“Is AI a Bubble or Not?” Is the Wrong Question

I’ve spent a lot of time lately thinking back to my dot-com days as a way to understand this moment. Is AI a bubble or not? I don’t think that’s the right question. It’s far too simplistic.

Bain put a number on the buildout last week: the AI industry needs roughly $6 trillion in annual revenue by 2031 to justify the data centers going up now, and over $4 trillion of that has to come from products and markets that don’t exist yet. Goldman expects hyperscaler capex to hit $1.2 trillion next year.

By 1999 I was building Handshake.com, convinced the internet would change everything. To double down, I also bought Global Crossing stock with the little cash I had (figured I’d win on the application layer and the infrastructure layer at the same time). I lost on both.

I wasn’t wrong about the internet. It was every bit as transformative as I believed. I was wrong about where the money would land, and when. Friends who graduated a year or two ahead of me sold to Microsoft or went public. Maybe they were smarter?! Mostly, I think, it was timing.

The real money from that era wasn’t made at the peak. It was made by companies disciplined enough to still be alive in 2002-2004, when capital had fled and everyone was traumatized. Global Crossing raised enormous sums and still went bankrupt.

I learned that one the hard way. At Handshake we spent half our money on consumer marketing before pivoting to B2B. B2B was probably the right call, but by then we’d burned the time and runway to see it through.

And nobody tells you how excruciating that patience is. You look like an idiot for a long time before you look smart, if you ever do. Back then, the big spenders got to be geniuses on paper for years.

I’m watching a version of this play out in our own portfolio. Whoop started in 2012 and Verkada in 2016, long before anyone called AI a wave. Neither was started because AI was hot. Now you can talk to your Whoop coach about years of your own data, and Verkada customers can search hours of footage just by describing what they’re looking for.

That’s what patience buys you: when the wave finally shows up, it comes to you. So I’m not trying to call a top. I couldn’t do it then and I can’t do it now 😉