The biggest constraint on AI infrastructure expansion isn't computing power or capital,it's electricity and the time required to connect data centers to the power grid. That's the stark warning from Greg Abel, who recently took over as CEO of Berkshire Hathaway, one of the largest energy providers in the United States.

What's Limiting AI Data Center Growth Right Now?

In a September 2 interview on CNBC, Abel laid out a perspective that challenges much of the conventional wisdom about AI infrastructure bottlenecks. While industry observers have focused on chip shortages, capital availability, and cooling systems, Abel pointed to something more fundamental: the physical power grid itself.

"I've sort of always had a strong view that energy would be the constraint. We can produce the energy. It's how long it would take to get the sites prepared and be in a position where they could serve the data centers. And I continue to see that as a big constraint," Abel stated.

Greg Abel, CEO of Berkshire Hathaway

This observation carries significant weight because Berkshire Hathaway operates Berkshire Energy, which serves millions of customers across multiple states. Abel reported that data centers already account for roughly 8% of Berkshire Energy's load in Iowa, with additional demand expected to grow. The company is actively working to serve major hyperscalers, including Alphabet (Google), which reported second-quarter 2026 capital expenditures of $44.92 billion and has guided toward $175 billion to $185 billion in total 2026 capex to build out AI infrastructure.

Why Is Grid Infrastructure the Real Chokepoint?

Independent research supports Abel's assessment. According to the Grid Strategies 2025 load growth report, data centers are now the largest driver of U.S. electricity demand growth. However, the infrastructure to support this growth is lagging significantly behind demand. The real bottlenecks aren't technical in nature; they're logistical and regulatory.

• Site Readiness: Preparing physical locations for data center construction requires environmental assessments, land acquisition, and infrastructure development that can take months or years.
• Permitting Processes: Local and state regulatory approval for new power infrastructure involves multiple agencies and can introduce substantial delays to project timelines.
• Interconnection Timelines: Connecting new data centers to the existing power grid requires upgrades to transmission lines, substations, and distribution networks, which are complex engineering projects with long lead times.

These three factors create a compounding delay that no amount of capital or computing power can overcome. A hyperscaler might have the money to build a data center and the chips to fill it, but if the power grid can't deliver electricity to the site, the project sits idle.

How Are Energy Companies Approaching Hyperscaler Partnerships?

Abel was explicit about Berkshire Energy's strategy for underwriting hyperscaler load. The company is interested in serving these massive computing operations, but only under specific conditions. "We are interested in serving these hyperscalers, if there was no impact to the rates of our other customers. And in fact, we've pretty much taken the approach that there has to be a net benefit to our customers," Abel explained.

This public commitment reflects growing political friction around data center-driven utility bills. Communities across the United States have begun pushing back against rapid data center expansion, concerned that the massive power demands will drive up electricity costs for residential and small business customers. Berkshire's approach suggests that utilities will increasingly demand that hyperscalers contribute to grid infrastructure improvements or pay premium rates that offset costs to existing customers.

Abel's rationale for Berkshire's recent $16.5 billion investment in Alphabet (a $6.5 billion equity offering at a 6.5% discount plus an additional $10 billion in open market purchases) underscores this strategic thinking. He noted that Berkshire recognized Google as a significant player in AI and understood the technology's transformative impact on American business. However, the investment also positions Berkshire to benefit from the infrastructure spending that will necessarily follow AI's growth.

What Does This Mean for AI Infrastructure Investment?

Abel's message reframes the investment opportunity in AI infrastructure. Rather than focusing primarily on chipmakers or cloud platform companies, investors and industry observers should pay closer attention to the utilities, grid equipment manufacturers, and power infrastructure companies that will enable data center deployment.

The supply chain for AI infrastructure extends far beyond semiconductors. Companies that manufacture power supplies, cooling systems, and grid equipment are experiencing accelerating demand. For example, Delta Electronics, a Taiwan-based power and thermal management company, supplies the power delivery and liquid cooling systems that keep hyperscale data centers running reliably. The company is already seeing record quarterly sales driven by data center demand.

Similarly, equipment manufacturers like ASML Holding, which produces the lithography machines required to manufacture advanced chips, face orders closely tied to capital expenditure plans at chip manufacturers like Taiwan Semiconductor Manufacturing (TSMC). As Microsoft prepares to reveal Azure's quarterly revenue in absolute dollar terms rather than just growth rates, the visibility into cloud infrastructure spending will likely increase pressure on these supply chain partners to expand capacity.

Abel's warning suggests that the next phase of AI infrastructure growth will be determined not by innovation or capital availability, but by the unglamorous work of upgrading electrical grids, securing interconnection rights, and managing the political and regulatory challenges of rapid energy demand growth. For investors and policymakers, that shift has profound implications for where resources should be directed and which companies will emerge as winners in the AI era.