Jack Collier on decentralized AI compute replacing centralized infrastructure

0:00I'm Ashen Addison from the Crypto Coin Show and today on Blockchain Interviews with Jack Collier, CMO of io.net, the decentralized AI compute network that aggregates aggregates GPU power from data centers, mining rigs, and consumer devices into a marketplace that provides up to 70% lower cost than AWS or Google Cloud, which can be very expensive. I personally know. Uh excited to see

0:27io.net back on the show and thank you so much for taking the time, Jack. Thanks, Ashen. Good to be here. Yeah, let's let's dive into this because uh AI compute is necessary. There's uh a lot of demand for it and it's great to have a blockchain-based project with decentralized compute but not stuck in the realms of crypto and stuck in the troughs of some of the hard times that

0:52people that are purely in crypto uh face because AI is just so big but combining AI and blockchain together actually has a lot of value and I io.net is right at the center of that. Maybe we can kick it off with sort of the latest on io.net and what you guys have been focusing on. It's been uh many months since we last had you on the show. Yeah, yeah, for sure. So, if anyone who

1:16doesn't really know about io.net, as you sort of elegantly mentioned, um we sort of work with um any of the data centers that aren't the big three. So, um as most people probably know, uh 70% of global compute is managed by three providers, you know, GCP, uh Azure, Amazon Web Services. And what we're trying to do well, I guess the issue with that is that a lot

1:41of those like those three providers are expensive. They often gate access to the people they want to have the latest tech. They give that to them first. Um so, it's like a bit of a a bit of a boys club. Um and it's really expensive for people. You know, we're talking over $30 an hour for an H200 device, for example. And um obviously, there are many other

2:02data centers across the world and even consumer people like retail people who actually have these types of devices. And we thought, well, what if we do this differently? What if we instead of having a centralized provider that provides all of this compute, what if we were to join all those other secondary tier providers together and offer that um offer that redundant compute that

2:20they they have, that spare capacity they have, to everyone? And what we realized was that we were able to offer um like the same performance, if not better, to consumers at a much cheaper price. Um and what we also find is that the fact that we're plugging in many different devices from across the world, we have over 138 uh different countries and we have tens of thousands of devices

2:42live. We're able to offer more flexibility and varieties for people to so that they can have the right device for whatever task that they need. So, you know, for example, if you're if you're a cost-sensitive, say, university wanting to train um like, you know, your an AI model, for example, you may not care too much about the performance of the device. You care about the cost. And

3:02we're able to provide, you know, more of that consumer-grade um hardware to those to those communities. Or if you want something that is high-powered, high-performance, we have all of that as well. And so, I think that's the novelty of what we offer. And as you mentioned, we're a blockchain-based project. Um we use uh we use the our own IO token to uh secure the performance of the

3:23network. So, suppliers that can come on and opt in with their supply, they stake some IO against that supply to say, "Yeah, this device does what it says it does." Um and that really creates this like trustless ecosystem whereby, you know, people can people can be rest assured that the device they get is the is is as advertised. Um and that's really the novelty of of

3:45using blockchain in this in this way. So, we're able to almost offer an open network of compute to anyone. And it's no longer gated by the top three, you know, the three big hyperscalers that are out there. It gives people that flexibility and that choice. Um and that's what we're really passionate about at at io.net. Yeah, it's a beautiful thing and the

4:04centralized AI infrastructure, although it's really expensive, they're continuing to build more data centers. It seems like uh that they're building data centers all over the US, at least >> [clears throat] >> we hear you know, in the US news, spending hundreds of billions for this infrastructure. But at the same time, there's a lot of delays or protests

4:24around these centralized AI data centers. What does that tell you about where centralized AI infrastructure is right now? Yeah, well, it tells me that um obviously, demand for these devices is going through the roof. Like, we know that AI is going to power most things in the future, right? There's going to There are use cases we haven't even thought of that um that AI is going to

4:47be relevant for. And that just means that we need to provide the food for the AI to consume. We need that compute for it to run. And so, you know, I don't deny that we need to build new data centers, but I do question the strategy that many countries and uh and organizations are adopting when uh when building and, you know, planning these data centers because

5:07centralization isn't always the answer, you know? What we find with a lot of these data centers is, you know, 80% of the capacity actually on the from those data centers isn't actually utilized, hm? Right? So, we all know use data centers we use our chips in spikes, you know, we we train a model and then it's you know, we don't use it for a little while. We don't use the chipset for a

5:28little while. Um or inference is spiky, you know? And obviously, there are ways which organizations try to increase the the efficiency of the devices they use, but in reality, you know, 80% goes goes underutilized. And again, that's where solutions like io.net really do help, right? So, if you if you have locked yourself in with one of these big hyperscalers or or with a device, um you

5:51know, in any data center, you can opt in that redundant supply into the network, you know? So, instead of having to build lots of data centers all over the world constantly, we we should be juicing the the data centers we have more effectively. And that's where solutions like io.net and other decentralized um infrastructure providers are trying to solve that problem so that we don't have

6:11to keep digging up, you know, rare minerals out the earth and, you know, putting up these massive data centers all over the shop and consuming all this energy. Um we can actually do it in a much more effective way. And again, it's only possible by solutions uh built on the blockchain. Mhm. You know, this couldn't really done with a with with a centralized provider.

6:31I love that. Uh and we Yeah, we don't need more inefficient data centers. Let's make the other data centers more efficient first and actually utilize the capacity before we make a bunch more that are that are not using uh you know, >> Yeah, it's also a security risk and it's it's a reliability risk, Ashen. So, we've seen it before, like AWS goes down. Uh-huh. If AWS goes down, all the

6:55services connected to that data center go down. You know? And and I think this is where again, like, we see we see people using io.net in very novel ways. So, it's like, you know, they they have chipsets in four or five different countries. And it's very rare, but if one data center was to have an issue, it could easily just be pointed to the to the second,

7:13you know? Hm. And so, it this flexibility becomes quite a a USP for people, you know? They no longer do they just their service goes offline for hours on end. You know, and it's the same with um with inference, you know? We have a product at io.net, IO Intelligence, which is our open-source models um which people can plug into through a single API. And you

7:35know, we have that hosted on services all across the world. And so, people can start, you know, start offering local inference to provide better, you know, better performance um for people in different countries. So, if I have a global product, I need inference, you know, I have customers in, say, Japan, I can serve them local inference from Japan. And if I then have customers in South

7:57Africa, I can provide local um inference and from South Africa. So, you know, it's it's that way of thinking to provide better services to their customers, which again is only really possible by using something that has more flexibility built into the system at its heart. Definitely. And it reminds me of a little bit of of like torrenting back in the early 2000s. You're getting a little

8:19piece from all of these different devices. If one goes off, the other devices probably have that piece. And then you you mix in the localization there. You know, if your if your friends all have it, you're probably going to get it faster in in your town than if it's on the other side of the world. Is that a fair analogy to translating, downloading that file

8:3830 25 years ago versus now getting AI compute in a similar way? Yeah, it's very it's it's very similar, actually. Um I've never actually used that analogy before, but I think it's a good one. You know, it's just obviously, the block the novelty of using blockchain with this as well as it's like it's the verifiability, you know? I guess with torrents, the issue of you

8:58don't always know that the file you're downloading is exactly what it is. Hm. You know, I think what's so great about um you know, deep in providers and and with with people like io.net is that, you know, that that um there is no incentive for people to lie about the device that's on the network, you know? It's not in their economic interest to do so, you know?

9:18And I think that's where these trustless systems, this not and building these novel solutions is going to be the answer because, you know, we know that centralization corrupts and we know that it it provides an inferior service over time or a service that is gated. It's not built for the people. And that's why people build things like, you know, torrents

9:36and um and uh you know, decentralized compute providers like io.net. We're here to provide power back to the little person so that they they can participate in what's what's on offer, you know? And so, people want to get involved in AI, people want to build in AI, it's becoming increasingly more expensive to do so because of these centralized hyperscalers. And we say, "No, that's

9:59that's not right. It doesn't have to be like that." And, you know, we know that um innovation comes when more and more people participate um and when things are built open or open source. And so, that's what we're trying to do. We're passionately trying to give power back to people so that they can participate in this opportunity. Definitely. And there's probably a lot

10:18of people that don't even aren't even aware of that they could contribute, whether it's through the supply side or through the demand uh in something like comp- AI compute power that they probably have lots of AI compute devices sitting around their house that are not being used where they could probably earn. Can Can you touch on that and who what kind of new users can get access on

10:39both demand and supply moving into a decentralized model? Yeah, for sure. So, um I'll just say if you have a 4090 uh lying around, um yeah, if you have devices like that lying around or, you know, you're you have a little server at home that has these divide that they these this hardware in, um you're more than you can go on io.net now and not that into the

10:59network, you know? And um and like I say, we we test we test the reliability and the performance of those devices. We do our own sense checks on them just so that that's that, you know, everything's cool. Um the the other the other side is that, you know, people can opt in if they rent if they're renting from a third party, you know, they can opt that supply into

11:18the network as well if they're, you know, renting space in a data center or or whatever it is the hardware in a data center. Um really what we care about is providing reliable um quality devices to the end consumer, right? So, as long as those devices do what they say they're going to do. And, you know, obviously the network has that incentive mechanism

11:37built in, so um so performance is is guaranteed to some extent. Um then then from the demand side, it's really can we offer a variety of devices that really suits whatever customers that come onto the platform? And And like I say, today we have over um we have uh tens of thousands of devices live across 138 different countries. Um and they're anything from, like I said, 4090 retail,

12:01you know, consumer grade chips all the way through to the the latest devices, you know? The H200s, E200s, all all of that, whatever it is. Um and uh and so, what we're trying to you know, our whole approach is to go out to people and just say, "Look, if you're a startup and you're consuming 40, 50, 60% of your costs on uh you know, on compute power, um

12:23actually it it doesn't have to be like that. You could save 70% by using a decentralized provider." And you know, and that's where I think, you know, flipping this to the cryp- the the crypto side is, you know, crypto can provide real value to to people, you know? And they don't even need to know they're using crypto at the end of the day. That's what I'm really

12:42passionate about. That's what I've been doing for the last 10 years in the industry is trying to forward that, you know? It's not just about hype cycles and narrative. It's about No, no, no, blockchain actually solves real world problems. And, you know, serious businesses um that are now coming into crypto and building these products and services on top of blockchain and

13:02providing them almost to way to to end consumers that they don't even need to know that it's powered on it. And And I think that's the beauty That will be the next wave in crypto is there are a lot of these There are a lot of things in other industries that have been built like this. Um and they'll start to be adopted. Um and that's where the, you know, token utility comes into play.

13:20Like tokens aren't just there as a, you know, investment vehicle. They're there to power a network, this trustless network. Um and um and so, yeah, we'll start to see more of that coming to life in the next few years as well, I'm sure. Mhm. It sounds so obvious uh for a startup that would want to save uh 70% of their costs. You know, I I've ran uh a SaaS company through AWS. And

13:45yeah, it's like a lot. It's uh it's really expensive. It's a major It's a major hurdle for especially for startups. Why are more non uh why aren't more non-crypto and and blockchain like Web2 companies just making this switch easily? What are the factors that are inhibiting this from exploding? I think it's obviously the the war chests of an AWS, Azure, GCP, like they're big, you know?

14:13So, >> to shut it off, actually. It You need a degree just to figure out how to to cancel. It's It's It's complex. >> Exactly. They're locked in in so many ways, right? They have so many like connective tissues in what they do with everything. It's hard, you know? Once you've Once you've built something on something, it's hard to extract it out, you know? And um and there's also that

14:31the the fact that they just got billions of dollars to to put in front of people. And when a lot of these stories that go to the press about GPU shortages and, you know, you need to lock in your supply now, etc. I mean, bet your bottom dollar a lot of that's pushed by all those suppliers as well, right? You know, so a lot of the hyperscalers as well. So, you know,

14:50that's what we're up against. We're up against the big gun narratives. We're up against big money, big tech. Um and so, you know, what we're trying to do is, you know, we're we're trying to come up with novel ways to get the message across, you know? We're We're obviously letting our product do a lot of the talking, you know? We We have, you know, H200 devices available for less than $2

15:11an hour on our platform today, you know? And And that is very different to say AWS where you're looking at what, 25, 30 dollars an hour as of today? So, you know, we're we're trying to let that do the talking. We talk through real life use cases and some of our case studies. We have some fantastic products building on io.net infrastructure. We have Leo- Leonardo.AI

15:35who do imaging in AI. Um and they've just been bought by Canva. And they they use us all the time, you know? They're they're a great customer. And they're innovating They're able to innovate faster, I believe, because they're saving so much money on their their inference costs, you know? So, we're trying to It might be a slower burn to start, but we're trying to tell

15:54those real life stories um of of how people have been able to do more with providers like io.net and through our services. Um and we're starting to see that tide turn, you know? Most of our revenue comes from Web2, like people who don't even know that they're building on crypto crypto rails, really. And um and that's that's just We've just got to keep keep at it,

16:16really. Definitely. Talk to me about AI agents. I'm sure there there's a factor of that in here. It's not just a SaaS uh software that's, you know, just switching from AWS. AI agents are the future. Uh and as AI compute gets more powerful, people want agents to take over their work. I saw that there was a agent compute uh update with io.net. Um how do AI agents fit into this network?

16:43Yeah, for sure. So, obviously it's for AI agents to run, they they need compute power, as I mentioned. It's like the fuel for them. Um and so, we just thought, "Well, wouldn't it be cool if an AI agent was able to go to the supermarket and buy its own food?" You know? And that's what Agent Cloud is, really. So, you know, if Agent if you have an agent running and

17:04um and it has access to a wallet of some sort, which you can control, um you can now use Agent Cloud on io.net for it to go and literally purchase its own compute power uh using Agent Cloud, which is a an MCP library that we've we've created. So, it can come onto our platform, automatically find the right device that it requires, and then automatically purchase that. Um so, it's

17:27just a way for us to, you know, it's a way for us to just try and automate um and and and improve the usability of our product and access to our product, you know? So, you you don't have to always just go through the whole payment flow. Obviously, we've tried to make it as possible, but, you know, AI agents are just making things so much so much simpler for us these days. And we wanted

17:46to like provide a better way to do that. I mean, our CEO talks quite passionately about a world where, you know, AI agents are being spun up themselves and are able to purchase their own compute power and and run entirely autonomously, you know? And obviously, compute is a part of that, you know, having their own bank accounts, identities, you know? And and

18:06participating in our whole economy um as an AI agent. And um and so, obviously, being able to go out and buy your own groceries and food is an important part of that. And that's what we're providing with uh with Agent Cloud. That's super cool. And I I would love for my agent to go buy just the groceries it needs. Um how do you prevent it from, you know, going

18:26to the casino or you know, once you give it your credit card, it just starts going wild? How do you prevent anyone from going to the casino, actually? It's the drill. Um well, I think this is I mean, this is where we we need to put guardrails and things in place. Like, you know, so that that comes on both sides. That's on the side of like you designing your agent in a way to

18:49give it a set of rules and, you know, give it a conscience. Um and I think that's that's really in the implementation of how the agent's created. But then also, obviously, we we we put guardrails in place. So, you know, we we we can't let We don't let agents just buy, you know, 20 20 devices straight off the bat for like a year, you know? It's um we have specific checks and balances in

19:12place so that so that, you know, they don't they don't go wild. Yeah, that makes sense. With regards to the payments, do AI agents equally work with fiat bank payments or are we sticking to crypto rails and and stable coins and this X402 protocol that's now allowing AI agents to start pay with crypto more? Uh it's both. So, you know, if it has access to to a bank account or

19:41you know, a card, a Stripe, you know, access to Stripe or something, it can it can it can use that. Um but we do try and build in crypto first, right? So, um so yeah, if you have access to a wallet, then then obviously um we would accept that. If they have IO tokens, even better. We We just Honestly, we're not We're We're precious in the sense that we we obviously want more and more

20:04people to participate in a permissionless world, you know, we we're Anyone who's in crypto is passionate about that, putting power back into people's hands, giving them control of their data, privacy, finance, whatever it is. Um but at the end of the day, we we realize that that is the that is going to be that is going to be a a transition, you know? And so that's why

20:23we still integrate in things with, you know, traditional providers. We We want We want people to be able to participate in this in a way that is that they're comfortable with. And so So yeah, we do continue to take um card payments and traditional payments as well. Just just to increase access. And I'm, you know, there are other organizations that are trying to solve

20:43all of the financial side uh of things. You know, I I used to work at Circle. I'm very passionate about what they're doing. Um uh I believe that in, you know, 5, 10, 20 years' time, we'll we'll the the the tide will flip and uh and we'll see more people on permissionless tech than than than permission tech, for sure. Definitely. It's It goes perfectly hand

21:02in hand with uh decentralized compute. You're already on a decentralized rails, might as well be paying uh with global currency and not have to pay those Stripe fees and all of these. But I can imagine as you alluded to, many of the customers are actually Web 2 that, you know, there's a lot more uh at least on the companies that are using compute and that have uh SaaS products

21:27that are hosting on AWS, there's a million times more of those than there are crypto companies. They're probably used to paying in fiat for now. And And once they've moved to decentralized networks like IO.net, I can imagine they're they're like, "We're saving 70%. What's another switch to stable coins here to save another 3% and increase our margin even more?"

21:47Yeah, for sure. And And it will be it will be tech that that they Again, they don't even know that they're they're on. You know, I I can imagine a world not so far down the line where people are holding all of their capital in what they think is a bank and it's actually their own wallet, you know? And they're they're purchasing like they would at a supermarket with a

22:08card and it's actually just trying two wallets transacting on on the blockchain. Like I genuinely And And that's not a blockchain problem per se, it's it's a software problem, it's a UX problem, it's a you know, it's all those things. It's the things that have been solved for before in a centralized way. We just now need to solve some of those problems

22:25now in the crypto space. You know, it's making things easy, it's making things accessible, it's making them usable, you know? And it's not as easy as just creating a nice fancy webpage. It's all the connections that go on underneath. It's making sure everything's stress tested. It's making sure that they're it's distributed in the right way, you know? It's

22:44it it's We're just at the start of what is going to be an explosion of of crypto tech that that um that that takes over many different industries. And like I say, we're we're here to solve the um the compute side of that. We're here to ensure that everybody can build in a way that is cheap, fast, and flexible. Um and critically that in a way where they can build in the open, you know? I

23:08alluded to our IO intelligence product. You know, we're really passionate about people using open-source models as well as as opposed to just, you know, relying on the closed products like ChatGPT and Claude and and and all the others. You know, we we want people to be in control of the data and and things that go into training their models and get that you

23:29know, that that these models are actually utilized. And so, you know, this is what Again, we're just trying to expand off the top of the GPU product that we have. Um we're trying to expand the suite of products so that people can participate more in that open economy um and not have to default to these closed centralized providers where we don't even know what they're doing with all of

23:48our data at the end of the day as well. So. Definitely. I'm all in. I'm all in I've been all in, but now that AI is in there and more companies are waking up to it, uh they're going to move over to the dark side or should I say the light side uh of blockchain where everything's more fair, faster, and cheaper. Amen. I agree. Can Can we touch on the new IDE uh

24:13fixing the the economics and the tokenomics for IO.net and how you guys have improved on that recently? Yeah, for sure. So, as I mentioned, IO tokens are the fuel of our network. So when someone when someone stakes their device on the network, they're staking IO tokens against that the performance of that device. That's the supply side. And then on the demand side, when

24:34someone comes and buys a device, whether they use, you know, Stripe or whatever or just, you know, USDC or something, we're that's essentially converting into IO and and fueling the purchase, right? So, the amount goes to the supplier, uh fee goes to the network, and and that's how that all works. Um how most DePINs operate is to incentivize suppliers to come and join

24:55the network, they uh the emissions of that network go to incentivizing them, right? So let's say, you know, just for argument's sake, there's 10,000 tokens that get minted every month. That gets split across suppliers that come onto the network. The issue with that is that it's somewhat extractive and it's somewhat um like like volatile. So, if the price of

25:17a token goes down, suppliers don't get much incentive to join the network. And if the price goes up, they're over-incentivized. And so what we've done at IO is we've we've released uh what we've called the incentive dynamic engine. And that that changes things from it being a supply-driven network to being a demand-driven network. So when um So instead of giving that fixed amount

25:38every month to suppliers, what happens is that um that that someone will come and purchase a device. Uh suppliers are made whole through uh through the network. That is paid in USDC equivalent IO. So they're always paid a stable incentive to to join the network. And then any ex- excess of that goes into a vault. So that if the price of IO goes up and down, um there is a

26:05mechanism to sort of make suppliers make suppliers whole and to ensure that they're incentivized. So it's So it basically creates a stable demand-driven network as opposed to being an extractive supply-driven network. And I suppose the the the third part of the change is that the um the vault that we keep where um there's the sort of the additional emissions go

26:28into, uh we've committed that uh 50% of those emissions will be um burned over over a period of time. And so instead of having an an inflationary inflationary type network where tokens are minted all the time, which is most DePINs and most projects in crypto, um ours will become a deflationary network over time. So that So that the supply isn't just uh you

26:53know, ever expanding. And so again, that just means that there's more value being driven into the network. It's more It's It's more hard-coded. So it just creates a sustainable a sustainable ecosystem for people to build on. Yeah. I I love that. That's the keyword I use with with every project I talk to. I'm like, "How do we create a sustainable ecosystem?"

27:13Uh because, you know, with the volatility of of Bitcoin and then moving into uh governance tokens underneath that, uh the liquidity and and just the volatility is it gets a little crazy. Um and well, part of the key is to have a a great company that has actual revenues, uh which you guys do. And then tying that into the economics properly, I feel like is super important

27:38if you're going to create something sustainable uh for the long term. Yeah, for sure. And And the beauty about this is like a lot like a lot of people that I've seen that have changed or amended their tokenomics over the last couple years, it's almost like, you know, it's just sometimes lazy. It's just like, "We're going to burn X tokens over Y years."

28:01You know? That still doesn't create a sustainable network, you know? What we've done at IO is we've tied the utility to the like the success of the the network. Like the the the value to the success of the network. So, you know, the more demand we drive that comes through the network and that is real usage, the more um value is created. You know? And I think that's the

28:26that's what a token should do, you know? Either it should be stable and and um you know, off off a consistent value or it should it should have more value based off the usage of it. It should have that utility built in. And I think a lot of projects out there are there's they claim to have that and it's really just a wrapper for something else or um

28:48just something that, you know, it's not tied to the success of what they're trying to do. Um so And to me, that's a litmus test. It's like, "Is this project serious, yes or no?" Um and one of those tests is, you know, is their token tied to the core proposition of what they're trying to build? Mhm. And if it's no, then, you know, I I would stay clear. It's

29:09It's It's simple as that for me. Mhm. Yeah. What's What's your projection on, you know, the demand that will funnel into decentralized networks like IO.net and the amount of people that are going to start needing more AI compute? You know, I I've seen the the charts in that it there was like 0.03% of people were actually using AI for like Claude code and

29:37actually building things. Most people are just asking questions like it's Google. And half the world's, you know, still using Google search. They're not even using AI. So, there's so so uh so many people that are are still that are waking up to hey, look at the what power we can have if we start utilizing the capabilities of actual compute in AI. What's the kind of trajectory and and

29:59how much of that can flow into these decentralized networks over the next 12 to 24 months? So what are we looking at? Like $650 billion is spent on building data center data center infrastructure. Um you know, I I owe you we're at 20 25 million dollars in in annualized revenue. Um and that's [clears throat] just the start, you know, we've only really been

30:24you know, serious about this for like the last year or so and we're seeing our growth trajectory sort of of taking off. I think the demand for compute is only going to go through the roof. I think the questions that everybody asks are always going to be there. You know, I think you know, AI search seems to be delivering a better experience than regular search for people like to get

30:44into their answers much more quickly that they need. So I don't think that's going to go away and as you alluded to like vibe coding these models are only going to get better and better at being able to do do do coding do software development. They're only going to get better and better at doing whatever they're trained to do and so what I foresee is that the

31:03demand for compute is just going to skyrocket still. You know, and again it's just going to put it's going to put more and more emphasis on like we talked about at the start of the call. It's going to put more and more emphasis on people trying to juice the existing supply more effectively and this is why I think decentralized networks for this is is

31:21inevitable. There's no other way that we can you know, get that utilization up of these devices without offering them in a flexible way to people. And and that that just benefits everyone. You know, I think I think you know, if we're talking specifically on cloud code and things like that, you know, people are going to be creating their own software soon. You know, what people

31:45won't be using off-the-shelf SaaS products unless it's something very particular. You know, my whole marketing team runs off tech that we built. You know, it is AI we don't use subscription software anymore. We do everything in house. It's it's all very novel. We we have a couple of third-party systems that help with like frameworks and context like helping

32:07with con shared context across the organization but that's that's really about it and I see that will go all the way down from business all the way to to individuals. You know, I vibe coded the other week a a an app which tracks all my friends run running in a competition style and we you know, it's now a competition between us who can run the most for the rest of

32:28this year. You know, I don't need to travel for that. I can just we it's done in a week, you know, and that's just one example and we're going to see that and that just means that the demand for all of for AI compute and inference is just going to continue to skyrocket as it becomes more accessible to people and more interesting to people. Definitely. It's very exciting.

32:48I'm glad that we have a decentralized network blockchain project that's really got a handle on AI compute and can and show people that blockchain version of AWS and and these big three that are gouging people and are also running it inefficiently. You know, there is blockchain is just it's a it's a better solution and better solution in many industries but this is a clear example

33:13and I think you guys are poised for a lot of growth in the future when as AI compute gets into the hands of more people they realize the potential. If there are people that want to utilize the compute, how easy is it to get on boarded? You know, do you need a degree like like AWS or how quickly can we get on boarded? You can literally spin up a cluster

33:35now in 2 minutes. Just go to io.net and choose the device that you want, the country you want, the setup you want. Pay for as for as long as you want and away you go. The device is is ready for you within 2 minutes. So yeah, no wait lists. You know, it's not like a ridiculous KYC process or anything none of that and and that's that's how it should be.

33:57You know, let's get compute accessible. Let's put power back into people's hands so that they can build what they want to build in AI without being gouged unfairly by these centralized providers. Well said, Jack. Thank you so much for taking the time. Wishing you and the team all the best at io.net. I'm excited to see more AI compute move into decentralized AI compute and let's

34:21follow up again in the near future. Sounds great. Thanks for having me, Ashton. Appreciate it.