You’ve watched the headlines. Big tech dumps billions into AI. Investors panic. Stocks wobble. The chorus of “bubble” gets louder every quarter. And if you hold a 401k, work in tech, or just pay attention to where the world is going, you feel that knot in your stomach — the same one you felt in 2000, in 2008, in 2022.
But here’s what nobody on CNBC will tell you: the sell-off isn’t about AI being a bubble. It’s about a fight over who gets to own the future.
Wall Street wants quarterly returns. Silicon Valley wants to own the infrastructure of the next century. These two things cannot coexist peacefully.
Look at Alphabet. The article everyone’s panicking about mentions them as the poster child for reckless AI spending. So I looked it up. Their stock is up 86% over the last 12 months. That’s not a company in crisis. That’s a company playing a different game than the one analysts think they’re watching.
The game is this: build so much compute capacity, at such scale, that nobody else can afford to compete. Microsoft, Google, Amazon, Meta — they’re not betting on a chatbot. They’re building power plants for intelligence. And the investors screaming about ROI are standing on the tracks, wondering why the train isn’t stopping for them.
When you control the compute, you don’t need to win the application layer. You tax everyone who does.
This is the part that should make you uncomfortable. The narrative says AI will democratize intelligence. The reality is that AI is concentrating it. The companies spending the most are creating artificial scarcity — limited chips, limited data center capacity, limited energy — and then renting access at whatever price they want. They’re not commoditizing themselves. They’re commoditizing everyone else first.
Think about it. A startup can build a brilliant AI product. But they still need to rent GPUs from Amazon or Microsoft or Google. The big tech companies don’t care if your startup succeeds or fails. They get paid either way. It’s a tollbooth, and they’re pouring concrete right now while everyone argues about whether the road leads anywhere.
The innovator’s dilemma doesn’t apply when you’re the one building the dilemma.
Now, will some of this spending be wasted? Absolutely. Billions will evaporate. Startups will die. Hardware suppliers like NVIDIA will feast and then starve when the cycle turns. Your pension fund might take a hit if fund managers over-indexed on AI hype without understanding the underlying power dynamics. That’s real, and it matters.
But confusing short-term capital allocation pain with long-term strategic failure is the kind of mistake that makes people sell Apple at $12 and buy it back at $180. The companies dumping capex into AI infrastructure aren’t gambling on a product. They’re ensuring that in ten years, the only way to build anything intelligent is to pay them rent.
Every tech bubble has a core of truth buried under the mania. This time the truth is simpler than anyone wants to admit: compute is the new oil, and big tech just claimed the wells.
So when you see the next headline about AI spending sparking a sell-off, ask yourself one question: who’s panicking, and who’s building? Because those are rarely the same people. And history has a very clear preference for which ones end up owning the future.
FAQ
Q: Isn't this just the dot-com bubble all over again?
A: No. In 2000, companies burned cash on websites with no revenue model. Today's spending goes into physical infrastructure — data centers, chips, energy — that generates recurring rental income regardless of which AI products succeed. The bubble risk is in AI startups, not in the infrastructure owners.
Q: So should I buy big tech stocks right now?
A: That's the wrong frame. The real implication is that infrastructure dominance creates a moat that compounds over time. If you believe compute is the new oil, the question isn't whether to buy — it's whether you can afford not to have exposure to the companies building the wells.
Q: What if open-source AI makes big tech's compute moat irrelevant?
A: Open-source models still need compute to run. You can download a model for free, but you'll still rent GPUs from Amazon or Microsoft to serve it at scale. Open-source democratizes the software layer while big tech tightens its grip on the hardware layer. It's a trap, not a threat.