Texas sounded a two-fold increase in electricity demand alert! AI infrastructure drives the wave of "grid grabbing + self-generation", and the bull market in power stocks is coming.

The latest warning from the operator of the Texas power grid in the United States indicates that by 2032, in order to meet the booming expansion of large artificial intelligence data centers and the electricity demand brought about by population growth, actual electricity demand may double compared to recent record levels. This increase will require generating capacity equivalent to nearly 300 new nuclear reactors. The Electric Reliability Council of Texas (ERCOT) stated in a presentation and public statement that peak electricity demand could reach 367,790 megawatts in six years, more than four times the historical peak of 85,508 megawatts set in August 2023. According to a presentation released by the organization on Wednesday, AI data centers will account for over 60% of this predicted increase. In fact, U.S. tech giants have already entered an unprecedented era of "megawatt-level power grabs" and "self-generation," and are now moving towards "grabbing power connection rights, dispatchable power source allocation rights, seizing long-term power system resources, and seizing fast deployable power generation capacity." This policy direction based on "self-generation" will effectively transform AI data centers from large electricity consumers into "electricity infrastructure investors," shifting demand from simply "grid connection capacity" to a full set of data center power equipment CAPEX, which also means that the unprecedented power demand brought on by the "electricity monster" AI will lead to a "super bull market" in the power sector. The surge in power demand from data centers is pushing Texas to its electricity limits. Texas is facing "exceptionally strong growth" across various industries, with CEO Pablo Vegas of ERCOT stating, "We believe that this latest data center forecast exceeds expected future growth levels." ERCOT's latest forecasts show that as the scale of data center projects continues to increase, grid operators, large utilities, and U.S. energy regulatory agencies are struggling to figure out how to respond. Some observers are skeptical of the committee's predictions. "This is not an actual forecast, because it is actually impossible to happen," said Travis Kavulla, policy director at home battery startup Base Power. "There needs to be a market-based approach to allocate scarce grid system access for supply and demand, which has not yet been proposed." Kavulla added that this massive scale of load "is actually impossible to happen entirely," because the real bottleneck is not only generation capacity, but also grid access, transmission capacity, equipment delivery cycles, and price tolerance. The Trump administration's recent "Ratepayer Protection Pledge" explicitly requires large-scale cloud computing providers and AI tech companies increasing their load to bear all the costs of the energy and infrastructure needed for their data centers, without passing on the costs to ordinary residents. This means that Trump is requiring large tech companies to "build power plants for data centers," which is no longer an abstract policy slogan, but is pushing these tech giants committed to building or expanding large-scale AI data centers towards a "Bring Your Own Power" model. Microsoft is in exclusive negotiations with Chevron and Engine No. 1 for a 2,500 megawatt natural gas power plant in Texas for a data center campus; Oracle has signed a up to 2.8 gigawatt fuel cell power supply agreement with Bloom; Google's cloud computing platform, Google Cloud, has also signed long-term power supply and co-siting agreements with AES, TotalEnergies, and others in Texas. Trump stated in his State of the Union address to Congress that tech companies will be required by the government to build dedicated-level power supply systems for their expanding AI infrastructure, rather than drawing additional power from local grids or significantly increasing loads. Chevron explicitly stated last year that its gas-fired power plant for data center construction will initially bypass existing transmission lines to reduce the risk of raising residential electricity prices, and U.S. tech companies are indeed entering the era of "power grabs" and moving towards the "contractual self-generation" era. Here, "power grabs" are not for immediate electricity, but for future scarce electricity access rights, transmission capacity, equipment delivery cycles, and freely dispatchable power allocation rights. The global process of building and expanding AI data centers led by Google, Microsoft, and Meta, the parent company of Facebook, is heating up, and this process is increasingly highlighting the importance of power supply resources, which is why the investment theme of "the end of AI is power" is becoming increasingly popular. What's more significant is that if the "self-generation" path is ultimately institutionalized in the entire U.S. and even other regions like Europe, it will undoubtedly shift a large part of AI capital expenditure systematically to power equipment and grid technology stacks. The reason why the global capital markets have recently pushed power equipment and grid chains to the forefront is also clear: the AI arms race has spilled demand from GPU/TPU/AI server computation clusters to power generation equipment (gas turbines), transformers/switchgear, transmission and distribution expansion, and grid engineering and scheduling software. From the breakdown of power system engineering, "self-generation" does not mean "off-grid operation," but more commonly involves Behind-the-meter power sources (large gas engines/gas turbines, renewable power continuous supply + stable storage, and even nuclear power PPAs) + dual redundant connections to utility grids. The most direct beneficiaries are power generators with long-term contract power, nuclear/gas asset holders, rapidly deployable distributed power sources and fuel cell companies, and transmission and grid technology equipment chains that can handle large capital expenditures. For example, European industrial giant Siemens Energy is clearly benefiting from the demand for AI-driven gas engines and grid equipment, with its performance and stock price performance directly linked to the "AI data center construction boom" driven by North American AI development processes. Cited as driving the preparation for the massive expansion of European AI data centers. Additionally, the European stock market's utility sector has become one of the beneficiaries of AI-related market trends.