The end of AI is electricity! After the merger of American utility companies, the AI data center's new "electricity stocks for sale" are under the spotlight.

When the global capital markets are still fiercely competing for GPU computing power and HBM memory bandwidth, a more hidden but disruptive thread is reshaping the energy landscape in North America and even globally. Electricity has become the most critical constraint for AI data center construction. However, the current issue is not the insufficient total power supply, but a triple mismatch - the speed of grid connection lags far behind the speed of data center construction, aging regional power infrastructure cannot support explosive growth, and traditional approval processes are almost ineffective under the time scale of the AI era. The "mismatch crisis" of power supply and demand: the underestimated systemic bottleneck This week, a major acquisition also revealed this signal. On May 18th, New Era Energy (NEE.US) acquired competitor Dominion Energy Inc(D.US) for $67 billion - the largest utility acquisition in US history. Driving this deal is not any traditional energy strategy, but the insatiable thirst of global AI data centers for electricity - Dominion Energy Inc holds more than 51 GW of data center contract demand, described by New Era Energy CEO John Ketchum as "unprecedented demand growth". This deal is the latest sign that power supply is becoming a key constraint for AI data center expansion. As IEA data shows, global data center electricity consumption grew by 17% in 2025 to reach 485 terawatt-hours, and IEA expects this number to double to around 950 terawatt-hours by 2030. This article comprehensively analyzes the deep logic, industry chain restructuring paths, and investment value of related listed companies on the AI power bottleneck from both supply and demand sides. According to industry research reports, the three-year compound growth rate of AI data center electricity demand in North America is as high as 73%, while the progress of grid connection and interconnection approval is sharply contrasting with this. On May 16, 2026, a report from Monitoring Analytics revealed the most core negative signal of this contradiction: the electricity price of the largest US power market PJM Interconnection suddenly increased irreversibly, with an increase of up to 76%. The report rarely used the term "irreversible", which means that the structure of power supply and demand is undergoing fundamental changes. This "mismatch crisis" is now spreading from the grid connection end and reshaping the entire energy supply chain. As pointed out by the Data Center World Conference in 2026, the expansion speed of power infrastructure has become the factual limit of AI deployment, and the combination of distributed on-site generation with large-scale real estate platforms is becoming a feasible path to bridge the supply-demand gap. From a longer-term investment logic perspective, energy research firm AInvest's analysis indicates that power infrastructure investors are seeing AI as a persistent demand driver - the investable asset category of power infrastructure is undergoing structural repricing. If the projected growth of renewable energy installed capacity along with the demand for grid infrastructure upgrades continues to grow in the next decade as forecasted by IEA, institutional fund inflows into the AI power supply field may still be in the early stages of structural development. Industrial ecosystem reshaping: from "behind-the-meter" supply to "energy as a service" Faced with grid constraints, the power supply model of data centers is undergoing structural innovation, and the emerging models of the power energy industry are particularly intriguing. "Behind-the-meter generation": a radical approach to bypass grid constraints Traditional data centers typically obtain power from the grid through utility companies, but faced with the reality of grid connection delays that could last for years, more and more operators are turning to behind-the-meter (BTM) generation - using natural gas or other energy sources to generate power on-site. The attractiveness of this model lies in several factors: it allows data center operators to avoid queuing for grid connection and gain a first-mover advantage in time; direct supply eliminates transmission and distribution costs, significantly improving cost predictability; and self-generated power can ensure the required electricity quality and reliability standards. But the cost is also evident - large-scale natural gas generation faces equipment supply shortages, emission reduction commitments, and community approval resistance. Data shows that the price of gas generators has increased by 10% to 15% annually due to the surge in demand from AI data centers, and this trend is expected to continue at least until 2028. The delivery cycle of gas turbines has lengthened to 1 to 3 years, with queues extending beyond 2028 - these core bottlenecks on the supply side of equipment are forming new constraints. Industry insiders further point out that even if companies adopt a behind-the-meter supply model, it does not mean that overall energy pressure will decrease; the related burden is simply shifted from the grid to the natural gas supply chain, and the tense state of the energy market may persist. Gigawatt-level "energy as a service": transitioning from assets to services On May 12, 2026, Japanese Hitachi Ltd. and US Energy Corp. infrastructure investment and operation company X LABS announced a strategic partnership to jointly develop gigawatt-level "energy parks" for AI data centers in North America, providing integrated generation, storage, and energy management solutions in an "energy as a service" (EaaS) model. These energy parks are designed as on-site power supply hubs, integrating a variety of generation sources, storage facilities, transmission and distribution infrastructure, and energy management systems, which can provide reliable and controllable large-scale power to data centers as the main power source while coordinating with regional grids. The collaboration plans to provide full lifecycle services covering design, development, operation, and power supply through project-specific special-purpose vehicles (SPVs) - X LABS is responsible for SPV operation, project financing, and site development, while Hitachi provides gigawatt-level transmission and distribution technologies, grid stability solutions, and BESS battery storage systems. The core innovation of this model is that it shifts power supply from capital expenditure to operating expenditure, allowing data center operators to quickly access stable power without having to bear massive upfront capital investments or manage complex energy operations. The two companies plan to complete the construction of the first energy park in the early 2030s, and although this is not a short-term solution in terms of timeline, the maturity of this model could be an important variable in addressing the power bottleneck of AI data centers. BESS: from diesel backup to AI load buffering In the power infrastructure ecosystem of data centers, battery energy storage systems (BESS) are playing an increasingly important role. Traditional diesel backup generation schemes have shown many limitations in AI load scenarios, and BESS is not only becoming a viable alternative, but also taking on a crucial new function - addressing the rapid and significant power fluctuations caused by AI workload. During the training and inference processes of AI chips, power loads can undergo drastic changes in microsecond-scale time frames, posing unprecedented challenges to grid power quality. Oracle Cloud Infrastructure AI Infrastructure Vice President Ram Nagappan eloquently explains: "Traditional grids and power plants cannot handle the huge power fluctuations generated by AI data centers, which can occur multiple times per second. BESS can act as a buffer to suppress these power fluctuations, which is becoming a new requirement for AI data centers." MarketsandMarkets predicts that the global BESS market will grow from $508.1 billion in 2025 to $1,059.6 billion by the end of 2030; BloombergNEF reports that the global addition of BESS installations in 2025 reached 112 gigawatts, a 48% increase from 2024. More significantly, BESS is increasingly being deployed in conjunction with fast-response gas generators and synchronous condenser cameras to form a hybrid architecture - in projects such as Carter's Incorporated's deal with US AIP Company in West Virginia and Baker Hughes' supply of NovaLT gas turbines to Frontier Infrastructure, BESS has become a standard component to enhance system transient response capabilities. Full industry chain benefit spectrum: from chips to concrete, the AI power investment matrix The power demand of AI data centers is penetrating every corner of the industry chain, from large independent power producers to traditional utility companies, to building materials suppliers and special electronics distributors, forming a multi-level benefit chain. Primary beneficiaries: independent power producers and nuclear power newcomers Talen Energy (TLN.US) is one of the independent power producers most directly linked to the power supply of AI data centers. The company owns and operates approximately 13.1 GW of American Electric Power Company, Inc. infrastructure, including 2.2 GW of nuclear capacity at the Susquehanna nuclear plant. In June 2025, Talen signed a $18 billion, 17-year power purchase agreement (PPA) with Amazon.com, Inc. AWS to supply up to 1,920 MW of carbon-free power. The agreement is scheduled to complete the transition from behind-the-meter to in-front-of-the-meter grid connection during the spring 2026 refueling outage of the Susquehanna nuclear plant. In early February 2026, Talen completed the acquisition of 2.6 GW of natural gas power assets in the PJM market from Energy Capital Partners, further enhancing its flexibility and bargaining power in data center co-location negotiations. In March 2026, Talen signed a memorandum of understanding with X-energy for the deployment feasibility evaluation of the Xe-100 advanced reactor in Pennsylvania and the PJM market. Oklo (OKLO.US) represents the cutting-edge "nuclear power directly supplying AI" energy model. The company recently launched a $1 billion "at-the-market" (ATM) equity financing plan, endorsed by a syndicate of top-tier Wall Street investment banks including Goldman Sachs Group, Inc., Bank of America, Citigroup, and JPMorgan Chase. The company's core projects include the Idaho Aurora nuclear plant (planned for commercial operation in 2028), a 1.2 GW power supply agreement with Meta (in Ohio facility, first phase as early as 2030), and a collaborative AI modeling project with NVIDIA, aiming to build a "nuclear + AI infrastructure" dual-empowerment system. In the first quarter of 2026, Oklo's net loss narrowed to $33.1 million ($0.19 per share), lower than the market expectation of $0.20 per share. Vistra Energy (VST.US) signed a 20-year power purchase agreement with Meta in January 2026 for the entire 2,176 MW capacity of the Perry and Davis-Besse nuclear plants to support their AI data center operations. The company achieved a record high adjusted EBITDA of $1.494 billion in the first quarter of 2026. Vistra has put forward the core strategic initiative of "speed to power", promoting co-location, demand response, and gas bridging solutions for mega-cloud clients to address grid connection delays. The company's full-year 2026 core profit guidance is $6.8 billion to $7.6 billion significantly higher than the $5.7 to $5.9 billion in 2025. Secondary beneficiaries: traditional utilities and power infrastructure On May 18, 2026, New Era Energy acquired Dominion Energy Inc for $67 billion, with the core DRIVE behind it being Dominion Energy Inc's demand of over 51 GW of data center contracts - equivalent to the installed capacity of about 50 large nuclear power plants. Dominion Energy Inc's service area covers the "data center corridor" of Northern Virginia, where an estimated 70% of global internet traffic flows through every day. American Electric Power Company, Inc. (AEP.US) and Idacorp (IDA.US) are traditional utility companies whose interconnection businesses are also related to this theme. AES Corporation (AES.US) signed a 20-year power supply agreement with Alphabet Inc. Class C in February 2026 to supply its new data center in Wilbarger County, Texas, and adopt a co-location power generation model. As of the announcement date, AES has signed energy agreements with data center customers for nearly 12 GW of capacity, with 9 GW being direct PPAs signed with mega-scale cloud service providers. In March 2026, AES announced that it would be taken private through a $33 billion transaction, expected to be completed by the end of 2026 or early 2027, reshaping the company's financial structure in expanding clean power to serve mega-scale data centers. Hitachi (HTHIY) and X LABS' gigawatt-level energy park cooperation put them at the forefront of the AI data center energy service market. Hitachi will provide high-voltage transmission and distribution technologies, grid stability, power quality, and other systemic solutions through Hitachi Energy, as well as integrate BESS storage technology and the HMAX energy AI optimization platform. Tertiary beneficiaries: building supply chain and specialty equipment The spillover effects of AI infrastructure investments are beginning to penetrate the building materials and specialty equipment supply chain. Prologis, Inc. (PLD.US), a traditional logistics real estate giant, is accelerating its layout of data center power access. In the first quarter of 2026, the company initiated a $2.1 billion new development project, with $1.3 billion allocated to two data center projects. Its data center power reserves - locked-in or in-progress power access capacity - have reached 5.6 GW, becoming a strategic pillar business of the company. Prologis is also collaborating with NVIDIA, EPRI, and InfraPartners to evaluate the deployment of 5 to 20 MW pre-fabricated micro-data centers near underutilized utility substations, with the goal of having at least five pilot projects in development by the end of 2026. Insteel Industries (IIIN.US) is a direct beneficiary in the steel supply chain of the data center construction boom. In the first quarter of 2026, the company saw revenue grow by 23.3% year-over-year to $159.9 million, with net profit soaring by 602.4% to $7.59 million. Management views data centers and IIJA infrastructure projects as the dual engines of strong demand in 2026 - "We believe data center projects will act as a timely transitional bridge before traditional private non-residential projects recover". The company plans to invest about $20 million in 2026 for factory and information system upgrades to support growth. Richardson Electronics, Ltd. (RELL.US) is in the semiconductor equipment upstream segment of this investment cycle. The company is entering a "long-awaited upswing cycle" driven by high-margin wafer manufacturing equipment business and long-term AI data center demand, and management expects this cycle to exceed the typical 6 to 12-month equipment investment cycle. In addition, other companies in this field include Alliant Energy (LNT.US), TransAlta (TAC.US), Capital Power (CPXWF), and Central Puerto SA (CEPU.US). Conclusion While the global capital markets are still cheering for thousands of GPUs added each quarter, the true bottleneck of the AI industry is quietly shifting - from chip foundries to transformers and power lines in the grid. NextEra's $67 billion acquisition of Dominion is essentially the largest vote of confidence that capital has cast on "power over chips" in the future. The reason why this investment thread is worth keeping an eye on is not only because of its scale - from AES with 12 GW of contracted capacity to Prologis with 5.6 GW pipeline - but also because it possesses unique countercyclical resilience. Regardless of how AI model architectures evolve, chip processes advance, or political waves in the industry, once a data center is built, its power demand becomes an almost rigid long-term contract revenue stream. In an industry of accelerated technological changes and intense competition, certainty itself is the most valuable scarce asset. Electricity - the least noticeable bottom layer in AI infrastructure - perhaps is the most certain long-term investment logic of the next decade.