The electricity demand of US data centers has been increased by 36%, while the soaring energy consumption of AI has instead exacerbated concerns of a bubble.

date
22:02 01/12/2025
avatar
GMT Eight
Research institution BNEF's latest forecast shows that by 2035, the electricity demand of data centers in the United States will soar to 106 gigawatts, a significant increase of 36% compared to the forecast in April of this year by the same research institution.
Research institution BNEF's latest forecast shows that by 2035, the power demand of data centers in the United States will skyrocket to 106 gigawatts (1 gigawatt is roughly equivalent to the power output of a nuclear reactor, which can power approximately 750,000 American households), a significant increase of 36% compared to the institution's forecast in April of this year. This forecast adjustment is mainly driven by a wave of new early-stage data center projects. BNEF points out that the proportion of artificial intelligence (AI) training and usage in data center capacity will significantly expand, expected to increase from the current 12% to nearly 40% by 2035. AI training and applications will also drive a significant increase in data center utilization rates (i.e. the proportion of actual capacity used). Red alert sounded for the US power grid However, the expected surge in electricity demand from US data centers is expected to collide with aging and dilapidated grid infrastructure, leading to a power crisis in the next decade. Schneider Electric's latest warning shows that if current reserve capacity levels are maintained, there will be a substantial shortfall in US power generation capacity during peak electricity demand by 2028. More alarming is that its models predict a power shortfall as high as 175 gigawatts by 2033, enough to trigger widespread blackouts and a systemic power rationing crisis. It is reported that American Electric Power Company, Inc.'s system is pieced together by regional grids and utilities, currently facing multiple pressures from data centers, new factories, and electric vehicles. Even before the rise of the AI trend in the past two years, frequent extreme weather events and the continuous increase in the proportion of intermittent renewable energy sources have already put a strain on the grid. Jeannie Salo, Chief Public Policy Officer for Schneider Electric North America, said, "This may mean that we may be lagging behind in the AI race. We are also sending a signal to the world that the United States may struggle to meet the electricity needs of this era, which may lead to reduced investment." Schneider predicts that the increasing peak demand will increasingly squeeze the so-called "reserve capacity" - the additional power buffer available during extreme weather or network attacks. If emergency power is diverted to cope with daily peak demand, it will threaten grid reliability and increase vulnerability. Salo metaphorically compares reserve capacity to emergency savings in a bank. She pointed out that in three years, the United States will have to regularly tap into these "reserves," by which time these reserves may be almost depleted, making it impossible to withstand the impact of serious emergencies. Despite a wave of battery installations, the actual amount of electricity that can be stored is still woefully inadequate, like a drop in the bucket. It should be noted that the grid must balance power generation and consumption every second with precision to avoid collapsing. Salo said, "At present, we are facing increasingly frequent peak demand and power fluctuations. With this trend, by 2028, we will have to use the reserved backup power in advance, and cannot keep it for emergency use. This situation poses a significant threat to the reliability of the power system, as the severity of power shortages may increase, and we must take measures to deal with it in a very short time." Information released by the North American Electric Reliability Corporation (NERC) in November shows that the impact of data centers has become apparent - they have led to a significant increase in winter power demand, with a significant increase in the risk of blackouts under extreme conditions. It is predicted that peak electricity demand this winter will be about 20 gigawatts higher than the previous winter, while electricity supply will only increase by 9.4 gigawatts during the same period. However, some technologies and strategies have already been applied to alleviate the congestion of the grid. Take Texas as an example, where battery capacity is rapidly increasing, easing concerns about the grid collapsing due to a surge in demand. Salo advocates prioritizing the deployment of these "grid-enhancing technologies" instead of large-scale construction of new power generation and transmission projects, as the latter may be unable to meet actual electricity demand in the coming years. US utility stocks decline Expectations of a surge in power demand driven by the AI trend had pushed the stock prices of US power generation companies and utilities to historic highs. However, these companies are currently facing concerns about the decline of the speculative bubble - investors will not wait indefinitely for returns to materialize. After reaching a historical high in October, the S&P 500 Utilities Index has performed poorly in November. The stock price of Constellation Energy (CEG.US) has dropped by 11% since its peak in October after its third-quarter earnings conference call did not disclose any details of new power generation projects. Similarly, Vistra Energy (VST.US) has fallen by 16% since mid-October, as analysts noticed the slow pace of its data center projects' announcements. These companies, which have recently set new valuation records, are gradually returning to a rational valuation range as the market realizes that their expected large-scale data center business may not be as substantial as anticipated, or that project progress is significantly slower than expected. This expectation gap is shaking investor confidence and prompting capital to reevaluate the real value and risk alignment of this sector. Mark Malek, Chief Investment Officer at Muriel Siebert & Co., bluntly stated, "The market is raising new questions - can these companies achieve rapid expansion as expected? Are they wasting capital on projects that will never come to fruition?" The massive power consumed by data centers running AI has already changed the global energy landscape, but AI as a profitable business model has not been fully validated to some extent. This uncertainty has made traders worried about investing in the construction and operation of data centers, sparking discussions about an AI bubble. Although the utility industry has traditionally been seen as a safe haven for funds, the large-scale construction of AI-driven data centers has sparked an investment frenzy in the sector. Now, as the industry's realities gradually emerge, investors are trying to clarify two key questions: which companies can truly deliver on their expansion commitments, and which ones can exit unscathed and unscathed when the potential trillion-dollar AI bubble bursts. Travis Miller, a utility analyst at Morningstar, said, "Concerns about the AI bubble have exacerbated the recent weakness in utility stocks. If the expected growth in electricity demand fails to materialize, utility stocks are overvalued at current trading prices." Many utility companies have already started to lower their expectations. Constellation Energy lowered the upper limit of its full-year earnings forecast last month. Vistra Energy also adjusted its EBITDA forecast downward. NRG Energy (NRG.US) maintained its full-year EBITDA forecast in November, when investors were expecting an increase. In addition, Tim Winter, utility sector portfolio manager at Gabelli Funds, said that another major threat facing the utility sector is that the AI market may see disruptive technological breakthroughs similar to the DeepSeek incident earlier this year, leading to a complete collapse of power growth expectations that sustain it. It is reported that earlier this year, there were reports that DeepSeek's AI model had power consumption only a fraction of that of American companies, causing widespread decline in utility stocks. However, some analysts believe that the recent pullback in utility stocks is not a cause for concern, as investors are simply taking profits after a strong rally in October. Sophie Capu, a utility analyst at KeyBanc Capital Markets, said, "There is no bubble in the utility sector. The market is consolidating, waiting for the next round of growth to come."