Contact Us
Contact Us
Featured Image

Gridlock: Lack of Power Could Slow AI Growth in US

The meteoric rise in valuations of artificial intelligence (AI)-related companies over the past two and a half years has been nothing short of historic. The AI ecosystem of high-performance processors, software to manage training and inference, massive data centers to house the processors and run the software and the infrastructure to power this capability have all benefitted from this virtuous cycle and massive growth and adoption of AI technology.

We may, however, be approaching an inflection point to this growth, as grid managers grapple with rapid demand growth while managing a grid infrastructure that is approaching capacity.  For example, PJM Interconnection, the largest US power grid serving 65 million people across Washington DC and 13 midwestern and eastern mid-Atlantic states, issued four "Maximum Generation Alerts" due to high electricity demand during June and July.

According to Anthropic’s recent policy paper, the US needs at least 50 gigawatts (GW) of additional electric capacity by 2028 to maintain global AI leadership.1 Adding this much capacity would equal 50 1GW nuclear power generation plants with the electrical power capacity to support around 35 million average US homes.

With electricity already the largest operating cost for data centers, accounting for 46% to 60% of total operating expenses,2 headwinds are emerging that could slow this growth.

Ohio – “Ground Zero” for Grid Access Regulatory Changes

Ohio, with its central location, low natural disaster risk, and inexpensive power and land costs, has experienced a surge of interest over the past several years from large tech companies looking to establish or expand hyperscale data centers in the state. 

After receiving inquiries and preliminary requests totaling more than 30GW of additional load for data centers from 50+ customers, AEP Ohio instituted a data center construction moratorium in March 2023.3 The sum of these requests are approximately one-half of the estimated 60GW global data center load inventory.4 

After a lengthy study, Ohio regulators adopted landmark regulations for large, new data centers in early July and ordered AEP Ohio to lift the moratorium. Data centers, which disproportionately benefitted from existing grid infrastructure, will now more fully bear the costs to build and maintain grid capacity to meet the growth in their power demands. These new rules have created a potential tipping point with both positive and negative implications across the data center ecosystem. 

Key features of these new regulations include:

  • Data Centers to Pay for 85% of Subscribed Capacity: New data centers must pay for at least 85% of their subscribed energy capacity, even if their actual usage is lower. This applies to facilities with loads above 25 megawatts (MW), ensuring that grid expansions aren’t left stranded or paid for by others.

  • Long Term Contracts: Requirements are set for up to 12 years, with a 4-year ramp-up and an exit fee if operators withdraw early or fail to meet obligations.

  • Financial Viability: Data center operators are required to demonstrate financial stability and offer collateral, making speculative projects less likely.

  • Sliding Scale: Smaller and mid-sized data centers will receive more flexible consideration and terms.


Conclusion

The Ohio regulatory ruling indicates that the US is still in the early stages of addressing supply and demand issues around AI data centers, available electric power and how this expansion will be financed. 

The combined impact of this decision, broader demand growth and a shifting national regulatory environment could alter the economics for participants in the AI/data center/power supply ecosystem.

Although this new regulatory framework model may create more equitable cost sharing for capacity build-out, it also introduces a new set of challenges and higher operating costs that could impede AI adoption, growth and, if more broadly adopted, US AI leadership. 

Important Disclosures & Definitions

1 Anthropic. (July 21, 2025). Build AI in America. Anthropic Policy Report.

2 IDC. (September 24, 2024). IDC Report Reveals AI-Driven Growth in Datacenter Energy Consumption, Predicts Surge in Datacenter Facility Spending Amid Rising Electricity Costs. IDC Press Release. 

3 Patel, S. (July 10, 2025). Regulator Approves AEP Ohio’s Landmark Data Center Tariff. Power Magazine, Topic: Data Centers.

4 Green Street. (July 11, 2025). Data Center: Shrinking the Queue. Green Street, Research Library.

AAI000971 07/29/2026

Recent Two Minute Tuesdays