The Bubble Talk Got Loud
Ray Dalio says we are in the early stages of a bubble, and one lab has reportedly put around $150 billion into the buildout against a fraction of that in revenue. Expectations are running ahead of reality. That does not make the whole thing fake.
The bubble debate asks the wrong question. The spending is real, the capability is real, and the revenue still has to show up. All three can be true at once.
This week the word "bubble" stopped being a whisper and started being a headline. Ray Dalio said it out loud: we are in the early stages of one. And the number making the rounds is hard to look away from. One lab has reportedly put something like $150 billion into the AI buildout against a fraction of that coming back as revenue. Read that again. That is the kind of ratio that makes serious people nervous, and they are not wrong to be nervous.
the mood this week
The vibe shifted. A few months ago the skeptics were the ones who had to explain themselves. Now the believers are doing the explaining. Every funding announcement lands next to a quiet question: when does this pay for itself? The honest answer from most of the industry is some version of "later," said with great confidence and very few numbers attached.
I get the nervousness. I have lived through enough hype cycles to know what the top feels like, and parts of this market smell like the top. The fundraising rounds are bigger, the timelines are longer, and the word "inevitable" is doing a lot of heavy lifting in pitch decks.
what the number means
$150 billion in, a fraction back. Strip away the commentary and that is a bet, not a business. Not yet. It becomes a business if the revenue curve bends up to meet the spending curve before the money runs out or the patience does.
Bets like this have paid off before. The railroads, the internet backbone, the cloud buildout. Every one of them looked insane on the spending side while it was happening. Every one of them also had a graveyard of companies that bet right on the technology and wrong on the timing. Being right about the direction is not the same as surviving the trip.
the part that is real
Here is where I part ways with the "it is all fake" crowd. The capability is not fake. I use these models every day on real work, and they do things that were flatly impossible three years ago. The data centers are poured concrete and humming hardware, not renderings. The cost of a useful answer keeps falling, and that falling cost is the most important fact in this entire argument. It is the thing the bubble debate keeps ignoring.
A bubble in prices is not the same as a fraud in the product. Tulip bulbs were never useful. This stuff is useful right now, today, for boring work that nobody wants to do. The question was never "does it work." The question is "does it work at a price the revenue can cover," and that question is still open.
how i am positioned
I am not rooting for a crash, and I am not buying the victory lap. My position is simple, and it is the same one I take with every release: I build on the part that is already true.
The part that is already true is the cost curve. Intelligence keeps getting cheaper, and cheaper intelligence changes what one person can ship. That holds whether the funding environment is euphoric or terrified. I do not need the $150 billion bet to pay off for my work to pay off. I need the models to keep getting better and cheaper, and that trend has survived every mood swing so far.
So let the debate rage. Watch the numbers, not the narratives. And if you are building, build on the capability, not the multiple. The capability is the part nobody can take back.