AI资本支出激增，电网更吃紧，高盛大幅上调全球AI用电预期：2030年需求暴增220%

Core Insights - The investment in AI is shifting from chips and servers to electricity, with major cloud providers increasing capital expenditures and R&D budgets, leading to a significant rise in data center electricity demand [1][2] - Goldman Sachs has revised its forecast for global data center electricity demand growth from 175% to 220% by 2030, with the U.S. expected to account for approximately 60% of this increase [3][5] - The report highlights the challenges in electricity supply chains, emphasizing that the focus has shifted from whether electricity is needed to whether it can be delivered on time [1][6] Group 1: Electricity Demand Projections - Goldman Sachs estimates that the global data center electricity demand increase will reach 905 TWh by 2030, with the U.S. contributing 60% of this growth [3] - The projected capacity for U.S. data centers is expected to rise to 95 GW by 2030, up from 32 GW in 2025, while overseas capacity is projected to reach 72 GW [3] - The increase in electricity demand is driven by higher expectations for AI server shipments and faster data center capacity expansion [3] Group 2: Investment Trends - Major cloud providers are expected to have a reinvestment rate of nearly 90%, with capital expenditures and R&D projected to double by 2029 compared to 2025 [4] - The report indicates that while investments are increasing, the free cash flow available to shareholders is being squeezed, leading to a greater focus on quantifiable AI revenue growth [4] - An example of quantifiable AI impact is in drug development, where AI has improved success rates and reduced development timelines significantly [4] Group 3: Electricity Supply Challenges - Goldman Sachs has raised its forecast for U.S. electricity demand growth to an annualized 3.2% by 2030, with data centers contributing 2 percentage points to this growth [5][6] - The report notes that a significant portion of new load is being met by behind-the-meter solutions, primarily natural gas, due to challenges in delivering electricity from the grid [6] - The efficiency of new servers is improving, but the overall power consumption is increasing due to higher demand for computational power [7] Group 4: Pricing and Policy Implications - The concept of a "Green Reliability Premium" has emerged, indicating that the cost of reliable clean energy for data centers is higher than the baseline, estimated at $40-$48 per MWh [8] - This premium could lead to an industry expenditure of approximately $37-43 billion by 2030, impacting the profitability of hyperscalers [8] - The report suggests that data center operators will need to provide clearer commitments regarding flexibility and infrastructure costs to mitigate reliability risks [9] Group 5: Labor Market Constraints - Goldman Sachs estimates that approximately 510,000 new jobs will be needed in the U.S. and Europe to meet electricity demand growth from 2023 to 2030, with a significant focus on transmission and distribution roles [10] - The labor market constraints highlight the attractiveness of behind-the-meter solutions, which bypass some of the complexities of grid connections [10] - Companies with advantages in labor acquisition, such as contractors and utility firms, are likely to be revalued in the market [10] Group 6: Broader Investment Themes - The report emphasizes a broader theme of reliability investments across electricity, water, networks, and supply chains, with an estimated annual capital expenditure growth exceeding $80 billion [12] - Data center electricity supply chain stocks have significantly outperformed broader market indices, indicating a shift in investor focus [12] - The report outlines conditions under which the current investment cycle may end, including a decline in AI competitiveness and reduced corporate returns [12] Group 7: AI Investment Cycle - Goldman Sachs places AI within an innovation cycle, currently in the "Appraisal / Hopes & Dreams" phase, with discussions about transitioning to the execution phase intensifying [13] - Three potential triggers for this transition include constrained financial flexibility, declining corporate returns, and oversupply of products [13] - The report concludes that while the infrastructure chain remains favorable in the short term, market scrutiny on AI revenue and cash flow will increase [13]