The U.S. Federal Energy Regulatory Commission (FERC) may be approaching one of the most delicate balancing acts of the AI era: reconciling two objectives that increasingly appear to be at odds—shielding utility customers from soaring electricity costs while dramatically accelerating “speed-to-power” for the data centers that will underpin America’s AI ambitions. Recent actions by the regulator suggest that a new framework is beginning to emerge, one that could simultaneously protect ratepayers, shorten interconnection timelines, and provide a clearer pathway for AI factories, semiconductor manufacturing facilities, and other energy-intensive industrial projects to connect to the grid.
The stakes could hardly be higher. Data centers have become one of the largest sources of incremental electricity demand in decades, exposing the limitations of a fragmented energy system that was never designed to accommodate such a rapid increase in consumption.
Recognizing the urgency of the challenge, FERC unanimously voted to accelerate policies aimed at bringing AI data centers onto the grid more quickly while simultaneously tightening oversight over who bears the cost of the necessary infrastructure investments. The commission has directed six regional transmission operators, collectively serving nearly two-thirds of the U.S. population, to demonstrate that they have credible plans to expedite data center interconnections without passing the associated costs on to households through higher electricity bills.
At the same time, FERC’s latest actions represent a significant milestone for large-load interconnections and could establish a new national framework for connecting AI factories, semiconductor fabrication support facilities, and advanced manufacturing projects to the grid. The objective is not merely to modernize the interconnection queue—the often lengthy approval process required for new projects—but to fundamentally rethink how large electricity consumers participate in the system.
Under this emerging framework, large industrial customers would no longer be passive entrants waiting for access to an already congested network. Instead, they would become active participants in building the infrastructure they require. Companies seeking access to substantial amounts of power would be expected to finance their own network upgrades, thereby reducing the burden on existing ratepayers. They would also be encouraged to bring additional generation capacity online, increasing supply alongside demand, while offering flexible load capabilities that would allow grid operators to better manage periods of peak electricity consumption.
Perhaps the most important innovation is the introduction of incentives for flexibility. Customers capable of shifting or curtailing electricity demand in response to grid conditions could qualify for accelerated approval timelines, with study periods potentially reduced to as little as 60 days. This would represent a dramatic departure from the multi-year interconnection processes that have historically delayed projects and emerged as one of the primary bottlenecks to AI infrastructure deployment.
The shift should therefore be viewed not simply as faster interconnection, but as smarter interconnection. FERC is attempting to create a system in which the largest power users bear a greater share of the associated infrastructure costs. In doing so, the regulator hopes to avoid a politically sensitive outcome in which households effectively subsidize the AI ambitions of some of the world’s largest technology companies.
From an investment perspective, these reforms could prove highly consequential for our Powering AI strategy. Direct funding of generation and transmission assets by hyperscalers could materially accelerate electric infrastructure capex, creating a favorable backdrop for power generation, transmission, and grid equipment suppliers. Several recent developments suggest this process is already underway. Anthropic has announced that it will finance 100% of the grid upgrades required to connect certain data center projects, while Google acquired power developer Intersect Power earlier this year in a transaction valued at approximately $4.75 billion. More broadly, the massive capital raisings announced by major AI players—exceeding $80 billion at both Google and SpaceX, which owns xAI—are likely to translate into a sharp increase in power infrastructure orders over the coming quarters.
One of the more interesting beneficiaries could be providers of on-site power solutions, particularly fuel cell companies such as Bloom Energy and the Ceres Power / Delta Electronics partnership. As data center developers are increasingly required to shoulder the cost of expensive grid upgrades, self-generation may become an economically attractive alternative. Deploying on-site power systems could allow operators to bypass lengthy interconnection queues, reduce dependence on constrained transmission networks, and accelerate deployment timelines. In other words, as the cost of connecting to the grid rises, the economics of generating power directly at the data center may become increasingly compelling.
