Data center construction across the Americas is outpacing power infrastructure development at a pace Wärtsilä calls "spe
A new analysis from Wärtsilä, the energy technology company, warns that the race to build data centers across the Americas has hit a critical infrastructure bottleneck. While demand for AI computing capacity is driving unprecedented construction, the power systems needed to run these facilities cannot keep pace with building schedules. This mismatch, which Wärtsilä terms "speed to powerless," represents one of the biggest hidden risks to AI infrastructure growth across the region.
The challenge spans all major markets examined. In the United States, approximately 2,600 gigawatts of generation and storage capacity were seeking grid interconnection as of the end of 2025, creating serious timing uncertainty for developers relying on traditional grid connection. In Brazil, projected data center load on the transmission grid in 2030 was revised upward by 60 percent—from 18.9 terawatt-hours to 30.3 terawatt-hours per year—just three months after the country's Annual Energy Operation Planning for 2026 to 2030 was published. In Mexico and parts of the United States, water scarcity is further narrowing viable power options, as cooling requirements for certain turbine technologies create significant water demand in already stressed regions. Chile faces corridor-level transmission saturation following lessons from its February 2025 national blackout, while Argentina confronts grid underinvestment and policy uncertainty.
Wärtsilä's research, titled Beyond the Grid: Building the Power System for AI in the Americas, draws on data from Lawrence Berkeley National Laboratory, the International Energy Agency, and national energy planning bodies across the region. Risto Paldanius, VP Americas at Wärtsilä, states: "The race to build data centers across the Americas is extraordinary in its pace and scale. But power is not keeping up. Interconnection queues stretch years, transmission capacity is saturating in the corridors where data centers are clustering, and equipment lead times are pushing delivery towards 2030 and beyond. The risk is not that the market slows down – it is that it builds faster than it can reliably power."
The company proposes a fundamental shift in how data center power must be planned. Rather than relying solely on traditional grid infrastructure, developers should consider macro-grids—fully or partially isolated systems requiring more than 100 megawatts of on-site peak load—as a logical step beyond microgrids. These retain the potential for grid interconnection over time, functioning as a bridge to future integration rather than permanent substitutes.
Wärtsilä's analysis compares reciprocating internal combustion engine systems (RICE) against aeroderivative gas turbines over a 20-year levelized cost horizon. RICE technology shows a 25 percent cost advantage, at approximately 86 dollars per megawatt-hour versus 111 dollars per megawatt-hour—an annual saving of approximately 178 million dollars for a 1-gigawatt data center. Beyond economics, RICE maintains full rated output across temperatures from minus 45 degrees Celsius to 45 degrees Celsius, with negligible process-water consumption for power generation. Aeroderivative turbines, by contrast, can derate output by up to 27 percent in high ambient temperatures.
Wärtsilä recommends that the most resilient data centers combine modular on-site RICE generation with flexibility, including grid supply where available, clean power contracts, and long-term planning for eventual interconnection. The company argues this represents a shift in mindset from securing power fast to securing power that lasts—built to remain efficient, resilient, and commercially viable for decades, not merely for launch day. Those planning for this hybrid approach now will remain efficiently operational in 2035, while those that do not risk finding themselves powerless at the moment the market expects them to deliver.