I’ll say it plainly: I don’t think the right question about artificial intelligence is whether there’s a bubble. The right question is who is paying for it while it lasts, and who’s left holding the bill when it’s over.
What’s actually real
On October 8 the Nasdaq 100 fell 1.4% —its worst one-day drop in seven weeks— after the Financial Times reported that OpenAI’s revenue run rate was $50 billion in September, $20 billion below what it had implied earlier (Bloomberg). Bank of America had already warned days earlier that the ten biggest AI names make up 41% of the S&P 500 —a concentration level that, by its own analysis, exceeds tech and telecom at the dot-com peak. Nvidia, Apple and Microsoft alone account for nearly 21% of the index. I’ve spent more than 15 years watching capital flows, and I’ve never seen such a large bet rest on such a small group of names.
What nobody tells you
Here’s the uncomfortable part: the four big hyperscalers —Amazon, Microsoft, Alphabet, Meta— have already committed between $600 billion and $725 billion in 2026 capex, growing 60-80% year over year, while their combined revenue grows only 15-16%. Amazon’s free cash flow collapsed from $38.2 billion in 2024 to $11.2 billion in 2025, and that gap is increasingly being covered with debt, not profit. There’s also roughly $1.5 trillion in remaining performance obligations still to convert into real revenue —a figure that depends on a tightly concentrated group of labs and suppliers, starting with Nvidia, continuing to deliver for each other.
Who wins if this bursts? Value investors who stayed on the sidelines, convicted short sellers, and anyone with fresh liquidity once multiples correct. Who wins if it doesn’t? Those who already built real capacity and are monetizing it —not everyone who merely promised to. As Ray Dalio put it a year and a half ago: “a great company that gets expensive is much worse than a bad company that’s really cheap” (Benzinga). That’s not a verdict on whether AI works. It’s a verdict on whether price has already outrun profit.
The opportunity —and what I’d do
If you invest from Colombia, the most common trap isn’t ignoring AI —it’s doubling your concentration without noticing. If you already hold a global index fund and then add a thematic AI ETF on top, you’re not diversifying, you’re betting twice on the same risk. I’d rather have broad exposure through a diversified index than bet on a single sector that already makes up 41% of that very index.
On the exchange rate: the Colombian peso closed at 3,209.78 per dollar on October 6, well below the 3,700-plus level of barely a year ago (Portafolio). A stronger peso is, objectively, a cheaper window for anyone planning to build dollar exposure calmly. But that currency window should never substitute for diversification discipline —they’re two different decisions, and mixing them up is like buying something cheap that you didn’t need two of in the first place.
Have you checked whether your portfolio is betting twice on the same AI story, just because it arrives wrapped in different names?
If you want to keep following this story day by day, I’ll see you in the ViSFI Daily Briefing.