases. And we decided to make them available, um, at scale in the cloud.

And we were one of the largest FPGA clouds previously. And so Host: VMXL?

Guest: Yeah, yeah. Yeah, so VMXL, um, was the company. And then we, uh, were working with a lot of the chip providers, Zylinx at the time and, uh, Intel who had acquired Altera and then recently kind of decoupled Altera.

I want to say realized how, uh, how challenging FPGAs can be. It's, uh, we started with that and so, uh, we were mostly focused on, you know, video transcoding and weather modeling and, and not really focused on AI at the time.

But it taught us what we needed to know to deploy, uh, cloud infrastructure, set up data center infrastructure, create a really, um, you know, um, easy experience as possible, uh, for the end user.

And so we were doing that for some time and we realized that, you know, as soon as the market shifted all into AI after ChatGPT was launched, um, all the attention shifted from any resources that were going into FPGAs into GPUs.

And, you know, in 2022 and 2023, um, there was huge demand for GPUs. Nvidia had supply shortages. You couldn't get access to anything.

And, uh, we had a friend come up to us asking, saying, you know, knowing that we were doing cloud, you know, we were kind of AI adjacent, asking if we could help him and his portfolio companies get access to some GPUs.

And, um, we said, you know, we're not really focused on, um, GPUs right now, but we do work with and we'd work primarily with Zylinx and Zylinx had gotten acquired by AMD about four plus years ago at the time.

And then our company, um, we were working with AMD. We became kind of their support internally with our FPGA cloud.

They would send us all of their latest and greatest FPGAs and we'd get them, uh, you know, we'd code develop and work and get them debugged for them in the cloud. And so we had already kind of built, um, and were embedded into, uh, AMD.

And then when they announced their GPU offering, uh, it made sense for us to make the shift to go all in and deploying their GPUs at scale. And so when our friend, my VC buddy, came up to us saying, hey, can you help us get access to some GPUs?

We said, would your portfolio companies consider going after AMD? And, uh, he said something that, you know, I never forgot and he says, if it works, um, we will definitely, you know, encourage them to use AMD. And the light bulb went off.

The next day, uh, frankly, we created Tensor Wave and, um, called AMD and said, hey, we need a significant allocation of GPUs and we're going to go all in to be the first and best to deploy your GPUs at scale.

And we were announced December 2023 as one of, uh, AMD's official launch partners for their, uh, first kind of data center chip, the MI 300X. And we've been off to the races ever since.

And so it started with, you know, the supply shortage and then the frustration that the customers were having around, um, you know, the the pricing and the profit margin that, uh, Jensen and Video were demanding.

And we kept having people coming to us saying, hey, we're tired of giving our, you know, profit and margin to Jensen. We need another solution.

And so helping bring optionality to the market and we, um, with, uh, our experience with AMD and, and seeing, you know, Lisa Su's vision and roadmap on where they plan on going, we knew that AMD was the next best solution to, uh, to go all in and, um, be, be that for them.

And so we are the premier AMD support. If you are considering AMD, um, we're your best solution to, uh, to make it as easy as possible deploy their ships at scale. So that's how it got started and, uh, we've been off to the races ever since.

Host: Um talk, talk a little bit about, you know, the now the front of a FPGA cloud, like, what was that business like? Uh, who were your customers? Uh, because I think that will help understand us, like, how did that help in transitioning towards building Tensor Wave?

Guest: Yeah, so, I mean with, with FPGAs, um, there's a lot of, you know, significant challenges around it. Um, there aren't compilers written or if they are, they're not very good and everybody was trying to solve that problem.

And one of the things with an FPGA is, yes, theoretically, it can be more flexible and more efficient and, and, you know, performant than a GPU, but the amount of complexity it takes to squeeze out that extra performance, um, we found out very quickly wasn't worth the squeeze.

And so even if we could get an extra 10 to 15% performance boost, um, the amount of extra work you had to do wasn't worth it.

So that's one of the lessons we learned very early on as we were, um, you know, one of very few companies working, you know, in this space and working with a small group of trying to to build Host: cloud for FPGAs, or?

Guest: There was a couple, there was a couple but not very many.

Um, but it was more so the ecosystem that, uh, we were working with, working with Zylinx and Altera and there were a few other, um, researchers that were also, um, you know, trying to solve this problem as well.

And so everybody was kind of leaning in. I'll never forget being at Intel, um, and Intel's like, you know, asking us the same question, like, who, who are your customers? Like who is actually, you know, buying these things?

And, uh, we did have, um, you know, a number of customers that, um, you know, that were building various products with like whether I said like weather weather modeling products and video transcoding products, a lot of simulation, um, products and so it, uh, it was, it was rough because there wasn't a lot of tools available and then the developers needed, you know, because FPGAs in order to re, you know, for those that don't know, FPGA is a field programmable gate array, um, that is, you know, very, very fast.

And, um, you can, you know, reprogram an FPGA, but before you had to be an electrical engineer to go in and, and reprogram, you know, and Verilog or whatever.

Um, but now, like the engineers that can create, um, you know, reprogram an FPGA and create the bit streams necessary to do it. They're very, very expensive, um, and specialized.

And so I ran into a lot of just challenges early on as we were, you know, putting all of this together, but, you know, it it taught us everything we needed to know as we were, you know, deploying the the, uh, you know, data center infrastructure to support this.

Um, you know, building our relationships with our OEMs and, um, you know, dealing with, uh, building a a cloud platform on top to make sure that it's as easy as possible.

So our initial goal was to create a cloud platform that abstracted away what was underneath the hood completely. And so they just could say, this is what we need, um, and we can spin it up and give them access to it.

And so, um, working on some of those problems, um, early on is what gave us the experience to it and it was funny when people were, you know, when we first went into doing, uh, the GPUs with AMD, everybody's like, you know, these are going to be very challenging, you know, the the first, first batch of AMD GPUs are going to be a lot of problems.

And we're like, man, you don't know the kind of problems we were dealing with previously. Like the problems we're dealing with on the on the GPU side is nothing in comparison.

So, um, yeah, it so building out the infrastructure, um, supporting customers and then making a platform, um, is easy for the customers to use as possible and where they very complex chip was our our first kind of strike at this.

Host: And for the FPGA cloud, uh, is it similar to building a regular cloud where you're working with like different data center vendors? Who who are your partners in that network?

Guest: Um, yeah, we, so we actually, we started off co-located when we first built. We just needed access to power and we needed to do it quickly. So we actually got set up in, in Cheyenne, Wyoming.

Host: co-located and this is for the audience, like, there's already working data centers and you basically, you know, take some space in a working data center and put your stuff or infrastructure there.

Guest: That's right. That's right. And so, yeah, they'd already built a data center, they were managing it.

All we had to do is bring our servers, deploy them and, um, yeah, they had a team working in there around the clock that would manage and maintain it. If we had any issues, they would go in and and and help us with it.

And so we started with that and then we were able to actually build out our own, um, data center.

And, uh, it's where we're able to get, uh, really creative and, and work on some some new and exciting efficiencies that that made it, um, even better for us to do it ourselves where we're able to to save a lot more money and and learn along the way.

So we, we learned the the whole stack from, you know, acquiring the power to building out the data center infrastructure, to managing and maintaining the servers, to building out the the software and cloud stack to support it, um, with a very small tiger team at the time.

So, um, yeah, I I still have PTSD from those days, but it's, uh, it was it was a great, a great experience and, uh, learned a lot of lessons about, you know, when you start a company, um, you know, there was a lot of incentives that Cheyenne, Wyoming had given us to to move out there.

And, you know, they'd given us some grants and different things, but and all of that was contingent upon us, you know, hiring, um, people for, you know, to fulfill those grants.

And if you're in a town or a place where people don't want to live, it's very hard to recruit.

And so we were trying to recruit from the coasts and in a town that didn't have access to the infrastructure, housing with amenities or or resources, um, it became, it became a challenge.

And so we thought, hey, like this is a place where we can, you know, be a big fish in a small pot, we can grow, we can work with the city and the government and make all these different moves and they're going to give us money.

But, you know, sometimes the giving, you know, when people give you money, it comes at a cost. And so you have to count that cost when you're starting your company and, um, that was one of the things we definitely learned, um, early on.

Host: See, in some sense, you were, uh, right time, right place with the right experience, deeply embedded, uh, into the, you know, um, infrastructure space and, you know, how to build data centers.

And you basically are in some ways position to start Tensor Wave if you'll if you look back at it. Um, so let's start about Tensor Wave.

So you're now building, uh, data centers with AMD GPU clusters that offer AMD GPU clusters and top of that, uh, you offer a cloud or a platform on which, uh, I can just deploy my own, uh, compute clusters, uh, to inference, training, uh, all that stuff.

Is that the right way to describe Tensor Wave?

Guest: That's right. So we, yeah, we, we will do the whole thing from identifying power, building the data center, retrofitting older data centers to support these, uh, GPUs and then, yeah, build out the whole stack for customers that need access to compute for both training and, uh, inference in production.

Host: What what what does, uh, you know, building data center today look like? Like how long does it take? You know, what are the challenges that we're seeing?

How much of the, um, you know, challenges that you see in the news about, you know, getting access to power, getting access, like how quickly people want to get these things running? How much of that is high process?

How much of that, uh, uh, that you see on ground?

Guest: No, it's a real problem. Um, power is the commodity that is the bottleneck right now. Uh, they can, they can, you know, make the chips. Uh, the supply chain is is is fine for now.

And, um, but there isn't currently enough developed critical power, uh, to deploy the GPUs.

And so there are a lot of, um, companies out there that have access to power that are trying to sell it, but, you know, you need to be able to get your substations and everything thing built and set up and the water and all the other pieces necessary in order to, uh, to build your data center on top and those timelines are a lot longer.

Um, take a lot longer than than people usually anticipate. And so, yes, like getting access to power is a major concern. Um, there isn't enough power available.

I think like in order to hit the current demand over the next few years, we need like 30 gigawatts.

And if you think like a it takes one nuclear power plant to power one gigawatt of of power and, you know, how fast are we putting up, uh, you know, nuclear power plants today. It's just not fast enough.

And so, um, one of the things that we focus on is, you know, we have a pipeline of um, opportunities for us to build and they could be, you know, uh, completely open, you know, greenfield opportunities that allows us to, you know, build on top and do all those things.

But for us, we have to optimize for speed. We have to be able to build, um, quickly and deploy quickly. We have customers that that want it now. They said optimized for speed.

They're willing to pay a premium, uh, for being able to deploy fast because, you know, this is, this is a race. Um, and they are the big AI labs need access to as much compute as they possibly can.

And, yeah, as I mentioned, the the biggest bottleneck is getting access to power.

And then, you know, we're seeing today, so you can, there's a lot of stranded power out in the middle of Texas and we have some of these other larger projects that we're seeing, um, that, yes, you can get access to power, you can build these data centers, but getting the people out to those data centers to act to build and support the data centers is the challenge because it takes thousands of plumbers and electricians and, you know, everybody else to support and build it.

So if it's out in the middle of nowhere, you know, you almost have to build your own little town around it to support it.

So that comes with its own individual challenges and there are some towns, as we're seeing, uh, that are running into permitting issues or having the towns or cities, you know, protest, having data centers in their town.

They don't want, you know, AI taking over their their jobs.

And so, um, there's a lot of moving parts that have to go into identifying data centers and making sure that the data centers and power and builds all happen within the proper timelines, um, for the end user.

Like that's, that's, you know, really what's most important to them. So, yeah, there's a lot of moving parts and then it's the financing those things that have to go into it as well. So. Host: Yeah, that was my next question.

Like how I mean, you guys raised, you know, 100 million in series A, uh, you know, looks like a lot of amount, but when you're thinking about constructing a lot of data centers, then, you know, it's it still looks like a small amount considering.

I think big tech is spending about 600 billion in 2026, uh, I'm Cafel expenditure of building data centers.

Uh, so how's how and you guys have built I'm assuming, I think two data centers already, uh that are up and running or, is the second one already up and running? Uh, but you have two data centers up and running, which is, you know, quite a costful.

Like how does the financing today work and how are you sharing uh, you know, costs?

Guest: Yeah. It is a lot. It is a lot of money when we got the $100 million it just comes in the bank and right out the the bank.

I always told people that we're still living on ramen even though we've raised, you know, uh mid nine figures currently, uh, for our, um, data center, you know, GPU deployments.

And so luckily we do have, you know, great partners when it comes to financing, um, with our investors, our lead investor is Magnetar. Magnetar, uh, is the fund that took Coreweave from seed to IPO and, uh, backed their financing.

And then, uh, they're able to help bring in some of the other larger banks, uh, to also help, uh, with the financing as well. And so, um, cost of capital is is an important factor in this. Uh, you know, we are a new company. We're only two years old.

And so, um, you know, being able to, number one, like get access to customers that have a great credit rating and the better the credit rating, the better our cost of capital is, um, that then we're able to, you know, lower the price of our deployment for the, uh, for the end user.

And so, um, we are in a great position with funding, uh, to do everything we need, but again, it still comes down to, you know, lining up, playing air traffic control to line up your power with your customer, with the financing and making sure all of those things land and can be, uh, built and deployed at the at the time necessary and coordinate all of that is is really the the the challenge behind it.

And, you know, one of the things you mentioned early on is that, yes, there are a lot of people attemping to do cloud, you know, before it was just the hyperscalers and then a couple of others.

And then all of a sudden, you know, one of my friends at Nvidia said, like, they have 150, you know, neo clouds, uh, you know, that are I don't even think that there are 150 AI labs, you know, in this ecosystem.

And so how, how is a, you know, ecosystem with 150 neo clouds actually able to to stay alive?

And I do think some people saw, you know, the the the profit opportunity to make money where people think all you need is a data center, some servers with some GPUs in them, you plug them in and you rent them back to Open AI and you're, you're fine when that's there's a lot more that goes into it that, um, a lot of people don't take into consideration as they are, as they're doing this.

And so I think some people just thought it as a financial arbitrage and, uh, I think, um, those that saw it as such will find out the hard way that there's a lot more that goes into it.

Host: I didn't I think the difference would be, uh, companies who on paper can just say, you know, go look at a company like Equinix or, uh, you know, uh Equinix, uh, that, you know, builds data centers and just partner with someone like that.

Um, and then if you can arrange financing and buy GPU clusters from Nvidia, then put that in the same data center and create a cloud on top of it and rent it.

Like that's that seems like a a doable version given if you're good at raising certain amounts of capital on paper.

Uh, I think there are a lot of companies like that, but I I would not consider them as, you know, clouds because, uh, there is still a technical challenge of bringing a new type of GPU into the data center and building the racks, the cooling systems, uh, you know, all the compilers, you know, uh NVDAsCUDA, now you have to make sure that, uh, if I'm a large AI lab, I'm running my models on both Nvidia and AMD, so that means there's a technical challenge of how do you make that happen for your customer who is running a large cluster and they already have Nvidia already in, now the one easy way to run on your cluster.

So right there's a technical challenge of, you know, building the actual rack and making it available to a customer through your cloud. Uh, I think there if we count that way, I think there are not that many neo clouds.

Uh, but I think the neo cloud market definition is slightly different because on paper a lot of companies can look like neo clouds, which are not actually neo clouds because there's no like technical differentiation or like there isn't there isn't a technical challenge that is being solved, uh, as a company, right?

Guest: No. It's it's all it's spreadsheets and financial arbitrage. And, and if you put all of that together, you can, you know, make it happen, but the rubber meets the road when you actually have to deploy it.

So, GPUs can be finicky and, uh, as you mentioned, like they're they're a lot of, um, there are a lot of challenges at at stake.

So, um, we love it though because we, you know, worked on some of the harder problems first in the FPGA world like seeing some of the opportunities that we're able to to work on with AMD to, uh, to truly bring this to market is really exciting for us.

Host: Uh, so I mean, you obviously talk a little bit about your role and you're chief growth officer. you're trying to grow this thing. Uh, talk a little bit about customers, right?

Uh, one thing is, you know, all the top AI labs want more and more compute. That's pretty obvious.

Um, but whenever I think of training workload, uh, how many this I think that larger market for fine tuning, I think a lot more companies are doing fine tuning with either smaller models or larger models.

But how many companies are, you know, going after towards big clusters and how do you see that demand shift enough?

Guest: Yeah. You're right. I I do think at especially at the enterprise level is where we see fine tuning is the the bigger opportunity.

And, um, there are a lot of companies out there that are working specific with enterprise where they can, you know, take one of the larger models and then fine tune it to the enterprise's specific, you know, use case or needs.

Um, and that being done at scale, I think is is a significant opportunity.

And still, I do feel like the enterprise is still trying to navigate the AI, you know, landscape and how they're going to integrate and implement it into, uh, what they're doing.

And so, uh, I do see a lot of pent up demand on the other side of of enterprise that hasn't fully broken yet, but there are a lot of people trying to solve that. Um, on the other front, yes, like, I mean, you, you asked how many?

I again, I don't think there's more than 100 that are doing significant like hero training runs that need more than like a thousand GPU clusters for years at a time.

I could be wrong, maybe there's a lot more hiding underneath the bushes, but I I cannot imagine they're being more than that.

And, you know, the primary focus is on being able to support the top 10, you know, hyperscalers and AI labs that need access to compute. Um, and so, for both training and for inference.

Now, with AMD, AMD started out optimizing their GPUs for inference. And, um, that was their their first use case and and making it optimized for that was was really important, um, to hit the ground running.

While Nvidia is focused more on training, uh, AMD was able to capture a lot of the inference market. You know, Meta announced that they used their, uh, you know, they host their Lama models, uh, on AMD.

Um, yeah, obviously Azure has AMD in their platform. OpenAI just announced, you know, a few months back that they're doing a significant deployment. Um, they said 6 gigawatts of AMD will be deployed in the future.

And so, um, we're seeing a lot more of the AI labs, um, interested and focused and they're still their primary focus is on inference.

But, as of recently, we are seeing more like if you look at our 8,000 GPU cluster that we deployed of the MI 325s, uh, it was built as a training cluster.

And so Lisa Su had given us the, um, the challenge and the mandate to create the most performance AMD training cluster and and we did in record time. And so, um, we've been able to to focus on that.

And so from from my perspective as as the chief growth officer, as I like to joke, the uh chief GPU officer, um, as I'm, you know, meeting with with the labs that need access to compute, um, you know, they have a lot of people banging down their door.

The amount of customers I have that are, you know, that announced that they've raised 50 to over $100 million.

So like, man, we're getting so many stake dinners from all of these AI clouds and getting flown on private jets all over the place and they have a lot to choose from when it comes to the Nvidia world.

Um, but if they want to make a bet on helping to diversify and democratize, you know, what they're using, they have to look for an AMD solution and, you know, we are the best with that.

And when we first started, I mean, it was it was the challenge of people say, oh, I had no idea AMD did that.

And so it was coming up with as many different creative ways to get the attention of the market to make sure that they knew that AMD, yes, has a GPU.

Yes, it, you know, can do inference, yes, it also can do training and and giving people options, uh, to try it.

And so I'll never forget when we first started the company, um, like three months in we started in December and then GTC in 2024 was in March.

And, um, we had just raised our first like 40 million bucks and we decided to get a, uh, one of those LED trucks and we would surround the San Jose Convention Center, uh, during GTC and we had, you know, a comparison of the H100 and the MI300 on the back and showcasing all the different specs and showing that the MI300X definitely is better.

And at the end, there was this robot that, uh, that had a red pill in its hand and it went like this to the audi, you know, to to the thing. And everybody loved it. It blew people's minds and knew exactly what we were trying to to communicate.

Um, and people were taking selfies in front of it and everybody, uh, a lot of the big companies that were at GTC at the time were like, yeah, you guys were all in our slack channel as a way of like getting the attention.

So that people just needed to know, right?

It was like kind of a grassroots gorilla move, you know, gorilla marketing, um, thing that needed to be done to just kind of shake people like, oh, there's a new player that is viable on the market and, um, we should consider it.

And so we got a ton of leads coming in and how we took like a lot of people are just curious, you know, initially, uh, we had a lot of people coming who had worked for all the national labs, which do, you know, a lot of those supercomputers are built using AMD.

And so they were already familiar with AMD.

Uh, it's kind of our initial audience that was coming over and a lot of we, we have some clients that have never bought an Nvidia GPU, like intentionally, just because they, they love the, uh, they love AMD, they love what they stand for and they're committed to it.

And so we still see that kind of transfer over from the consumer side and to those that are now developing in the AI world who want only AMD.

And then, it's it's then it's, you know, we still have to go out and prove ourselves that we, you know, that it does work, it's just as good if not more efficient than an Nvidia GPU.

And, um, and we can show that, you know, we are the best at supporting them, um, at that process at scale. So we typically will bring in a customer, we'll analyze, you know, what are they doing? What's their use case?

What frameworks are they building on?

And then we will then go and validate to make sure that everything that they're using our internal AML team will validate what they're doing and then make sure that there aren't any, you know, gotchas or bugs or any issues that they're going to have. uh run it and then once we've validated it and sometimes there are issues that we have to go back to AMD, we have to go back to, uh, one of the frameworks and and fix something, um, before we let the, um, the prospect on.

And then once we have all that ironed out, um, we then get them on a POC to let them, you know, take a test drive, sit by the wheel for themselves and and show them that, yes, it does truly work and it's just as efficient and better than, um, than Nvidia in a lot of ways.

So, um, Host: What are the key architecture advantages people get by using AMD? Like is that a differentiation, um, that AMD does better than Nvidia?

Guest: Yeah, and so like right now, AMD still has an advantage and landed with the advantage of having more more memory. They have more VRAM significantly than, um, the, uh, than Nvidia does.

And so if you need to host some of the larger models, um, like for instance, if you're hosting like a 70B model, um, you need, you know, two GPUs in order to do it on an H100 that has only 80 gigabytes of VRAM versus the MI 300X that has 192 gigabytes of VRAM.

So you're able to host more without having to kind of split it up over across more GPUs, um, effectively. And then, um, AMD has their, uh, chiplet architecture that, um, that Lisa bet on early on.

And as a result of the, uh, chiplet architecture, um, yeah, it's it's starting to pay off showing that now they can they can break a chip down into, um, you know, more pieces and they can take advantage of a chip even more so than what you can do on Nvidia.

Host: Uh, the one of the biggest advantage Nvidia has is obviously CUDA provides libraries, compilers and debugging tools for GPUs and sort of like their proprietary system that will run on GPUs, I sorry, Nvidia GPUs designed for Nvidia GPUs and that has been long argued as one of the biggest sort of Nvidia moat out there.

Uh, how, how does that affect, you know, customers who are trying to adopt AMD? Like their development team might be already familiar with, you know, building on top of Nvidia.

---