원문 보기
Previously, an AI agent could keep its crypto wallet empty for about two and a half months before spending a single cent. By the third quarter of 2026, this wait had been reduced to 18 minutes. These numbers are taken from a Chainalysis report published on October 14, 2026, which tracked agents’ wallets from the day they were first funded.
Chainalysis also counted over 138 million payments made through Coinbase’s X42 protocol —a system that allows software to pay for services online with stablecoins. Over 87% of these payments were worth less than 10 cents. Such small payments almost do not exist anywhere else in the financial sphere.
Tom Lee is the co-founder of Fundstrat and chairman of Bitine, the largest corporate holder of Ethereum in the world . At the Token 2049 conference in Singapore in October 2026, he stated that Ethereum could easily reach $25,000 or $50,000 per token in this cycle. The excerpts you will hear are taken from a Fundstrat research session on agent economics.
A small note: this session was commissioned by Virtuals, an AI agent project that is a Fundstrat client, so please be aware of that. Jordi Vissa is a macro investor with over 30 years of experience in the markets who writes the research newsletter Vissa Labs. In an interview with Anthony Pompliano in October 2026, he explained what AI agents mean for banks and how he sees the future of Bitcoin.
In this video, we'll look at payment systems that already use AI agents, what they can do to banks, where Ethereum fits in, and what this means for the price of Bitcoin, which was about 35% below its record high in mid-October. Now let's begin. Thank you, Sean. I think this is a very important point, because today, when people think about how to position themselves in relation to AI, they are making two very different bets.
You know, chips and Nvidia, which is actually a correct statement because they produce something that no one else can do. And people are also buying up bottlenecks, like memory. In the mobile space, people also tried to buy bottlenecks and success stories, for example, which is why Motorola and various infrastructure companies were successful in the early days.
But during that cycle, there were two biggest winners in mobile communications. The first were the tower companies, as they had separated from the telecom operators, but took advantage of the capital shortage and had a very scarce asset. So they owned the most boring property—the tower— but they had tenants who always paid, demand was steadily increasing, and they actually became the most performing asset during that cycle.
And the second winner was Apple. So, they stole the ecosystem from operators in the AI world . I think people should start thinking ahead, you know, what are the natural end markets for the AI world? Well , one of them would definitely be financial services. For many reasons, because as AI capabilities grow, it's going to be easy to say, "Hey, listen.
Can you book me a ticket?" And later, you may have an agent who will not only book your tickets, but will actually start a business booking tickets for other people. So, a world of agent interaction will emerge. It is a mistake to think that this will be built on traditional financial rails. So I have six reasons to think about why traditional financial rails won't become a platform on which agent systems will operate, humans interacting with AI agents, and then AI agents interacting with other AI agents, and eventually machines—whether they're robots or drones or whatever— interacting with other machines and other AI agents.
See how much a small payment costs in the system you use every day. Stripe's standard rate for card payments in the US is 2.9% plus 30 cents. Make a 1 cent purchase through it and the commission will be 30 times greater than the actual thing you bought. This is why a working market for micropayments has never been created on the internet.
And that's why those 138 million agency payments went through stablecoins on the Base network, which Coinbase built on top of Ethereum. Also, an important announcement for everyone watching, especially in the US. If you look at the first link at the top of the description, you will see Kalshi. We have partnered with Kalshi, and if you sign up using this link and trade $50, you will receive $25.
That's why it's important. Now you can trade perpetual contracts in the US on a fully regulated platform. Many viewers from the US cannot use Bybit or any offshore exchanges. You can use Kalshi. It is regulated by the CFTC, and you can open long or short positions. Think of it as Bybit, but made for Americans.
The interface is clean and simple, and one of the biggest questions from viewers in the US is how to pay sales taxes. With Kalshi, you get the tax documents you need for the Internal Revenue Service (IRS) , with all your profits and losses in one place. They are also constantly adding new markets, including metals, and many more to come.
If you are outside the US, you can still check it out at the same link, but for US traders, this is a real breakthrough. Links are . Most of this activity is American. The same Chainalysis report noted that wallets linked to the US made 88.1% of payments for AI models that could be traced to a specific country.
Wallets linked to China performed 0.1%. The X42 system is currently supported by Visa, Mastercard, and Stripe, and Coinbase allows its business clients to accept payments from AI agents. Card networks are not fighting this. They are buying themselves a seat at the negotiating table, which raises a fundamental question about the banks themselves.
What have they been doing for us all this time? I think historically we have used traditional financial payment systems for four reasons, and while the order doesn't matter, the first and most important one was trust. In the past, imagine a pre-technological era , a bank was a place where you could leave money, even if you didn't know the person on the other side; they acted as mediators of trust .
Second, banks offered you proof of funds: if you made a payment or received it from someone, the bank stepped in, confirming that there were funds for this transaction. And this is really important, because with trust and confirmation of liquidity, the economy can function normally. Third, they actually provided credit, as they stood in the middle between receiving your money and transferring it to another person.
This is actually a credit transaction, but the calculations took a long time. I'm sure in ancient times, I don't know what the payment cycle was like. It could have been 30 or 7 days, but for those 7 days someone was using your money, and it was a loan. And fourth, tax collection, because by keeping records of these transactions, they could collect taxes.
Do we need this in the 21st century? You know these four things. Obviously, there are many different ways to implement these four mechanisms. So, in a sense, the only reason traditional banking systems are still successful is because all their customers are human. I think two things will happen related to the needs of agents that humans don't have.
Of course? First, I believe that money is becoming software now that it i bits of information. And that means that other things can start to act like money, and we can actually transact in them . We may think that there is fiat, like dollars and euros, but there are other things like gold, stocks, cryptocurrency, and, you know, reputation, but over time, they will all become monetary units.
Banking systems are not ready for this, because literally the entire accounting book of the banking system must have a currency, such as the dollar or the euro. For example, if someone created a new currency, calling it, say, “Wisconsin State Currenc truly rebuild the entire payment infrastructure of the US and global banking systems.
The second is micropayments; because in the future, if everything happens a thousand or a million times faster than today's transactions, the smallest unit will no longer be a cent. Micropayments will be everywhere. I mean, even payment installments. And cryptocurrency is actually a good fit for this, because, as you know, many crypto assets transactions costing less than a cent.
This also works well es can be collected easily because you have the necessary units to do so. A lot of money is already on these new tracks. According to DeFi Llama data as of mid-October 2026, the stablecoin market was around $306 billion, with Tether’s USDT accounting for around $184 billion and Circle’s USDC accounting for around $73 billion.
The current record is $322.4 billion, set on May 17, 2026. The legal framework has also changed. The Genius Act, signed into law on July 18, 2025, provided the United States with the first federal set of rules for payment stablecoins, including a requirement that each token be backed one-to- one by safe, liquid assets such as cash and short-term Treasury bonds.
The dollar on the blockchain has transformed from a crypto experiment to a regulated product with a single signature. If you want to push aside the fear and hype and understand where we really are in this cycle, " water", only important signals directly to your mail. Jordi Vissa has a name for what awaits the banks next.
And he is very careful about what he means . I believe that banks are facing a problem now—they will face a compression of multipliers over the next year due to two factors. The first is the development of AI agents, which will make consumers no longer, let's say, lazy . Some small businesses keep money in banks that do not bring them any income.
This is very profitable for banks. When banks can take your deposits, which are many, pay you zero, take that money and invest it in short-term Treasury bonds at 4%, not like SVB—in long-term ones, namely in short-term ones. You can get a spread from this. We saw Torsten Slok's article about the " agent" raid on banks.
He came from the other side, but that's only part of the story. And I don't want to say that a raid will definitely happen, because it is inevitable. I mean, agents won't make mistakes. They will look for the cheapest way to do something. This means that when agents control more financial decisions, banks will have problems.
The second thing is that regulatory barriers will also put pressure. And this is where crypt comes in. So, when users, consumers are no longer people but digital agents, and at the same time, financial barriers appear that are more integrated into the system where there are same-day settlements, which means stablecoins and tokenization.
We are moving towards this, although long-term settlements are beneficial for banks, for Visa, for Mastercard. The longer the process takes, the more someone gets this "float", the chance to invest this money for a day or three, and with instant settlements this is no longer the case. It was the belief that banks would still make money.
But the spread, the advantage they had, their future multipliers will suffer again because of terminal value. So this isn't some kind of end-of-the-world story. I played around with the “SaaS apocalypse” a bit to create a “ banking apocalypse.” But in my opinion, this will definitely happen with the development of consumer agents.
The article Jordi mentioned is by Torsten Slok, chief economist at investment firm Apollo. On September 27, 2026, he published a note titled " Is an Agent Raid on Banks Coming?" His numbers were straightforward. The average checking account in the US yielded about 0.1%. Fintech accounts from companies like Revolut, SoFi, and Wealthfront offered rates ranging from 3.3% to 5%.
Sllock called AI assistants, such as Meta's Muse, a tool that could automatically transfer household funds to accounts with higher rates. Then, on September 16, 2026, the Federal Reserve raised the target range to 3.75–4%. Each raise widens the gap between what the bank pays you and what your money could earn elsewhere.
For decades, this gap existed only because transferring money was a cumbersome affair. So, what will replace the bank as an intermediary? Tom Lee looks to his own cryptocurrency history for an answer. The first is stablecoins. You know, the stablecoin, or digital dollar, was originally created because owners of cryptocurrencies like Bitcoin didn't want to constantly exchange them for other cryptoassets, because sometimes they just wanted to keep their money in cash.
So the digital dollar was created. It was a dollar token backed by a reserve confirmation. Today, these stablecoins are becoming the future payment infrastructure for all financial services, as they are truly extremely efficient for making payments. The second major breakthrough in crypto was perpetual futures—these are futures contracts on assets that never close; The first such market was BitMEX.
It was created by Arthur Hayes over ten years ago; It allowed you to buy Bitcoin with a small amount of collateral, and the leverage fee was charged , it seems, every 15 minutes. But today, perpetual futures have become a common product on Wall Street. That is, a real breakthrough. Oil futures were traded even on weekends as people tried to hedge against the risks of a war in Iran , and now the Intercontinental Exchange, the world's largest exchange , wants to create its own futures markets .
As for interaction between agents, I think we are still at the very beginning of the journey. It's like giving kids lunch money and sending them to the store—they might get robbed along the way, but that will change over time. Therefore, I believe that the development of agent-to-agent interactions will likely be very similar to the path of stablecoins and perpetual futures.
This will be a real innovation, and in ten years it will be huge, but as for developments in the meantime—I'm not sure. I mean, for those using AI today: just remember that ChatGPT isn't even five years old, and it's already changed and transformed lives. So, is a highly intelligent payment system a matter of less than five years?
I mean, I have to say, I don't think time is slowing down. I think innovation is accelerating. Therefore, I believe it will take less than 5 years. The exchange Tom mentioned is the Intercontinental Exchange, the company that owns the New York Stock Exchange. It has a stake in the OKX crypto exchange .
And in May 2026, they announced perpetual futures, the prices of which are tied to ICE's Brent and WTI oil benchmarks. The first contract of this type was BitMEX's Bitcoin swap, launched on May 13, 2016. Most of the agent version is built on Ethereum. On January 29, 2026, a standard called ERC804, dubbed “Trustless Agents,” went live on the Ethereum mainnet.
Engineers from MetaMask, the Ethereum Foundation, Google, and Coinbase helped develop it. It gives each AI agent an on-chain identity, a reputation history based on feedback, and a way for outsiders to verify its work. Simply put, it is the way one machine decides whether to trust another machine. Whom she had never met.
Tom Lee has staked his company on this future . BitMine held 6,164,414 ETH at the beginning of October 2026 , about 4.9% of all Ethereum in existence. And at the Token2049 conference, Tom said that the company would stop buying once it reached 5%. On October 14, Reuters valued ETH at around $2,430, about half its record high of around $4,950.
Tom's figure of $25,000 would be about 10 times that price. Jordi Vissa approaches the same opportunity from a different angle. He wants to know who gets the money. That's the point, and that's where we can end our conversation. Um, I believe there are only two things worth focusing on. Those who collect toll fees (toll collectors).
This is cryptocurrency. As for me, I want digital money barriers, those who collect fees. I don't want to invest in Muse and then watch seven other agents get released. And that's before we have agents in Europe and Asia. That's the thing. We have competition on a global scale . I just want speed in collecting payments.
I don't want an attention economy. For me, these are all things from the last 17 years. I want to bet on things that will always be in demand beyond supply. We will not have enough computing power for agents. In my opinion, this is the best way to implement what you are talking about. Competition will only lead to complete commoditization.
I disagree with Anthropic and OpenAI because I believe there will be endless funding for superintelligence, and they will get it. Open source models will never reach this level of intelligence because they do not have such computational resources. All these companies simply do not have the necessary capacity.
They can continue to copy and distill models, but sooner or later that's what will be needed, because we won't be releasing the best models to the market. They will go to the business segment, and that's where it's impossible to turn it into a regular product. So at some point people will understand: if you have the computing power, you get the best model.
If you have the best model that solves all mathematical problems, then every scientific or physical field, including humanoid robots , will want it, but you won't release it into the public domain. And if you don't release it to the public, how are you going to distill all the information if the model isn't publicly available?
To understand why a toll collector is a good business, look at the one we already have. Visa generated $40 billion in net revenue in fiscal year 2025 and processed 257.5 billion transactions. Their entire business is a small fee for every transfer of funds through the network, and they make their profit regardless of whether the stores on either end of the payment win or lose.
The toll road for the machine economy was still under construction, and the price was far from favorable to those who held it. According to Reuters, on October 14, 2026, Bitcoin was trading near $81,600, about 35% below its October 2025 record of $126,080. The higher-than- expected inflation report sent the entire crypto market down 4.5% in 24 hours, and Coinlass recorded $331 million in liquidations, 82 % of which belonged to bullish traders.
In the week of October 5-9, about $679 million was withdrawn from spot Bitcoin ETFs, before $175.6 million was returned on October 13. So, if the toll road is being built, why hasn't the price changed? Let me continue, I want everyone to open the Micron stock chart. You remember, we talked about Micron when the shares were worth $100.
We talked about them when they were $150 each. If you look at 2024, we started this thing then. Many people in your office bought Micron. Should I stay with them at a price of 200? Should I go for 200? It reached 1400. If you look back, it broke the 200-day moving average, like Bitcoin, like Ethereum.
And then rolled back to her. He protested her. I'm not sure what will happen with Bitcoin, Ethereum, and others. But you have to figure out why this happened. When it starts, the theme behind it is AI. I want people to understand: when it comes to cryptocurrency, for me it's 100% about agents. And I'll tell you , the number of traditional players who keep knocking on my door to talk about tokenization, about how to make money from it — they're doing their job right now.
I think next year will be the year of the biggest shifts in crypto. Through agents. So I just want people to realize: I've been watching this a lot, especially in the semiconductor space. It was a long time ago when it was obvious to me that AI would require a huge amount of memory, but very smart people didn't see it.
And now a lot of smart people are running around saying that we are in a "bubble." This prevents people from investing in anything . They see that the rates are rising. Everyone should be grateful that with such a rise in rates, crypto has somehow gone up. Despite the fact that gold has fallen so much, crypto has somehow distanced itself and is growing.
So, considering all of this, I really believe that people should be patient. This is a bull market, and I can say this with confidence because I said it when I was bearish: we are above the 200- day moving average. The 200-day moving average is pointing up. Just be patient. These things never happen as quickly as we would like.
And what will accelerate this process the most is the arrival of traditional financiers. And for this to happen, tokenization volumes need to increase. We need the volume of tokenized assets to grow. We need stablecoin volumes to grow. That's why I talk so much about Stripe, because Stripe talks about it every day .
This is happening. Jord's thesis about gold is supported by the numbers. Gold futures closed at $4,157 per ounce on October 5, 2026, 21.8% below their record close of $5,318 on January 29. Gold was supposed to be a winner in a world of rising rates and nervous investors. Instead, it lost more than a fifth of its value in about 8 months.
Jordi also mentioned Stripe. Stripe and crypto investment firm Paradigm have created a blockchain called Tempo, designed for payments in stablecoins. Its public testnet opened on December 9, 2025. In October 2025, Fortune reported that Tempo had raised $500 million at a valuation of about $5 billion, and development partners included OpenAI, Shopify, and Visa.
So what does all this mean for the coins you hold? If you combine these two facts , the situation becomes simple. Tom Lee describes what machines need from money. Micropayments, instant payments, and bank-free trust. Jordi Vissa describes who gets the money when they use it. Both arguments point to the same thing: blockchains and stablecoins are at the heart of an agent economy, with Ethereum taking a large share of that traffic and Bitcoin being a neutral reserve asset that is not controlled by any company.
None of this is guaranteed, and the risks are very real. Agent payments in dollar terms are still meager. In July 2026, X42 processed about 75 million payments worth about $24 million in 30 days. Some analysts claim that a significant portion of this activity is bots trading among themselves. When Bit hits its 5% limit, it takes one of Ethereum's biggest buyers out of the market.
And rates that continue to rise could drag any risky asset down for longer than anyone expects. But two and a half months reduced to 18 minutes—that's what the initial stage of the adoption curve looks like. Coinbase, Stripe, Visa, ICE, and Ethereum developers are laying the rails, while as of mid-October 2026, the price was about a third below its high.
If you want to follow this without the panic and unnecessary noise, . Click . Anyway, .
Comments 0