AI交易“被忽视的风险”：万一，天量资本开支“花不出去”

Core Viewpoint - The article discusses the growing bipartisan political consensus against the rapid expansion of data centers in the U.S., highlighting the risks posed by both political opposition and physical limitations of the electrical grid [2][4][15]. Group 1: Political Environment - The political landscape has seen unusual alignment between figures like Bernie Sanders and Ron DeSantis, both advocating for a slowdown in data center construction due to public concerns over noise, water usage, and rising electricity costs [4]. - New York has become the sixth state to consider a moratorium on new data center construction, reflecting a broader trend of community protests and regulatory scrutiny across various states [3][4]. Group 2: Financial Implications - The anticipated capital expenditure of approximately $600 billion by 2026 faces skepticism regarding its feasibility, especially given the current limitations of the U.S. electrical grid [7][10]. - Major tech companies, including Microsoft, Meta, Amazon, and Google, have planned AI infrastructure spending of $670 billion this year, raising concerns about whether these funds can be effectively utilized [7][9]. Group 3: Energy Demand and Supply Constraints - Data centers are projected to double their energy demand by 2035, increasing from 34.7 GW in 2024 to 106 GW, which is equivalent to the electricity consumption of 80 million households [9]. - The Texas electrical grid operator, ERCOT, has proposed a review of projects consuming approximately 8.2 GW of power, indicating a significant regulatory slowdown that could hinder the expansion of data centers [10]. Group 4: Market Reactions - The financial markets have reacted sharply to the risks associated with unspent capital, leading to significant sell-offs in momentum stocks, marking one of the largest single-day declines in a decade [11]. - Concerns over the inability to meet new electricity demands have negatively impacted independent power producers, with some stocks experiencing declines of up to 27% year-to-date [13].