San Francisco — The tech industry’s rush to integrate large language models into everything from search engines to cloud services is threatening to derail the sector’s own climate commitments, according to utility planners, energy researchers and company documents reviewed by SourceRated.
Microsoft, Google and Amazon Web Services have all promised to run entirely on carbon-free energy later this decade. Yet the sudden popularity of generative artificial intelligence — which requires multiple times more computing power than conventional search or storage — has added an unexpected jolt of demand. Grid operators in the Pacific Northwest, Texas and Virginia told SourceRated they have received data-center interconnection requests totaling more than 15 gigawatts since mid-2023, roughly the output of 15 large nuclear plants.
“Nobody saw AI workloads scaling this quickly,” said a senior engineer at a Western U.S. utility who was not authorized to speak publicly. “The clean-energy projects in the pipeline aren’t enough to cover it, so we’re modeling additional natural-gas turbines to keep reliability up.”
Corporate filings show Microsoft’s global power consumption jumped 21 percent last fiscal year, while Google’s rose 17 percent, reversing a decade-long trend of efficiency gains. Both companies declined to provide AI-specific figures but acknowledged the technology is a major driver.
Environmental groups worry the pivot could lock in fossil-fuel infrastructure for decades. “A new gas plant built today will still be running in 2045,” said Kelly Ramirez, an analyst at the non-profit Carbon Plan. “That is incompatible with science-based climate pathways.”
Tech firms counter that AI will eventually optimize grids and accelerate the rollout of renewables. In a statement, Amazon Web Services said its investment in advanced batteries and offshore wind “will more than offset any short-term rise in thermal generation.” Google pointed to its Moonshot wind-prediction software, which it says increases the value of clean power to the grid by 20 percent.
Policy makers are watching closely. The U.S. Energy Information Administration last week raised its 2026 electricity-demand forecast by 4 percent, citing data centers as the fastest-growing load. Several state regulators, including those in Georgia and Ohio, are weighing whether to fast-track gas permits or require tougher efficiency standards for new server farms.
Analysts warn that how Big Tech balances AI growth with decarbonization will influence national emissions trajectories. “The sector could become a clean-energy catalyst or a carbon wildcard,” said Rebecca Howard of consultancy Ember. “The next 24 months will set the tone.”