You’ve been warned about the AI bubble so many times you can probably recite the script by now. Valuations are insane. Revenue doesn’t justify the spend. OpenAI is burning billions. Anthropic is burning billions. Every startup with “AI” in its pitch deck is burning millions. The crash is coming, and when it hits, it’ll make the dot-com bust look like a gentle correction.
And Silicon Valley’s response? Good.
Not “good” in a performative, let-them-eat-cake way. Good in a cold, calculated, we’ve-seen-this-movie-before way. Because the people who actually build things in this industry understand something that almost everyone else misses: the bubble isn’t a bug. It’s the feature.
The bubble is how you pay for infrastructure that no rational market would ever fund incrementally.
Think about what happens when billions of dollars pour into AI. Sure, most of it goes to companies that will die. But where does the money actually go? It goes to NVIDIA. It goes to TSMC. It goes to data center construction crews in Arizona and Ireland. It goes to power grid upgrades. It goes to thousands of engineers who spend three years learning how to train large language models — and who will carry that expertise forward regardless of which logo is on their paycheck.
The companies die. The infrastructure stays.
This is exactly what happened in the dot-com era. Pets.com evaporated. Webvan evaporated. Excite@Home evaporated. But the fiber optic cables they laid? Still in the ground. The data centers they built? Still humming. The engineers who learned TCP/IP and HTTP and database scaling? They went on to build Google, Facebook, Amazon, and everything else that defined the next two decades.
The bubble paid for the foundation. The survivors just collected the rent.
Every great technology era is built on the corpses of overfunded companies that died trying to run before anyone could walk.
So when you hear that venture capitalists are pouring money into AI startups at absurd valuations, and you think “this is irrational” — you’re right. But you’re also missing the point. The irrationality is the mechanism. Rational markets don’t build trillion-dollar infrastructure on speculation. Irrational markets do, because they’re too drunk on the upside to price the downside.
And here’s the part that should make you uncomfortable: the people complaining loudest about the AI bubble are the same people who’ll benefit from it. The enterprise CIO who says AI is overhyped today will be running his entire operation on the data centers and models that the bubble built — five years from now, at a fraction of the original cost.
The bubble subsidizes the future. The future just doesn’t send a thank-you note.
The winners of any technology revolution are never the ones who paid for it. They’re the ones who showed up after the bill was settled.
Now, this doesn’t mean bubbles are painless. People lose money. Careers get disrupted. Promising companies get cut down before their time. The human cost is real, and it’s ugly. But if you’re in tech, in investing, or in policy, and your job is to understand where value is actually being created, you need to stop looking at the companies and start looking at the substrate.
Watch the data centers. Watch the power deals. Watch the talent migration. Watch the supply chain. That’s where the bubble’s money is actually landing — and that’s what will still be there when the hype collapses.
The bubble will pop. Companies will fail. Billions will evaporate. And underneath all that wreckage, a new infrastructure layer will be quietly waiting for the next generation to build on top of it.
The bubble isn’t a mistake the market makes. It’s the tax the future charges the present to exist.
FAQ
Q: Isn't a bubble just wasteful speculation that destroys capital?
A: Yes, it destroys capital — but it also builds physical infrastructure that rational markets would never fund upfront. The dot-com bubble burned billions and still gave us the fiber backbone of the modern internet. Wasteful and foundational aren't mutually exclusive.
Q: So what should I actually do with this insight?
A: Stop tracking which AI startups will win. Start tracking where the money is physically landing: data center construction, power grid deals, chip supply chains, and engineer hiring patterns. The substrate outlasts the logos.
Q: Are you seriously saying we should cheer for financial destruction?
A: No. I'm saying the destruction is coming regardless, and the smart move isn't to prevent it — it's to understand what it builds. The bubble will pop whether you like it or not. The question is whether you can see what's being built beneath the noise.