Data Centers and the AI Power Crunch: The Infrastructure Investment Case

Every AI model runs somewhere. Behind the chatbots, image generators and AI agents sits a physical infrastructure of data centres, chips, cooling systems and, above all, electricity. As AI adoption has accelerated, so has demand for that infrastructure, and power has become one of its tightest constraints.

The International Energy Agency has projected that data centre electricity use could roughly double by 2030, driven largely by AI. Large technology companies are spending hundreds of billions of dollars a year on infrastructure, and some have signed agreements to secure power from nuclear plants. For investors, this creates opportunities well beyond the chipmakers. This guide explains the investment case, the ways to access it and the risks.

1. Why Power Has Become the Bottleneck

• AI workloads are energy-intensive. Training and running large models requires dense clusters of powerful chips that draw far more electricity than traditional computing.
• Grid connections take years. In many regions, new data centres face long waits to connect to the grid, while transmission upgrades lag behind demand.
• Cooling needs are rising. Denser chips generate more heat, pushing the industry towards liquid cooling and new designs.
• Clean energy commitments. Many large technology companies have climate targets, increasing demand for low-carbon power.

2. Signals From the Largest Buyers

• Nuclear agreements. In 2024, Microsoft agreed to buy power from a restarted unit at the Three Mile Island plant in Pennsylvania, and Google and Amazon announced agreements tied to small modular reactor developers.
• Massive capital spending by cloud providers on data centres, chips and networking.
• Private capital deployment. Infrastructure investors have made large bets on data centre platforms, such as Blackstone's 2021 acquisition of QTS.

3. Ways to Invest

Data centre real estate and operators

Publicly listed data centre REITs and private developers that build and lease facilities, often under long-term contracts with large technology tenants. See real estate syndications vs. REITs.

Infrastructure and private credit funds

Funds that finance construction and operation of data centres, backed by contracted revenue. See private credit.

Power generation and grid

Utilities, independent power producers, nuclear developers, renewable projects with storage, and grid equipment suppliers such as transformers and transmission infrastructure.

Venture opportunities

• Cooling technology, including liquid and immersion cooling.
• Power management and efficiency software that reduces energy use per computation.
• Energy-efficient chips and networking. See semiconductor startups.
• Grid and energy software that speeds interconnection and balances demand.
• New power sources, including advanced nuclear and fusion. See fusion energy startups.

4. Risks to Consider

• Overbuilding. If AI demand grows more slowly than expected, or models become far more efficient, capacity could outpace demand.
• Tenant concentration. Many facilities rely on a small number of very large customers.
• Technology change. Rapid shifts in chips and cooling could make some facilities less competitive.
• Power and permitting. Projects can be delayed or cancelled by grid constraints, permitting and local opposition to energy and water use.
• Interest rates, which affect financing costs and valuations for capital-intensive infrastructure.

5. What to Diligence

1. Secured power capacity and grid connection timelines.
2. Tenant contracts, their length and the credit quality of customers.
3. Construction costs and the ability to deliver on schedule.
4. Cooling design and readiness for high-density AI workloads.
5. Energy sourcing, including exposure to carbon and water constraints.
6. Exit options, from long-term holding to sale to infrastructure investors.

Frequently Asked Questions

Is data centre investing only for large institutions?

No. Public REITs, listed utilities and infrastructure funds offer access for many investors, while direct development is mainly for institutions.

Why are technology companies buying nuclear power?

Nuclear provides large amounts of steady, low-carbon electricity, which suits data centres that run around the clock and companies with climate commitments.

Could AI efficiency gains reduce data centre demand?

Efficiency gains could slow demand growth, though cheaper computation can also increase overall usage. This uncertainty is a key risk to weigh.

Where do startups fit?

In technologies that make data centres cheaper, cooler, more efficient or faster to power, often selling to operators and cloud providers.

The Bottom Line

AI's growth depends on physical infrastructure, and power has become one of its tightest constraints. That creates investment opportunities across real estate, energy, grid equipment and enabling technology. Investors should weigh the strong demand signals against overbuilding, concentration and power risks, and diligence secured power above all.

Global Capital Network connects investors with infrastructure, energy and AI opportunities through our events and investor network. Get in touch to learn more.

This article is general information, not investment advice.