The constraint on the AI economy just shifted from money to muscle—nobody knows how to build fast enough.
The Summary
- Brookfield Asset Management is heading toward a record fundraising year, driven by massive demand for AI infrastructure investments
- CEO Connor Teskey says the bottleneck isn't capital availability, it's the shortage of operators who can actually build data centers and power infrastructure at AI-scale speed
- The implication: institutional money is pouring in faster than the physical world can absorb it
The Signal
Brookfield manages over $1 trillion in assets. When its CEO says they're on track for a record year and the constraint isn't money, you should listen. Connor Teskey told Bloomberg that institutional investors are fighting to get allocation in AI infrastructure deals. The capital is there. The problem is execution capacity.
This is the infrastructure phase of the AI boom playing out in real time. Data centers need power. Power needs grid upgrades. Grid upgrades need permitting, construction crews, transformers, and people who know how to string it all together without blowing timelines. None of that scales like software.
"The biggest challenge isn't capital—it's finding operators capable of building fast enough."
Teskey's framing matters because Brookfield isn't a venture fund chasing the next model. They build airports, toll roads, utilities. Physical assets that take years and move slow. If they're calling speed the constraint, it means the gap between AI compute demand and infrastructure delivery is widening, not closing.
Consider what this means for the agent economy buildout:
- Training runs need megawatt-scale data centers that don't exist yet
- Inference at scale needs edge compute closer to users, which requires distributed build-outs across dozens of markets
- Both need power infrastructure that's decades behind where AI companies need it to be today
The money is ready. The shovels aren't. That's a different kind of scarcity than Silicon Valley is used to pricing in. Software economies assume infinite scale. Infrastructure economies assume friction, timelines, and the physical world saying no.
The Implication
If you're building in AI, your 2027 growth assumptions might be capped by someone else's construction timeline. If you're allocating capital, the alpha isn't in finding the best AI company. It's in finding the teams that can actually deliver the power and compute those companies need to run.
Watch for acquisition premiums on infrastructure operators, not just infrastructure assets. The humans who know how to build become the scarce resource. Brookfield is signaling that shift early.