The AI boom just ran into concrete walls—literally—and now your old cubicle farm might become someone's apartment.
The Summary
- PGIM's real estate chief says AI data center buildout is hitting physical infrastructure limits, while housing supply remains a "serious" issue
- Empty office towers are getting second lives as residential buildings as remote work permanently reshapes commercial real estate
- The collision of AI infrastructure demand with construction capacity constraints is forcing capital allocation choices between data centers and housing
The Signal
PGIM manages over $200 billion in real estate. When their global COO says the data center boom is testing buildout limits, that's not speculation. That's someone watching capital hit walls.
The constraint isn't money. It's transformers, fiber, construction crews, and the years it takes to build utility infrastructure for power-hungry GPU farms. Every hyperscaler wants rack space yesterday. The physical world doesn't move at software speed.
"AI infrastructure demand is colliding with the reality that you can't magic new electrical substations into existence."
Meanwhile, offices sit empty and housing stays expensive. Marcus sees office-to-residential conversions accelerating. Not because of clever policy, but because the math finally works. When Class B office buildings trade at distressed prices and housing supply is tight, conversion pencils out.
This is three trends colliding:
- Remote work made permanent by AI tools that make distributed teams more productive
- Housing undersupply that's been building for a decade
- Capital desperate for yield finding new use cases for stranded commercial real estate
The Implication
Watch where the capital flows next. If data center construction is capacity-constrained, developers will chase the next infrastructure play that AI enables. Retrofitting office buildings for residential use is labor-intensive, local, and hard to automate—until it isn't.
The agents don't just need data centers. They need the humans building them to live somewhere. The AI infrastructure boom creates its own housing demand in secondary markets where land and power are cheap. Your next apartment might be in a former insurance company HQ in Columbus, powered by the same grid feeding the GPUs next door.