The surge in artificial intelligence is fundamentally altering the United States' power landscape. According to a new forecast, data centers are projected to consume approximately 20% of the nation's electricity by 2035, a significant increase from current levels, placing imminent pressure on the power grid.
The latest analysis indicates that US data center power demand is expected to reach 194 gigawatts by 2035, representing a substantial 83% upward revision from a previous forecast made just months ago. This significant adjustment reflects the rapid expansion of AI infrastructure development pipelines and strong market expectations for the deployment of large-scale data centers over the next decade.
The demand surge is already having a tangible impact on the electrical grid. In regions with high concentrations of data centers, such as Virginia and Texas, their share of electricity consumption will far exceed the national average. Concurrently, some local governments have imposed moratoriums on new large-scale data center projects, with political resistance and environmental concerns emerging as new variables that directly impact the pace and layout of related energy infrastructure investment.
Revised Forecast Highlights AI-Driven Acceleration
Data shows data centers currently account for 5.9% of total US electricity consumption. This share is projected to rise to about 12% by 2030 and climb further to 20% by 2035. This growth trajectory implies that data centers' share of power usage will more than triple in less than a decade.
An analyst involved in the report provided a stark illustration of this scale: "For every five units of electricity generated by every US coal plant, gas plant, and solar farm, one unit will flow to data centers. This is on the same order of magnitude as the energy used to power electric vehicles and light cities."
The major upward revision to the 194-gigawatt demand forecast is primarily driven by the continuous expansion of the AI facility construction pipeline. One gigawatt is roughly equivalent to the capacity of a traditional nuclear power plant, meaning data centers alone will require power equivalent to nearly 200 such plants by 2035.
Grid Under Strain as Supply Struggles to Keep Pace
At the grid level, the pace of expansion on the supply side is struggling to match the explosive growth in demand. Historical data shows the single-year record for data center grid interconnection capacity is 71 gigawatts. Even maintaining this peak rate, a supply-demand gap of approximately 19 gigawatts is projected by 2035 under a baseline scenario—a calculation that already factors in on-site natural gas generation and assumes demand is met by on-site gas pipeline supply.
The analyst noted, "We are already seeing the grid scrambling to catch up, which is driving fundamental changes in how assets connect to it."
Regionally, the PJM Interconnection grid, covering 13 states including the Northern Virginia "data center alley," and the area managed by the Electric Reliability Council of Texas (ERCOT), are expected to see data center electricity consumption shares above the national average by 2035, making them focal points for grid pressure.
Industry Expansion Faces Policy and Environmental Hurdles
Despite the robust demand outlook, external resistance facing data center developers is increasing. Some local authorities have halted new large-scale data center projects, citing reasons including the projects' growing energy consumption and environmental impact.
In this context, data center operators are being compelled to explore new power supply pathways. Combinations of battery storage and natural gas generation have become important means to accelerate grid interconnection. However, even with these measures, baseline scenario calculations suggest that a combination of natural gas and grid power alone will still be insufficient to meet total demand.
For investors, these trends indicate that long-term capital requirements in power infrastructure, energy storage technology, and distributed energy sectors will continue to expand. Meanwhile, policy uncertainty and grid interconnection bottlenecks present major risk exposures for project implementation.