The AI Race for Power and Land

The race to build the physical backbone of artificial intelligence is spilling out of server rooms and into politics, land markets and power grids.

Across the United States, proposed AI data centers are becoming a campaign issue, with candidates and strategists in multiple states treating them not simply as symbols of technological progress but as flash points over electricity, water, tax subsidies and local control. At the same time, in Northern Europe, a very different response is taking shape: a scramble to attract the facilities, with Nvidia reportedly helping connect customers seeking computing capacity with data-center operators in the Nordic region.

Together, the developments underscore a shift in the AI boom. What was once framed mainly as a contest over chips and software models is increasingly becoming a fight over where massive computing campuses can be built, how they will be powered and who bears the costs.

From zoning fights to campaign trail

Until recently, resistance to data centers often played out in county board meetings and local land-use disputes. Residents complained about noise, transmission lines, diesel backup generators, water consumption and tax incentives granted to facilities that, after construction, often create relatively few permanent jobs.

Now those arguments are migrating into mainstream electoral politics.

With the 2026 midterm season approaching, opposition to AI data centers has begun appearing in political advertising and campaign messaging in a number of states, turning a once-technical infrastructure issue into a broader voter concern. The pushback has drawn support from both parties, reflecting how the issue cuts across ideological lines: conservatives who object to subsidies and top-down development have found common cause with progressives and local activists worried about environmental strain and pressure on public utilities.

That shift matters because large data-center projects depend on lengthy permitting, grid upgrades and, often, some degree of political goodwill. If candidates begin seeing opposition to these projects as a winning message, developers may face a tougher path even in places that had previously welcomed them as engines of growth.

The concern is not abstract. AI facilities require vast amounts of electricity to train and run advanced models, and operators often seek access to the cheapest, most reliable power they can find. In many communities, that has raised fears that households or other businesses will ultimately shoulder higher utility costs or that local grids will be stretched to accommodate a handful of energy-hungry campuses.

The Nordic appeal

As resistance grows in parts of the United States, the Nordic countries are emerging as one of the industry’s most attractive alternatives.

According to CNBC, Nvidia has been informally introducing GPU customers to data-center operators in the region, where abundant renewable energy, cooler temperatures and relatively available land have made large-scale expansion more feasible. The matchmaking role, while informal, points to how deeply the company is moving into the broader AI infrastructure ecosystem. Nvidia is no longer merely supplying chips; it is increasingly positioned at the center of decisions about where computing capacity gets built and how quickly it can come online.

The Nordics have long held appeal for data-center developers because cold climates reduce cooling costs, and electricity systems in countries like Norway, Sweden and Finland offer access to significant renewable generation. For AI operators, those advantages have become more valuable as power demand rises and environmental scrutiny intensifies.

The region’s momentum has been building. atNorth, a Nordic data-center company, announced a new mega-site in Norway in June, and commercial real estate advisers including Savills and JLL have recently identified the Nordics as one of Europe’s clearest growth zones for AI-oriented data centers. The pitch is straightforward: more room to scale, cleaner power and fewer immediate siting conflicts than in more crowded markets.

Even there, however, limits remain. Transmission capacity, interconnection queues and broader power-system constraints can still slow projects. The Nordics may be attractive, but they are not infinite.

A new geography of AI

The diverging trajectories in the United States and Northern Europe reflect a broader reordering of AI’s geography.

For years, the technology industry’s center of gravity was defined largely by talent clusters, venture capital and semiconductor supply chains. But as generative AI drives a surge in demand for computing, the map is being redrawn by more old-fashioned considerations: industrial land, permitting speed, access to substations and the politics of energy.

That has elevated a new set of players. Utilities, transmission planners, state regulators, local boards and data-center landlords now have a greater say in how fast the AI economy can expand. Companies that can coordinate among them — whether developers, cloud providers or chip makers like Nvidia — stand to gain influence.

It also raises a harder question for policymakers: who should benefit from the build-out, and who should pay for it?

Communities weighing data-center proposals often hear promises of construction jobs, tax revenue and economic prestige. But critics argue that the public is too often asked to absorb the downsides — heavier demands on water and power systems, expensive grid upgrades and years of local disruption — for facilities that may employ only a modest permanent workforce once completed.

That tension helps explain why the backlash has broadened beyond environmental groups or nearby residents. In some places, it has become shorthand for a larger unease with a development model in which high-value digital infrastructure consumes large physical resources while offering uneven local returns.

What comes next

Whether the American backlash results in lasting policy changes remains unclear. It could produce moratoriums, stricter permitting rules, limits on tax incentives or new requirements that developers pay more directly for upgrades to water and electric systems. But it is also possible that the issue peaks during campaign season and recedes once elections pass.

In the meantime, demand for AI capacity is not slowing. That reality is likely to push companies toward regions that can offer power, land and political support with fewer delays.

The result may be a more fragmented global AI build-out: contested and politicized in some U.S. states, accelerated in parts of Europe and increasingly shaped by infrastructure brokers capable of aligning chips, capital and geography.

The next phase of the AI race, in other words, may depend less on who designs the smartest model than on who can secure the megawatts.