Everyone is racing to build better AI models.
Almost no one is talking about the physical constraint underneath them:
Electricity.
AI doesn’t scale on GPUs.
It scales on megawatts.
The Reality
Training and inference loads are exploding.
Data centers are demanding gigawatts.
Utilities are facing 5–10 year interconnection queues.
Transmission buildouts move at political speed.
Gas peakers are coming back online.
Nuclear is being reconsidered.
Solar + storage is accelerating.
But here’s the key:
The grid was not designed for this load profile.
AI workloads are:
• Dense
• Constant
• Power-hungry
• Geographically concentrated
That changes infrastructure economics.
Where The Real Opportunity Is
The market is focused on:
NVIDIA
Model wars
Chips
Software
But the second-order trade is in:
• Distributed generation
• Utility-scale storage
• Substation upgrades
• Interconnection acceleration
• Orphaned solar systems
• Behind-the-meter batteries
AI is creating demand shock.
Infrastructure moves slower than software.
That gap creates capital opportunity.
What This Means
Three things are happening simultaneously:
Utilities are being forced into modernization cycles.
Private capital is flowing into grid-adjacent assets.
Distributed energy is becoming strategic, not optional.
The AI boom is accelerating the decentralization of power.
And decentralization changes market structure.
The Bigger Question
When energy becomes a bottleneck:
Who controls the power?
Who finances the upgrades?
Who owns the distributed edge?
The next decade isn’t just about artificial intelligence.
It’s about the infrastructure underneath it.
I’ll keep tracking this shift daily.
– Steve
Powercord