Core Insights - The report emphasizes that the competition for computing power in the AI era is fundamentally a competition for electricity, highlighting the critical role of electricity supply in supporting AI-driven growth in the U.S. [1] Group 1: AI's Impact on U.S. Electricity Demand - The overall electricity consumption growth in the U.S. appears slow, but electricity demand from data centers, driven by AI, is rapidly increasing [1] - Data centers currently account for 4.4% of total U.S. electricity consumption, with projections indicating this could rise to between 6.7% and 12% by 2028-2030, necessitating an additional power generation capacity of up to 100 gigawatts [1][14] - In 2023, U.S. data center energy consumption reached 176 terawatt-hours, a significant increase from 76 terawatt-hours in 2018 [14] Group 2: Current Electricity Supply Challenges - The U.S. electricity supply system is under significant strain due to limited new capacity in natural gas, insufficient interconnection capacity in the grid, and the concentration of data centers in a few states like Virginia, leading to localized grid pressure [1][3] - Natural gas is currently the primary source of electricity generation in the U.S., with a generation capacity of 207.2 GW, accounting for 43% of the total generation mix [20][23] Group 3: Solutions to Electricity Supply Issues - Short-term solutions include optimizing existing energy sources and rapid deployment of nuclear power, which is seen as an ideal stable and clean power source for data centers [2] - Gas-fired power generation is also a practical choice due to its short construction cycle and flexibility in meeting the rapidly growing electricity demand [2] - Long-term innovations such as small modular reactors (SMRs) and controlled nuclear fusion are being explored, with SMRs already in early project stages [2][3] Group 4: Future Energy Landscape - The energy transformation driven by AI demand is reshaping the competitive landscape of the electricity industry, focusing on stability, cleanliness, and innovative energy solutions [3] - The race for reliable and low-carbon energy sources is intensifying, with advancements in traditional energy optimization, nuclear technology upgrades, and exploration of fusion energy [3]

Group 1: AI's Impact on US Electricity Demand - The overall electricity consumption growth rate in the US is relatively low, with a projected total electricity consumption of 41,104 billion kWh in 2024, reflecting a year-on-year increase of 2.5% [13] - The electricity elasticity coefficient for the US is expected to be 0.77 in 2024, indicating a long-term trend of decreasing energy consumption per unit of GDP, with a compound annual growth rate (CAGR) of only 0.38% from 2008 to 2024 [18] - The share of electricity consumption from data centers in the US has been increasing, reaching 4.4% in 2023, with a significant rise from 1.9% in 2018 [23] Group 2: Supply Side Dynamics - Natural gas is currently the primary source of electricity generation in the US, with a projected share of 43% in 2024, while coal and nuclear power are on a downward trend [29] - The installed capacity growth rates for various energy sources in 2024 are projected as follows: coal (-2.3%), natural gas (-0.2%), nuclear (2.9%), solar (34.1%), and wind (10.5%) [24] - Recent policies have shifted from supporting renewable energy to tightening regulations, particularly affecting solar and wind energy projects [36] Group 3: Future Projections for Data Center Electricity Demand - By 2028, US data center electricity consumption is expected to range between 325-580 TWh, accounting for 6.7%-12% of total US electricity consumption [42] - The compound annual growth rate for data center electricity demand in the US is projected to be around 23% from 2024 to 2030, with an expected total consumption of 606 TWh by 2030 [52] - The additional power capacity required for data centers by 2030 is estimated to be between 9-100 GW, depending on growth scenarios [58] Group 4: Regional Disparities and Challenges - Data centers are highly concentrated in states like Virginia, Texas, and California, leading to significant regional disparities in electricity load, with Virginia's data centers consuming 25.6% of the state's total electricity [71] - The current electricity supply system is a major source of data center outages, with power system failures accounting for 52% of global data center incidents [76] - The interconnection capacity of regional power grids in the US is insufficient to match the growing electricity load, particularly in the eastern regions driven by AI data centers [86]