Everyone is investing in AI. Who is investing in the electricity to power it?

The digital economy has a physical bill.

The International Energy Agency estimated global data-centre electricity use at roughly 460 TWh in 2024 and projected it to more than double to about 945 TWh by 2030 in its base case. That is a forecast, not a guaranteed outcome; efficiency gains and AI adoption could move it in either direction.[1][2]

Every new compute campus needs far more than chips. It needs a reliable grid connection, switchgear to distribute power, transformers to change voltage, backup power, and cooling to keep equipment within operating limits. The investment question is not simply “will AI grow?” It is “which scarce, necessary inputs can suppliers deliver profitably?”

The grid is not a software update.

New data-center demand meets an electricity system built over decades. Grid connection queues, siting, labor, and equipment procurement all move on different clocks than software releases. The U.S. Department of Energy's 2022 supply-chain report described distribution-transformer lead times rising from a pre-2022 range of three to six months to one to two years, while large power transformers often took more than two years. These historical figures describe that report's period, not a live quote for today's orders.[3]

A later DOE report to Congress identifies long replacement lead times for large power transformers as a resilience issue.[4] More demand can support pricing and backlog for suppliers, but a backlog only becomes an investment payoff if companies execute, protect margins, and convert orders into cash.

Three ways to meet the load.

These are research subjects, not buy recommendations. The financial references below use FY2024 filings; check the newest filings and current prices before making a decision.

What if the bottleneck disappears?

The strongest case against this thesis is that AI's electricity demand grows more slowly than expected—or becomes easier to serve. More efficient chips and models, slower spending by large customers, faster grid upgrades, or excess new capacity could undercut equipment demand. Even if the grid investment cycle persists, higher competition and capacity expansion could push prices and margins down.

What could $10–$50 a month look like?

You do not need the price of a full share to learn from this theme. Some U.S. brokerages allow dollar-based fractional purchases of eligible stocks and ETFs for as little as $1 or $5; eligibility varies by provider and security.[8][9] A thematic infrastructure or utilities ETF may spread single-company risk, but can still be concentrated and charge an expense ratio. Check its holdings and fees before buying.

A practical framework: start with an emergency buffer and high-interest debt priorities, choose an amount you can repeat without strain, and consider a diversified core before putting a small portion toward a theme. $10–$50 is an illustrative monthly budget, not a prescribed allocation. Fractional shares can have different transfer and voting rules; verify them with your broker.[9]