Goldman Sachs(US:GS) Zhi Tong Cai Jing·2025-10-11 13:10Core Insights - Goldman Sachs reports that the primary barrier to unleashing AI potential is not capital but electricity, predicting a 50% increase in global data center electricity demand by 2027 and a 160% increase by 2030 [1][2] Group 1: Urgent Demand for Electricity - After a decade of stable electricity demand, the rise of high-energy AI data centers is expected to drive a 160% increase in electricity consumption by 2030, necessitating multi-layered solutions [2] - Collaborations between power companies and tech firms are emerging, such as Enterg and Meta's agreement to develop power generation and transmission assets for reliable electricity supply to data centers [2] - The approval process for new natural gas plants can take 5-7 years, highlighting the need for federal policy support to alleviate delays in electricity grid improvements [3] Group 2: Future Electricity Sources for Data Centers - Approximately 60% of the increased electricity demand from data centers will need to be met through new capacity, with sources projected to include 30% from combined cycle gas turbines, 30% from gas peaking plants, 27.5% from solar energy, and 12.5% from wind energy [4][5] - Renewable energy is currently the fastest and most efficient way to obtain incremental electricity, although it cannot meet the 24/7 power demands of data centers [5] Group 3: Embracing Nuclear Power - Nuclear power is seen as a key asset for providing 24/7 zero-carbon baseload electricity, essential for decarbonization and grid stability, aligning well with the needs of data centers [6] - The revival of nuclear energy faces challenges such as cost overruns and construction delays, exemplified by the Vogtle Unit 3 project, which exceeded its budget by over $17 billion and was delayed by about 7 years [6] - Small modular reactors (SMRs) are being explored as a reliable source of zero-carbon electricity, with large enterprises considering investments or long-term power purchase agreements to meet their energy needs [6] Group 4: Exploring "Behind-the-Meter" Power Solutions - Tech companies and data center developers are increasingly considering "behind-the-meter" solutions, acting as their own power suppliers to secure baseload electricity [7] - Many operators are exploring on-site microgrids or locating data centers near existing power plants to expedite electricity access and reduce reliance on the grid [7] - Companies like Solaris Energy Infrastructure and PowerSecure are providing distributed energy solutions to enhance reliability and reduce emissions amid growing demand [7] Group 5: Controversies Surrounding "Behind-the-Meter" Solutions - The "behind-the-meter" approach has sparked public debate regarding local cost burdens and potential environmental issues, as seen in complaints from communities near AI facilities [8] - Plans to co-locate data centers with nuclear plants have faced regulatory scrutiny due to concerns over increased electricity prices for local users [8]