What if 5 million or even $10 million Bitcoin isn't as crazy as it sounds? Today I'm talking with Avichal Garg about why Bitcoin could eventually become bigger than gold.
Because Bitcoin is so much more accessible than gold. Like anybody on a phone can get it, right? The size of these markets is just so enormous when you can get the entire world on them that I don't think it's crazy that Bitcoin ends up being bigger than gold in the fullness of time.
And why AI agents may create an entirely new wave of demand, but almost nobody is pricing in yet. These things are going to be sitting on assets and a huge use case is going to be AI agents trading. To me, very clear that agents will have wallets because an agent can't KYC. It has to have a crypto wallet.
We also get into autonomous AI companies, the future of on-chain capital formation, which crypto assets could actually accrue real value, and why the next generation of global companies may be created, funded, and operated almost entirely on blockchain rails. The convergence of AI and crypto can be much bigger than we're prepared for. Let's go.
Today's video is brought to you by Kalshi. You probably know Kalshi for prediction markets, but now you can also trade crypto pairs directly on the same platform. That means Bitcoin and other major crypto assets with leverage on a platform that's US regulated by the CFTC. Kalshi has the lowest pair fees in the United States. And you don't need to be trading millions of dollars to get exclusive benefits if you sign up through my link. Trade just $100,000 in volume per month and you unlock a bunch of VIP perks like one basis point taker fees, private invites to Kalshi events, and more. You can sign up using my link in the description and you'll get $25 when you trade your first $50.
Because you made some waves recently by saying that Bitcoin going to five or 10 million isn't crazy, right? So, walk me through the assumptions behind that kind of number because those really get people going.
Yeah. Well, the intention... Well, by the way, we've been saying this for a long time. So, the intention is not necessarily to get people going. It's just it's one of these things where I think if you look at the basics and the math on this stuff, you can kind of reason your way through it. And the way we've always thought about it is this is digital gold. It's a fixed supply asset. It's easy to transport. It's easy to subdivide. It's a global liquid market. I think people at this point have probably heard the arguments for Bitcoin. And I think people in the United States or people in the West sometimes forget how easy it is for us to acquire gold. You can go to the store and buy some or you can get some gold futures or whatever. In a lot of the world, you can't get gold that easily. And people want it. People want dollars in gold. And so what you finally have is the ability to get a digital gold equivalent. Not literally gold, but it meets all the same requirements and actually in many ways is superior. And so I think as institutions have figured this out, they're coming in. I think eventually the sovereigns will figure it out. They'll come in. But this is just that much more accessible to retail all over the world.
And so I tend to think as many people will have Bitcoin as hold gold. And if you just look at the market cap perspective from a net dollars perspective, how much money can come in, if Bitcoin and gold get to a similar market cap, you're talking well in excess of a million dollars per Bitcoin if they're at the same market cap. And I think there's a very straightforward argument that because Bitcoin is so much more accessible than gold, like anybody on a phone can get it. One of the lessons of the internet is when you put things on a phone you actually get TAM expansion. You get addressable market expansion. Like back in the day people used to talk about hey, how could Uber possibly be worth more than $10 billion because the entire taxi industry is only worth $10 billion without sort of realizing that well we put Uber on one of these things all of a sudden 10x as many people can do it. And so Uber is worth a hundred billion dollars, 10x bigger than the taxi industry used to be worth. And I think that sort of TAM expansion is underappreciated.
Like if you went back 15 years and you told your friends, hey, there are going to be multiple companies worth more than a trillion dollars. There are going to be 10 companies worth more than a trillion dollars. There are going to be some companies worth $5 trillion dollars. Like nobody would have believed you, right? The size of these markets is just so enormous when you can get the entire world on them. But I don't think it's crazy that Bitcoin ends up being bigger than gold in the fullness of time and so very quickly you can do that sort of one-to-one math and say, wait a second, if this is even bigger...
And that's only addressing humans.
So the TAM truly expands with AI agents. You know, we get right into that part because I know even at Electric Capital, I think you guys talk about AI now as much as you talk about crypto and certainly about where the two meet. But if Bitcoin becomes digital gold to AI agents or a savings account, we're talking about even exponentially larger potential numbers.
Yeah, you know, it's tough because these numbers can get so large that you have to be a little careful not to sound like a crazy person. But yeah, I mean you look at agents and they certainly have to use crypto. I mean those are certainly the rails that agents will have to use. Now whether or not on the front end, I think today it probably starts with you delegate a Ramp card or a Visa card or something to your agent and it can go buy flights for you. But slowly, like really what Stripe is doing on the back end that's actually going to move to stablecoin rails is actually what it'll move to. So at least the infrastructure will be crypto.
But I think to me it's very clear that agents will have wallets because an agent can't KYC. It has to have a crypto wallet, right? A cryptography-based wallet and that will have US dollars in it. And then very quickly you start saying, well, what other assets should I own? And I think owning things if you're a non-signal digital entity that doesn't have a physical presence in a particular jurisdiction, having non-jurisdictional money, I think, makes a lot of sense. Parking your assets in non-jurisdictional money to some degree makes sense.
Or even if you think about, most of the trading of the world is already bots, right? It's already AI agents. It has been for some time. Just that's what happens with electrified trading. So, these things are going to be sitting on assets and a huge use case is going to be AI agents trading. And these are very natural assets for them to trade as well. Also, I think invariably you end up with some percentage of these assets being held by AI agents in the fullness of time as well. And that's not even getting at the really weird stuff which you might be getting at Scott is like 10 years from now these things are smarter than humans in many respects.
But they don't exist in a physical place, right? So it's not like you and I exist in a physical jurisdiction. Even a company which is not a physical thing has a jurisdiction like a Delaware C-Corp. These things don't really. It's just a piece of code that can run anywhere. And so does it live in the Cayman Islands? Is the jurisdiction of the server that it's running on? Does it have to have a Delaware embodiment of some sort? And that opens up a lot of friction with the legal system. So KYC is just one of them. There's all sorts of questions around said liability or free speech. Like what happens if these agents are just saying stuff?
And so this whole category of agents, which we haven't quite gotten to yet, but I think we will in the next 5 to 10 years around autonomous agents, which are basically like humans, but they don't have a physical body and they don't live in a jurisdiction. I think it opens up a whole new can of worms. And I think those entities certainly will start to think about how do I think about where to park my assets and I think digital assets are almost certainly where they'll park some percentage of money.
Exactly that. So, I think that this first iteration that everybody's excited about is the one that you presented first, which is they're obviously not going to plug into JP Morgan. So, they're going to transact in stablecoins or some sort of crypto or on blockchain rails, but they're doing that on our behalf. Book my flight, pay for this thing, accept the payment for this thing. When they become truly autonomous, like you said, you have to start thinking of them as humans would think and they're going to also want to store value and have a savings versus checking versus yield versus all these things and they're going to start doing the same things that we do. I'm actually surprised you take five, 10 years, maybe faster. I don't know.
Yeah. I mean, at the fringes it will happen first. And you know, for all we know, it's already happening somewhere on the internet. Somebody's already released some code and it's autonomous for all intents and purposes. And these days, you can write code, you can ship it, you can create an app, you can get paid for it, you can run the inference on something like Venice does. You're pretty close. I don't think you're quite there yet. There's still some real friction points in terms of actually receiving money and things like that, but we're pretty close to having autonomous agents. So, yeah, it could be sooner than five years too. We'll see.
I wonder at what point they start acting in their own interest and not in ours.
Yeah. Well, I mean, that's motivation. What does it mean to have an agent that's not attached to a human and doesn't answer to you and doesn't have your KYC behind it to go out in the world and succeed, right? I mean, what does success look like to an autonomous agent? Does it care about making money? What does it care about?
Yeah. Well, I mean, that's a big open question, I think, in AI safety and alignment is how do you try to do these things in a way that at the very least isn't misaligned with humans. But yeah, this idea of like what is your objective function for the agent or what motivates you I think is a pretty gnarly one.
I would note we kind of are seeing this already, not in the public context but inside the enterprise context. Like there's the first wave of a lot of enterprise AI agents were sort of agents that acted on your behalf, like you gave it your credentials, they could see what you could see, and we're starting to see the second wave. And the second wave is agents that are basically employees, right, that you can sort of tag the agent, you can give it a... and it has its own permissions. It has access to documents. It has access to drive. It can create stuff. It can create tasks in Linear and just uses all the APIs to do this. And you can give it an objective function. So you can create like an invoice bot and its job is to process all the invoices at your company. So in a sense we're actually starting to see the beginnings of the autonomous agents just sort of scoped inside the enterprise because then you don't have these harder jurisdictional questions and legal questions and so on. But we actually already have autonomous agents inside companies.
How much have you started using them?
A lot. A lot. We have a team of engineers in-house and so we play with all this stuff when we build stuff. So, we actually have both systems. We built an in-house system that actually we'll probably open source pretty soon, which makes it really easy to build an agent that has the same permissions that you do. You can delegate permissions to it as a co-worker, as a co-pilot, and then you can create jobs and it'll do things but using your permissions. And the hard part of that is getting all the permissions piece of it right, like how do you really control access and make sure that the agent can't accidentally do something nefarious or blow up or accidentally send emails to your contacts. There's a bunch of safeguards you want to put in place for that.
And we also have a system which is a different system that creates essentially AI co-workers and has access to the CRM and to... you can come to a meeting, it can record the meeting for you but literally we have names for the agents and so you can just invite Marshall to a meeting, it'll record stuff for you. And so yeah, we actually have all these systems in production in-house already.
It sounds like you're largely using them to sort of amplify the power of your humans and not necessarily to replace them yet. Have you already gotten to the point where you can fire people because the agent can do their job?
Well, what we've noticed is there's a lot of tasks the agents can do and then there's a lot of stuff that they can't do. And so what it does is it shifts the job, so we don't have any less work. We're just able to do more now. So, for example, we had this invoice processing thing and it was a lot of work previously. Get this invoice from somebody, you want to make sure it's a real invoice. Double check the amounts, double check it against the contract, put it into a row in a database somewhere that says we have processed this thing. And then, there's some, typically in companies there's like some sort of compliance or workflow like somebody tees up the wire, somebody else checks the wire. And you can get rid of 90% of that work using an agent. You can go read the contract, you can go check the invoice, it can read the invoice, put it in the database. And so, you did all that work. And so now this human that previously had to do that has other work that they can go do.
So we haven't, at least for us, it hasn't resulted in us replacing humans. It's just changed the nature of the work. And so now all this work that was getting done by humans we don't have to do and so then you just shift your people to do other stuff. A very interesting question by the way of like what does it mean then for what is the valuable work that humans can do? And we have some ideas on that but yeah, we haven't really seen less work. It's just shifted the work.
So it just makes you more efficient and get more done as a company, right? And that's what I found as well generally, right? Mostly replaced, but it still requires a human to step in and manage it, which still is work. It's actually I would say for me it's been more work for me, but I'm doing 10x what I could in a strange way. So I did get rid of a person or two and the AI agent can do their work, but it requires me to do more management than it would have if it was a human myself.
Yeah, that's right. And I think it changes the nature of the work. It's a lot more management. Also I think it changes what's valuable work. So like the communication layer, what is the goal, have we thought about the goal properly, becomes a lot more important because these things are autonomous. They just do a bunch of work really quickly. So knowing that you had the right objective is important. But also I think a lot of the human-to-human stuff becomes more important we've noticed.
So if you can just create a document or a memo or whatever, the real value is like did you get on a plane and go shake somebody's hand or did you go have dinner with them, like it sort of shifts the human dynamic a little bit, I think is what we've noticed. Yeah, I think that the dynamic might be different at bigger companies. Like I think a bigger company like a Meta or something has just a lot of, there's a lot of people in the mix and so there may be a lot of fat and you want to cut that fat.
But I think that at the same time what we also see on the startup side is the efficiency of some of these companies. You can do so much more with few people and I think one of the properties of the internet is like when you get these kinds of technological unlocks, people talk about unlocking the long tail, like think about iTunes because you got digital distribution, because you got cost of distribution going to zero, you unlocked all of these musicians that previously you had to go to the record label and the fixed cost of producing music was very high.
And I think now there will likely be a lot of products and services where the fixed cost of producing that good or service previously was cumbersome. It was too high. And so you couldn't have a sustainable business. But now that the cost of producing that good or service has dropped by 100x because an agent could do 90% of it, a human can step in and like a two-person plus 10 agents business could be sustainable in a way that wouldn't have been sustainable before. So, I actually think we'll get this kind of, the internet created entirely new categories in music and art and social media and streaming. You got these podcasts, right? You just got entirely new stuff that you couldn't do before. I think we'll actually get entirely new businesses that are like one to 30 person businesses that couldn't have been sustainable before and are actually fantastic businesses.
So hopefully all these people that it doesn't make sense at the big tech companies will rotate out and go start a bunch of small companies which I think is better for the world actually.
Yeah. The notion of a single founder unicorn, you know, like one guy spins up some bots and creates a billion dollar company. I think that we're close to that and probably we'll see quite a bit of that. The music industry example you gave is great. It's very close to my heart.
That's what I did for two decades before I was here. But it's interesting because the cost of production goes to almost zero because anyone can make music and it's accessible via technology and I'm sure AI is helping with that. I haven't experimented with it yet. But actually the distribution piece is maybe the bigger one which you mentioned, right? Even if you spent the money to create an album and it was spectacular, you needed a record label to distribute it and put it on the shelves. And obviously that changed with iTunes and all those things and the SoundClouds of the world but you still couldn't necessarily get seen.
Now you can have your agents go out, market this thing, distribute it, post it automatically to every single platform, tweet about it. I mean when you think about that in the context of music extrapolated to everything it's really endless possibilities for people to create things.
Yeah, 100% 100%. And I think the create portion is the key portion from my perspective too. So like the internet took the cost of distribution going down to zero. But now you have the ability to create stuff using an agent at 100x the velocity. And so what does it mean if you, like as a musician, if you had some idea for some sample, some beats that you wanted to do? You fire up a hundred agents and you get 100 versions of it.
And but that human taste element is still really important, I think, on top of that. It's like, okay, well, which of these beats or melodies or whatever sounds great and is what I'm trying to express. But the throughput that you get now in terms of being able to produce just goes up 100x, which is phenomenal.
Yeah, it's insane. So, we're talking about all of these use cases, all these things happening, all the things that will theoretically happen with it using blockchain and finance. Are you actually using it in any way where it's transacting for you using crypto? I have not yet.
Yeah, we've seen a lot of agents that are doing the trading or, you know, can act on trades for you, but I personally have not. I'm always a little skittish about...
Same. It's funny. Anything that uses the money directly.
Yeah. It's like Steve Jobs never allowed his kids to look at an iPad. You know, I'm not letting this thing out into the wild.
Yeah. Yeah. That's right. You got to be a little careful, I think.
So, but it's interesting to see people using it in this way. I think one of the lessons with this stuff is it tends to start on the fringe. I guess it tends to start with young people or people in the developing world or the ultra wealthy or the ultra poor. Like, you got to look at the fringes. The artists, they're probably the first people that will figure this out and use it in interesting ways. I think kind of the mainstream people are probably a little bit slow to use this. I think that's why it's still a few years out.
Kind of hearing us talk about it, it reminds me a bit of our Web3 bubble in the last cycle. You know, we talked about art and music and gaming and all the ways that blockchain would innovate that. I think AI is obviously much more obvious how that will happen, by the way.
It's a good pivot though because I don't hear anyone talking about Web3 anymore and I haven't seen a massive resurgence of any of those previous ideas yet, right? I would think that we would have seen music royalties blow up using blockchain, right? Or NFTs come back in some more meaningful form. I certainly thought gaming was going to be a thing. So, do you think any of that comes back? Or do you think that we've moved on to greener pastures?
I think it does come back. I think it's a few years out. So, it's interesting you mentioned some of those use cases. It's almost like the first time people come up with these ideas, the ideas are right but it's too early. The tech is not there yet. The world's not ready for it. The markets aren't ready for it. And it takes almost 10 years for people to digest it, for the tech to catch up. You saw this with the internet too, right? Instacart is a huge business, but Webvan was one of the biggest flames from the late 90s of grocery delivery. And so it took those 10 years in between for the infrastructure to catch up.
And that's all infrastructure, not just technical infrastructure. It's like how do you actually acquire users, right? That's an important part of the infrastructure. So, when I look at the Web3 ideas, I think we're kind of in like 2017 ideas at this point, right? We're approaching 2027. And so if you go back to that era, people were talking about a lot of straightforward financial use cases. Things like how does insurance work or reinsurance or stablecoins or why doesn't Wall Street run on this stuff and how do you do private credit and real estate and these kinds of transactions on-chain. And we didn't get to the art and music and gaming stuff I think until like 2020. Like yes, we had NFTs earlier, pretty early.
And so I think that stuff will happen but I think it's a couple more years out. I think it's like 2028, 2029, 2030. We're just starting to see the very beginnings of it. Kind of like we saw the very beginnings of NFTs and gaming in like 2017. So I was just talking to a company actually earlier this week and they're playing with skills-based games in crypto. And so literally because the AI can now vibe code a game and you can create puzzle games or you can create games of skill. And you could allow crypto to be the money backbone for that. They're experimenting with this new genre of game where you can go in, invent a game, put it into the marketplace, a bunch of other people can come play the game. If the game is really popular, you get some cut of the money from the fees of the game, and everybody that's playing the game can have prizes inside the game.
So, it's like a really interesting marriage of AI, codegen, and crypto because all the rails have to be crypto for this stuff to work for micropayments. And so the fact that people are experimenting with this stuff, I think is really interesting, but it's probably going to be a couple years before people really see it take off, I think. But I do think it will come back.
It's interesting. I... you have a longer time frame than I kind of mentally modeled it because I always kind of say, you know, we have a DeFi bubble and you have an NFT bubble, metaverse bubble, all these things, and maybe three or four later or the next cycle they come back. But you're actually right. I think it's more of an 8 to 10 year time horizon for those things. I wonder if that can be expedited a bit by AI or any of these things. But we're amazing though in crypto of identifying those things exceptionally early and speculating heavily on them and creating these massive boom and busts. I mean, it's unbelievable.
It's really wild. Yeah, and I mean, if only for that reason, I think if people paid attention to what the speculators are doing, they're actually often directionally right. Like they're not close enough to the tech or the markets or the users to really understand if it's possible. But a lot of times these guys are actually very correct about this thing is gonna happen. They're just seven years too early.
