AI capex just broke through the dot-com peak. US spending on info processing equipment + software hit 4.9% of GDP — higher than the 2000 bubble's 4.5%.

This is the part where everyone either screams "bubble!" or "this time is different!"

Here's what I'm watching: capex itself isn't the problem. The question is whether the ROI shows up. In 2000, companies spent billions on infrastructure that eventually worked... just not for the companies that spent it.

Right now? Hyperscalers are betting the farm. Startups are burning cash on GPUs. Enterprises are "piloting" AI projects with unclear payback periods.

The winners won't be the ones spending the most. They'll be the ones who figure out how to turn this into actual margin improvement and revenue growth — not just cool demos.

I'm staying in boring businesses with real cash flow until the AI capex cycle proves out. Let others chase the infrastructure build. I'll look for the picks-and-shovels plays that survive when half these projects get written down.