The House That AI Built
On the nothing economy
Let’s talk about the AI economy, since there’s little else top of mind lately. Basically, seven companies are holding up the U.S. economy right now, and they’re all betting on the same thing. Strip them out of the S&P 500 and the picture looks pretty stagnant.
AI has become the economic story we’re telling ourselves instead of dealing with the actual economy. It’s the explanation for why everything’s fine when nothing feels fine.
How We Got Here
Capital is concentrating in a way we haven’t seen since the railroad boom, except this time it’s all going to variations of the same technology. Nvidia makes the chips. Microsoft, Google, Amazon, and Meta build the models and run the infrastructure. OpenAI and Anthropic are the research darlings. That’s it, that’s the whole ecosystem. These companies have become so valuable that they’re carrying the market while everyone else watches from the sidelines, trying to figure out how to get a piece of it.
The machinery of finance has decided this is fine, actually! The concentration is treated as inevitable, even desirable. The fact that a handful of companies control the infrastructure, the talent, the data, and the distribution gets framed as maturity rather than risk.
We’ve seen this movie before. Different technology, same plot. Except usually there’s at least some pretense of competition, some gesture toward antitrust, some acknowledgment that putting all the eggs in one basket might be a problem. Not this time. The argument goes: AI is so expensive to build, so resource-intensive, so technically demanding that of course only a few companies can do it. Economics of scale, they say. Natural monopoly, they say. Cue shifty eyes.
Holding Pattern
Meanwhile, the actual infrastructure required to keep this going is becoming impossible to ignore. Data centers are going up at an alarming pace. They’re power-hungry in a way that makes cryptocurrency mining look quaint. They require water for cooling, land for construction, chips that depend on supply chains stretched across the Pacific. When we talk about AI “scaling,” we’re talking less about intelligence and more about this physicality, and it is a heavy, heavy load.
It’s all being financed by the belief that AI will eventually generate returns that justify the investment. Those returns have yet to be proven out, yet (therefore?) that belief gets stronger and stronger. The economy is reorganizing itself around a technology that’s still experimental, still unreliable, still fundamentally unproven at the scale we’re betting on. This would be less worrying if the bet were distributed. Instead, everyone’s building the same thing because everyone’s funding the same thing because everyone believes the same thing. It’s a monoculture, and monocultures are incredibly fragile.
The honest read on what’s happening is that AI has become our way of not dealing with harder problems. The economy doesn’t know where growth is supposed to come from anymore. Housing is a political minefield. Manufacturing is complicated and global. Consumer spending is tapped out. The internet already happened and turned into infrastructure. So we’ve decided that intelligence itself is the next resource to extract. If we can just make the machines think, everything else will sort itself out. It’s appealing because it sounds like progress without requiring us to change anything fundamental about how things work. We don’t have to redistribute wealth or rethink consumption or build different systems. We just have to build smarter computers.
But technologies don’t solve structural problems. They interact with structural problems, usually in ways that make them more entrenched. Newsflash: AI isn’t going to fix inequality; it’s going to accelerate it. The companies building AI are already the most powerful entities in the global economy. Making them more powerful, more central, more essential actually creates brittleness, not resilience.
What Breaks First
I’m not an economist, and I don’t know what the correction looks like. Maybe it’s slow, a gradual deflation as AI settles into being useful but not transformative. Maybe it’s fast, a moment when the market realizes the returns aren’t coming and everything shifts at once. Either way, we’re not having honest conversations about what happens when the story stops working. We’re not asking what it means for the economy to depend on a handful of companies building something most people don’t understand. We’re not talking about the infrastructure costs, the resource constraints, the geopolitical vulnerabilities. We’re not asking whether concentrating this much power and capital in one technology, one narrative, one bet is wise, even if the bet pays off. A rising tide that only lifts seven boats isn’t a tide at all.
The phrase “too big to fail” was supposed to teach us something. We learned it in 2008 when we discovered that institutions had become so central to the economy that we couldn’t let them collapse even when they should have. We said we’d be more careful next time, that we wouldn’t let anything get that systemically important again. And yet we’re here again. AI is being positioned as the activity, the engine, the source of value. When something that experimental and narrative-dependent becomes central to how the economy functions, all I can say is: look out.
And you know, maybe the bet pays off. Maybe AI delivers on enough of the promises to justify the infrastructure and the hyper-concentration. Maybe in ten years this all looks obvious and necessary, and we wonder why anyone was worried. But standing here now, watching the global economy reorganize itself around something this unproven, feels like we’re hoping really hard that this works out because we’re not sure what the alternative is anymore.
Hope makes a terrible economic policy. But it’s what we’re running on.