AI Is a Bubble. And the People Warning You About It Are the Ones Inflating It.

You’ve felt it. That quiet unease every time another trillion-dollar AI valuation gets announced with the confidence of a weather forecast on a sunny day. Everything looks fine. But you can’t shake the feeling that something is off.

You’re not crazy. You’re just paying attention.

Torsten Slok, Apollo’s chief economist, just said what everyone in the room is whispering: AI markets could face a “painful repricing.” Translation? The money pouring into AI infrastructure — data centers, chips, models — is racing far ahead of anything resembling actual productivity gains. The returns haven’t shown up. The receipts don’t exist yet. And the gap between promise and proof is stretching into dangerous territory.

The market isn’t pricing AI. It’s pricing a prayer.

Here’s what makes your stomach drop: this isn’t some outsider sounding the alarm. This is Apollo. The same Apollo that has been aggressively applying eye-watering valuations to AI infrastructure investments. The same firm sitting at the center of the capital flow machine that’s fueling the very dynamic its economist is now calling unsustainable.

So what’s really going on here?

Let me walk you through the uncomfortable logic. When the firm inflating the balloon also publishes the warning that the balloon might pop, you’re not looking at honest analysis. You’re looking at positioning. Apollo isn’t naive. They know that near-term economic returns on AI are uncertain at best, illusory at worst. They also know that being the first to say it out loud — loudly, publicly, with their name attached — gives them a strange kind of power.

If they help shift sentiment toward caution, they can acquire distressed AI assets at a discount when the repricing hits. If the hype continues, their existing positions still benefit. Heads they win, tails they also win.

The most profitable seat in a casino isn’t at the table. It’s the one where you own the building and also write the odds board.

Now, you might be thinking: maybe Slok is just being honest. Maybe Apollo is genuinely concerned. And sure, that’s possible. Economists can be right even when their employers have skin in the game. The analysis itself — that productivity gains haven’t materialized to justify current valuations — is sound. It’s what any clear-eyed observer would conclude.

But here’s the twist that should keep you up at night: it doesn’t actually matter whether Apollo is sincere or strategic. The outcome is the same either way.

The capital is committed. The infrastructure is being built. The valuations are locked in. And when the gap between expectation and reality finally closes — whether through a slow deflation or a violent crash — the people who positioned themselves early will walk away fine. Everyone else will be holding the bag.

That’s you, by the way. If you’re an investor with AI exposure in your portfolio, a tech professional whose compensation is tied to AI-adjacent equity, or a policymaker betting on AI-driven economic growth to solve structural problems — you’re in the blast radius.

Bubbles don’t kill people who see them coming. They kill people who confuse a warning for reassurance.

And that’s the real danger of Slok’s commentary landing the way it did. It feels like transparency. It feels like the system working. A major firm’s economist publishes a cautionary note, and we all nod and say, “Good, at least someone’s being honest.” But honesty without action is just theater. If Apollo believed their own warning, they’d be deleveraging. They’d be marking down their AI positions. They’d be doing something — anything — other than continuing to pour capital into the same assets their economist says are mispriced.

They’re not. And neither is anyone else.

The AI investment cycle is entering its most dangerous phase. Not the beginning, where everything is speculative and everyone knows it. Not the end, where the crash is obvious and the damage is done. No — we’re in the middle. The phase where the hype has been going on long enough that people have stopped questioning it, but not long enough for the returns to have proven anything. This is the phase where beliefs calcify. Where “it’s probably fine” hardens into “it’s definitely fine” through sheer repetition.

This is exactly when repricings happen. Not when skepticism is high. When it’s lowest.

The market doesn’t crash when everyone is afraid. It crashes when everyone has stopped being afraid and started being bored.

So what do you do with this? You stop treating AI investment as a monolith. Not every AI company is overvalued. Not every infrastructure play is a bubble. But the aggregate numbers — the hundreds of billions flowing into chips and data centers and model training — are being priced as if the productivity revolution has already arrived. It hasn’t. And the distance between “hasn’t yet” and “won’t” is shorter than anyone wants to admit.

Apollo knows this. Now you do too. The question is whether you’ll act on it, or whether you’ll read this, feel a flicker of anxiety, and then go back to assuming someone smarter than you has it handled.

They don’t. They’re just better positioned than you. And in a repricing, that’s the only thing that matters.