Damian Maggio: Why Power, Not Just Chips, Is Shaping the Data Center Buildout

Originally published at damianmaggio.com

When people talk about AI infrastructure, the conversation usually starts with chips. After about 30 years in securities and private investment, I have learned to ask a plainer question about any capital-intensive business: what does it run on? For data centers, the answer is electricity.

The International Energy Agency examined this in its report "Energy and AI." Here is what it found.

The scale today. The IEA estimates data centers used around 1.5% of the world's electricity in 2024, or 415 terawatt-hours. The United States accounted for the largest share, 45%. The IEA also notes that a typical AI-focused data center consumes as much electricity as 100,000 households, and that the largest ones under construction will consume 20 times as much.

The projection. In its base case, the IEA expects data center electricity consumption to more than double to around 945 terawatt-hours by 2030. It stresses that uncertainty is large: by 2035 its scenarios range from 700 to 1,700 terawatt-hours, depending on how fast AI is adopted, how quickly efficiency improves and whether energy bottlenecks are resolved.

Why location matters. Data centers are geographically concentrated. The IEA says nearly half of U.S. data center capacity sits in five regional clusters, which can strain local grids even when national totals look small.

The grid bottleneck. The report says grid connection queues are long and complex, new transmission lines can take four to eight years to build in advanced economies, and wait times for transformers and cables have doubled in three years. It estimates around 20% of planned data center projects could face delays unless these risks are addressed.

Why forecasts differ. The IEA's scenarios assume different paces of AI adoption and efficiency gains. In its high-efficiency case, 2035 demand is 20% lower than in the base case.

Possible responses. The IEA points to building in areas with spare grid capacity, running servers or on-site power and storage more flexibly, and a mix of supply that includes renewables, natural gas, and over time nuclear and geothermal.

For anyone studying technology companies, public or private, this is a useful reminder. A company's growth plans can depend on things outside its control, such as grid access, equipment lead times and local permitting. Those dependencies are worth looking for in a company's disclosures. They are also a reason to treat any single forecast with caution, including the ones above.

Chips get the headlines. Power decides how quickly the buildings that house them can be switched on.

Damian Maggio writes about private markets and technology. This article is for informational purposes only and is not an offer or recommendation to buy or sell any security.



Comments

Popular posts from this blog

Damian Maggio on Tender Offers: How Late-Stage Private Companies Create Liquidity

Damian Maggio on Direct Listings vs. IPOs: Two Ways a Private Company Goes Public

Damian Maggio: Why Cybersecurity Has Become a Core Diligence Question