The Hidden War For Compute Power... With Akash Network

[00:00:00]By 2029, [music] we'll need a gigawatt data center to train the latest AI. A gigawatt data center is a nuclear reactor. And we know that's not sustainable. So, the alternative is >> Today's guest is Greg [music] Osuri, founder of Akash Network and CEO of Overclock Labs, who previously founded AngelHack, [music] the world's largest hackathon organization, and now is building a decentralized cloud marketplace powering AI [music] and cloud computing.

[00:00:23]>> Where did the original idea come >> It was becoming more evident. We started seeing cracks in the cloud model where the cost is no longer affordable.

[00:00:32][music] On top of that, cloud also dictates what kind of hardware you get built, what kind of software you get built. It was born out of a need to create an open cloud.

[00:00:40]>> Maybe we could speak a little bit about techno feudalism as an idea.

[00:00:43]>> Techno feudalism is where [music] few players control most of the resources that normal people need. [music] So, we have these cloud companies as kingmakers now.

[00:00:52]>> Is that how you see Akash Network as a competitor to the likes of Google Cloud, [music] Amazon Web Services, and so on?

[00:00:58]>> I don't see ourselves as competition, but rather a as a distribution channel for these folks.

[00:01:03]>> It very much seems like the world that we're moving towards is one where you're going to get humans being replaced by machines. I often wonder if owning machines is the way to stay relevant in that world.

[00:01:14]>> It's heading there for sure. We saw that coming a few years ago, and we actually built a product called Akash Home Node, which lets a home computer contribute the compute to the cloud.

[00:01:24]>> Why aren't more people taking advantage of this?

[00:01:26]>> Only 2.4 million people that even know how to use agents. So, there's an enormous information gap.

[00:01:32]>> I know for a fact there's people out there that have hardware today. How could they get involved?

[00:01:40]Welcome back, ladies and gentlemen, to the All In Crypto podcast. I am your host, Ruper Pickering, and today I am delighted to be joined by Greg Osuri, the founder of Akash Network and CEO of Overclock Labs. Akash is building a decentralized cloud marketplace giving developers and AI builders access to compute and GPU resources outside of the traditional cloud giants. And at a time when AI is creating massive demand for compute, the big question is simple. Can decentralized infrastructure become a real alternative to Amazon Web Services, Google Cloud, and Microsoft Azure, and some of the other market participants?

[00:02:23]Greg, welcome to the show. How are you?

[00:02:25]>> Great doing. Thanks for having me. Great to be here.

[00:02:28]>> The pleasure is truly all mine. Before we get into Akash, where did the original idea to build an alternative compute resource come from? You know, what was the kind of problem that you saw in the world that you sought to solve?

[00:02:46]>> Yeah, so say a flashback to 2013.

[00:02:51]The origin story goes back way long. We were um 2013, if I had to paint a picture, it's the renaissance of cloud infrastructure. It's really where of course you had Amazon, Google's, and Microsoft's earning record revenues and profits.

[00:03:10]There was a notion that all compute will go to the cloud and will have basically thin clients or light clients which are mobile. Mobile was getting very popular in 2010s early. And uh cloud gave a very flexible way to scale companies. So, famously you saw this case study with Netflix saving a lot of money and whatnot switching from switching to cloud from their own data centers. So, it was very promising. It was uh infrastructure and API, you call it. So, it's revolutionary. And that kind of was very attractive for developers in San Francisco. Uh you know, usually we tend to be the trendsetters, right? So, we were very deeply involved and I was scaling companies. I was working with a company called Demandbase, which we were scaling to about We were doing about a a billion transactions a day, low latency. So, my background is in low latency high performance computing.

[00:04:05]And um and I was also a big open source contributor. So, I was fascinated by this new technology space called container native uh infrastructure, which is a more nimble way to deploy things, a more flexible way to deploy things. And it's um responsible for scaling Google.

[00:04:23]And Google, you know, is probably the biggest infrastructure company in the world.

[00:04:29]Um so, the technology that helped Google was getting more democratized.

[00:04:33]And that's where I uh found uh a lot of interest. And um as we dug in, we realized that cloud, as we know it, was being flexible and being very capable, is also is antithetical to what the internet, the principles of the internet, was. Because the more bigger the cloud got, it was dictating market.

[00:04:58]>> Yeah.

[00:04:58]>> Just like how you see the GPU market today, it's kind of kind of has a little parallels for the cloud market back in the day. And it was to a point where cloud was increasingly [clears throat] becoming closed. So, the innovation was happening in the open source space, but cloud companies would come, take the open source software, and and build massive business on top of it with while returning very back to open source community.

[00:05:22]So, there was an enormous um imbalance between what open source offered versus what cloud, leveraging open source, could offer back.

[00:05:31]Um that didn't sit well with the industry and academia. And people were concerned that cloud would, you know, become this uh bottleneck for people to innovate. And it was becoming more evident. We started to seeing cracks in the cloud model where the cost is no longer affordable. It's funny how it has parallels to AI right now.

[00:05:55]It was exciting and then the cost became exuberant. You could see public filings from Asana to Netflix and all these companies that went public. You can see their financials.

[00:06:05]Turns out half of every dollar you pay to an online service goes to a cloud provider.

[00:06:09]>> Yep.

[00:06:10]>> So this you pay an invisible tax to Amazon without even realizing. You pay an invisible tax on top of their cloud also dictates what kind of hardware gets built, what kind of software gets built.

[00:06:19]So it's increasingly not only getting uh economically powerful, but also having quite a lot of influence in the future of computer hardware.

[00:06:29]That didn't sit with a lot of folks and um there was an effort in academia around building something called a supercloud, which is hey look, there's computer is not exclusive to the cloud.

[00:06:40]You have computer everywhere. We have computers at home. We have computers, you know, devices. There's a lot of computer available. And there was an effort to build a cloud that could leverage this compute that's latent across the world and build a system like a cloud that acts and behaves like a cloud to a normal person.

[00:07:00]Um but computer comes from everywhere.

[00:07:02]And so that's the genesis of Akash Network. And Cornell wrote a paper uh introducing the supercloud. We were excited about the paper. We took the paper and actually implement We wrote another paper on how do you do this practically in in real life. I published that paper in 2017 and got notoriety and people were extremely um excited about this notion and then uh we built and launched in 2020 and rest is history. So that's how Akash got started. It was born out of a need to create an open cloud.

[00:07:34]And turns out that it now in particular when you see Akash responsible for Akash bootstrapped to Venice AI, which you know, open now. Akash bootstrapped Hermes Agent, which is News. The company News Hermes is the number one agent today.

[00:07:50]Akash bootstrapped Prime Intellect. Um Brev. Brev is another company that got acquired by by Nvidia. I mean, yesterday, if you were the AI engineering summit here, you could see the local AI taking off because of Akash.

[00:08:04]So, if you have to attribute the successes today of a lot of incredible companies that came out, they can all be attributed back to Akash. And the reason for that is open source, open cloud.

[00:08:16]And so, with a lot of pride, I can say that, "Hey, look, we achieved what we set to achieve, achieved a bit of what we wanted to achieve." And and and the proof point is these incredible companies being built and being becoming extremely successful, and raising billions of dollars valuation, right? That wouldn't be possible if the cloud was was the way it was. So, that's that's what Akash brings to the table.

[00:08:40]>> Yeah. I think this idea is absolutely fascinating. Like you say, it's it it it's far more than an idea. It's it it's proven. It's practical. It's here.

[00:08:48]I think a lot of people in the crypto space understand the need for a kind of alternative financial system, an alternative money, an alternative store of value, and that you can use distributed ledger technology to enable that.

[00:09:01]I think those same principles very much apply to the cloud computing world, the computing world more broadly, the the the the kind of hardware marketplace more broadly. It's very interesting that you bring up the kind of necessity for this to be born as a result of the kind of techno-feudalism that goes on.

[00:09:19]Maybe we could speak a little bit about techno-feudalism as an idea, because a lot of people don't understand just how powerful these companies truly are. And that only grows infinitely in the world that we're moving towards unless we have an Akash.

[00:09:35]>> Uh absolutely. I mean, it's a good way to put it. Techno-feudalism is essentially a modern-day version of feudalism where few players control most of the resources that that normal people need.

[00:09:49]And that control is exercised either through squeezing in and selecting players. If you look at Amazon's and Anthropic's history, and you know, of course Anthropic's an incredible company that builds incredible products, but you know, there's mounting evidence that points to the Anthropic and Amazon deal that was done a few years ago that gave Anthropic the compute they need to build a company. Amazon chose Anthropic.

[00:10:17]So, we have these cloud companies as kingmakers now.

[00:10:21]It's very visible, right? It's very obvious.

[00:10:24]OpenAI didn't get the similar deal.

[00:10:27]So, it's And And if you look at small companies, you know, in especially around 2020 uh Let's say 2022-2023, it's impossible for smaller companies to get get access to GPUs if you have them.

[00:10:44]So, it's very clear these lords, you know, these digital lords are selecting who the winners are going to be, who's going to get the bread, and who's going to And you know, who And these these lords benefited immensely. Look at Google's stake in you know, OpenAI's or Anthropic. Look at Amazon's stake in Anthropic. Now they're trading another company. Now Amazon got involved at a very, very early stage. And like like any savvy company businessman, they instead of just giving compute for dollars, they took stake in the company.

[00:11:18]You know, because they knew this is such a powerful thing. So, I would want to have a share in the company. So, so that it's happening right in front of our eyes. And as we move from an age of internet information to age of intelligence, right?

[00:11:33]Now, how what's the thing can what you do will be influenced by this company?

[00:11:37]It's not It's very popular and traffic now. It controls a massive share of the market.

[00:11:44]Um if you look at if you ask it any political challenging question, it will be extremely biased into one side. So, now these companies will influence how we think.

[00:11:55]And what does that going to look like uh to the future of our planet, right? So, very important that in order to have a more unbiased uh thought process, you need to have an biased computer platform. Be- because compute is the food for this thought, right? It is It is the spice that runs the runs the gamut. Uh without which you will not have intelligence. So, if you can democratize the very access, the fuel that powers this intelligence, that will be the foundation for a just world.

[00:12:27]And that's what I truly believe in.

[00:12:29]>> I think we just perfectly hit the nail on the head there in terms of what's at the heart of this mission.

[00:12:35]And that is exactly that. You know, it's a very um it's kind of two ways it seems to want to go, right? There's this kind of dystopian future that might be ahead of us or this kind of utopian future. And I think that depends on the fight that's taking place right now and actually uh kind of democratizing I don't really like using that word, but at least breaking down um some of the powers that be by introducing competition, which I think Akash is to some of these other cloud providers. Is that how you see Akash Network as a competitor to the likes of Google Cloud, Amazon Web Services, and so on?

[00:13:08]>> I don't see myself as a competitor. I see them I mean, Akash is a supercloud.

[00:13:13]Supercloud is a cloud of clouds. So, as foundation of Akash, I mean, I I would imagine um Google and Amazon and Azure of the world will become part of Akash Network at some point, you know?

[00:13:26]So, the point of Akash Network is to remove bias, right? Remove bias in terms of compute provider selecting the winners, right?

[00:13:36]So, and create a a flatter world where the prices are transparent. So, you come to Akash today, as we're seeing more and more companies training towards new cloud businesses. I mean, we just XAI using their excess compute to provide to Anthropic. We're seeing Meta calling out for possibly becoming a cloud provider.

[00:13:58]We're seeing, you know, All Burst rename themselves to Smart Burst and becoming a GPU cloud provider. We're seeing companies that are like gab.com, which is a social network that, you know, having their own AI cloud.

[00:14:13]Every company will have their own AI cloud. And that's a prediction I made a long time ago, and that prediction still stands true today, right? Because ultimately, you have to own your AI in order to get any a sense of economic flexibility because the cost of inference is just not slowing down and real dollars, billions of dollars are being spent on Anthropic's of the world and companies like Facebook and and Ubers, you can see them talk publicly about their cost need to optimize their costs.

[00:14:46]And the public companies, you actually see their balance sheet. So, our P&Ls, rather. Um so, compute um cost is becoming a real concern. The obvious answer is to own your compute.

[00:14:58]So, when that happens, you are not using the compute 24 hours. So, the answer when you own compute, you have to optimize for efficiency. You got to, you know, we go back to the old world where everybody has compute, but then you're going to address what you do when you're not using the computer, right? So, it's like I think in a world of such um extreme ownership, which we're training towards a marketplace that can connect this leading computer supply becomes a lot more important.

[00:15:26]Uh and that's why I today we have about 100 120 new clouds in our in our database. I mean, that's what we we we research and most people maybe know about 10. So, I don't see ourselves as competition, but rather a >> a distribution >> Yeah, it's a distribution channel for these folks, you know, so uh Airbnb how Airbnb was was basically a distribution channel for extra space. I mean, >> you can maybe look at Airbnb as a competition to to the traditional hotels, but there's no stopping from traditional hotels to become Airbnb rooms, right? I mean, if the users are there in Airbnb and regular hotels should be there in Airbnb as well. So, it's a it's a distribution platform for uh for uh you know, unused space.

[00:16:09]>> Do you know something I always think about and I'm going to throw this before we really get into some more specific questions about a cash, but a cash is relevant to the question I'm about to ask. It it it very much seems like the world that we're moving towards is one where you're going to get humans being replaced by machines. I often wonder if owning machines is the way to stay relevant in that world, i.e. being a contributor to a cash network and and and and staying relevant through being able to provide that resource. What do you think about that? Do you think that's somewhere we could end up?

[00:16:43]>> Absolutely. I mean, in a recent tweet I tweeted I said only more you really have is owning computer and a way to distribute it.

[00:16:49]>> Yep.

[00:16:49]>> Yeah. A cash gives you the distribution capabilities. Uh I think only more really is is you got to own your source of intelligence, right? I was just talking to a a group of uh folks innovating in the local AI space. Local AI is having to run your own AI in your own machines.

[00:17:06]>> Yep.

[00:17:07]>> Yeah. Last week we had um the um AI engineering summit uh in San Francisco. It is considered to be the largest AI conference in terms of AI practitioner conference.

[00:17:20]Uh it's the most exciting conference, I think, for for AI engineers. It's a It's not a research conference in the sense it's more on the how do you use AI and how do you fine-tune and whatnot. It's more on this side. And you can see the excitement for local AI there in that room, which was filled by It was organized by Nvidia, uh but it was filled by folks like Ahmed Osman, who's getting really popular now on on Twitter for his uh stance on local AI or Alex Shima, who's a part of Exa Labs and the We saw a photo with him and Jensen Huang talking about uh local AI. So, Nvidia is getting extremely excited about local AI. They're actually building machines you can buy to uh GPUs. They're called DGX Box. They're beautiful machines you can can put them in your house and you can cluster them. You can actually run GLM 5.2, which is a very capable model, uh with a decent performance. I I saw the demos yesterday. I was I was blown away.

[00:18:15]Um so, there's this movement, uh you know, on on the ground that's brewing.

[00:18:21]And I was surprised how packed that room was. It was I I couldn't even find room to stand. It was such a packed room. So, you can clearly see the kind of questions people were asking and the kind of interest people are showing towards running their own models locally uh is uh is a sign that people are going to own their own compute. And these are normal engineers, they're normal people.

[00:18:42]These are not companies we're talking about, right? So, um it's heading there for sure. Um we are we saw that coming a few years ago and we actually built a product called Akash Home Node, which um which lets a home computer, like a laptop, contribute uh the compute to uh to the cloud, to the Akash Hour Cloud. Uh Granted, we are focused on GPUs more than computes and more than CPUs there because GPUs are um you know, something like a Razer laptop, which is a gaming laptop, has a very powerful GPU, uh a 5090, which is extremely attractive for a lot of developers and extremely capable when it comes to running different models.

[00:19:21]So, we're seeing quite a lot of folks that already have GPU compute you know, if you can do it in a cash all or people that want to get developers that want to get local AI and running on running agents on them are, you know, sitting on latent supply, right? So, it's a really exciting time to be in for local AI.

[00:19:42]>> Very exciting indeed. You know, and I I being selfish, I always look at, you know, how am I going to be able to partake in the economy that we're moving towards? I think creators will be all right. I don't think everybody wants to listen to AI, but I could be horribly mistaken. I think people like a personal touch. I often think maybe I should start to get some hardware so that I can partake in what clearly forming a new economy, you know, an already existing economy, but it's evolving to the everyday individual, kind of like industries have previously. In regards to AI and that kind of broader build out, where do you see the bottleneck right now? Do you think it is in the hardware? Do you think it's software or do you think it's kind of energy, you know, that the three kind of components there?

[00:20:22]Where do you see the bottleneck right now? I think that's going to be very relevant to a cash.

[00:20:25]>> Oh, yeah. This is a This is an area I spent a lot of time. This is the area that I spend most of my time right now because our challenge is compute, not our demand.

[00:20:34]Uh we have four times more demand than supply right now.

[00:20:37]>> Yeah, wow.

[00:20:38]>> Uh it's incredible. I mean, the demand is that hot. It is just getting getting incredible. And before I get into supply, just to give you a little little idea what that demand looks like, right?

[00:20:48]So, so there are two ways we use AI, right?

[00:20:51]You can go to a chatbot and you can type in a bunch of questions and then get answers. So, that's one inference call for every time you chat with it. So, right? You use ChatGPT, obviously it's it's very obvious now.

[00:21:03]Um but a better way to use compute is through agents.

[00:21:08]Agents are where the chat is question answer querying the AI, whereas agents is having AI do the work for you.

[00:21:19]So, you know, if you do coding agents at cloud cloud code code or if you do cool work, which basically opens a browser and and you can just ask it to do anything you want on the browser, like write you an email or you know, um you know, go file your taxes. I mean, it can do a lot of things. It can do it's very, very capable. So, it's just like acts like a human being.

[00:21:40]Uh and thinks you feed needs to resolve their board challenge, right? So, so it's very capable and it using agents you can achieve incredible amounts of efficiency gains, right? So, in our company every engineer it has to use agents.

[00:21:53]>> Yep.

[00:21:54]>> At this point, we don't really work with engineers that don't do the agents because their productivity gain is like 10x.

[00:21:59]>> Crazy. Yeah. So, that said, so there are about a billion people that use AI.

[00:22:05]Out of which, how many I mean, what that do you use agents in your business or your >> No, we use them in the business. I don't personally use them. Adam that you met off camera there, he's the technical brains. I think I would be dangerous if I was left to that. I'd probably get it to do all sorts of things it's not supposed to. But, yeah. So, we do use it in the business broadly, but not I I personally don't use it. Should be, right?

[00:22:26]>> Yeah. So, out of a billion people today, how many people do you think it use agents?

[00:22:34]>> Less than probably 1%, I would say.

[00:22:38]>> So, 1% is what? Um like 10 million?

[00:22:41]>> Something like that. Probably. Maybe not even. You know, it'd be it'd be very small.

[00:22:46]>> Or we are.

[00:22:47]>> Or less.

[00:22:48]>> Less.

[00:22:49]>> Wow.

[00:22:50]>> be 2.4 million people.

[00:22:51]>> And I shot low there.

[00:22:52]>> Yeah, it's 2.4 million people.

[00:22:55]>> Wow.

[00:22:56]>> So, that 2.4 million people. So, and to give you an idea, using regular chat, you make one call per per message. You type in a message, you hit that's one call coming back.

[00:23:08]That's one inference call.

[00:23:09]Using agents, you thousand x that. Every time I use agents, you make a thousand calls back to the inference.

[00:23:16]>> Okay.

[00:23:17]>> So, the usage is enormous. I mean, the reason why these companies are complaining what ChatGPT or cloud tokens is because agentic usage is out of control.

[00:23:28]In our company itself, we spend as much as we pay people in their base salary on AI agents uh per person.

[00:23:35]>> Okay.

[00:23:36]>> So, in some cases it's much higher. Like uh I think my last month I spent like $11,000 on on inference alone uh from a from a cost standpoint. Not last month, but before I was before I switched to a cloud, I mean, that's a different story, but uh we are spending a lot of money on agents itself.

[00:23:53]That 2.4 million alone and another interesting metric is there are about 80 million people that write code.

[00:24:00]We call them programmers today. 80 million people.

[00:24:04]Out of that, 50 million [clears throat] do professionally.

[00:24:06]I mean, that's all they do.

[00:24:08]We We're talking about not even 10% of professional software engineers using agents.

[00:24:17]>> Wow.

[00:24:17]>> So, the penetration we're so early in terms of agentic usage.

[00:24:22]That alone is creating such a crisis today with AI supply chain.

[00:24:28]All the way from GPU access >> to >> foundational energy access.

[00:24:34]>> Yep.

[00:24:35]>> If we start out with GPU access, there is no depreciation in GPUs today.

[00:24:41]Like H100s, which are the workhorses of, you know, Nvidia chips, and these are not the latest and the greatest, but most most widely used. H100s have increased in price. They're nearly double in price. We're talking about used H100s. That's not supposed to happen with that economics was to go down in price.

[00:25:00]>> Yeah.

[00:25:01]>> You don't depreciate, you appreciate cuz the demand outweighs supply so much because of gigantic usage, the economics don't make any sense right now. Talk about 4060s which are professional grade GPUs for home usage. Now they they introduced them for $4,000, but they're available on the open market for about $12,000.

[00:25:22]>> Wow.

[00:25:23]>> A lot more than what the retail you know what you can get from Nvidia, right? So because the demand is so high.

[00:25:29]Um if you look at the components that make these GPUs, the most important component in GPU besides the chip is memory, high dynamic memory HBM. A company called Micron makes them. And if you follow Micron stock, it doubled in a day.

[00:25:43]>> How how can you not, right? I mean Yeah.

[00:25:46]>> Right.

[00:25:46]Uh and the reason because they're out, they stopped supplying HBM memory to consumers. They only do um enterprises now, companies like you know, and whatnot. And uh there are all kinds of speculation whether they actually are There's a US government investigation now, I think, on are they purposefully um uh squeezing out the supply, but that's a different discussion, but there's a enormous supply chain problem there.

[00:26:12]>> Yeah, are they trying to be Rolex?

[00:26:14]>> Uh >> Yeah.

[00:26:15]>> memory.

[00:26:16]So my memory costs four times more more than what it used to pay. All right, that's probably why the GPUs went up because memory uh in them uh and well Okay, so let's say you you you solve the bottleneck somehow and now we got to build data centers, right? Because the demand is so much, we're running out of data center space. Well, data centers need energy, right? And then energy and water. In order to get energy, so in US, well, you can do renewables, right?

[00:26:43]Solar, uh wind and whatnot.

[00:26:45]And you can do non-renewables, right? So fossil.

[00:26:48]Renewables are impossible to scale. You can build quickly a let's say a 100 MW uh data uh 100 MW capacity because there are all kinds of challenges with renewables.

[00:27:00]They're non-continuous power. That means they need storage. Our storage is extremely expensive cuz most of the storage of batteries are made in China.

[00:27:08]>> Yeah.

[00:27:09]>> And now there is geopolitical uh unrest that is creating all kinds of imbalances with trade. So, there's that you know, trade risk that's happening, but also just in general the mad is all hyper with the batteries.

[00:27:22]And the amount of cooling you need solar hyper batteries because usually these solar panels are installed in extremely hot areas.

[00:27:30]Um it just just not efficient to put solar in a data center. So, a lot of folks and wind and all those are not practical at scale. So, a lot of new data center build-outs are all LNG.

[00:27:42]Liquid natural gas. And dirty as hell, but besides environmental um impact to burn LNG, to convert LNG, you need something called uh uh generators, right? You need to you you know, generators have something called turbines. So, these things that thing. The lead time for a turbine today is 14 years.

[00:28:07]>> What?

[00:28:08]>> 14 years.

[00:28:10]The lead time for a transformer is 4 years. Transformer is the most basic thing you need to have a power plant.

[00:28:18]>> Wow.

[00:28:20]>> 4 years.

[00:28:21]Uh the And that's why you're seeing all these projects uh especially this target project Open AI announced last year, this half a billion-dollar as a half a trillion-dollar investment from SoftBank and US government was a big big you know, fanfare with Trump or whatnot.

[00:28:38]They paused the build-out because they can't get transformers. They can't they can't get um you know, turbines.

[00:28:46]>> Yeah.

[00:28:46]>> Uh so, so the situation is unsustainable to a point that people like Elon Musk so the world are thinking about space.

[00:28:56]Uh there's this new notion of space data centers because the advantage it has in space is continuous power from from uh solar from the sun and you can uh have a orbit that can fixate to the sun's like direction. So, uh on Earth, I mean, >> They're probably >> in a way the a lot of innovators trying to solve this problem. There's nuclear that's coming out. There's a lot of lot of work happening in the energy sector and and and Trump administration declared energy as a national security priority now. So, you get regulations cleared very quickly. Even with that, we can't move as fast as we we want to.

[00:29:32]China is the only probably only country that's in a way that's building up. Uh you know, being a communist power actually has some advantages here because >> No attitude. Yeah.

[00:29:42]>> move move at speeds that democracies cannot.

[00:29:46]>> Um >> So, yes, there is a massive crisis now for energy. I testified before Congress about this uh last year uh talking about the energy crisis that's coming and my testimony included an innovative approach as a solution where we have where we want to do distributed data centers. Happy to chat and I think that's really where decentralization has incredible opportunity to play.

[00:30:08]I think we got our ducks in a row, right? So, yeah, it's really bad, right?

[00:30:12]>> Yeah, you've got all these bottlenecks, you've got all these problems.

[00:30:16]Surely we're getting to a point where people are going to look for alternative ways to facilitate this and I think this is where a cache is going to start to shine.

[00:30:24]>> Uh absolutely. I mean, this is the exciting part. So, to explain you how where where the crisis is coming from, right? So, there two main ways uh AI uses compute. One is training, another one is inference.

[00:30:38]Now during training, it's a very resource intensive task.

[00:30:45]Particularly, it needs lots of chips connected together in the least amount of latency, that means the fastest way possible. And because a lot of them are talking to each other, right? They're rapidly talking to each other, and there is a lot of information that they exchange between these chips.

[00:31:08]And the bigger the model means the higher quality of the model, more amount of chips you need. So, as and I I think we're doubling every uh every 2 years in terms of how much energy you would need to train the latest and the greatest model.

[00:31:21]>> Yep.

[00:31:22]>> Um so, you know, as you evolve, fable, for example, needs bigger bigger computes, right? The next model's going to need much bigger compute.

[00:31:31]And that's how it's been done so far.

[00:31:33]So, of course, that's not scalable because you're going to hit a wall, right? By 2020, I think by 2028 or 29, we'll need a gigawatt data center to train the latest AI. A gigawatt data center is a nuclear reactor.

[00:31:47]>> Yeah, literally.

[00:31:48]>> Um I mean, that's nuclear reactor is the only one that can produce uh not only one, but that usually were the one that produces a bunch of my energy.

[00:31:56]And we know that's not sustainable. We can't build nuclear reactors fast enough. I mean, it's really sad what happened in Europe, uh taking down the nuclear reactors. So, the alternative is can you do distributed training? Can you actually have clusters distributed across the world to train? It's not like we don't have energy. We have a lot of energy. I mean, we have sun, we have, you know, we have the biggest nuclear reactor that's on at least half a day.

[00:32:21]Uh uh you know, fully full day if you take the whole globe into account.

[00:32:26]Um but transmitting energy is very challenging, right? So, you can't transmit energy from, let's say, UK to to to US cuz you You can't. Even amongst you in the transmitting energy from one city to another city is very challenging. Cuz there's a fixed amount of energy you can transfer through a copper wire.

[00:32:44]>> Yep.

[00:32:44]>> And in order to transfer more, you need bigger wires. And that's impossible to upgrade wires. Nobody wants It's a lot of challenges to upgrade the grid.

[00:32:52]So, instead of bringing power to where the computer is, how about you take compute to where the power is?

[00:32:58]>> Yeah.

[00:32:59]>> That's That's the area of research that's very exciting in in AI right now.

[00:33:04]Uh that's called distributed training.

[00:33:06]There are a lot of companies working on it.

[00:33:09]Somehow achieved more than others have achieved. Companies like PluralSight and whatnot are able to do uh distributed training for um I believe uh or even news research to do for about 100 billion parameter model, which is a decent-sized models.

[00:33:25]Granted, it's not a trillion parameter model, which the latest models are that are coming from Anthropic, but it's proven that you can do this level of training on a distributed computer network.

[00:33:36]Um PluralSight is another company that's innovating quite a lot. And another important innovation uh that's critical for distributed training to work is heterogeneous compute. So, in your regular training, you the compute has to be homogeneous in the sense you're If you're using H100s, you got to use H100s through and through and through. You can't mix and match your A100s or other GPUs.

[00:33:57]Uh so, most of them would prefer to use the latest GPU, right? Because that's the fastest and fast means lower cost and energy and more efficient whatnot.

[00:34:06]So, So, there's a lot of like older GPUs that sit in idle that nobody really uses them. So, if we crack heterogeneous compute, that means you can mix and match older GPUs and newer GPUs, then you have significantly little more more supply available, right? To to train.

[00:34:22]So, companies are innovating on it.

[00:34:23]Heterogeneous uh training is a is a big area of excitement as well as um This is cutting edge, by the way. This is extremely hard science, right?

[00:34:33]There are a lot of very powerful, very very smart people working on this, including Anthropic, including Meta. So, it's it's industry wide sort of like effort. As that becomes real, I think you're going to see decentralized network come to light in ways that was not obvious before because these decentralized networks, especially like Akash, has access to a latent compute, which is usually very heterogeneous in nature.

[00:34:57]You get compute everywhere, right? So, and the beauty of distributed computing or decentralized computing is you can tap into a distributed grid. Yep. So, I'm talking about a home. My home itself has solar capable of hosting good eight H100 nodes. I can tap into my solar.

[00:35:13]It's dual purpose, right? So, I have my solar that I can power my compute and I can use that heat that's coming from my compute to power my home.

[00:35:23]So, you have this incredible harmony between compute, data centers, and my my my life.

[00:35:30]Which I think what we need right now considering there's a massive backlash against data centers and water usage. So, the heat exchange is extremely attractive and that's why I'm a big proponent of having computer home. That's what I I testified in Congress about having computer home as we get this training to be real. And second more and more even more exciting part of having computer at home is inference.

[00:35:56]>> Yep.

[00:35:57]>> So, like I talked about how local AI is getting very strong, that means you can actually run GLM 2 5.2, which is very capable open source model.

[00:36:07]For context, GLM 5.2 is as good as Opus 4.6.

[00:36:12]Cloud and Anthropic Opus, which is at the previous model. So, it's very good model. I use it personally. Um I use GLM for coding a lot.

[00:36:21]But, having GLM run in your home gets you privacy advantages because if you have to offload your life to AI, you have to be concerned about how the data is being used. I wouldn't I never trust an open AI or never trust Anthropic because they publicly said they're using my data to train.

[00:36:38]I don't want to become a sample in their in their training run, right? I don't want my private information, my personal information to be exposed to internet in any way. And I have cameras and I have sensors and I have microphones in my house that's listening all the time. I can no way in hell that that that information is going to leave my home.

[00:36:57]So, you know, but for me to leverage AI to talk to my AI, right? In in my home or having sensors that detect danger, right? What what not in my home, um that's very intimate. It can track my heart rate and whatnot. Uh that should be in um that should be controlled and in uh the intelligence for which should come from my local GPUs and that's it. That's the big vision.

[00:37:20]So, uh that means inference is going to move to the house. And now, I'm not using the inference uh all the time, right? I you know, I can offload that inference to say you want to use it, right? Say um you know, and we have technologies like TEE or or or trusted execution environments which let let privacy uh optimized networks like Akash, for example. If you use inference on my computer, there's no I can't see what you did in the inference. So, we have technologies that can that can achieve that now with privacy guarantees.

[00:37:51]So, if you ask me the future of inference is going to be home?

[00:37:55]>> Yep.

[00:37:56]>> Um it's going to be a local and uh and that is an is an amazing uh what you call proof point or amazing um solution to the energy problem that that is looming.

[00:38:10]And uh and it's dual purpose. You can actually use that to heat your home and you're using excess capacity anyway. In California, for example, we we produce about 60 gigawatts of solar energy that most of us don't use and try to sell that back to the grid, you get peanuts on the Yep, on the dollar. It's useless.

[00:38:27]Uh so, use that excess energy to power your computer. I think it's a incredible symbiotic relationship between your data center, your entire your AI, and your life.

[00:38:35]>> This is what I love I mean, I love my job and I love talking to people like yourself because I don't think people quite grasp and I love spreading the word on it just how important the likes of what a Akash Network is trying to do.

[00:38:48]And I've interviewed professors at MIT.

[00:38:51]I've interviewed I think I think just the other day we had an individual from the University of Sydney on. Uh both these people their their their sort of field of academia is decentralized systems.

[00:39:03]And they've kind of merged to crypto now as a maybe practical way of of of implementing decentralized systems. But one thing that was concurrent was just the efficiency benefits with decentralized systems.

[00:39:15]Actually, it goes further than efficiencies. We've spoken about how this can solve some of the current issues facing our world.

[00:39:21]It can also solve a lot of the social issues and these kind of techno-feudalistic issues that we've spoken about throughout. So, networks like Akash and I'm not just saying that cuz you're here on the show. I've always been very interested in this kind of topic of deep in decentralized physical infrastructure are extremely important towards that direction between dystopia and utopia.

[00:39:44]Effectively, you know, which which which one gets chosen is is going to really depend on having the right competitors, I think. And I see Akash uh as that and I think it's very exciting the prospect of this, even for myself as an individual, maybe being able to take advantage of this new economy. There we go. We've got some some fireworks there.

[00:40:03]But I think hats off to you guys, you know, very very interesting from the inception to where we are today and the problem that it's it it is being addressed. I mean, this is kind of at the forefront of more than just efficiency. You know, the world as we know it from a social point of view.

[00:40:19]>> Sovereign do you like? So, it really comes down to I am big proponent of owning your your own stack. I mean, that includes energy to water to compute. Like, having solar, having batteries means you are free from the grid.

[00:40:35]>> Yeah.

[00:40:36]>> And you know, if you think if you look at the grid as much as I have looked into it, you will want to own your own computer because you have West is built on a very fragile grid.

[00:40:48]It's very fragile. I wouldn't call myself a prepper, but I do have um I feel like the more I look into it, the more anxious I get >> Yeah.

[00:40:57]>> about how fragile we are.

[00:41:00]So, owning uh energy is extremely important. Uh owning intelligence is extremely important.

[00:41:05]Like, famously we saw what happened when US government you know, shut off Fable.

[00:41:11]>> Yeah.

[00:41:12]>> Now, I was actually using the model and it went away.

[00:41:14]It was incredible model. So, if I had to depend on AI to do my work, to live my life, I mean, I use AI for everything now, right? Like, I don't I don't I barely use Google. I barely use any other tools for knowledge, you know, or gaining knowledge. I mean, AI is what it is. So, if you had to take that away from me, I mean, that's not something that I want um to be taken away. So, I want to own my intelligence and having GPUs at home, that's your own intelligence. Akash gives you the economic uh case as well.

[00:41:45]Like, it's not only for rich people now.

[00:41:48]You can actually have a GPU.

[00:41:50]Uh for example, you can actually have a 5090, which is a very powerful GPU, and your home. Put the Akash home node, and if you lease it out, you can you you can amortize your entire investment within 8 months.

[00:42:04]>> Wow.

[00:42:05]>> Get back all the money you spent on your on your compu- chip alone in 8 months.

[00:42:11]>> Wow.

[00:42:11]>> So, after which it's your own chip. It's fully yours, right?

[00:42:15]making a an economic case for owning computer I think is also very important if you want to get my mass adoption. So I can see a future where you know a state-sponsored loan for your computer will is not is not something out of a out of fiction. I think it can be real I mean can be a reality where the government can sponsor your computer.

[00:42:39]Get a loan and use a computer to train larger national car clusters and your contribution is your roof space. If you have roof space that's your biggest value add because you can put solar on the roof and that can power your computer and then that computer can be used by by let's say a government for for doing and you own the computer and they you know they advertise the thing for you.

[00:43:02]So totally can see a future.

[00:43:04]>> Oh oh not even a good I mean it doesn't even have to be a government. You had Larry Fink on stage recently talking about he believes in new asset classes kind of future compute, right? You mean maybe they did maybe they give you the loan for that. It's a viable business for or >> It's a viable business actually now. Now you think about it like you there there there new lending programs that you know a bunch of crypto as well lending for computer markets, right?

[00:43:25]There there new lending programs that are popping up here and there to power power your homes.

[00:43:30]>> Yeah.

[00:43:31]>> Yeah.

[00:43:31]>> Why aren't more people taking advantage of this? I mean >> We got to spread the word.

[00:43:36]>> Is it an educational thing? Is it because people don't know about this? I mean we I certainly feel I live in a bit of an echo chamber being in the cryptocurrency space and regardless how many times I talk about some of the properties or something people just don't get it. Is is that why more people aren't getting on board with the cash and taking advantage of an opportunity that's arising >> only 2.4 million people that even know how to use agents. So there are enormous information gap in terms of what people know, what people can use, what people cannot.

[00:44:07]So, I think that has to be solved, it will be solved, right? And it's getting extremely popular. Akash, our usage in for Akash email, for example, is doubling every month. Like we are doing Akash email, for those of you that don't know, is an inference as a service on Akash. So, like I said before, I use GLM more than Opus now because it's open source and it's much cheaper on Akash. Yep. It's very advantageous to use Akash. Our usage, our tokens, our average tokens per day was about 5 billion last month, and now we're doing about 10 billion uh tokens a day.

[00:44:42]Uh which is a significant amount of tokens, right? So, it's it's very clear the usage is is gaining, and it's very clear the usage is coming from agents.

[00:44:50]Um so, it's getting there, and definitely in the early early phase now with with compute. And people, as they realize, you know, the government control, you know, you saw what happened there. What the US government did last month with with Claude was an eye-opener. A lot of people have the local crypto um AI meetups. I can see the interest.

[00:45:10]These are people that you would consider to be fairly knowledgeable. These are These are people at AI conference, and you can see by the questions they're asking, they had no idea you can do this stuff.

[00:45:20]>> Yeah.

[00:45:21]>> The word is spreading, so it's not it's very early stage, right? It's very cutting edge right now. The word is definitely spreading for sure.

[00:45:27]>> I guarantee there's people watching this that have got hardware. I know I I I I I I know for a fact there is cuz we we get asked about it quite a lot.

[00:45:35]Let's maybe be a bit more layman. I think we've covered a quite a high-level overview of a lot of things, the world at large, and and and how all this technology is relevant. But I know for a fact there's people out there that have hardware today. How could they get involved? And I guess in explaining that, we're going to really explain Akash and that whole of ecosystem.

[00:45:56]If you go to home.node.akash.network, we built a consumer grade product.

[00:46:04]And if you have a Windows machine, you can download it and you can run the installation and you don't have to touch a thing. There is a wait list because we're on boarding compute >> Of course.

[00:46:15]>> slowly as we as we get more demand and demand is increasing. This is for 5090s.

[00:46:20]Um it's very easy. It's home based product and we also announced a partnership with Razer account the laptop company and they are also you know, this product is designed for Razer nodes as well. So it's very optimal if you have a Razer compute, but you can do it on any computer really doesn't matter.

[00:46:39]Any gaming machine can become part of your part of Akash network and the way it works is you install a program. You don't have to touch a keyboard. It's like click click click.

[00:46:48]And then when you're ready to let's say sleep, you can give the machine to Akash network.

[00:46:55]And the machine, you know, Akash takes over the machine for the time it's on the network. And when you're ready to you know, get your machine back and click a button to reclaim and takes about 10 minutes to reclaim your machine during which time if there's something running on the machine, that workload gets transferred out and safely like transferred what not.

[00:47:15]And you can reclaim and you're earning money for the time that's on that's on the network. And right now the economics are incredible. I mean right now per 5090s, I think they're paying out and this could change cuz we're in the process of doing that.

[00:47:29]Um I think they're paying like 20 cents per 5090s. So that means you can earn your investment back within 8 months on your 5090.

[00:47:39]This provider you have 24-hour you know, contribution. If you're not doing 24-hour, something takes longer because you're playing. The more you play, the less money you earn.

[00:47:49]>> Yeah.

[00:47:49]>> Not saying playing is bad, but but there's an economic cost. Now how much you want to play play your games.

[00:47:54]>> Isn't that interesting? Yeah.

[00:47:55]>> Yeah, it's funny. Yeah.

[00:47:57]So So yeah, I think I think you know, well, there's another another case to be made like hey, less you play, more games you can buy.

[00:48:05]>> Yeah.

[00:48:07]I mean I'm I'm a big gamer. I feel like there's a lot of us. The amount of people that I come across in the crypto space that were just massive in wow and stuff like that is is brilliant for me as somebody that was. Where does the token play a role in all of this? I feel like we're coming to a maturity in the crypto space where actually now when I have conversations, the token comes up more and more, whereas before it was like shh, don't mention the token. Now we can actually talk about it and it's important. Where does the token come into all of this?

[00:48:35]>> AKT is a new token for Akash and it's the only way to pay for compute.

[00:48:40]>> Brilliant.

[00:48:41]>> It implements something called a burn mint equilibrium pattern mechanism where every token is burnt to use compute on Akash.

[00:48:52]So the more more demand we have, which is a function of more supply, means more burning of AKT tokens that creates higher demand for AKT.

[00:49:03]When you only way So in order to use AKT or use Akash in general, you have to go market buy AKT.

[00:49:11]And and use that to to pay. So the more demand we have, the more market buying pressure AKT gets and the economics are inter-twined into Akash, right? So it's uh um it's it's fully mature token, it's been around for 6 years and it's a very liquid, very well distributed token.

[00:49:29]Uh I think it's one of the oldest tokens that are that are liquid right now. Um and and the tokens also have a governance kit. So all the changes that happen on Akash network is is voted by the token holders. There's a take rate as to how much percentage that it goes to the network, that's also voted by the token holders.

[00:49:48]And you know, with more usage, the more scarcity it gets, very very straightforward economics.

[00:49:53]>> Yeah, I love that we're at a point where honestly I've been interviewing people in the crypto space for a long time, many, many founders over the years, and I feel like it's only now that we're at a stage where we can actually talk about the token without any kind of like oh, Gary Gensler might be listening or, you know, any of this kind of stuff. So that that for me is really liberating and I feel like um >> Yeah, you mean Gary's listening? You mean you don't >> Yeah.

[00:50:18]>> It's all well distributed. Fine. No one even controls the network. Like there's no single player that can control the board. Not Overclock Labs, not no one I know can control the board actually because it's very distributed.

[00:50:32]Um it's very decentralized, so you know, Gary's listening, good luck to you. Go you do your research, it's all public, right? So So yeah, I mean that's what we want to achieve with like distributed I mean and the beauty of using Akash, if you use the Akash model, the all the token mechani- mechanism is entirely hidden. Yeah, entirely abstracted.

[00:50:52]Cuz how it should be.

[00:50:54]You know one cares about crypto as much as I want them to care, but no one really does.

[00:51:01]People always choose comfort over you know, or I say values as opposed but people all they want is to earn money and all they want is to spend money.

[00:51:15]Uh and then money comes from fiat. Yes, we do all the conversion in the background. You don't even deal with any of that stuff as a user. As a provider you have a lot more to be dealing with compute because with with AKT because that you know, you need the economics are designed in a way the providers need to have some AKT in order to to bid what not. There's a lot of spam protection that happens, but as a consumer you don't uh have to deal with AKT.

[00:51:42]>> Yeah. Yeah.

[00:51:43]>> And it's very important to to to a good user experience.

[00:51:47]>> For sure, I think history's taught us. I mean, PayPal's always a great story about how they won. It was really down to UI versus anything.

[00:51:54]Um you know, there there was other competitors, other people have tried that. It was the UI that basically made that uh the success that it has been.

[00:52:02]And of course, um still in many regards is today. And my word, the time, it's it it's 2026.

[00:52:08]A lot's gone on around the world and with the world. And it's it's it's a kind of crazy world that we live in right now. But where's your focus? You know, what what are you looking to do with a cash this year? And what's kind of the top of your priority and push?

[00:52:23]>> Yeah, so we've done a lot this year. In fact, we should be ME. We shipped uh gosh, I'm all we shipped God. So, I mean, our road map we we've evolved the protocol in terms of provide flexibility. We're we're shipping very soon we're shipping uh TEE or trusted execution environments of privacy give privacy guarantees that that we the cash you get much stronger privacy guarantees than what we have.

[00:52:51]>> Yep.

[00:52:51]>> Um from a product standpoint, there's enormous amounts of features that are really focused on making it lot more private and lot more uh easier for providers to jump on. Like I just talked about home node.

[00:53:09]It's also we ship this year.

[00:53:10]Home node is a just a tip of the spear in terms of what user experience uh um um um improvements we can achieve by abstracting away. Like home node is designed for home computers, but we're designing a similar experience for data center computers as well.

[00:53:27]Now, I don't want to get in too much in the data center, you know, compute because the way that comp the way data centers are managed is not of course not uh how you manage your laptop, right?

[00:53:37]So, it's a very different set of tooling, is a different set of technologies, different set of um uh enablement we need to have for improving or rather reducing friction for uh for data centers to come on board. And the reason I'm saying that is right now the biggest bottleneck for Akash growth is supply, not demand. Not demand.

[00:54:01]Demand is growing. It's I I don't I don't know when it'll stop, but it doesn't look like it's going to stop anytime soon. So, demand is of course we keep improving the products to get more demand, but I we don't have enough compute to just service our demand.

[00:54:15]Uh we need compute. Um you know, to get compute it's not incentives alone because there's so much demand for these compute players. So, it's really the user experience. It's really building a product for data centers and and compute providers that uh matches the quality of a consumer product. So, there's a lot of like user experience, a lot of like product building that that happens uh on the computer side um that we're very excited. So, one of the we we have a new marketplace protocol that's coming up that allows for bare metal compute, uh allows for virtual machines, allows for um micro VMs, all all different stuff type of compute that's out there to come on board Akash Network, right? So, there's a lot of uh provider diversity we're going after. Um also from a user standpoint, Overclock Labs is a company that that builds Akash Network is also going through a big evolution.

[00:55:05]So, we we we don't depend on Akash to as a business. We actually have our own business that we're running, and we are looking to acquire computer ourselves because the demand is so much, and the economics are just doesn't make any sense for not to own computer.

[00:55:20]So, Akash Overclock Labs is transforming to a uh enterprise company. Uh we have multiple signed contracts now with with large players, multi-million dollar contracts uh that are signed with enterprise companies for for compute through Akash Network. And also there's when you when when you have to go to get more demand, you have when you're dealing with enterprises, it's a different mechanism that you you you go through versus dealing with uh regular consumers. Enterprises want SLAs, they want clients, they want cloud integration with their internal systems.

[00:55:53]There's quite a lot of stuff that is not open source. You can't really open source this stuff, right? Um so, Overclock Labs is handling the heavy lifting in terms of the legals and the compliance and the requirements large enterprises need, the hand-holding, the support. They want call somebody in the middle of the night and have that somebody pick up the call, right? So, there's certain things that enterprises need and that's what Overclock is providing. That in turn goes back to Akash uh in terms of economics, right? So, uh we have an amazing symbiosis between Akash and Overclock Labs, where Labs, I believe, should be profitable as a business based on Akash Network that doesn't rely on tokens.

[00:56:32]>> Yeah.

[00:56:32]>> Right? That's very very very important.

[00:56:34]>> Yeah.

[00:56:35]>> networks to thrive. So, a big problem today with with crypto foundations are not designed as for-profit entities.

[00:56:43]So, when you don't have for for-profit entities, it becomes a sustainability problem.

[00:56:49]>> Yeah.

[00:56:49]>> Yeah, you got a few years of runway, but if your runway involves selling tokens, you're not going to sustain yourself in the long term. Look at Ethereum Foundation, look at pretty much any foundation.

[00:57:01]>> Yeah.

[00:57:02]>> You look at why these prices are being suppressed or why you know, it's seems like a charity more than a company.

[00:57:12]>> Yeah.

[00:57:12]>> It's because there is incentive misalignment. There's no incentive for Ethereum Foundation to actually go build a business. So, but we have an incentive to build a business uh because this market is incredible. AI is incredible.

[00:57:22]I don't know what they will do in the future.

[00:57:24]>> Yeah.

[00:57:25]>> Yeah. It's amazing to be in this business, right? So, we're not going to drop that opportunity. And being on Akash gives us a leg up. Yeah, they're an advantage at this business.

[00:57:34]So, that's our moat. Akash Network is our moat and I we have a lot of inbound because of Akash Network because it's a household brand at this point, especially in AI.

[00:57:44]Uh so, a lot of the deal flow comes from Akash Network, right? So, we have this amazing deal flow, amazing brand, amazing um moat that we we're building with Akash Network, and that's contributing to our our traditional business. So, I'm very excited. I think this year is going to be uh a an Akash will be an amazing uh growth.

[00:58:02]We're seeing incredible like it's a SaaS-type model. We're we're seeing subscription revenue coming up going fairly high there. So, you know, by end of the year, we'll be uh we'll we'll be much larger. We're hiring quite a lot. And uh and it's it's nice to pay people with money we earn through outside of crypto than than inside crypto.

[00:58:22]>> That's so refreshing. We deal with a lot of foundations, and honestly, I can tell you that some of them are interesting in terms of how things are funded, and all all raising that question, thankfully, in terms of look, this is not sustainable. Like you mentioned the EF Foundation or the Ethereum Foundation reducing trying to reduce uh emissions.

[00:58:43]So refreshing to have that take.

[00:58:46]Sounds like you've got a busy 2026 ahead of you. Sounds like you've got a busy lifetime ahead of you, I think. Forget this year in terms of building this out.

[00:58:54]Um where can we send people? A lot of people are going to be very curious after watching this. Where can we send people? And we'll leave links to all of that in the description.

[00:59:03]>> Akash Network on Twitter is the most active uh place. And Greg Osuri is my Twitter account.

[00:59:10]Reach out to me. Um I'm looking for compute providers. If you are interested in making money, reach out to me. We are hiring quite a lot. Uh we're hiring engineers. We're hiring DevOps folks. I mean, we're Like I said, we're a good business, right?

[00:59:22]So, we're hiring uh forward deployed engineers, we call them. These folks are folks that go into companies on behalf of Overclock and Akash and you know basically deploy them.

[00:59:35]We're looking for technical folks and we're also looking for marketing content folks. We're hiring a lot so yeah so reach out to me on on Twitter.

[00:59:44]>> You know what to do ladies and gentlemen we could have gone down so many avenues.

[00:59:50]I feel like we've given a really great introduction into a cash but more broadly the problem in the world that it's trying to solve which I think is important. There's a lot of cryptos out there that have created their own problems and trying to solve them and not doing a very good job with that but it feels like a cash is really trying to address a real world dilemma in a way that actually gives back sovereignty to the individual etc. etc. Which is always great to see so truly love my job it's been an absolute pleasure interviewing you Greg. Thank you very much for your time.

[01:00:21]>> Thank you so much this was a lot of fun.

[01:00:23]>> [music]