AI_数据中心全球电力激增_推动上涨_下跌的五个因素以及可靠性投资的助力_28页_2mb

Report Summary: Global Data Center Power Demand and AI Impact

Introduction

Goldman Sachs' January 12, 2025, report analyzes the surge in global data center power demand, driven by artificial intelligence (AI) and digital transformation. It projects significant growth, with data center electricity consumption expected to increase by over 160% by 2030 compared to 2023 levels. This demand is linked to broader power grid and infrastructure challenges, with AI being a key catalyst.

Five Key Drivers of Upside/Downside Risk

The report identifies five primary factors influencing data center power demand:

• AI server shipments: Constrained by data center capacity; market is tight through 2026, with sufficient capacity in the base case.
• Data center power capacity: Constrained by power infrastructure; 50% of demand may need new capacity, with natural gas CCGTs and renewables as key sources.
• Green optionality/cost: Constrained by low-carbon sourcing; Green Reliability Premium exists, even if renewable costs are lower for intermittent solutions.
• AI chips and power demand: AI innovation can reduce power intensity, but demand growth may outpace efficiency gains, driven by compute power needs.
• AI results/innovations: May constrain or boost demand by enabling efficiency solutions in sectors like healthcare, agriculture, and climate.

Data Center Power Demand Outlook

Globally, data center power demand is forecast to surpass 1,076 TWh by 2030, contributing 3–4% to global power demand, equivalent to a top-10 consumer. Exhibit data shows:

• U.S. demand growth at ~1% CAGR, with data centers influencing power mix more intensely.
• Europe sees an inflection due to data center connections, with ~170 GW pipeline and potential 10–15% demand boost.
• Efficiency gains have decelerated, with AI driving rapid power consumption despite improved efficiency per server.

Reliability as a Multi-Year Investment Theme

Rising power demand, aging infrastructure, and extreme weather risks amplify the need for Reliability investments. Outcomes focus on sustaining power and water systems:

• Grid upgrades require $720 billion in U.S. capex by 2030.
• Policies and corporate actions prioritize low-carbon, reliable power sources.
• Confluence of AI/Data Center growth and climate adaptation drives sustained investment tailwinds.

AI's Impact on Sustainability and Emissions

AI significantly boosts power demand, with emissions projected to double by 2030. While Green Reliability Premiums support low-carbon options, renewables cover ~40% of growth via power purchase agreements (PPAs). AI's potential in Sustainable Development Goals (SDGs) includes healthcare, agriculture, and climate optimization, but risks include societal destabilization and job displacement.

Key Conclusions

The convergence of AI, data centers, and power demand creates substantial investment opportunities in Reliability, while risks arise from supply constraints, climate challenges, and AI's broader societal impacts. Upside potential exists for companies supporting efficient and sustainable power solutions.