The race to build AI data centers has hit an unexpected wall, and it has nothing to do with chip shortages or funding. According to executives at Brookfield Asset Management, the real bottleneck is electricity. The U.S. power grid simply cannot deliver enough power to meet AI infrastructure demands, even as companies race to build massive computing facilities.

Why Is Power the Limiting Factor for AI Data Centers?

During Brookfield's second quarter 2026 earnings call, Sikander Rashid, the company's Global Head of AI Infrastructure, laid out the math starkly: the United States will need 100 gigawatts of power for artificial intelligence over the next decade, but the existing grid can only make 30 gigawatts of that available. That gap represents a fundamental constraint on how fast the AI infrastructure buildout can actually proceed, regardless of how much capital investors are willing to deploy.

This power shortage is reshaping how companies approach data center development. Rather than relying on the public utility grid, major infrastructure investors are now pursuing "behind-the-meter" power solutions, where electricity is generated and consumed on-site without relying on grid transmission. Brookfield has expanded its partnership with Bloom Energy, a company that manufactures fuel cells and power generation systems, from a $5 billion commitment to $25 billion to finance quick-to-deploy power solutions specifically designed for AI infrastructure.

The challenge extends beyond just generating more electricity. According to reporting on the power sector's constraints, the binding limitation is no longer permits or capital; it is workforce capacity. The utility industry faces a severe labor shortage, with 2.4 utility workers nearing retirement for every worker under 25 across advanced economies. In the United States specifically, 56% of the utility workforce has fewer than 10 years of service, meaning experienced technicians are aging out faster than replacements are being trained. For companies waiting on grid interconnection approvals to connect their data centers, this workforce shortage means the queue problem will worsen before it improves, even with signed power purchase agreements in place.

How Are Major Investors Responding to the Power Constraint?

• Nuclear Power Partnerships: Brookfield announced a $17.5 billion commitment from the Department of Energy to support the development of up to 10 nuclear reactors, recognizing that large-scale, reliable power generation is essential for AI infrastructure.
• Repurposing Federal Sites: The company is developing the Paducah American Energy Hub in Kentucky, which will repurpose federally owned industrial land to support 2 gigawatts of compute capacity while bringing new power generation directly to the site, bypassing grid constraints.
• On-Site Generation: The expansion of behind-the-meter power solutions through partnerships with companies like Bloom Energy allows data centers to generate their own electricity, reducing dependence on the public grid.
• International Expansion: Brookfield increased its France Development Framework from 20 billion euros to 30 billion euros to support sovereign AI infrastructure initiatives, diversifying power access across geographies.

Brookfield's strategy reflects a broader recognition that AI infrastructure is becoming as much an energy business as a technology business. The company raised a record $77 billion in the second quarter of 2026, with significant portions allocated to AI infrastructure and energy partnerships. The company also launched a dedicated $10 billion AI Infrastructure Fund intended to anchor a broader investment program of approximately $100 billion in AI-related opportunities, with a specific focus on large-scale AI factories and behind-the-meter power solutions where competition is thinner and returns are more attractive.

"Energy is the primary bottleneck for AI," stated Sikander Rashid, Global Head of AI Infrastructure at Brookfield Asset Management.

Sikander Rashid, Global Head of AI Infrastructure and Head of Europe, Brookfield Asset Management

What Does This Mean for the AI Industry's Future?

The power constraint is reshaping investment priorities across the entire AI infrastructure ecosystem. Rather than competing primarily on computing hardware or software efficiency, companies are now competing for access to reliable, large-scale power generation. This shift has profound implications for where AI data centers will be built, how they will be financed, and which regions will emerge as AI computing hubs.

The Paducah project exemplifies this new paradigm. By bringing power generation directly to the site, the facility can support 2 gigawatts of compute capacity without waiting for grid upgrades or interconnection approvals. Brookfield expects the project to attract up to $100 billion of private investment, signaling that investors see power-constrained sites as strategic assets.

However, the workforce shortage in the utility sector suggests that even aggressive investment in new power generation will face execution challenges. Building nuclear reactors, upgrading transmission lines, and training new utility workers all require time and skilled labor. The gap between AI infrastructure demand and power supply availability will likely persist for several years, creating a structural advantage for companies that can secure long-term power contracts or develop on-site generation capabilities.

For hyperscalers like Google, Microsoft, and Amazon, which are already committing hundreds of billions of dollars to AI infrastructure expansion, the power constraint means that capital spending alone will not solve the problem. Strategic partnerships with energy companies, government support for nuclear development, and investment in alternative power generation technologies are becoming as critical as securing the latest GPU chips.