That's crazy! The biggest bottleneck for AI isn't the chips, but energy (or its scarcity), and that's already showing up in electricity prices.

It took 20 years for electricity prices in the US to rise by 68%. Then, it only took 5.5 years for a 40% increase.

https://m.stacker.news/130696

And that's no coincidence. The IEA says the world has entered the "Electricity Age" after years of slow growth.

Between 2025 and 2030 alone, global demand is expected to grow by something like ~5,500 TWh. Now comes the question: who is driving this?

The obvious answer is: USA, China, and India. But there's a new "black hole" of demand that's growing too fast:

DATA CENTERS

According to the IEA, they consumed ~415 TWh in 2024.
And they could reach close to ~945 TWh by 2030.
That's almost double. By 2035 it should triple.

In the US, data centers have become the number one industry.

The biggest increase in electricity demand in the next decade isn't coming from industry. It's coming from AI.

The IEA projects that the annual increase in global electricity generation should be close to ~1,100 TWh/year in the coming years.

And most impressively: China accounts for ~40% of this growth. And most of it comes from solar power.

In the US specifically, the largest share will come from Nuclear (2026 and 2027), followed by Natural Gas (2028-2030).

https://m.stacker.news/130697

And there's a detail that almost no one has noticed: data centers aren't in the middle of nowhere.

They're located near large cities. In other words, they start competing with residences for the same available electricity, causing prices to skyrocket.

That's where excellent transmission lines (grids) become necessary.

Today the global grid is old, congested, and lagging behind demand growth:

👉 The IEA says that more than 1,500 GW of renewable projects are stuck in the connection queue because the grid cannot absorb them.

👉 Grid investments need to be >US$600 billion/year by 2030 to keep up with demand.

👉 Without upgrades, the risk of overload, rationing, and blackouts increases. Not due to a lack of energy, but due to a lack of grid that delivers energy where it is needed.

If the grid doesn't keep up, approximately 20% of planned data centers could be delayed.

And the bottlenecks are already here:

• New transmission lines take 4 to 8 years in advanced economies
• Delivery time for transformers and cables has doubled in the last 3 years
• Gas-fired power plant turbines already have lead times of several years (many only after 2030)

And this is already having real-world consequences:

Overloaded grids + poor planning = blackouts.

This is what happened in the Spain/Portugal blackout in 2025.

And here comes the most important part: the grid problem is NOT just about building more lines.

The IEA says it's possible to unlock 1,200 to 1,600 GW of capacity in the short term just by improving the existing grid.

• 750 to 900 GW via "non-firm connections"
• the rest via technological upgrades (DLR, flow control, re-conductoring, etc.)
In other words: the bottleneck is engineering + regulation.

https://m.stacker.news/130698

What is a "non-firm connection"?

It's basically this:

You get access to the network faster, but you accept an agreement where, during peak hours, the operator can temporarily reduce your load.

It's the fastest way to get data centers (and renewables) up and running before the physical expansion of the grid is complete.

And this is where the opportunity arises. The AI race won't just reward chips.

It will reward those who sell what's missing in the real world: transformers, cables, substations, circuit breakers, engineering, and construction.

The AI "boom" has become a super cycle of capital expenditure for the grid.

https://m.stacker.news/130699

To address this topic, there are 3 very different "buckets," and these are already flying by 2026.

• Grid / electrification (equipment + engineering)
• Regulated utilities (capex with permitted return)
• Gas + midstream (firm energy for data centers)

Each reacts differently to interest rates, regulation, and growth.

https://m.stacker.news/130700

This is the part that almost no one is pricing in:

AI has turned into a battle for physical infrastructure.

And that completely changes where the opportunities lie in the coming years.