The AI Sell-Off
The recent slide in big AI stocks follows weeks of growing doubt about whether the AI growth story can last. That fear is pushing investors to take money off the table in the weeks before August. The old saying was "sell in May and go away." The same thing is happening now in July. Ahead of the low-liquidity summer months, it looks like a good moment to pull cash out of the trade.
Some selling is also happening before these firms report second-quarter earnings. A pattern keeps repeating during AI fear: AI-linked companies beat earnings estimates, then sell off anyway. Whether this two-year bull run lasts will likely be decided by these earnings. Massive beats that surprise to the upside could deliver another leg up. If not, this could turn into a deeper structural shift in how money is invested.
How Good Is Good Enough?
Companies have posted strong numbers, beat-and-raise quarters, better-than-expected guidance, and topped whisper numbers across the board, yet still drew negative reactions. The goalposts keep moving. To prove themselves now, these firms need to be great and then better than great. Expectations are so huge that, to make the trade work on valuation, beats must run well past what these companies would normally have to deliver.
Sometimes even that is not enough. Nvidia has posted great results that should have pleased the market and did not. How much of a beat is enough to keep the bull market going is the big open question, answerable only by watching how stocks react when results land.
There is a second worry spooking markets and the investment banking world: the sense that things simply can't get better than this. Even with results 2x better than expected and records everywhere, the argument becomes "this is peak, so sell it anyway." When Alphabet did its major capital raise earlier this year, a lot of market chatter said it was doing it because there would never be a better time. That same feeling seems to be driving the rush to market, including some big IPOs. If you're going to do it, do it now, because there's real fear the rally won't last much longer.
Nvidia as the Proxy, SpaceX as the Warning
Nvidia has always been the proxy for the whole sector. How it trades sets the perception around AI. As of yesterday's close it was down about 14% from its last high, which signals where the market's head is. It has dropped sharply before and come back, so the coming results and any sentiment shift matter.
SpaceX is a telling case study. It's a rocket business tied to orbital data centers and possible life on Mars, but if you look at its total addressable market, it's a bit of an AI play, and it got hit especially hard in the sell-off. IPOs always have lower liquidity right after listing than an established mega cap like Nvidia. Still, the reaction in SpaceX is important. Its TAM and earnings projections suggest the market may have been near peak AI exuberance when SpaceX came to market. The stock sits at 12270, more than $10 below its IPO price.
What It Means for OpenAI and Anthropic
The SpaceX drop is a real problem for the unicorn IPOs still coming, OpenAI and Anthropic. Everyone who bought SpaceX is now sitting on large losses, which makes them more skeptical about backing another unicorn AI IPO. But the danger runs the other way too: if things get worse in 2027, listing then won't be possible either. So do they come to market now at a lower valuation and leave more money on the table? That cuts against their whole reason for listing, which is to raise huge amounts of capital, so they don't want to give up much on valuation. If the sell-off continues, the discussions happening inside these firms would be fascinating to hear.
Cheap Models as a Direct Threat
New competition makes the picture harder. Alibaba put out its updated Qwen 3.8 Max, claiming it's second only to Fable 5. Last week Moonshot AI released Kimi K3; test results show it topped only by Fable 5 and the latest ChatGPT. Cheaper models like these come right on the heels of the premium providers who are also planning IPOs.
This feeds the debate that has run through the corporate world all year: what return am I getting on my AI spend? If a model is just as good, or good enough, compared to premium providers like OpenAI and Anthropic, but costs almost nothing by comparison, that's the model most people will choose.
AI still needs more human involvement in many processes. It hasn't reached the level of being a real decision-maker; it remains a process- and tool-driven technology. A few months ago there was a story about Uber scaling back its AI spend over return-on-investment and a "token maxing" problem. That kind of thing will keep happening. AI is getting more expensive from premium providers. Sam Altman has talked about metering, which many companies won't want. If they can get a cheaper alternative with similar results, they'll pursue it. Cheaper open-source and Chinese models are a big threat. There's even talk of SpaceX doing an open-source model. Anthropic and OpenAI have to account for this when modeling their IPOs and total addressable markets, and investors will study it before buying into those stories.
Where the Money Is Going
A lot of flow is moving into Europe, especially into large European equity selldowns, with strong demand for big block trades on European exchanges. This is a genuine diversification play away from AI, because Europe has very limited exposure to the sector while the US has led it.
Consider bank stocks. If an AI bubble bursts, it could realistically hurt the earnings of JP Morgan, Goldman Sachs, and other major Wall Street banks whose pipelines lean heavily on the sector. A European bank, where AI matters far less to client flow, becomes an alternative. There has already been an uptick in the Dow versus the Nasdaq, with a clear performance edge for the Dow Jones over the Nasdaq 100. Notably, the Stoxx 600 has outperformed both indices during this AI selldown wave. Top-level indices make it hard to see exactly who's moving money where, but the trend points to money leaving big US tech and heading to regions with less AI exposure.
Buy the Dip, or Rotate?
The hard part with buying the AI dip is knowing the true value and where the dip actually is, the level that's fundamentally attractive to catch the falling knife. For an investor who studies balance sheets and discounted cash flows, the traditional tools we were told to ignore when SpaceX listed, it's very hard to call the right level to buy.
Two paths exist. This could be a general correction that strips out excess valuation and marks the start of a truly pronounced sell-off, in which case you go fully into traditional defensive and cyclical assets. Or, if you think the drop is overdone, you buy in. Take Nvidia: a fantastic business with great revenues, but if AI growth over the next five years falls short of what was projected at the start of the year, what happens to its equity story and valuation? How far does the valuation have to fall before the dip becomes fundamentally attractive?
Every investor cycling out of AI has to weigh whether this is a true market reset, meaning we were wrong about how we valued the trade from the start and the stocks have much further to fall. It might not be. Good results could bring buyers back and send stocks off to the races again, at least until the next quarter. It's very hard to call right now. For anyone wanting to pull money out of the trade, sitting in cyclicals, consumer, and more defensive names may be the more sensible play.
Nvidia won't report until the end of summer, but several names between now and then should offer clues about what to expect from it.
