AI Stocks Could Crash in 2027. Three Ways to Not Get Wiped Out

Alan Greenspan gave the most famous speech of his career in December 1996. He asked, out loud, whether "irrational exuberance" had taken hold of asset prices. The Nasdaq was trading around 1,300.

It didn't peak until March 2000, at 5,048.

The Warning That Came Four Years Early

That's a near quadruple, and it took 39 months to get there. Then the index fell 78%, bottoming near 1,114 in October 2002.

So yes, Greenspan was right. He was also four years and roughly 280% too early, which is the part most bubble Cassandras conveniently skip. Investors who sold the day after that speech watched the greatest bull market of their lives from the sidelines, and plenty of them never got back in.

Timing the top isn't a strategy. It's a coin flip with a story attached.

Why Everyone's Circling 2027

Here's why this matters now. The AI capex cycle has a shape. Hyperscalers are booking depreciation on GPU clusters over four to six years, and the revenue those clusters are supposed to generate is, in most cases, still a projection. Capital spending from the big cloud providers blew past $200 billion in 2024 and guidance keeps climbing. Concentration in the S&P 500 sits near multi-decade highs, with a handful of AI names responsible for most of the index's return.

So when people say 2027, they're not pulling it out of thin air. They're counting the depreciation cliff and the point where capex growth has to flatten. That's a reasonable thesis. I'm not entirely convinced it's a schedule.

Markets don't burst on a calendar. They burst when credit tightens, when a big buyer blinks, or when earnings miss by a dime and the multiple does the rest.

The question worth asking: what's your plan if the music stops in March 2027 rather than December?

Three Ways to Protect the Money

First, cap your single-name exposure. If one AI stock going to zero would dent your net worth by more than 5%, you're holding a bet, not a portfolio. Trim into strength. It's boring advice and it works.

Second, keep dry powder. Cash yielding north of 4% is a real option, not a cop-out. The people who made fortunes in 2009 weren't geniuses. They just had liquidity when everyone else was forced to sell.

Third, hedge instead of exiting. Buy puts, set trailing stops, or sell a slice every time the position runs up 25%. You keep the upside if the rally continues and you sleep through the drawdown if it doesn't.

Nobody rings a bell at the top. The investors who sold in 1997 looked smart right up until they looked broke.

Color me skeptical, but the entire 2027 crash prediction business exists because it sells. The useful version of it isn't a date. It's a process you can run whether the crash comes in 2027, 2028, or never.

Watch hyperscaler capex guidance and high-yield credit spreads. Those two numbers will tell you more about the next two years than any forecast on your timeline.