06/01/2026
AI Has A Power Problem
Most investors still think the AI race is about models and GPUs.
But according to NVIDIA CEO Jensen Huang, the real constraint sits much lower in the stack.
Huang recently described AI as a five-layer system:
• Energy
• Chips and compute
• Infrastructure (data centers, networking, cooling)
• Models
• Applications
The key insight is simple: every layer above depends on the layer below.
For years, the market focused on semiconductors because they were the bottleneck. Today, that bottleneck is beginning to shift toward power generation and grid infrastructure.
This is a notable conclusion coming from the CEO of NVIDIA itself. Huang's argument is that every token generated by AI ultimately requires electricity. Without sufficient power, the rest of the stack cannot scale.
The implications are significant:
• Future data centers may be power-constrained rather than chip-constrained.
• Grid interconnections and substation upgrades can take years, while compute capacity can be deployed much faster.
• Large technology companies may increasingly invest directly in energy assets and generation capacity.
• AI factories may need to dynamically manage workloads around available electricity instead of unlimited computing resources.
This aligns with a broader trend emerging across the industry.
As AI adoption accelerates, demand for electricity is becoming a strategic issue. Goldman Sachs estimates data center power demand could increase by roughly 175% this decade.
The market narrative is evolving.
The first phase of the AI boom was about chips.
The next phase may be about who can secure the energy required to run them.
In my latest article, I examine one of the most interesting U.S. companies operating at the intersection of AI infrastructure and energy, providing mobile natural gas-fired generation and balance-of-plant equipment for data centers, hyperscalers, and industrial customers.
The companies powering AI may not be the only winners.
The companies powering the power may matter just as much.
This is bigger than headlines. Follow if you want to understand what comes next.
