The $720 Billion Grid Crisis Stalling AI
Here's the real problem that the AI revolution is facing today.
The AI revolution has a power problem. While tech giants pour hundreds of billions into AI development, they're hitting an unexpected wall that has nothing to do with chips, algorithms, or talent. The problem is far more fundamental: America's electrical grid can't handle what's coming.
Goldman Sachs Research has recently estimated that about $720 billion of grid spending through 2030 may be needed to meet the surge in AI-driven power demand. To put that in perspective, that's roughly equivalent to the entire market capitalization of JPMorgan Chase (JPM) and 45% more than what the U.S. spent building the Interstate Highway System (adjusted for inflation).
Clearly, this isn't a theoretical problem for some distant future. It's happening right now, and it's already reshaping where and how the AI revolution unfolds.
The Supply Crunch Is Already Here
The scale of this capacity crunch becomes clear when you look at the numbers. My former employer, Goldman Sachs also estimates that data centers will go from 92% occupied in 2024 to 97% by 2026 – it was 84% as recently as 2023. Just take a look at the graph below.
Note: Data center occupancy is the percentage of available data center space that's already rented or used. Think of it like an apartment building—97% occupied means almost no space left for new tenants.
The problem? That narrow 5% cushion represents the difference between a functioning AI ecosystem and complete infrastructure gridlock. When data centers hit 97% capacity, there's virtually no room for the explosive growth everyone is banking on.
What this means on the ground is that data center developers now have to wait longer to hook their projects up to the electric grid. According to Dominion Energy Virginia president Edward Baine, "It could be as quick as two years, it could be four years depending on what needs to be built."
Four years. In AI terms, that's an eternity. And these wait times are about to get even longer as we approach 97% capacity.
The Industry Is Building, But It's Not Enough
What’s surprising is that data centers are running out of capacity even as new construction accelerates. The next chart illustrates this.
According to IEA research, because of AI, the U.S. economy is poised to consume more electricity in 2030 for processing data than for manufacturing all energy-intensive goods combined, including aluminum, steel, cement and chemicals.
Morgan Stanley recently estimated that U.S. data centers alone will need about 57 GW of power over the next several years. If you have been following along here at Prinsights for some time now, we’ve mentioned that this Morgan Stanley analysis projected a massive 36 GW shortfall – simply because there isn't enough power to go around.
Now, Goldman Sachs' latest projections suggest the shortfall could be much worse.
And this isn't just a U.S. problem. The International Energy Agency (IEA) projects that global data center electricity demand will more than double by 2030, reaching ~945 TWh – that’s more than Japan's entire electricity use. The infrastructure strain is becoming a worldwide challenge.
So, why the Bottleneck?
The core problem is timing. It takes six or seven years and thousands of approvals to secure power for a major project. Meanwhile, AI development cycles are measured in months, not years.
The extreme case is West London, historically considered a hub for data centers, where new facilities now have to wait until 2030 to connect to the grid.
We don't expect the situation to become this severe in the U.S., but the bottleneck is very real here, too.
The Hidden Winners in America's Grid Crisis
While tech companies scramble to find power solutions, there's a category of companies that most investors are overlooking: the firms that actually control and optimize the electrical grid itself.
We’re not talking about utilities. We’re talking about the companies that make the technology that runs the utilities.
Think about it. Every new power plant, every grid connection, every load balancing decision that keeps the lights on during peak AI demand – it all depends on industrial automation and control systems. These aren't glamorous businesses, but they're absolutely essential infrastructure.
The companies that make the "nervous system" of America's power grid – the sensors, software, and control systems that manage everything from nuclear plants to renewable energy integration – are about to become incredibly valuable. They're the hidden layer that makes the entire AI revolution possible.
For these companies, the $720 billion grid crisis isn't just an obstacle to AI development. It's creating one of the largest infrastructure investment cycles in American history. The firms that provide the critical automation and control systems for this buildout are positioned for extraordinary growth.
This is the kind of undervalued company we’ll be covering with our Prinsights Pulse Premium analysis for our monthly issue due out this week. So, if you’re a paid subscriber, keep an eye out for that! We’re delivering research and a recommendation on an overlooked tech leader that couldn’t be more relevant or timely. It’s a dominant player in the critical control systems behind America’s electrical grid, and it is ideally positioned to capture outsized value from the AI infrastructure boom.
If you’re not a Premium reader yet, you can upgrade here now and join our growing community!
Wouldn't this be great for the poor millions of white collar workers waiting for their heads to be
chopped at the AI guillotine ? Amen.
What's getting little/no exposure is "reconductoring" ..... i.e., replacing current tower to tower grid transition lines with updated versions having at least double the current transition capacity/capability. MMM is the only publicly traded company that manufactures such lines. There are at least two privately owned companies that also make such advanced cables. The Dept of Energy knows of and and has written releases on the topic. Restringing the grid now would increase current capacity and anticipate future generation (nuke/SMR) expansion.