AI Stock Selloff: Why Tom Lee Thinks the Bottom Is In
Tom Lee says the AI stock selloff may have bottomed. Learn how supply constraints, Intel’s outlook and crypto signals support—and challenge—his case.
Has the AI Stock Selloff Hit Bottom? What Tom Lee's Case Gets Right
When AI stocks drop, it's easy to treat every red day like a verdict on the whole idea. But stock prices can weaken for lots of reasons - demand can cool, costs can rise, or a company might simply struggle to turn customer interest into revenue quickly enough. Those distinctions matter if you're trying to figure out whether this selloff is a pause or whether expectations ran ahead of reality.
In a CNBC interview, Fundstrat co-founder and Bitmine chairman Tom Lee argued that the AI stock selloff has reached its bottom. His main claim was that AI demand is still strong while the supply situation remains tight. He also pointed to crypto's reaction to a major Binance security breach as a sign that investors may have already cut some risk. Still, the case isn't clean-cut: Intel's shares fell sharply even with strong growth in its data center and AI revenue, and Bitcoin fund flows have also looked anything but calm.
Here's how to think through Lee's reasoning - what the market evidence can and can't tell you - and a more grounded way to assess AI-related stocks without treating a confident "bottom" call as something you can bank on.
What does Tom Lee mean when he says the AI selloff has bottomed?
Lee's core idea is about supply versus demand - specifically, the difference between an industry with too much capacity and one where customers want more than providers can deliver right now. In an overcapacity setup, companies build more than the market needs. That often means lower prices, postponed projects, and weaker returns. In a supply-constrained setup, customers may still be willing to buy more than the industry can supply.
Lee said the AI market was "very tight" and not an overcapacity market. He did acknowledge that spending constraints - including political pushback - are slowing growth. But his interpretation was that slower growth doesn't automatically mean the AI boom is over. In his view, demand can stay healthy even if infrastructure build-outs and corporate expansion move more slowly due to friction.
That distinction is useful. It doesn't prove, though, that stocks have stopped falling. Markets don't just react to current demand; they price expected future earnings. If investors already started assuming rapid growth, then a stock can drop even if the business is "doing fine" in a demand sense - especially if revenues come in later than expected or margins underwhelm. A tight chip or data-center supply can support long-term demand while still limiting near-term sales.
Lee also made a point worth remembering: calling bottoms in real time is hard. In his words, "Bottoms are dangerous because you don't know how much more downside you have at the bottom." He's right. Prices can keep sliding after selling pressure starts to ease, and the "bottom" usually looks obvious only after the fact.
The distinction between a boom and a stock price
It's totally possible for an industry to grow while its stocks decline. That often comes down to expectations, valuation, and timing. The AI investment story can be intact even when individual stocks struggle, and the two aren't interchangeable.
For a "bottom" call to really hold, investors need more than a sign that customers want AI. They need evidence that companies can deliver it - turn that demand into revenue - and earn a sensible return on the money required to do it. Lee's supply argument addresses the first part, not the whole checklist.
How can AI demand stay strong while AI stocks fall?
The clearest example in the reporting is Intel. The company said its data center and AI revenue rose 78% year over year to $7.3 billion. Yet its shares fell nearly 30% to $65.76 after it issued a disappointing fourth-quarter revenue forecast. Intel projected revenue of $15 billion to $16 billion, with the midpoint about $1.6 billion below analysts' expectations.
Those numbers can sound contradictory: strong AI revenue growth alongside a sharp drop in the stock. But they're answering different questions. Intel's revenue growth suggests customers are buying and using its products - at least in that segment. The forecast and manufacturing constraints raise the next question: how much of that demand can Intel serve, and when. On top of that, the company's weaker PC business added another source of pressure.
For investors, it helps to separate demand that exists in the market from demand a specific company can capture profitably. A customer might want more computing capacity, but a chipmaker still has to produce the right products, at the right economics, and deliver them on schedule. If production constraints prevent it from filling orders, then strong demand doesn't automatically show up as the earnings growth investors expected - on the timeline investors expected.
This same issue applies to AI infrastructure, especially data centers. These facilities are where the compute power for training and running models comes from. Building them takes capital, power, equipment, and time. Shortages can leave customers waiting, but they can also delay revenue for companies building and operating the systems. And if companies invest before capacity becomes usable, the risk shifts from "not enough supply" to underused assets.
What Intel's results do - and do not - tell us
Intel's numbers don't prove the entire AI market is weakening. They do show why you can't evaluate a single stock by looking only at the general "AI demand" story. A company can be in a real demand tailwind and still miss expectations due to execution limits, product mix, or weakness elsewhere in the business.
A better read-through is to follow the chain: customer interest → orders → production/deployment → recognized revenue → earnings. Any step can introduce delays or problems. When you're assessing an AI stock, ask which step the company is currently hitting - and what evidence would support the next step.
Does crypto resilience offer a useful signal for AI stocks?
Lee's second argument comes from crypto. He pointed to the market reaction to a Binance security breach, which the exchange initially estimated would affect about $6.3 billion. He suggested that if crypto trades like a leading indicator for risk appetite, then a limited reaction to serious bad news could mean many investors who wanted to reduce exposure already did so.
That's more of a market positioning argument than direct evidence about AI earnings. The idea is straightforward: if some selling has already happened, the next negative headline may have less incremental impact. But it's hard to measure from one event, and it doesn't answer whether the underlying AI businesses are actually doing well.
The reported data also complicates the picture. Bitcoin fell about 4% to roughly $76,400 on Friday, and more than $1.3 billion in leveraged positions were forcibly closed. U.S. spot Bitcoin exchange-traded funds - funds that hold Bitcoin and trade on stock exchanges - recorded $643.2 million in outflows that day, their largest daily outflow of October. Binance said about $3 billion of the affected assets were frozen, and that it would cover losses using its emergency reserve and company funds.
A 4% Bitcoin move isn't, by itself, proof the market is breaking down. Crypto traders can see moves like that even when conditions are improving. Still, the forced liquidations and ETF outflows don't indicate that risk has disappeared. They suggest investors were still reducing exposure - and that leverage can intensify market moves.
Why the comparison has limits
AI stocks and crypto both respond to shifts in risk appetite, but they're different assets with different drivers. Crypto price action can offer context about investor behavior. It can't establish that semiconductor orders, data-center spending, or AI software revenue will stay strong.
Treat the crypto signal as one possible clue about positioning - not as proof that an AI-stock bottom is in. It only becomes useful if it lines up with what the companies are doing: steady orders, better delivery capacity, and forecasts that match what they're actually able to deliver.
How should investors assess the next phase of the AI market?
A practical approach is to split the question "Is AI demand real?" from "Are these stocks attractive at today's prices?" The first is about customer adoption and the infrastructure required to support it. The second is about whether a company can capture that demand - plus valuation and the risks already reflected in the share price.
Instead of betting on a single headline or strategist's call, use a short evidence checklist:
- Demand: Are customers continuing to order, deploy, or use the product? Look for reported activity, not just broad claims about "interest."
- Capacity: Can the company deliver what customers want? Manufacturing constraints, construction delays, and power availability can all affect timing.
- Conversion: Are orders turning into revenue - and revenue turning into earnings? Growth in one segment can be offset by weakness elsewhere, as Intel's PC business showed.
- Expectations: What were analysts and investors already expecting? A "good" result can still disappoint if it misses the bar the market set.
- Funding and returns: Does the business have a credible path to returns on the capital required? A tight supply can support pricing, but capital-heavy expansion also brings execution and financing risk.
- Risk appetite: Are investors cutting exposure elsewhere too? Crypto flows can provide context, but they shouldn't replace company-level evidence.
A decision framework for separating signals from conclusions
Start by identifying what would change your mind. If your thesis is that the selloff is temporary because supply is tight, then evidence of persistent order growth and improving delivery should support it. If orders weaken, customers delay projects, or capacity expands faster than demand, the thesis needs revisiting. Listing the counter-evidence upfront makes it harder to reinterpret every negative update as "good news for the next buy."
Then separate business quality from market timing. You can believe an AI company has a durable position without knowing whether the stock has already reached a short-term low. Lee's warning about bottom calls captures the problem: you can be right about the industry and still be wrong about the timing.
Finally, follow the chain of consequences. If supply is tight, which companies benefit? Can they produce more and deliver it? If output increases, does profit improve - or do higher costs and heavier investment drag returns? And what if demand grows more slowly than planned? Those questions turn a broad narrative into something you can actually test.
Conclusion
Tom Lee's bottom call rests on a sensible distinction: constrained AI supply doesn't necessarily mean AI demand is fading. That helps explain how a market can keep growing while some tech stocks fall. But it still isn't enough to prove the selloff has ended.
Intel's results highlight the gap between customers wanting more AI capacity and companies delivering it on terms investors expect. Crypto's reaction to the Binance breach may tell you something about positioning, but it can't confirm whether AI earnings are improving. If you want the more reliable path, follow demand through capacity to revenue and returns - then compare it to what the market has already priced in.
Use Lee's thesis as a hypothesis to test, not as a signal to follow blindly. If you're assessing AI exposure, focus on whether companies can turn demand into sustainable earnings - and be clear about what evidence would actually change your view.