Congressman Suhas Subramanyam calls for AI data center construction to be spread across the US

In the United States, a dispute is heating up over how to build infrastructure for AI. While Bernie Sanders and Alexandria Ocasio-Cortez are preparing a bill on a moratorium on new AI data centers, congressman from Virginia Suhas Subramanyam proposes a different approach: not to slow development, but to distribute new capacity across the country.

Why the dispute intensified

The trigger for the new discussion was growing concerns about the impact of data centers on the power grid and electricity bills. The more actively companies expand computing capacity for training and deploying AI models, the more noticeable the physical side of this boom becomes: server clusters require enormous amounts of energy, grid connections, and new facilities. When such objects concentrate in a few regions, local residents, utility companies, and politicians start asking more insistently who exactly will pay for this expansion.

Against this backdrop, Bernie Sanders and Alexandria Ocasio-Cortez are preparing to present an initiative that introduces a moratorium on the construction of new AI data centers. Such a step looks like an attempt to sharply slow the pace to assess the consequences for the energy market. At the center of the dispute now are not abstract discussions about the future of artificial intelligence, but quite tangible questions: will there be enough capacity, how quickly can networks be modernized, and won't costs be passed on to ordinary consumers through higher tariffs?

How Virginia responds

Subas Subramanyam's position is important because he represents Virginia — a state that is home to one of the largest concentrations of data centers in the United States. In other words, he's speaking not from theory, but from a region that already lives alongside this infrastructure and sees its pluses and costs in real time. That's why his response to the moratorium idea sounds like an attempt to find a politically and economically more flexible option than an outright ban on new projects.

Based on the discussion on Bloomberg Tech, Subramanyam doesn't dispute that the strain on the power grid has become a real problem. But instead of stopping construction, he proposes spreading the development of AI infrastructure more widely across the country, rather than concentrating it all in the same nodes. The logic here is simple: if new projects continue to flow into a few already overloaded locations, resistance from local communities and politicians will only grow.

More even distribution could reduce pressure on individual networks and reduce the risk of sharp public backlash against the entire industry.

What measures are being discussed

Essentially, Washington is now facing two models of response to the AI boom. The first is to hit pause and temporarily stop new construction while authorities understand how they affect the energy market. The second is to not abandon expansion, but to change the map of where facilities are located and more strictly tie new projects to the capabilities of regional networks. From this discussion, several practical directions are already visible, and it's around this choice that political conflict is now being built.

• Temporary pause on new AI data centers pending assessment of impact on energy tariffs
• Relocation of some future projects from overloaded hubs to other states
• Stricter alignment of construction with available capacity of local power grids
• Political balance between the AI race and consumer dissatisfaction over bills

This fork is important also because the issue is quickly extending beyond Virginia. If demand for computing continues to grow at current rates, similar conflicts will emerge in other states that want to build large data centers. Then the dispute over infrastructure placement will cease to be a regional issue and become part of national industrial policy: where acceleration is possible, who gets investments, and what conditions the state is willing to impose on technology companies.

What this means

The dispute over AI in the United States is increasingly shifting from models and chatbots to electricity, land, and networks. For the market, this is a signal that the next phase of competition will depend not only on the quality of algorithms, but on where companies can quickly and politically safely launch new capacity. For business and investors, this means a simple thing: the strategy for AI growth now increasingly comes down to infrastructure geography, not just software.