For the last two years, the AI conversation has been dominated by capability.
Which model is better.
Which company has the most compute.
Which application reaches consumers first.
I think the more interesting question today is different:
What happens when demand for AI stops being constrained by software and starts being constrained by physical infrastructure?
The market still talks about AI as a technology story.
Increasingly, it looks like an industrial story.
Hyperscalers continue committing enormous amounts of capital toward AI infrastructure, even as investors debate whether spending has peaked. Recent estimates from major market participants suggest AI-related infrastructure spending could remain far larger and more durable than many expected.
That matters because the constraint appears to be moving.
A few years ago the bottleneck was chips.
Today, the bottleneck increasingly looks like power.
The U.S. Energy Information Administration is projecting record electricity demand in 2026 and 2027, driven in part by AI-related data center growth. Commercial electricity consumption is expected to exceed residential consumption for the first time on record.
That's a subtle but important shift.
Markets tend to extrapolate visible winners.
They are often slower to price the supporting ecosystem.
The AI narrative is still largely centered around models and semiconductors.
The underlying buildout increasingly revolves around grid capacity, transformers, generation assets, permitting, cooling systems, and financing. Multiple industry and policy analyses now identify power availability—not capital—as one of the primary constraints on future data center deployment.
That's where second-order effects begin to matter.
When a market transitions from a software bottleneck to a physical bottleneck, timelines lengthen.
Projects become harder.
Capacity cannot be created instantly.
And capital starts rotating toward assets that previously looked boring.
Another dynamic worth watching is psychology.
Most investors understand the AI opportunity.
Far fewer have fully internalized the infrastructure requirements needed to support that opportunity at scale.
The market may still be treating power demand as a side effect.
What if it's the core story?
Several forecasts now suggest data center electricity demand could rise dramatically over the next decade, putting increasing pressure on local grids and energy systems.
That doesn't automatically mean every infrastructure-related asset wins.
In fact, crowded positioning is becoming a risk in some corners of the trade.
We've already started to see situations where stocks tied to AI power demand have rallied aggressively and then faced valuation scrutiny.
The distinction matters.
A good narrative is not always a good position.
The opportunity often emerges where the market acknowledges the trend but underestimates the duration.
That's what I'm watching.
Not whether AI adoption continues.
The market broadly agrees on that.
I'm watching where the next bottleneck forms.
Because the highest-conviction opportunities often emerge one layer beneath the headline narrative.
Right now, the market is still talking about intelligence.
The bigger question may be whether it has fully priced the infrastructure required to sustain it.
— Connor
Alpha Before It Prints
© 2026 Alpha Before It Prints
Unsubscribe