Just been digging into something that's been nagging at me lately -- the whole AI valuation story doesn't quite add up when you look at actual usage numbers.

So here's what caught my attention: Yeah, 41% of American workers have experimented with AI tools, but only 13% actually use them daily. That's a massive gap. People are trying this stuff, sure, but they're not really integrating it into their workflows yet. The average American spends less than 6% of their workday with AI. Compare that to how the market is pricing these companies, and you start to see the disconnect.

Now, there's definitely a bull case here. Meta, Microsoft, Alphabet, and Amazon are collectively planning to drop over 500 billion on AI infrastructure this year. Companies don't commit that kind of capital without believing there's real opportunity ahead. Plus, unlike the dot-com era, these aren't unprofitable startups -- they're cash-generating machines. And the tech itself is legitimately improving. We're getting close to agentic AI systems that can handle complex workflows autonomously. If that pans out, the economic upside is genuinely enormous.

But here's where it gets uncomfortable. Stock valuations tied to AI are at extreme multiples -- we're talking cyclically adjusted P/E ratios that only spiked higher during the dot-com peak and the Covid crash. There's also this weird circularity: most of the AI revenue right now is companies selling to other companies. Real end-user adoption and external revenue streams? Still pretty thin. That's the foundation everything needs to rest on, and it's not there yet.

What really concerns me though is the debt factor, which honestly doesn't get discussed enough. Companies like CoreWeave are leveraging up massively to build out data center infrastructure, betting that AI demand will inflate fast enough to cover those obligations. This playbook never ends well when the credit environment shifts. We've seen this movie before -- 1929, 2000, 2008. When rates were cheap and capital flowed freely, everything looked fine. But the moment lending tightens or rates stay elevated, the overleveraged players crack first.

Here's the kicker: a Bank of England survey found that 9 out of 10 senior managers said their AI initiatives haven't produced measurable improvements in productivity. That's a staggering reality check against the trillions in market value supposedly justified by these projects.

I'm not saying panic and sell everything -- that's never the move. But I think we're closer to a meaningful correction than we are to early innings. The smarter play right now is using this as a portfolio stress test. Look hard at what you own. Are these companies that can actually survive a major drawdown? Will they come out stronger on the other side? That's the question I'm asking myself.