When the Building Is the Product
$100 billion. That’s the number being floated for Vantage Data Centers going public. Backed by Silver Lake and DigitalBridge. A partner on the Stargate campus in Wisconsin, alongside Oracle and OpenAI. If it lands anywhere near that number, it would be the largest data center listing in history.
Switch is reportedly chasing something similar, around $80 billion.
No data center company has ever gone public at this scale. Not close.
Here’s the thing that should stop you. We’ve spent three years being told the story of AI is a story about models — who trains the biggest one, who has the best benchmark, whose chatbot writes the better email. The model companies are the ones doing press tours. Sam Altman is a household name. Nobody at Vantage is.
But look at where the capital is actually going. It’s going into concrete, steel, transformers, cooling systems, and long-term power contracts. It’s going into square footage. The model companies — the ones we’ve been trained to think of as the protagonists of this era — are, financially speaking, tenants. They rent compute. They sign leases, sometimes decade-long ones, on buildings owned by someone else.
Software eating the world was the story of the last two decades. This is close to the inverse. The world — in the most literal sense, dirt, permits, substations — is what the software now depends on to exist at all.
Neil Postman understood that a technology’s economics reveal its actual center of gravity faster than its rhetoric does. The rhetoric of AI is intelligence, alignment, the singularity. The economics of AI, at least right now, are industrial real estate with better marketing.
That’s not a knock. It’s just a more honest description of what’s happening. When Oracle and OpenAI show up as partners on a campus in Wisconsin — Wisconsin, not San Francisco — you’re watching the frontier of AI get relocated to wherever there’s cheap land, available power, and a grid that can take the load. The compute has to live somewhere. Somewhere turns out to be the whole ballgame.
A $100 billion valuation for a company whose product is essentially “a very well-built shed with excellent electrical service” tells you something about where the real scarcity is. It isn’t algorithms. Algorithms are, increasingly, commoditized — open weights, leaked techniques, fast-following labs. What’s scarce is power. What’s scarce is land near power. What’s scarce is the industrial capacity to pour that much concrete and run that much cable in a timeframe measured in quarters instead of decades.
Landlords have always made money off other people’s ambition. What’s new is the landlord being the single largest bet in the sector.
If Vantage prices anywhere near $100 billion, watch what happens to how people talk about AI companies afterward. I’d bet the conversation shifts, at least a little, from “whose model is smartest” to “whose lease terms are better.” Because that’s the conversation the money is already having. The public narrative just hasn’t caught up.
Software ate the world for twenty years. Now the world is sending an invoice — for power, for land, for the physical plant none of this runs without. The model companies built the mind. Someone else built the body, and it turns out the body was the expensive part.
What happens when the tenants realize they don’t own the building?