Nick / Compute Labs

Compute Labs creates AI-Fi ecosystem for decentralized computing

InterviewFebruary 24, 202520:19

In this episode

Ashton Addison speaks with Nick, Co-Founder of Compute Labs, about creating the AI-Fi ecosystem, an innovative intersection between decentralized finance (DeFi) and artificial intelligence (AI). Nick explains how Compute Labs is helping investors gain direct exposure to the rapidly expanding AI industry by investing in underlying compute resources essential for AI's growth.

They discuss the exponential demand for computational power driven by advancements in AI, highlighting how every new generation of large language models (LLMs) significantly increases hardware requirements. Nick dives into the differences between centralized cloud providers like AWS and CoreWeave versus decentralized physical infrastructure networks (DePIN), explaining the potential benefits and limitations of each in meeting AI’s enormous compute needs.

Key takeaways
  • Compute hardware demand is growing at approximately 30% annually, with projections reaching $420 billion by 2030.
  • Centralized providers like AWS and CoreWeave serve medium to large businesses, while DePIN projects better serve small to medium-sized businesses with specific use cases.
  • Each new generation of large language models requires significantly more computational power than previous versions.
  • AI-Fi refers to infrastructure enabling investors to gain exposure to AI growth through underlying compute resources and investments.
  • DePIN networks using consumer-grade hardware may be more economically viable for developing markets than first-world economies.

Chapters

Read the interview

Compute Labs' Nick: Tokenizing GPUs so investors can own a slice of AI growth

Nick, co-founder of Compute Labs, explains how the company raises capital to buy AI hardware, leases it to data center operators and tokenizes it for investors. The conversation also covers surging compute demand, where compute comes from and how DePIN fits in. It was recorded in February 2025, and the plans described are as stated at that time.

Why compute demand keeps climbing

Ashton: How fast is demand for compute growing, and what is driving it? 00:45

Nick: The need for compute is growing exponentially. Based on several sources, aggregate growth in compute needs is about 30% a year. By 2030, only five years away, demand for compute hardware could reach about 420 billion. You might think there's a limited number of use cases and we'll hit the curve, but it doesn't work that way. Compute demand only grows more and more, because innovation and engineering require more and more hardware. xAI is building out data center infrastructure, and AWS and CoreWeave are scaling their GPU data centers. We're even getting requests from gigantic companies asking us to help them get hundreds of millions of dollars' worth of hardware. We don't talk about it as much as we did six months or a year ago, but it's still growing.

Ashton: Where does all that compute actually come from? 02:33

Nick: There's the centralized side and the decentralized side. On the centralized side you have AWS, Google Cloud and CoreWeave, plus others like Lambda. Most of them own their hardware and provide it to customers. Others, like RunPod, aggregate from different sources, similar to a DePIN but Web2-based. Then the Web3 DePIN projects aggregate from as many sources as they can, sometimes even consumer-grade hardware from retail people like you and me. What matters is which provider has the specific hardware you need. For a really heavy training model, even AWS or GCP may not have everything. A DePIN pulling from varied sources may not sustain the speed and efficiency you need. So customers building heavy AI products often end up with custom-built servers. They go to an operator with a very specific set of requirements, and the operator builds it.

Read the full interview →

Transcript

Read the full transcript 3,193 words, auto-generated and lightly edited

I'm ashon Addison from the Crypto Coin Show and today on blockchain interviews we have Nick co-founder of compute Labs Nick welcome to the show and thank you for taking the time asan thanks for having me glad to be here there's so much to talk about in terms of AI the computational power needed decentralized compute networks and I think people are underestimating

just how much computational power is going to be used in the next couple of years I'm excited to dive into all of that today with you I would love to start out with a little bit on what you and your team been building at compute labs and then we can dive into all those details yeah that sounds like a fun topic I'm happy to be a part of it awesome all right let's kick it off cool

Should we start with what is compute and what's happening with the industry and the growth of it I think so that'd be great so you definitely hit the nail on it the needs for compute is growing exponentially what we know is that there's a aggregated growth rate for compute needs based on several different sources out there about 30% a year and what that means is that in

general by 2030 and only 5 years the demand for compute Hardware is going to be just exponentially growing at that same rate so about like 420 billion by then and what we know is to your point when we think that okay there's probably like a limited amount of use cases there's not going to be like that much demand like we think we probably

like hit the curve there's that much compute needed no it only grows more and more because Innovation and Engineering requires just more and more Hardware I mean we have xai and their coin gigafactories or infrastructure data centers and then we have AWS and core weave scaling their GPU data centers there's just more and more Hardware being used and we're

seeing this every day so for example like we're getting demand from gigantic companies asking us to help them get like hundreds of millions dollars worth of Hardware so yeah the point is that this industry is growing we don't talk about as much as we did maybe six months ago or a year ago but it's still growing definitely I it seems like every new

version of llm that comes out is requiring 10 times more power and computation and I think a lot of people don't really know you know where does that compute power come from does it just come from an Amazon warehouse is it actually coming from decentralized networks maybe you can elaborate a little bit more on Gathering that compute power and then getting it to the

right place yeah that's a really good point where are people getting their compute Hardware from so you're right that there's of course AWS and then there are the deep in projects so you can think of it of course as like one is centralized one is decentralized and un centralized aspect of it we have the big players we have AWS we have gcp or Google cloud services we also have

core weave and then we have smaller players like I don't want to call them small but you know in comparison WS I'm sure we have Lambda I guess there are a ton of cloud marketplaces out there and the way they work is most of the time they have their own Hardware that they own and they provide it to customers that need cupu needs and then there are going to be others who

aggregate kind of like runp pod who they just getting it from different sources similar to a dein project except it's Web Two based and then we have the very web 3 deepen projects which are simply aggregating from as many variable sources as they can sometimes even from consumer grade like retail people just like you and I the difference is

going to be when you have a use case that requires compute which of these providers has more of the specific requirements you have in terms of Hardware so for example if I was going to be building out like a really heavy training model that needs super efficient computational needs I can go to gcp or maybe AWS but even sometimes they don't necessarily have all the

hardware that I need I can go to a deepend project that maybe is pulling from all of these variable sources but again because they're pulling from different sources altogether it might not even mean it'll be enough for me to sustain the speeds and efficiency I need so a lot of times actually what happens is these customers who are building these heavy

AI products end up getting custom built server structure so they'll go to an operator who'll say he like hey I have a very specific set of requirements can you get it for me and then the operator go and does their thing it's just custombuilt servers based on very unique cases it's very interesting and do you think that there is a growing role in

the dpin networks and with decentralized physical infrastructure networks they're not always to do with just Gathering resources for AI you know there's a lot of different use cases but with Gathering compute power these deepin networks it seems like 2025 could be a year for major growth in physical infrastructure networks as compute the

compute power problem seems more prominent more people are aware that AI specifically requires a lot more compute and hey maybe there's potential to tap into phones or home computers or there's a revenue generating opportunity for people do you think that's something that will like increase dramatically in 2025 it's a good question so I think of these different types of compute vendors

or compute providers serving a different customer profile almost when I think of deepin projects I think they're really excellent at serving small businesses small to mediumsized businesses when I think about those other centralized players that we've been talking about I think about them serving U medium to large size businesses and the difference is

simply because of where they're pulling their resources from oftentimes on the deepins depends on what type of deepin project we're talking about so we have Aether hyperbolic and IET these are really awesome deep in projects and like all good friends and they're good for solving the specific use cases that they're built for athers are really good

at solving video game needs hyperbolic is really good at serving Ai inferencing and training needs IET is a good combination of both then we talk about you what kind of customer profile can they help with from small to large when it comes to those aggregated resources from like you know my phone this laptop or your devices those have their needs me

personally I don't know what problems they solve but what I would gather is that they would solve economies a scale and they would probably make certain resources become way cheaper I mean you can get like 409s on an hourly 40 like consumer grade RTX 490s pretty cheap pretty affordably on an hourly basis if that's too expensive for certain businesses or certain you know

people practicing and trying their AI use cases yes these other devices can help but I guess my question is like how much cheaper do you need to get it to maybe it's not built for first world economies or like you know the United States or Canada maybe those would be more effective for Latin American demographic or the subsaharan African market so some different markets

but I don't know if they'd be really effective for like the markets that we're in right now yeah that's a great point and I feel like the ethos in part has always been to help the unbanked or to help the ones that couldn't normally access these kinds of services whether it's just getting a bank account or if it's accessing AWS and they charge

a lot you know for anyone who's run a business they know so I definitely could see hopefully the rest of the world and the developing countries can tap into and be the one to grow these deepin networks so that they can and hopefully there's a revenue generation for the people in those countries that are contributing that would be great yeah I

mean any more business for growing economies is always great and you know I'm specifically have always been interested in Defi and decentralized finance and now ai seems to be merging ever more so this year with blockchain and defi specifically and I've seen this new term that's coined aifi or and also defi which an AI at the end do are is compute Labs

involved with aii working with that at all and what can you tell me about that yeah we started using the term ai5 back in March I won't lie a little bit of a little bit of it is just marketing but effectively what we're trying to do is build the infrastructure allowing investors to get into the growth of AI and AI Investments and

that's why we coined it AI fi because our product is built around making sure that people can get into the investment growth of AI interesting and is that just for investing in AI related projects or does the AI actually help you choose manage any of those kind of functionalities of investing it's a good question so why aii wider usage of that in the name really it's

part marketing and then the other part is simply like making sure that people know that by investing in compute resources they are enabling the growth of the AI industry so it's it's kind of like a commodity approach like invest into the core asset that builds powers and enables AI very interesting and with Gathering the compute power for AI how is compute

labs functioning efficiently in doing this can you talk a little bit more about the business relationship coming in and going out yeah certainly so this is a for me this is the fun part basically what we do is we raise Capital we raise Capital through institutional U traditional Finance but also through retail investment from markets that we can take capital from what we do with

that is we then buy compute Hardware so mostly Enterprise grade Hardware like the b200 and h20s however we also do double in the consumer grade products like 490s and 59s now what we do when we have that Hardware we actually lease it out to operators you can call them data centers or operators or data center operators you can even call them like cloud

service providers the point is we lease our Hardware to them so that they can continue to scale and grow their businesses most of the time they already have a customer lined up or demand lined up to use their Hardware which is why they need more Hardware to serve their customers so that's where we help them that's really interesting and I keep hearing that you know there's

so much demand that's growing for AI that as you said they C they already have a customer that's demanding it they need more is there a shortage of Supply or is there inefficiencies in supply chain in getting compute Hardware to actually start running and bringing that power for computing AI it's a really good question there's the short answer is that there

are inefficiencies in Le time so often times the provider that can get it most efficiently or fastest is going to be the winner to actually get the contract with the customer and the point is that because you know first you make an order to buy hardware for that custom built cluster it's going to take like 6 weeks to get it then you need like four weeks to actually run it test

it make it operational that's a lot of time already for a AI company that needs their thing done tomorrow I mean a lot of Enterprise businesses they think they need something tomorrow and now often times there are going to be projects that know that they're planning for which is great but there's also the opposite another area of inefficiency I

would say though is actually in the energy a lot of data centers simply do not have the energy or power to actually continue scaling or growing sometimes it's legislative so for example governments might actually forbid or prevent a data center from you know using more resources than they can and other times data centers just simply are there aren't enough of them in

demographics to actually put Hardware in a specific area H very interesting and with an example some of the clients that you're working with are are those data centers all in America and what are the potentials of tapping into Alternative forms of energy or maybe mining Bitcoin on the side or renewable energies so that they can be able to run bigger

operations yeah it's a really good question so comp labs in general we work globally we work with as many data center operators as we can whoever needs compete resources whether it's the capital to buy it or the actual Hardware will serve them as long as you know they're not from a saction country that's one and then on to answer I think your later question which

is like using the run time that isn't being utilized or the underutilized U processing power yeah so there's a lot of like algorithms and products out there that can help you get the most or maximize your compute availability and so basically if your customer is locking in 80 or 90% And reserving most of your Computing

availability then you can have like 10 to 20% that will either be used for mining but also that's kind of why there's these deepen projects out there you can also offset that allocation to those providers like Aether IET hyperbolic so some providers do that as well very interesting and we've had discussions with ather and IET on the show before but I so AI compute seems

to move so quickly that it's hard to keep up if we already trying to keep up on top of blockchain as well and with compute Labs do you guys tie into the blockchain equos system or have a token and that merging of AI and blockchain yeah it's a good good question so the short answer is yes effectively what we're doing is we are tokenizing our Hardware so when we back

up and I talk about how we're raising Capital we're raising Capital through those institutional investors and those retail investors but the way we're doing it is that we are issuing a structured product we're issuing an investable asset to these investors and what happens is we're giving the rights to the Hardwares Revenue the profits and all the liquidation proceeds to those

investors and that's what a token means so when you're an investor and you invest into one of Compu lab structure products you get a onetoone representation of your Hardware so if you're investing in h200 okay you'll get a GPU nft of a h200 product what that means is that now that you own this one token of h00 you'll get all its Revenue you'll get all

its proceeds when we sell it or liquidate it you'll get all of those proceeds as well and that's what blockchain is doing here tokenizing the actual Hardware that's really interesting sorry yeah yeah no with the yeah it's like real world asset tokenization of if you're looking at buying and selling assets that are in a warehouse you know you're not really

you're not moving them if somebody else wants to own it's still going to be computer the same thing in that warehouse then nfts and real world asset tokenization is a great use case for this it's very interesting and what are some of the next steps for growth for compute labs and keeping up with the growth of AI in the industry for 2025 some of those road map plans

yeah great question so our priority right now is helping operators get more res sources whether it's the capital to buy Hardware or just give them the hardware they need to serve their customers so we've already done a pilot with the b200 with our partner in Japan that's going to go live in late February early March we're really excited about that asset because the

b200 is very limited in Access Compu Labs is one of eight of the largest suppliers in the world for it because that Hardware won't be available to the Mass Market until like Q2 Q3 2025 but as we wait for that pilot to go live we're constantly looking for more operators and data centers to help them we have one lined up that we're going to be soon fundraising for we're we're

going to be needing a ton of capital to serve that one customer but we also are looking for small clusters you know 1 mil 3 mil 5 mil these types of Hardware clusters or operators that need that much happy to help those as well mhm that's great and what's the best way to learn more about compute labs and follow along with the work that your team is doing yeah we have a guy on our

Twitter just killing it so follow us on Twitter we have a lot of fun there we also have our own Discord Channel as well sounds great Nick thank you so much for the insights into Computing AI where everything is going I will leave a link in the show notes to the compute lab site the Twitter and the Discord as you mentioned and wishing you and your

team all the best to moving forward and I would love to follow up in the near future thanks so much Ash great to be here appreciate it

More interviews

Browse all 1,089 interviews

Get new interviews firstCCS Insider, the free newsletter from Ashton Addison. Twice a week.