Governments straining to keep pace with surging electricity demand are increasingly curtailing data center construction—but restricting these facilities risks handing an edge to competitors in the AI race. That is the warning from Serhiy Tokarev, co-founder and general partner of Roosh Investment Group, who argues that expanding power supply and grid capacity, rather than blocking new facilities, is the better path.

A growing wave of restrictions

In Ireland, data centers consume nearly a quarter of the country's electricity, according to the Central Statistics Office. New grid connections around Dublin were effectively frozen for three years, and under revised rules from the Commission for Regulation of Utilities, new projects must provide generation or storage capacity onsite or nearby. Amsterdam previously imposed a moratorium on new facilities, while Frankfurt regulates their location through zoning.

In July 2026, New York State introduced a moratorium of up to one year on certain environmental permits for new large data centers. State officials said the pause would allow them to assess the facilities' impact on the power grid and the environment. As of May 2026, data center projects representing nearly 12 GW of potential demand were in the state's grid interconnection queue—the lineup of projects awaiting approval to connect to the grid.

Tokarev has been investing in technology companies for more than five years. His portfolio includes AI infrastructure, and energy is now a key focus. That experience, he says, has shown how access to energy infrastructure influences investment decisions: reliable power is increasingly becoming a decisive factor in a company's ability to scale and attract capital.

As businesses rely more on AI, control over its models and infrastructure matters more. A foreign provider's decision to change terms, raise prices, or restrict access could affect entire industries. Domestic computing capacity, in this view, gives countries more control over the infrastructure their businesses depend on.

Control over AI is becoming a source of economic power

This helps explain the intensity of U.S.–China competition in AI. American companies hold leading positions across the AI stack, from advanced models and chips to major cloud platforms, and Washington views the power supply for data center infrastructure as a national security issue.

Beijing is pursuing a state-led strategy. Its "Eastern Data, Western Computing" program directs the development of computing capacity toward western China, where energy and land resources are more readily available. The country is also expanding power generation at scale: according to the International Energy Agency, China added nearly 500 GW of renewable capacity in 2025—more than 60% of the global increase.

Europe, meanwhile, risks falling further behind. Brussels plans up to seven AI gigafactories—large-scale computing hubs intended to train AI models—and at least a tripling of EU data center capacity within five to seven years. Yet some countries are slowing construction, while European businesses still rely primarily on American cloud providers. Restrictions, Tokarev argues, do not remove demand for AI; they risk pushing investment abroad and reinforcing that dependence.

The bottleneck is local

Data centers' hunger for power should be understood in the context of wider electricity demand. According to the International Energy Agency, data centers accounted for only around 70 TWh of the 800 TWh increase in global electricity consumption in 2025—less than a tenth of the total growth. Industry, air conditioning, heat pumps, and electric vehicles contributed far more. The challenge is rapidly growing demand concentrated in a handful of locations—and power must reach new computing capacity in time.

What is holding up connections?

Grid hardware is one obstacle. Transformers—the equipment that steps voltage up and down so electricity can move efficiently from power plants to end users—are in especially short supply: in the U.S., lead times for large transformers now exceed two years, compared with roughly 50 weeks before the pandemic, according to a federal report on transformer shortages. Morgan Stanley estimates that U.S. data centers could face a power shortfall of around 49 GW by 2028. Microsoft CEO Satya Nadella has acknowledged that power has become the company's binding constraint—some of its GPUs are sitting in inventory with nowhere to plug them in.

Dedicated gas-fired plants offer no quick fix. GE Vernova, Siemens Energy, and Mitsubishi Power produce around two-thirds of the world's gas turbines, and their factories are booked years ahead. GE Vernova reported 116 GW in gas power equipment orders and slot reservations, while targeting annual turbine output of 20 GW in the third quarter of 2026.

The hardware itself adds another challenge. A server rack once required 7–12 kW. Systems built around NVIDIA's GB200 consume 120–132 kW per rack—more than a tenfold increase, and beyond what conventional air cooling can handle—so new facilities must be designed for liquid cooling from the outset.

Tech giants are securing power themselves. Microsoft agreed to buy electricity from the 835 MW Three Mile Island reactor, which is being restarted to serve its needs. Google's agreement with Kairos Power calls for small modular reactors—a new generation of compact, factory-assembled nuclear units—with the first expected online by 2030.

Governments cannot rely on corporate initiative alone, Tokarev contends. They must plan energy and digital infrastructure together: expand grids, accelerate transformer production, streamline connections, and encourage long-term power purchase agreements, generation near data centers, and storage.

How war accelerated distributed generation in Ukraine

Ukraine's experience shows how a crisis can accelerate the deployment of local generation. War and repeated attacks forced the country to rapidly deploy local power sources where the centralized grid could no longer guarantee a reliable supply.

Companies in Ukraine have long had to manage energy risks, maintaining reliable electricity supplies, backup power, and plans for potential outages. That resilience also matters to investors: a company's ability to operate during power outages affects how they assess its growth potential.

Ukraine already has around 1.8 GW of distributed gas-fired generation, with more than 500 MW under construction, and is streamlining grid connection and permitting procedures. The country's experience cannot be directly applied to large AI data centers. But it illustrates how making energy resilience a priority can accelerate investment in additional generation and changes to connection procedures, even under extreme conditions.

How to power the data center build-out

Powering new data centers will require coordinated action by governments, grid operators, and businesses. Governments can plan sites for new facilities. Grid operators can reserve capacity for them in advance. Businesses can invest in their own power plants, storage, and cooling, and reduce electricity use during peak hours in exchange for faster grid connections. A combination of grid supply, gas-fired plants, nuclear power, and renewables can help meet demand where no single source is sufficient. Several milestones already on the calendar will serve as checkpoints: New York's moratorium runs for up to one year, the EU has set a five-to-seven-year horizon for tripling data center capacity, GE Vernova targets 20 GW of annual turbine output in the third quarter of 2026, and Google's first small modular reactor is expected online by 2030.

The ability to build and power data centers quickly is likely to become an increasingly important factor in investment decisions. An advantage in the AI race will depend not only on access to models and chips, but also on a country's ability to expand computing capacity, generation, and grids together. Restricting data center development may temporarily ease grid pressure, but it can also deepen dependence on foreign infrastructure and technology.

The opinions expressed in Fortune commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune. This article was originally published on Fortune.com.