io.net tackles growing demand for GPU resources in AI

0:01I'm Ashen Addison from the cryptocoin show and today on blockchain interviews we have Tori green CEO of IET Tori welcome to the show and thanks for taking the time thanks for having me Ashton very welcome uh excited to dive into all the updates with IET um some big news uh major announcements I would love to start off our conversation first of all by hearing a little of your

0:23background a bit of your background uh in in your fields of expertise and and how you got involved with IET and then we can dive into all the latest details sure um so started off my career uh in a pretty traditional fashion was uh Investment Banking uh PE corporate strategy BC as a VC I was really more of an operating partner so I took on you know

0:52coo cro CMO CFO roles for compies this was a stable job I liked it but you know I I I always kind of knew that I wanted something more um and that's when I really started to get into crypto in 2016 I went really really deep down the rabbit hole and was you know like most people was just totally Blown Away by the technology I thought like hey this is the you know perhaps the

1:18most important technology invented in the last 100 years that said I was a little bit hesitant to go into the Space full time most notably because I felt like most protocol s were effectively Solutions looking for problems that is they weren't generating real demand that all changed in 2019 when I was the COO of a machine learning fintech and I got to understand

1:43firsthand some of the problems facing AI most notably the shortage of compute and uh that's when I had a realization that turned into an obsession that decentralized networks were not only the best way to solve this problem but probably the only way um that was kind of the point of no return for me and um you know that eventually led to the creation of iron now um it's

2:09been a very interesting Journey along the way it's uh you know there were uh it wasn't an easy Journey there were a lot of naysayers um a lot of people didn't believe in the tech um but fortunately we you know we had a a good group of allies as well we had a great set of investors and we had a great team and you know quite frankly together we worked our asset off um we I think we

2:32were in 14 countries over one year um none of the team uh took a day off for a year and um I think we ended up building something pretty special we we raised over 40 million from Top tier investors like multi-coin hack six-man Deli um air Foundation got a day one listing on binance and uh we built the world's largest AI decentralized compute Network

2:56wow that's amazing and I'm so glad that you did jump in to the industry um because yeah it is it is so exciting and uh just the the breakthroughs that this technology and it it resonates with me what you said about yeah it's sort of like looking for a problem but I feel like as AI gets bigger uh there's going to be a lot of problems that uh this kind of technology is going to solve and

3:20maybe the solution was ahead of of its time of the problem um but I feel like now it's it's all catching up in like the coming years here absolutely yeah and now could you I I love the name by the way IET it's like as simple as it can get I I know a lot of people have seen it around but I think some people have a hard time grasping you know what exactly it is because you know AI can be

3:44confusing and these compute networks could you try to break it down a little bit in layman's terms of what Iona is capable of doing and what the team has been working on sure so I think the best way to describe this is we have to sort of um if we really want to break this down like understand a GPU is um a GPU is a type of computer um yeah I'm vastly

4:06oversimplifying this but it's the type of computer that is used to basically create AI so it's used to build AI models it's used to optimize them and it's used to run them um now as we all know AI has been like pretty much the fastest growing technology ever um I think uh chat GPT was like the first app to 100 million users it blew everyone else away um and this is great but the

4:34problem is AI uses a ton of compute power um in fact if you look at the growth of compute power needed it literally doubles every 3.4 months so that's a 10x increase every 18 months and we've seen this you know Zuckerberg just came out with a new llama and he's like this will require 10 times more processing power than um than the last version and so you have this just

5:00enormous demand for compute and quite frankly the supply is just not there um at least in the traditional Cloud because uh people go to get most most people go to get their gpus from um like Amazon um Google or Microsoft now there there there are problems with this there's multiple problems with this centralized model but most notably they can't scale um you know number one uh

5:26you know there's shortages of they have to buy new G gpus there's always shortages uh number two they have to consistently build new data centers and that's a problem that requires construction which is a long lead time it requires permitting and there's so there's all these Capital expenditures that they have to face and so this persistent shortage really leads to

5:45multiple problems for the consumer um it leads to low availability long wait times and high costs um what ironette is is effectively we're we're completely rethinking how people get compute and I think the best way way to describe it is we're building an Airbnb for gpus right because there's actually all these underutilized sources of compute um there are thousands of data centers in

6:09the us alone that are running at like 12 to 18% capacity um every consumer has a GPU like I have a GPU in my Mac you you know you have a GPU in whatever you're using um and so that's another source and then eventually you know I think that mobile devices will be able to contribute to this too so we're able to go out and aggregate all of these Global

6:31Sources into a single platform and this just creates multiple benefits for consumers um you know it it creates access to nearly unlimited computing power it's um highly scalable um it's censorship resistant in some cases it can perform better and perhaps most importantly it is uh much much much cheaper yeah that's that's great to think about but I feel like there's a

6:55lot of factors in that you know when a company goes to traditionally goes to Amazon and they're like yeah we need a lot of compute power um you know Amazon is trillion dollar company that they can trust and now they're like okay we're going to you know the decentralized version or let's go get compute power from all over the world is it as strong is it as cheap is

7:17it as reliable I feel like there's a lot of things that I'm sure your team has thought through maybe you can uh sort of do that compare and contrast of how is it better and what do you need to think about with the reliability and the cost of a decentralized network for compute sure that's a great question um the best way to think about us is kind of like a a price line or a kayak

7:41or booking.com um we're a network of different sources of gpus and so you know just like you can go to Price Line and you can get um Emirates first class or you can get you know spirit for super cheap you can do that with iron um you know a lot of our data centers are actually like they are like large legitimate data centers they've got all the bells and

8:04whistles that Amazon have they've got the the latest h100s A1 100s they've got Envy link and so basically you know you can go to IET and you can literally fully recreate what you get even up with sock 2 compliance at AWS for maybe you know 30% cheaper um there are some companies that say Hey I just need like really cheap costs and you know maybe for them we use um we go to Consumers we

8:32we use consumer cards like 409s um you know maybe they're a little slower um but there they can get um 80% off and there's other services like interruptible Computing which I won't go too much into but there's a lot of different ways the user can customize so I think the short answer is that it it's really about choice I mean you can you can go really cheap um you could go for

8:56same quality at cheaper and then there's some cases is where you could even actually get better performance and I'll give you a couple examples of this um let's say you have um an image generation app right and uh you have a user from Vietnam that says hey you know draw me a picture of a dog with a hat well if you're using AWS that may have to go from Vietnam to the us back to

9:23Vietnam so little latency there we could set a user up with a data center in Singapore which is a little f even more importantly let's say that you are a hedge fund in New York same idea right like are you going to use Amazon's data centers in Texas or maybe we can find a data center in Upstate New York for you which is just a little bit faster which if you're a hedge fund and

9:46you're using AI that could make you billions for just those couple seconds of latency yeah no that's a great that's a great point so with the the decentralized network you're really able to get compute from all over the world and now is any you mentioned like consumer devices I'm sure there's lots of people that would be interested and potentially earning if they were able to

10:10contribute to the network is that how it works for say I have some extra Max laying around in my studio or can people contribute extra GPU power overnight that they're not using they can okay anyone um you know the network supports consumer cards like 490 um that said there's obviously more demand for the Enterprise cards but I think that um I personally believe that

10:38will change as you know technology continues to grow you're going to see a lot of um you're going to see a a lot more applications for Consumer cards and eventually I don't know when it's going to happen I think mobile devices as well yeah that would be great I'm sure everyone has extra mobile devices laying around uh you know since the new iPhone 16's just coming out and people are like

10:59just want to keep getting the next device every year you have so many devices laying around that could be used to you know power AI or mine Bitcoin or something um I feel like there's a severe lack of uh of utilization of these Technologies yeah and that's why I think that um this is really the future um you know they like the demands for compute are so immense

11:22that and you know there's all this underutilized Supply laying around it's just going to be really exciting to see how this grows definitely and it's hard it's probably hard to quantify because it continue continues to grow and it's such a large number um but you mentioned at the beginning you know one of the problems is these AIS need more and more compute um and maybe

11:44Amazon and these other centralized parties alone can't solve it in comes IET brings in more power I'm guessing that it's still a balancing act to keep up because AI is only getting more powerful and requiring more computation are you guys is the industry going to ever be able to keep up and have an excess Supply that you know AI there will be extra computing power laying

12:09around or will AI just keep growing I I I don't think so um I think Chris Dixon said something very insightful that um you know the demand for compute always outpaces the supply um you know as we said uh compute requirements for AI are doubling every 18 months um so even with AI you know that's like I that's theoretically a hundred times every three years this demand is just

12:34immense and AI is really only one use case for gpus um gpus are also used in video transcoding they're used in rendering they're used in iot they're used in zero knowledge um and especially you know as you start to think of like metaverse being built um yeah like the the compute requirements for for rendering are going to be huge so I think that we are going to consistently

12:58see a shortage of Compu M yeah that metaverse uh you know I think the hype was ahead of its time but eventually there will be a lot required U where there's actually a good metaverse that has good graphics and you know powering real life HD in in in real time is going to be a insurmountable amount of of GPU power required oh 100% yeah and now with IET tying into sort of like these

13:29decentralized network is there specific blockchain technology that's involved as well are you working with certain blockchains or um is that sort of separate yeah um we are a uh you know we're we're a crypto company we right now we run on salana um you know and salana has been a fantastic partner we chose them because we do uh an enormous volume of transactions you can actually

13:55get all of our transaction figures on our website but um you know salana provides a couple different benefits for us uh number one um they're hypers scalable and number two they're low cost and um actually if I would highly recommend you check out an app called bc. it's um it's an app we built I believe it's the first fully decentralized um aid app and that means

14:21that uh both the payments are made fully on chain and the compute is fully decentralized but if you look at this app every inference that user makes and an inference is just like you know it's an image generation app so going in and saying like you know give me a picture of a cat with boots um there's actually four separate transactions there U number one you have to pay the app

14:42creator number two you have to pay the model Creator uh number three you have to pay the uh you have to pay IET and four you have to pay the compu provider so every inference is like four to five transactions um and these are all microtransactions right they're like a thousandth of a Cent and so you know we needed a network like salana to be able to both uh support this high volume and

15:04this low cost that's really interesting and yeah I think M you know it's it sounds scary when you have to pay all this stuff but it's so cheap uh that um and for it to all work in the background uh it's nice to see that salana is able to do that and I feel like the salana ecosystem I'm continuing to talk to projects that you know despite the second half of 2024 you

15:27know Bitcoin sort of going sideways or you know the people that aren't in the industry sort of like Get not not getting excited I feel like the salana ecosystem continues to grow I think I just saw today the amount of active addresses and transactions is still at all-time highs um despite you know sort of the boringness of the trading market right now well and that's the great

15:46thing um I think another reason that we're very happy with salana is uh both the team and the community have been great I mean we actually officially launched a breakpoint in Amsterdam in November and you know not only has the team been super helpful and supporting us but uh the community as well it's just it's it's a great Community um and you know I think perfect place for IR

16:07that's really cool and now with the with the transactions as say for example we're we're lending some GPU power and then there's a maybe you're being rewarded for it um or say you're paying for compute power is that all powered by uh stable coins on salana or the io token what are the options for that sure so um primarily through the io token now it is interesting users can

16:34choose to pay in Fiat because a lot of our clients or web to they don't want to deal with crypto um so they can pay with feat on the front end but then that gets converted and then all the suppliers are paid out in IO um and really you know when you talk about IO there are a couple different utilities to the Token um the first is it is a utility token

16:54it's it's used to pay for goods and services um perhaps more importantly um it provides network security so IO is used for staking and so basically what happens is a user uh sorry a compute provider comes in they stake a certain number of IO and it effectively acts as collateral and so um if if everything runs well they get that back they get rewards in terms of IO for

17:20providing compute but if they act maliciously you know if they uh or if they don't perform if they shut down too much then we can slash the IO and so this basically contributes to the Safety and Security of the network um finally IO is um uh it's it's you know we have plans for it to be more of a governance token our long-term vision is always to make this uh you know we're following

17:43Progressive decentralization so we want to move to more of of a dow based structure we're already setting up things like a grants Dow um token based voting and so that will be another use for iio in the future that's great to hear and yeah I feel like the mechanisms for uh incentivization and disincentivized actors is it those have to be put in place because how do you

18:10you know how else can you monitor especially in a decentralized nature uh whether you know somebody comes with a lot of compute power but then they use it maliciously or their up time they keep going down and it's like well how do we how do we lower the ranking of of these people and make sure that the people who have the greatest compute power are at the top

18:30well how do you know they're a GPU right it sounds trivial but we actually have something called proof of work um where we literally when we get a GPU in we like run it for a second to see if it is a GPU and then the second thing is how do you know what they say they are um you know it's it's very easy for someone to come on with a consumer card and say

18:48hey we're an interpr card so we our Tech team has all kind of different um strategies we have our proof of work system to try to mitigate this um we also are going to have a program that we're releasing where we actually release data to the community and so they can go in and they can find uh people that are spoofing different cards uh and they'll get a piece of those

19:07block rewards and so that's why the staking and the toonomic security is so critical and with that staking for the collateral for compute providers that makes total sense but if there are other people that are just interested in the mechanisms of it can anybody stake and help secure the network uh so right now it's just suppliers but we will be rolling out

19:29something very soon that allows anyone to stake okay that's good to know you can stake on different machines and with the the the progressive movement towards a dow I feel like that it only makes sense there's always there there's so many faucets of running a company especially one that needs to support you know trillion dollar companies of of AI as it

19:52continues to grow um where do you start you know and and how can you ensure that it sort of fair and also doesn't limit the the fast growing nature of a company uh you don't want to have more bureaucracy or more barriers to entry by introducing a dow structure sure and that's that's kind of why we chose this idea of progressive decentralization um you know in order to disrupt the market

20:16you do have to move fast and so we you know we're we we started off as more of a centralized entity um obviously you know we were we were just IET Inc um and then with the token launched we took the step in decentralization which is creating the foundation um and then you know over time as we like the idea is move fast start to capture market share and then eventually over time um you

20:41know start with things like the grand sty and then token based voting and then eventually just move it to a fully decentralized Network that's great and you know we've covered a lot uh about the actual compute Network the token the Dow um what are some of the major Milestones leading up to the of 2024 and and into next year that IET is looking to get it get

21:05done sure so I would highly encourage anyone who's interested in learning more about our long-term plans to check out our stateof the network address it's located on our website and it basically covers where we've been um what our goals are for the rest of the year and uh what our goals are in the long run um but you know to really describe this in a nutshell I'd say

21:29what we are trying to do is pretty simple um we're trying to disrupt AWS as the leading supplier in the multi-train on our Cloud market now how do we do that um simple not easy but it's just build a business that can compete and exceed with AWS and gcp and Azure um and I think to do that there are a couple things that we're always going to be focused on um number one is continuing

21:58to drive aggressive we have this inherent competitive advantage that we can effectively undercut all these people and so we just need to go out there and continue to steal share from the major providers uh number two is continuing to maintain uh Enterprise grade um service um like our platform has to be able to meet or exceed the standards of what's available

22:21on the current cloud and so our Tech team is always working to uh make the platform more robust to release further products to keep that up um third and this is a unique advantage that we have that a lot of cloud providers don't is we want to keep leveraging our community and I actually think that our community is one of our biggest sources of competitive Advantage you know we can

22:42gain insights from them um that help us keep us Innovative keep us agile um and then finally it's um it's continuing to build a world class team um you know if we want to compete with the big guys I think we're going to need the best talent in the world and actually I'm interviewing for a lot of positions right now and I'm talking to some pretty senior level people at the

23:07aws's the core weaves the runp pods the Lambda labs and the idea is get the people there who are sort of disenchanted and bring them in and you know they'll give you the Playbook to growth and so I think that you know like if you look at the State of the Union you can kind of see the iterations of these goals um but really it all sort of pushes towards the same long-term

23:26Vision yeah no that's great and feel like there's probably a a lot of people that are disenchanted or fed up with these these web 2 Cloud providers um and would be eager to jump into something like this so uh that sounds great I I I can also leave a link to the state of the union and I'm going to check it out in the show notes so people have easy access to it and you mentioned the

23:48network sorry state of the network um yeah that's uh a little different the um you also mentioned about the community uh what where is the the main communities for ION and if I want to go talk to people about this and the community you know where are you guys located mainly sure um so the first thing I would say if you want to learn about IET like you know go to the website and

24:10actually like create a cluster um it's very very easy to do um almost anyone can do it in you know less than 90 seconds um heck it's it's it's even easy to connect your GPU so play around with but um in terms of community the best two sources will always be like every web 3 project Twitter and Discord um we have a you know a huge community in Discord we have support in over 25

24:35languages so definitely recommend getting out there talking to people wow incredible and yes uh I'll leave a link to the the the Twitter Discord state of the network uh iio site um I'm looking forward to these following updates and it's great to be in an industry that has an Ever growing increase in demand they have to keep up with um you know there

24:56the problems only going to get bigger and I feel like the solution I'm hoping uh the blockchain side and the decentralized nature of it will keep will up hopefully over one day overtake uh these traditional guys and make it more fair and rewarding for everyone so thank you so much for taking the time Tori all the best with IET everything your team's working on and let's

25:17definitely follow up in the near future definitely thank you so much for having me sir