At the end of artificial intelligence is truly electricity! The United States' largest power grid has sounded the alarm for upgrades and renovations, and the power stock bull market is roaring in.

The CEO of PJM Interconnection, the largest grid operator in the United States, David Mills, stated that the country's largest grid system and widespread grids need massive upgrades to meet the unprecedented power resource demand caused by the construction boom of super-scale AI data centers. Multiple market research reports indicate that the power demand of global AI data centers is skyrocketing, surpassing the elastic growth pattern of traditional IT loads. In the coming years, utility companies and grid developers in the United States may need to spend trillions of dollars to upgrade outdated grids on a large scale and provide the strongest power capacity support for the unprecedented construction and expansion of AI data centers led by tech giants like Microsoft, Google, and Amazon. Mills stated in an internal letter to stakeholders that the current structure of the grid system PJM Interconnection LLC currently serves over 67 million people in 13 states cannot ensure an adequate power supply while protecting residents from soaring bills. "The current situation is clearly unsustainable," Mills wrote in a letter released on Wednesday. "In addition, the pressures currently visible in prices, grid system reserve capacity margins, and investment pipelines reflect some fundamental and substantive issues that need to be recalibrated." The crisis facing PJM includes the operator expecting severe power shortages in the grid as early as next year and the possibility of one of the largest utility companies in the United States, American Electric Power Co., exiting the grid system. Therefore, in this power stock bull market narrative, the most beneficial will undoubtedly be those who can deliver power resources to data center server rooms quickly, reliably, and affordably hot power stocks with high load growth exposure, recoverable capital expenditures, regulated utilities, beneficiaries of power grid upgrades, core power equipment such as natural gas/nuclear power/distributed power that can be delivered in large quantities quickly, and data center power chain companies that provide extremely flexible load management and demand response capabilities. Grid upgrades are imminent! The Trump administration urges tech giants to take "self-generation" measures, but it may be difficult to alleviate high electricity bills for residents in the short to medium term. The surge in household electricity bills and the influx of energy-intensive AI data centers has become a key election issue in some areas. According to a report released by the U.S. Chamber of Commerce on Tuesday, electricity prices in the PJM region have risen by 51% in Maryland and 41% in Illinois over the past five years. "The urgency of upgrading and transforming the grid is paramount, as the region has only a few years, not decades, to consciously make these significant choices," Mills wrote. Senior analyst Ryan Levine from Citigroup expressed concerns about the delays in finding solutions at PJM and the devil being in the details of each proposal. Levine wrote in a report, "We are concerned that the continued hesitancy and back-and-forth stance on proposals may cause PJM to miss out on opportunities. If it truly takes years to figure out these issues, the project leaders of data centers may 'choose to move to other regions in the world.'" The Trump administration earlier in the year called on large tech companies in the "State of the Union" address and subsequent policy pushes to self-generate or independently secure power supply for their rapidly expanding AI data centers and to sign so-called "ratepayer protection commitments" to pledge to build, introduce, or purchase electricity themselves, rather than passing on the increased electricity burden to regular residential consumers. This commitment was signed by tech giants including Amazon, Google, Meta, Microsoft, OpenAI, Oracle, and xAI. The government emphasized that this approach could prevent AI data centers from causing a rise in residential electricity bills. The focus of the measure is for tech companies to bear the costs of increased power due to AI growth rather than relying on existing public grids to spread these costs. However, the feasibility and implementation details of this "self-generation" policy are very limited. Most of the plans proposed by Trump and related officials are political frameworks or non-binding commitments without clear legal enforcement mechanisms, construction standards, or specific timelines. Additionally, while tech companies have pledged to bear the costs of electricity, there are no specific rules on how to build power plants, obtain environmental assessment approvals, and grid connection schedules. Many energy laws and regulatory frameworks still require utility oversight and grid interface standards, which cannot be circumvented by a mere presidential declaration. Experts and the media have also pointed out the significant challenges in implementing this proposal, suggesting that it may be more of a political statement rather than a feasible policy. Therefore, even with the Trump administration's "self-generation" policy, PJM still faces fundamental supply and market design issues, inevitably leading to the continued expansion of electricity costs for residents. The core challenge for PJM is not whether tech companies are willing to build their power plants, but the outdated grid structure, slow growth in power generation capacity, inadequate long-term market mechanisms and price signals, and the pressure of large-scale new loads (such as data centers, electric vehicle charging, etc.) on existing transmission and distribution capacity and reserve capacity. Even if tech companies build their power facilities, they still need time to coordinate scheduling with the grid, obtain transmission access permits, and comply with federal and state energy regulatory rules, rather than simply solving the entire supply shortage issue in the short to medium term by "embedding power plants." Recent reports indicate that PJM is still considering reforming market mechanisms, introducing long-term contract mechanisms, and long-term resource investment strategies to address supply shortages, rather than relying solely on companies to self-generate power. This underscores the urgent need for a transformer in the U.S. electricity grid. The power-hungry trend of AI data centers has ignited a super bull market in power stocks Undoubtedly, whether it is the upgrade and transformation of the grid system or the "self-generation" policy direction, it is driving the global electricity assets into a structural revaluation cycle driven by the construction boom of AI computing infrastructure. The construction and expansion of global AI data centers led by companies like Google, Microsoft, and Facebook's parent company Meta are in full swing, highlighting the importance of power supply for these data centers. This is why the investment theme of "the end of AI is power" is becoming increasingly popular. More importantly, if the "self-generation" path is institutionalized across the United States and even Europe and other regions, it will undoubtedly shift a significant portion of AI capital expenditures to power equipment and the grid technology stack. Power assets with reliable power generation, transmission access, long-term pricing agreements, and the capacity to handle data center loads are garnering valuation premiums similar to bottleneck assets in the AI supply chain. PJM's warning is crucial serving 13 states and over 67 million people, the existing market structure is unable to simultaneously ensure "adequate power supply" and "protect residents from soaring electricity bills." Statistics show that PJM's power capacity prices have increased by over 1000% in the past two years, yet market mechanisms have failed to stimulate sufficient new power generation. Power shortages in the United States may occur as early as 2027. From a power engineering perspective, AI data centers are not just ordinary commercial loads but are high-density, round-the-clock, highly reliable, low-interruption-tolerant "industrial-grade base load requirements." Recently, Wall Street behemoth Goldman Sachs revised its forecast for massive electricity demand driven by global data centers to be 175% expansion by 2030, compared to the previous prediction of +165% by 2023. The analysts at Goldman Sachs emphasized that the end of AI's large models is electricity the institution stressed that the "power-hungry beast" of AI will bring about an unprecedented global electricity "super-demand cycle" and a "super bull market" in power stocks. The Electric Reliability Council of Texas (ERCOT) recently predicted in a presentation and public statement that six years from now, peak power demand could reach 367,790 MW, a significant increase compared to the historical peak of 85,508 MW set in August 2023 more than a fourfold increase. The latest warning from the Texas grid operator indicates that by 2032, to meet the soaring demands of large AI data centers and population growth, actual electricity demand may double from recent record levels, requiring the equivalent of nearly 300 new nuclear reactors' worth of power generation capacity. The recent warnings from ERCOT and PJM jointly highlight the complete inability of the United States' aging grid system to keep up with the massive power demand of the AI era, and that U.S. tech giants have entered an unprecedented era of "megawatt-scale power rush" and "self-generation." This also indicates that the ongoing construction boom of AI data centers is accelerating the most critical power supply links in the data center power chain towards a new round of super bull market trends in power stocks. The global capital market, in pushing power equipment and the grid chain to the forefront as the "new mainline," follows a clear logic: the AI arms race has overflowed the actual demand and the main axis of AI computing construction from GPU/TPU/AI server clusters to power equipment (gas turbines), transformers/switchgear, transmission and distribution expansion, grid engineering, and scheduling software. From a breakdown of power system engineering, "self-generation" does not mean "off-grid operation," but rather common Behind-the-meter power sources (large gas engine sets/gas turbines, renewable power continuous supply + stable storage, and even nuclear power PPAs) + dual redundant connections to public utilities. In this power expenditure cycle, the most obvious beneficiaries will be those who focus on transmission system upgrades, power source agreements, self-generation equipment, joint energy sources, demand response, and high-level PPA/capacity contracts assets that truly address power supply bottlenecks in this round of the super bull market script for power stocks. --- Please note that the translation may not be perfect as it was generated by AI.