AI Data Center Power Hunger Spurs Nuclear Revival: Microsoft's 20-Year Small Purchase Power Agreement Behind the Investment Logic

Microsoft announced a major development on August 15, 2026, signing a 20-year power purchase agreement with nuclear energy company TerraPower to provide clean electricity for its AI data center located in Washington State, USA. This agreement not only addresses the growing power demand of Microsoft's AI data center but also marks the resurgence of nuclear energy as a stable, efficient power source amid the explosive growth in AI computing power demand, injecting a new dimension into the investment logic for AI data centers. ## I. Power Hunger of AI Data Centers: The Bottleneck of Traditional Power Supply In recent years, with the proliferation of large models and generative AI, the power demand for AI data centers has been growing exponentially. According to International Data Company (IDC) statistics, global AI data center power consumption reached 120 billion kilowatt-hours in 2025, a 300% increase from 2020. It is projected that this figure will exceed 200 billion kilowatt-hours by 2027, accounting for over 40% of global data center power consumption. The power demand for AI data centers primarily comes from high-energy-consuming computing chips like GPUs; a single H100 GPU can consume up to 700 watts, and a large AI data center might deploy tens of thousands of such chips, resulting in enormous power consumption. However, traditional power supply faces multiple bottlenecks. Firstly, the intermittency of renewable energy (such as wind and solar power) leads to unstable power supply, which cannot meet the 24/7 uninterruptible power needs of AI data centers. Secondly, the carbon emissions from traditional energy sources (such as coal and natural gas) conflict with global carbon neutrality goals, forcing tech giants to seek clean power solutions. Furthermore, the pace of upgrading power infrastructure lags behind the expansion rate of AI data centers, leading to power shortages in some regions and driving up energy costs. ## II. Microsoft's Nuclear Energy Strategy: Strategic Considerations Behind the 20-Year Agreement Microsoft's cooperation with TerraPower is a significant move in addressing the power needs of AI data centers. TerraPower is a company specializing in Small Modular Reactor (SMR) technology, and its developed Natrium reactor features high safety, short construction cycles, and stable power output. According to the agreement, TerraPower will provide power to Microsoft's Washington State data center by 2028, with a contract term of 20 years, ensuring long-term power stability. The reasons for Microsoft's choice of nuclear energy are threefold: first, stability, nuclear energy can provide 24/7 uninterruptible power to meet the high load demands of AI data centers; second, cleanliness, nuclear energy has near-zero carbon emissions, aligning with Microsoft's carbon neutrality goals (Microsoft aims for carbon neutrality by 2030 and net-zero emissions by 2040); third, cost-effectiveness. Although the initial investment in nuclear energy is high, in the long term, its power costs are lower than traditional energy, especially when electricity prices fluctuate significantly. Furthermore, Microsoft's nuclear energy strategy is not unprecedented. As early as 2023, Microsoft partnered with TerraPower to plan the construction of two small nuclear reactors by 2028 to supply power to data centers. The signing of this 20-year agreement further solidifies this strategy, demonstrating Microsoft's confidence in nuclear energy as a source of power for AI data centers. ## III. Industry Impact: The Resurgence of Nuclear Energy and the Collaborative Development of AI Data Centers Microsoft's move attracted widespread attention from the industry, and other tech giants are following suit. Google announced a partnership with nuclear energy company NuScale Power in 2025 to use small nuclear reactors in its data centers; Amazon signed a contract with TerraPower in 2026 to provide power for its data centers. These collaborations indicate that nuclear energy is becoming a crucial power source for AI data centers, driving the revival of the nuclear energy industry. For the nuclear energy industry, the power demand of AI data centers presents a new market opportunity. The traditional nuclear power industry faces pressure from decommissioning, while the development of Small Modular Reactor (SMR) technology allows nuclear energy to meet the needs of data centers of various scales more flexibly. According to the International Atomic Energy Agency (IAEA), over 20 countries are researching SMR technology, and the market size for SMRs is expected to reach $50 billion by 2030. For AI data center operators, the stability of nuclear energy can reduce electricity costs and improve the operational efficiency of data centers. For example, Microsoft's data center in Washington State is expected to save 20% in electricity costs while reducing carbon emissions. Furthermore, long-term contracts with nuclear energy can lock in electricity prices, avoiding the risks associated with price fluctuations. ## Four, Investment Logic: Synergistic Opportunities in Nuclear Energy and AI Data Centers From an investment perspective, the synergistic development of AI data centers and nuclear energy brings new investment opportunities. First, nuclear energy companies (such as TerraPower, NuScale Power) are direct beneficiaries, as their technological advantages and long-term contracts provide a stable source of income. Second, the optimization of electricity costs by AI data center operators (such as Microsoft, Google, Amazon) will enhance their profitability, thereby boosting stock prices. Additionally, companies in the nuclear energy supply chain (such as reactor manufacturers and power equipment suppliers) will benefit from this trend. Specifically, investors should pay attention to the following types of stocks:

• Nuclear Technology Companies: Such as TerraPower, NuScale Power. These companies possess SMR core technology and have signed long-term contracts with tech giants, ensuring stable income expectations.
• AI Data Center Operators: Such as Microsoft, Google, Amazon. The optimization of their electricity costs will increase their profitability, especially when electricity prices fluctuate significantly.
• Nuclear Energy Supply Chain Companies: Such as reactor manufacturers and power equipment suppliers. These companies will benefit from the revival of the nuclear energy industry, especially the proliferation of SMR technology.

## Five, Future Outlook: The Power Revolution in AI Data Centers As AI technology continues to develop, the power demand of AI data centers will continue to grow, and nuclear energy, as a stable and clean power source, will play an important role in this. It is projected that by 2030, the share of nuclear energy in global AI data centers will reach 15%, becoming the second-largest power source after renewable energy. However, the nuclear energy industry still faces challenges such as technological maturity, regulatory approval, and public acceptance. However, the positioning of tech giants like Microsoft will drive progress in nuclear energy technology and improve the regulatory environment, laying the foundation for the power revolution of AI data centers. In summary, the 20-year agreement between Microsoft and TerraPower marks the revival of nuclear energy in response to the power hunger of AI data centers, providing new investment opportunities for investors. Against the backdrop of continuously growing demand for AI computing power, focusing on the synergistic development of nuclear energy and AI data centers will be an important direction for future investment.