Technology giant Microsoft (MSFT.US) is engaged in exclusive negotiations with U.S. oil and gas behemoth Chevron (CVX.US) and the well-known investment fund Engine No. 1. The parties aim to finalize a long-term agreement to provide substantial support for a major energy complex in West Texas, which would power a massive AI data center campus led by Microsoft. According to media reports citing informed sources, this planned ultra-large natural gas power plant in West Texas is estimated to cost approximately $7 billion. Its initial power generation capacity is expected to reach 2,500 megawatts, positioning it as one of the largest projects of its kind in the United States. The sources requested anonymity as discussions are still ongoing. The companies stated in a jointly released declaration: "Chevron, Microsoft, and Engine No. 1 have signed an exclusive agreement concerning a proposed power generation system and associated power offtake arrangements. Commercial terms have not yet been finalized, and there is no definitive agreement at this time." Chevron and Engine No. 1 had previously disclosed some details about their plans for a large power plant but had not identified the end-user. A final agreement with Microsoft would secure a major, long-term customer for the plant's electricity, aiding significantly in financing its construction. The project could commence operations as early as 2030, pending approvals from U.S. tax and environmental authorities, as well as final agreement on commercial terms by the three companies. The Trump administration had publicly urged large technology companies in February to "meet their own power needs." Subsequently, the White House pushed Microsoft, Google, Amazon, and Meta, the parent company of Facebook, to sign commitments requiring major AI data center operators to fully bear the costs of related energy infrastructure and to develop new power supplies themselves. This U.S. government policy direction, encouraging tech giants toward self-powering, is actively fueling the popularity of the investment theme that "the ultimate frontier for AI is electricity." The American AI technology sector appears to be entering a new phase characterized by "capital actively securing power" and "self-supply." For global tech giants like Microsoft, Google, and Amazon, access to power resources is increasingly becoming a core competitive barrier in the AI era. On one hand, the U.S. power grid faces challenges including interconnection queue backlogs, transmission bottlenecks, and shortages of large-generation equipment. PJM, one of the nation's largest grid operators, indicated that data center demand growth is outpacing new supply additions, potentially leading to power shortages as early as next year. On the other hand, technology companies closely linked to AI are evolving beyond simply signing Power Purchase Agreements (PPAs) to directly securing underlying generation assets. Microsoft, through an agreement with Constellation, is facilitating the restart of the Three Mile Island nuclear power plant to supply a large data center it is developing. Meta, in a revised agreement in Louisiana, agreed to cover the full service costs for its hyperscale data center and is advocating for the grid connection of over 5,000 megawatts of clean energy resources in a Texas contract, while leading the development of its global largest data center in Richland Parish. The recent actions of these tech behemoths suggest that whoever can more quickly secure baseload power, underlying large-scale transmission pathways, and generation resources will be better positioned to expand the next wave of AI computing power. In practice, "building your own power system" typically does not mean completely disconnecting from the grid or creating an isolated network. More commonly, it involves tech companies funding the securing of power sources, bearing the costs of grid interconnection and supporting infrastructure, and, when necessary, developing dedicated or semi-dedicated power supply projects. A $7 Billion Bet Ignites the Self-Power Era! Microsoft Partners with Chevron on Texas Energy Mega-Project As a longtime supporter of ChatGPT developer OpenAI, Microsoft, despite recently posting its worst quarterly stock performance since the 2008 financial crisis, continues to ramp up data center construction. This effort aims to compete with Google parent Alphabet Inc. and Amazon for dominance in cloud computing and supremacy in cloud-based AI training and inference platforms. Securing reliable baseload power is becoming a critical challenge. The collaboration between Chevron and Engine No. 1 is expected to address this challenge by leveraging their extensive natural gas production in West Texas and contracts for large turbine projects. New-generation large data centers are increasingly being built away from major population centers but closer to fuel sources due to their immense power requirements. The site selected by Chevron and Engine No. 1 is near the city of Pecos, close to the Texas-New Mexico border, in the heart of the Permian Basin, the largest oil and gas producing region in the United States. The Permian Basin produces vast quantities of natural gas as a byproduct of oil extraction, often exceeding pipeline capacity. Consequently, some gas is flared off as it cannot be transported to demand centers, making the region an ideal location for power plant construction. Landbridge Co. LLC, a major landowner in the area, stated that over the past two years, tech companies have proposed at least nine large data center projects in northern and western Texas. A key strength for Chevron's entry into the AI power supply sector is its longstanding partnership with Engine No. 1. The latter is the activist investment firm that successfully campaigned against Chevron's rival ExxonMobil in 2021. This partnership has already secured an order for seven large natural gas turbines from GE Vernova Inc. Demand for such equipment is extremely high, with new customers often facing wait times of several years. Chevron indicated that the large power project in Pecos is expected to begin operations in 2027, ramping up to its full 2,500-megawatt capacity over three years. According to U.S. Department of Energy data, this capacity is equivalent to more than two typical nuclear reactors. If successful, the plant could potentially be expanded to a maximum of 5,000 megawatts in the future. Under the entity name "Energy Forge One LLC," Chevron has applied for several tax abatements from local governments in West Texas. Its air emissions permit application submitted to the Texas Commission on Environmental Quality was deemed administratively complete in October. AI's Ultimate Frontier is Electricity! A "Super Bull Market" for Power Stocks is Underway As the Trump administration pushes U.S. tech giants into the "self-power" era, companies like Microsoft, Google, and Meta are increasingly using long-term offtake agreements, co-located gas generation, nuclear restarts, associated transmission investments, and cost-sharing arrangements. These actions are transforming electricity from a public utility into a dedicated means of production tied to super-sized AI data center campuses. This policy-driven shift towards "self-powering" effectively turns AI data centers from major electricity consumers into "investors in power infrastructure." Consequently, demand expands beyond mere "grid connection capacity" to encompass the entire suite of power equipment CAPEX for "self-generation + grid interconnection + campus-owned distribution systems." This implies that AI, often described as a "massive electricity consumer," will drive an unprecedented "super bull market" for power-related stocks. During a State of the Union address, Trump stated that tech companies would be required by the government to build dedicated, utility-grade power supply systems for their expanding AI infrastructure, rather than drawing additional power from local grids and significantly increasing load. Chevron explicitly stated last year that its gas-fired plants for data centers would initially bypass the existing transmission grid to mitigate the risk of raising residential electricity rates. American tech companies are arguably entering an era of "scrambling for power" and moving towards a "contractual self-power" model. The global construction and expansion of AI data centers, spearheaded by Google, Microsoft, and Meta, is proceeding rapidly, increasingly highlighting the critical importance of power resource supply. This is why the investment theme "the ultimate frontier for AI is electricity" is gaining significant traction. More significantly, if the "self-power" pathway becomes institutionalized across the United States and other regions like Europe, it would undoubtedly systematically shift a substantial portion of AI capital expenditures towards power equipment and grid technology stacks. The recent market trend of elevating power equipment and grid-related stocks to a "new main investment theme" in global capital markets has a clear logic: the AI arms race is spilling demand over from GPU/TPU/AI server clusters to generation equipment (gas turbines), transformers/switchgear, transmission and distribution expansion, grid interconnection projects, and dispatch software. From a power systems engineering perspective, "self-power" does not equate to "off-grid operation." It more commonly involves Behind-the-meter power sources (large gas-fired units/turbines, renewables plus storage, or even nuclear PPAs) combined with dual-path redundancy via interconnection with public utilities. For instance, European industrial giant Siemens Energy is clearly benefiting from AI-driven demand for gas turbines and grid equipment, with its financial performance and stock price directly linked to the "AI data center construction wave," specifically citing demand for data center power infrastructure driven by North American AI development. E.ON, Europe's largest grid operator, is also directing increased investment towards "preparing for the massive expansion of AI data centers in Europe," and the European stock market's utilities sector has become one of the beneficiaries of AI-related market trends.
Comments