高盛：人工智能数据中心电力激增与可靠性 - 成本上升及美国政策转变如何影响绿色可靠性溢价

Investment Rating - The report highlights 44 Buy-rated stocks that are expected to benefit from the AI/data center power surge, including Quanta Services, GE Vernova, Xcel, Cameco, and NextEra [3][54][57] Core Insights - Strong demand and government actions are driving an average cost increase of 23% for new power generation capacity additions in the US, with a focus on reliability and affordability in power sourcing [1][12] - The Green Reliability Premium, which accounts for redundancy in power solutions, is projected to rise significantly if IRA renewables incentives are eliminated, impacting the overall cost structure for data centers [2][42] - The report anticipates a 160% increase in global data center power demand (AI + non-AI) by 2030 compared to 2023, equivalent to adding another top 10 power-consuming country [21][27] Summary by Sections Green Reliability Premium - The Green Reliability Premium is estimated to be $40/MWh with current IRA incentives, increasing to $55/MWh if these incentives are removed, representing a modest impact on hyperscaler EBITDA [42][44] - The report emphasizes that the cost of power is not seen as a constraint to data center growth, although regional power reliability and affordability will influence siting decisions [21][46] Power Demand and Supply - The US power demand is expected to grow at a CAGR of 2.5% through 2030, with data centers contributing approximately 100 basis points to this growth [22][24] - The report outlines diverse drivers of upward cost pressure across power generation, including tariffs and demand for natural gas, which could lead to significant increases in levelized costs [12][19] Investment Opportunities - The report identifies a bullish outlook for infrastructure contractors, regulated utilities, and industrials due to the rising demand for reliable power amid aging infrastructure and extreme weather events [3][54] - It highlights the importance of an all-of-the-above approach to power sourcing, considering the variability in supply availability and time to market [1][46]