The Next Bottleneck in the AI Boom Isn’t Chips

Two years ago, Elon Musk predicted that electricity shortages would be the next big problem facing the advancement of artificial intelligence.

In a discussion about infrastructure constraints, he said: “I think next year, you’ll see they just can’t find enough electricity.”

Most people ignored Musk’s prediction because AI had a different bottleneck at the time. GPUs were scarce and prices were rising fast. AI training runs were expensive enough to make compute feel like the only constraint that really mattered.

After all, electricity was available whenever it was needed, and it almost never showed up as the thing slowing a project down.

But that’s starting to change now.

Musk might be a year late, but his prediction is coming true because of how fast demand is growing and how slowly our infrastructure is able to keep up.

And if this power gap keeps widening, it could become a huge issue for the U.S. in the race for artificial superintelligence (ASI).

AI Growth Meets an Aging Power Grid

Electricity demand in the U.S. is rising faster than the systems that generate and deliver it can expand.

Obviously, this hasn’t always been the case. It used to be that demand growth was slow and predictable. Utilities planned years in advance and built accordingly.

But that planning model is now being tested.

One of the clearest signs is the growing backlog for large power transformers.

These are the heavy industrial machines that convert high-voltage transmission power into electricity that data centers, factories and large commercial sites can use.

Lead times for large power transformers that once measured a few months now regularly stretch to one or two years. Some orders have been pushed even further out.

Transformer production is concentrated among a limited number of suppliers. The equipment is largely custom-built, and capacity expansion takes time.

So these delays are now cascading through the system.

Grid hookups are taking longer than expected, and new power plants are sitting idle. Even projects that are otherwise fully permitted face postponements because the hardware required to deliver electricity hasn’t arrived.

Meanwhile, electricity demand is accelerating in ways that the grid wasn’t designed to handle.

Goldman Sachs estimates that global data-center power demand could grow by more than 150% by the end of the decade, driven largely by AI workloads.

Morgan Stanley warns that the U.S. could face a material power shortfall by 2028 if generation and transmission investment doesn’t speed up. Their analysis puts the potential gap at roughly 13 to 45 gigawatts, depending on the pace of infrastructure build-out.

That amount of capacity is comparable to the electricity consumption of around 33 million U.S. homes.

And it’s because data centers now run nonstop and draw huge amounts of electricity. At the same time, cars, homes and factories are all being electrified, which adds steady demand across the economy.

Add in more manufacturing moving back to the U.S., and the grid is being asked to grow much faster than it was designed to.

But the grid’s biggest challenge isn’t just reliability. It’s speed.

Power plants take years to finance, permit and construct. Transmission projects take years to approve and build. Equipment orders must be placed long before demand materializes.

The system has historically favored caution. But the tradeoff is that caution limits how quickly supply can respond.

And that constraint is showing up today as delays.

Utilities are pushing connection dates further out and developers are reworking projects around available power. Sometimes, companies have to choose sites based on where electricity is easiest to secure.

And in some cases, power access has become the deciding factor for whether a project moves forward at all.

This is changing the behavior of the companies building the infrastructure behind AI.

Large data center operators are signing long-term power agreements earlier in the planning process. Some are even investing in on-site generation rather than relying entirely on the grid.

For example, Google recently acquired Intersect Power for about $4.75 billion to gain more direct control of generation and storage assets.

Other companies are prioritizing regions with excess capacity, even if it means higher upfront costs or less attractive incentives elsewhere.

This might feel familiar to you.

Recent chip shortages showed how fragile supply chains can become when demand outpaces supply. The U.S. responded by investing heavily in new factories, adjusting policy and scaling production to close the gap.

Something similar needs to happen now to meet the surge in electricity demand being driven by AI, electrification and industrial growth.

Utilities are already increasing their capital spending today. Equipment manufacturers are expanding capacity, and new generation projects are moving forward across a range of technologies. <

But there’s still a lot of work to be done so we don’t fall further behind.

Here’s My Take

Elon Musk has a track record of spotting problems before most people do. Two years ago, he saw that AI would grow faster than the power systems needed to support it.

Today, that gap is becoming a real problem.

The next phase of the AI era will be shaped by how quickly the U.S. can build enough energy capacity to support the larger models ASI will require. Especially since China has been building electricity capacity at a breakneck pace, adding the equivalent of an entire European Union’s electric demand in just six years.

Some might see this as an insurmountable problem. But I see it as a constructive moment.

Now that electricity is no longer being taken for granted and is starting to factor into planning decisions, it should push companies and investors to make better, more realistic choices about where and how they build.

That means more money flowing into the power systems that make growth possible. Grid upgrades, new generation and essential infrastructure equipment will start getting the attention they’ve needed for years.

This will create new opportunities for the U.S. to strengthen its energy backbone while maintaining its lead in the race to ASI.

And it will give investors like us the chance to get ahead of this energy buildout.

Regards,

Ian KingChief Strategist, Banyan Hill Publishing

Editor’s Note: We’d love to hear from you!

If you want to share your thoughts or suggestions about the Daily Disruptor, or if there are any specific topics you’d like us to cover, just send an email to [email protected].

Don’t worry, we won’t reveal your full name in the event we publish a response. So feel free to comment away!