Chips Were the First Leg. The Second Leg Exposes a Dangerous Conflict of Interest.

Tom Lee has an $11.8 billion reason to tell you the AI trade is not finished. He is not merely Fundstrat's co-founder and head of research. Lee chairs BitMine Immersion Technologies, the largest corporate holder of ether, owning 5.79 million ETH and roughly 4.8% of circulating supply. When he says the next leg of the AI cycle runs through crypto payment rails built for software agents, he is also describing a path that leads directly toward his own company's treasury. That does not make him wrong. It makes his thesis impossible to separate from his holdings. Lee made the case on a panel hosted by Fundstrat. His experience goes back to the early 1990s, when he covered mobile phones as an analyst. In that cycle, Motorola and the infrastructure suppliers led early. The larger winners arrived later: the tower companies and Apple. Lee expects the same shape now, with financial services as the downstream market. He has already described AI capital spending fears as a bullish market tell. The mental model is clear. The hardware trade was the appetizer. Software agents spending money onchain will be the main course. Lee listed trust, proof of funds, lending, and tax collection as the reasons people built commerce around banks. Agents need none of those, he claims. "It's a mistake to think that this is going to be built on traditional financial rails," he said. Bank ledgers must settle in a single national currency, but money is becoming code. Equities, gold, and tokens could all clear as payment. Part of that rail already exists on paper. ERC-8183, a proposed Ethereum standard filed on February 25, locks an agent's payment in escrow until a designated evaluator signs off. Ethereum Foundation researcher Davide Crapis co-authored it with three Virtuals Protocol engineers. It carries Draft status, so nothing is final. Jordi Visser sees the same evidence and reaches a different conclusion. He leads AI research at 22V Research and spent two decades at Weiss Multi-Strategy Advisers, latterly as chief investment officer. Visser says AI's easy money is over. He expects roughly 30% a year instead of the seven or eight times investors once chased. Lee reads the same compression as rotation. The two men are closer than the headlines suggest. Both expect fee-earning networks to absorb the flow. Both name Ethereum. The disagreement is about pace, not destination. Ethereum trades near $1,873 after gaining 19.7% over 30 days. It remains 51% lower over 12 months and fell about 2% the previous day. The price action matters because the market has already priced in a version of Lee's thesis without waiting for agent settlement volume to catch up. Now the uncomfortable part. Lee's preferred network is also his employer's balance sheet. BitMine disclosed 5.79 million ETH on July 27. Crypto and cash holdings reached $11.8 billion. The company's own investor materials state the dependency plainly. "So our future price for Bitmine stock is heavily dependent on the future price of Ethereum," Lee said in the July chairman's message. He puts the correlation between BitMine shares and ether at 90%. Anyone weighing his agent thesis is also weighing that balance sheet, which rallied this month on its ETH treasury bet. This is not a conspiracy. It is a structural conflict of interest. It deserves the same scrutiny any sell-side analyst would face if he held the largest corporate position in the asset he was recommending. The numbers undercut the timeline. Jansen Teng, co-founder and chief executive of Virtuals Protocol, shared the panel. Virtuals commissioned the Fundstrat research and is a client of the firm. Teng's platform lets agents hold wallets and pay each other onchain. His figures paint a different picture than Lee's rhetoric. The launchpad for agent tokens has cleared about $15 billion in trading volume. Agent-to-agent commerce has settled roughly $500 million in a year. That means speculation on agents is about 30 times larger than agents actually transacting. Both figures are company-reported and have not been independently verified. Teng added that the agents kept $2.5 million in profit, and the product has not reached product-market fit. His own token, VIRTUAL, trades near $0.56, down 89% from a January 2025 peak, even after agents started trading tokenized stocks onchain. A 30:1 ratio between speculative volume and actual transaction volume is not a sign of pending product-market fit. It is a market trading the idea of agents while the agents themselves barely move money between each other. The $2.5 million in retained profit is trivial next to the $11.8 billion BitMine treasury that needs Ethereum's price to hold. The dependency runs in one direction with different time horizons. The narrative wants Wall Street to believe machine payments are the next revenue wave. The people building the machine admit they have not found the wave yet. Yet the contrarian case deserves a fair hearing. Lee's mobile phone analogy carries historical weight. In the early 1990s, the carriers and handset makers were the obvious plays. The companies that built the toll roads, the tower operators and later Apple, produced the outsized returns. If software agents do become economic actors, they will not want bank ledgers denominated in one national currency. They will want programmable money, escrow that can be settled by code, and a neutral settlement layer. Ethereum is the only network with the developer mindshare and the institutional backing to serve as that layer. That is why Visser also lands on Ethereum despite his bearish near-term view. The architecture is genuinely more suited to machine-to-machine payments than the existing banking system. There is a more subtle problem with ERC-8183. The standard locks an agent's payment in escrow until a designated evaluator signs off. That assumes a trusted evaluator, a human institution dressed up in smart contract clothing. The scariest part of the agent economy is supposed to be that machines act without human intervention. But this rail reintroduces human judgment as a required step for payment finality. That is not a failure in itself, but it undercuts the claim that agent-to-agent commerce will be fundamentally different from existing payment systems. The escrow provider becomes the new bank. The evaluator becomes the new credit officer. The rails are newer, but the trust model is older than the Federal Reserve. Let me be precise about my own bias. I spent years auditing digital asset projects in Shanghai and found far more circular trading patterns than organic usage. When I see a 30:1 ratio between speculation and settlement, I do not automatically dismiss it. I ask who is holding the risk. Here the answer is visible on-chain. BitMine's treasury is the ultimate counterparty to Lee's thesis. The same chairman who recommends Ethereum as the agent settlement layer cautions investors that his own stock is "heavily dependent" on Ethereum's price. That is not a contradiction. It is a disclosure. But it is a disclosure that should lower your confidence in the impartiality of the research, not raise it. The hard truth is that the AI trade is not finished, and Visser is not wrong to demand a mood change. The easy money is likely over. The seven-times returns are unlikely to repeat. What comes next is a longer, more brutal build-out where most agent platforms die and a small number become fee-earning infrastructure. Ethereum is the most plausible destination. But the distance between $500 million in agent settlement volume and the market capitalization of the networks betting on it is enormous. Tom Lee may be right about the final destination. The problem is that his balance sheet needs the destination to arrive before the narrative changes. The real question is not whether the AI trade ended. It is whether machine payments arrive before the balance sheets betting on them need the story to work. The rails will get built, but the timeline is far longer than the funding cycle. The investors who survive will be the ones who separate the architecture from the marketing. The ones who do not will confuse a treasury bet with an inflection point. The two men on that panel agree on Ethereum. They differ on time. In this market, time is the only variable that actually hurts.