The Grid Still Isn't Ready for AI

Editor's note: America's power grid wasn't built for the AI boom. As electricity demand surges, the country is leaning on an energy system that's becoming less reliable. In this issue, Joel Litman, chief investment officer of our corporate affiliate Altimetry, explains the hidden cracks forming beneath the U.S. energy market – and the major implications that could have for investors...

The U.S. is facing an issue that we haven't seen in decades...

There's not enough power.

For the first time in decades, our energy supply is under pressure. So prices are on the rise. And we're racing to build capacity to keep up.

Today, we'll explain why America's power grid is becoming less reliable at the exact moment AI is creating a historic surge in energy demand...

And why that combination could reshape energy markets in the age of AI...

Our energy grid is built to handle times of peak energy demand...

Think the hottest days of summer, when people crank up the air conditioning. For the rest of the year, the grid operates well below its maximum capacity.

This relationship creates a common way to measure the strain on our energy system: comparing peak energy demand with the grid's total power-generation capacity.

Peak demand has been effectively flat for the past two decades, averaging around 750 gigawatts ("GW"). Total capacity also stayed largely stable, around 1,000 GW.

That was, until 2021... when energy production ramped up following the pandemic.

Said another way, peak demand only required 74% of our grid's total capacity from 2005 to 2020. And it has called for just 64% since then.

It might seem like our energy-generating capabilities are better than ever. But that's not the case... because not all power-generation capacity is created equal.

The Energy Squeeze Behind the AI Boom

Back in 2005, about half of our power came from coal. Another 20% came from natural gas, and the rest came from nuclear (20%) and renewables (10%).

Environmental concerns have led the U.S. to phase coal out of our energy solutions. It only accounts for about 17% of power today. Natural gas has largely filled this void, rising to roughly 40% of capacity. Renewables have jumped from around 10% to closer to 24%.

Renewables are the long-term goal. But they're not ready yet. Solar and wind farms can't operate 24/7 to supply the grid with constant ("baseload") energy.

And battery technology isn't advanced enough to store reliable power in the event of a surge... or even to support steady growth in energy demand.

On top of that, not all natural gas and remaining coal plants deliver constant energy. "Peaker plants" are designed to come on line only for short periods when demand is surging.

So while overall power-generation capacity has increased since 2020... the reliability of that power has declined.

In other words, our grid is less prepared to take on today's rising baseload power demand than it was 20 years ago.

That means our system is less resilient to shocks. And shocks might be exactly what we're about to get...

AI is putting a huge strain on what's already becoming a less-reliable power grid. Data centers are always on. And their power demand is insatiable.

The numbers aren't final yet. But experts believe this shift will have raised residential electricity prices across the U.S. from 3% in 2024 to 4.7% in 2025.

As for that extra reserve of energy supply we built up?

It's starting to decay again. This year, peak energy demand is expected to get back to 70% of our total capacity. By 2030, it's expected to surpass 80%.

And that's still the middle innings of the AI boom. Data centers will drive the largest increase in AI-related electricity demand... But they won't be the only one.

In other words, this isn't just an AI story anymore. It's the beginning of a long-term energy squeeze.

Regards,

Joel Litman

Editor's note: AI's biggest bottleneck is no longer chips... It's electricity. The scramble for power is already reshaping the market as Big Tech races to build what some are calling a new "shadow grid." That's why our corporate affiliates Altimetry and Chaikin Analytics teamed up to identify the lesser-known companies emerging as a critical group for the future of AI.