Facebook and Instagram's parent company, Meta Platforms (META.US), announced three significant agreements on Friday, planning to secure a massive nuclear power supply system of up to 6.6 gigawatts (GW) for its large-scale AI data centers by 2035. This represents one of the most extensive and ambitious collaborations between a major U.S. tech giant and nuclear power suppliers to date. The global construction and expansion of AI data centers, spearheaded by Google, Microsoft, and Meta, is proceeding at a breakneck pace, increasingly highlighting the critical importance of power supply. This underscores why the investment theme "AI's ultimate frontier is electricity" is gaining immense traction. Among the currently surging power demands, nuclear power needs are rising most rapidly, with its clean, stable, and efficient attributes making nuclear systems the most favored power resource for tech behemoths.
Meta announced a deal with U.S. power giant Vistra (VST.US) to purchase electricity from three existing nuclear plants in Ohio and Pennsylvania. Furthermore, Meta will support a new type of nuclear project planned for large-scale construction within the next decade by U.S. emerging nuclear developers Oklo (OKLO.US) and TerraPower—specifically, Small Modular Reactor (SMR) projects. These transactions closely follow a major nuclear power agreement announced in June 2025, where Meta aimed to secure robust nuclear power supply from a Constellation Energy nuclear plant site. What is the scale of 1 GW? It can approximately power 750,000 to 850,000 U.S. households (based on average consumption). According to EIA (U.S. Energy Information Administration) data, the average monthly residential electricity consumption in the U.S. for 2024 is about 865 kWh per household (approximately 10,380 kWh annually). A rough calculation using the annual electricity output of "1 GW continuous supply" (about 8.76 TWh) equates to around 840,000 households; considering real-world operations are not always at full capacity, a more conservative range of 750,000–850,000 households is more accurate in an engineering context.
Benefiting from Meta's newly announced massive nuclear power cooperation agreements, Vistra's stock price surged over 17% in pre-market trading, while Oklo's stock price skyrocketed over 22% pre-market. The latest 20-year power purchase agreement signed between Meta and Vistra will procure electricity from Vistra's Perry and Davis-Besse nuclear plants in Ohio, and the Beaver Valley plant in Pennsylvania, collectively providing over 2,600 megawatts (MW) of nuclear energy. This includes 2,176 MW of operational generation capacity starting later this year, plus an additional combined 433 MW of output enhancement. Vistra stated this constitutes the largest nuclear capacity expansion project in the U.S. supported by a corporate customer.
Vistra indicated that these Power Purchase Agreements (PPAs) provide certainty for proceeding with subsequent license renewals for each reactor, extending each license by an additional 20 years. Currently, the Beaver Valley Unit 1 license is valid until 2036, Davis-Besse's until 2037, Perry's until 2046, and Beaver Valley Unit 2's until 2047. Oklo stated that this significant agreement with Meta substantially advances its plan to build a 1.2 GW power campus in Pike County, Ohio, and introduces an innovative mechanism allowing Meta to make advance payments and provide funding, enhancing certainty for the deployment of Oklo's Aurora "powerhouse" project; specific financial terms were not disclosed.
Meta will also invest heavily to help fund TerraPower's development of two additional reactors, which could begin power generation as early as 2032, with a combined capacity of up to 690 MW, and grants Meta priority rights to procure power from up to six additional flagship TerraPower reactors by 2035; specific financial terms were not disclosed. These newly announced deals from Meta closely follow CEO Mark Zuckerberg's "AI ambition" super blueprint, which involves investing hundreds of billions of dollars by the end of this decade (before 2030) into AI large model R&D/training and the AI computing infrastructure needed to support massive AI inference workloads.
Since the emergence of ChatGPT ignited the global AI wave, Meta under Zuckerberg's leadership has sought to fully integrate an AI software and hardware ecosystem, aiming to become a dominant leader in AI large models, agent-centric AI applications, and AR+AI smart glasses. The essence of the global AI race is a competition in AI computing infrastructure, and the core foundation powering AI computing clusters is a stable and massive electricity supply system. Consequently, power demand from AI data centers is soaring at an unprecedented rate, with AI effectively transforming into a "power glutton."
The exponential expansion of high-energy-consumption AI data centers, driven by intense demand for computing infrastructure like AI chips, is fundamentally dependent on power supply, giving rise to the market view that "AI's ultimate frontier is electricity." Wall Street giant Goldman Sachs recently revised its forecast for massive electricity demand driven by global data centers by 2030 upward to a 175% increase from 2023 consumption (up from a previous forecast of +165%), equivalent to adding the power load of a "top-ten global electricity-consuming nation." In the view of Goldman Sachs' strategy team, the ultimate constraint for AI large models is electricity—the firm emphasizes that AI, a veritable "electricity devourer," will trigger an unprecedented global "super demand cycle" for power and a "super bull market" for utility stocks.
Following Google's major launch of the Gemini 3 AI application ecosystem in late November, this cutting-edge software quickly gained global popularity, causing a sudden surge in Google's AI computing demand. The immediate release of the Gemini 3 series products generated an enormous volume of AI token processing, forcing Google to significantly reduce free access to Gemini 3 Pro and Nano Banana Pro and impose temporary limits even on Pro subscribers. Coupled with recent South Korean trade export data showing持续强劲 demand for HBM memory systems and enterprise SSDs, this further validates Wall Street's assertion that the "AI boom remains in the early construction phase where computing infrastructure supply cannot meet demand," implying that power demand will continue its surge.
A forecast report from the International Energy Agency (IEA) indicates that global electricity demand from data centers will more than double by 2030, reaching approximately 945 terawatt-hours (TWh)—slightly higher than Japan's current total electricity consumption—with AI applications being the primary driver of this growth. The IEA expects overall electricity demand from AI-focused data centers to increase at least fourfold by 2030. Therefore, within the current and long-term global decarbonization trend, nuclear energy, as an efficient and stable clean power source, has become the most favored energy choice in recent years for tech giants like Amazon, Google, and Microsoft.
This energy source, combining cleanliness, stability, and efficiency, is expected to provide robust, 24/7 power support for their immense data centers. Consequently, support for nuclear energy and nuclear plants from global politicians and tech companies is currently stronger than at any time since the 1970s. Examples include Microsoft signing a long-term PPA with U.S. energy giant Constellation to facilitate the restart of the Three Mile Island Unit 1 (approx. 835 MW), Amazon securing about 1.92 GW of nuclear power from Talen for a large AI data center supporting its massive AWS cloud platform, and Google planning with Kairos to deploy 500 MW of advanced nuclear power by 2035 to match new data center loads.
Most notably, the U.S. government's attitude towards nuclear reactors has undergone a comprehensive shift, especially with Trump's return to the White House vigorously promoting a U.S. nuclear power revival. President Trump has signed multiple executive orders to advance reforms in the U.S. nuclear industry, including expanding nuclear scale and the supply chain, and shortening approval cycles for nuclear projects; this move may signal the clarion call for a U.S. nuclear renaissance. Influenced positively by Trump's signing of multiple so-called "nuclear revival orders," this explains why U.S.-listed stocks like OKLO and other nuclear-related equities have experienced sustained significant gains.
