Brookfield CEO: AI Infrastructure Supply Bottlenecks Constrain Computing Power, Slowdown Comes Before AI Companies Acknowledge It
nashnova research
Brookfield CEO Bruce Flatt told investors the AI slowdown isn't a choice by AI companies — physical bottlenecks in power and construction were already throttling compute expansion before anyone said a word about pulling back.
AI companies say they're slowing down — but what's really hitting the brakes?
Flatt's core judgment: the slowdown is not a strategic decision by AI labs. Infrastructure simply cannot keep pace with demand.
His exact words: "As an industry, we don't build enough. We can't even build a fraction of what everyone thinks they need."
This means → The "go slow" statements from Anthropic's Dario Amodei and OpenAI's Sam Altman look more like narrating a physical fact that already exists than announcing a genuine strategic pivot.
Where exactly is the bottleneck — and why can't money fix it?
Flatt was explicit: the constraint is not capital — it's physical supply, above all electricity.
Uptime Institute — a data-center research body — called power the "defining constraint" on data-center growth for 2026 and beyond.
In plain terms = The money is there. But grid expansion and power-plant construction run on multi-year timelines that no amount of spending can compress. The bigger the AI workload, the tighter the grid — and large-scale power deployment is inherently slow.
How extreme are the demand-side numbers?
Per Statista-compiled guidance, Meta, Microsoft, Alphabet, and Amazon plan combined 2026 capex of roughly $760 billion — nearly four times the ~$200 billion they spent in 2022.
By company: Amazon ~$220 bn, Alphabet up to $205 bn, Microsoft ~$190 bn, Meta ~$145 bn.
This reflects a demand side flooding in capital at accelerating speed — while the supply side's power and construction cycles operate on a completely different timescale. The gap is widening, not closing.
How does Brookfield itself read this situation?
Brookfield estimates AI infrastructure will require over $7 trillion in capital investment over the next decade.
Flatt called the slowdown a good thing — it "brings more discipline to the entire system."
This means → For an infrastructure asset manager like Brookfield, the wider the supply-demand gap and the longer the cycle, the stronger its pricing power as the "pick-and-shovel" seller. A slowdown isn't bearish — it's an order reset.
What variable should investors watch?
Flatt's own test: whether the supply-demand gap narrows over the course of this capex wave — that is the key to judging AI infrastructure investment returns.
In plain terms = Don't just track how much AI companies spend. Track whether power and construction delivery can keep up with the spending. If the gap keeps widening, compute stays scarce and infrastructure players profit. If it closes fast, returns face a haircut.
市场有风险，内容仅供研究参考，不构成投资建议。