, which already has hundreds of millions of users. And OpenAI’s flagship app won’t be the only one that will boost data center use. Among other projects, Amazon will support the development of a small modular rewriter (SMR) facility in Washington State to provide carbon-free power for its data centers. The initial phase will include four units with a capacity of 320 MW, expandable to 12 units with 960 MW. Construction will begin later this decade, with operational implementation expected in the next. In addition, technology companies have signed 20-year power purchase agreements (PPAs) with companies that generate electricity from nuclear power plants. For example, Microsoft and Meta have signed contracts with Constellation Energy to restart the Three Mile Island plant and expand the Clinton Clean Energy Center, respectively. Google has also signed a purchase agreement with Kairos Power to acquire energy from its SMR reactors, which are scheduled to come online in 2035. The momentum behind these AI-related projects has sparked widespread debate. A recent report published by Barclays asserts that a global “nuclear renaissance” is underway, driven by the sector’s position at the intersection of megatrends such as energy security, decarbonization , geopolitics, and technology. For these analysts, a favorable policy environment is developing as governments identify nuclear power as a clean, stable, and energy-dense form of energy. Furthermore, Barclays says this process is being further accelerated by demand from large cloud service providers or hyperscalers for their data centers, including Amazon Web Services, Google, Microsoft, and Meta. These companies have multi-billion-dollar investment plans underway due to AI. It believes it provides revenue certainty for developers of next-generation reactors, designed to address concerns about dispatch speed and unit economics. In general terms, they emphasize the increasing energy needs and long-term demand of hyperscalers . Thus, the International Energy Agency projects that global electricity demand could double by 2050, while data center demand alone could reach this threshold as early as 2030. According to Barclays, hyperscalers increasingly identify energy procurement as the main obstacle to AI development, and current projections suggest that grids are unlikely to be able to meet this growth in demand. For this reason, hyperscalers have opted to obtain their own power “behind the meter.” These analysts believe that nuclear’s characteristics as an energy-dense, emissions-free, and stable power source have made it an important part of the solution for hyperscalers, many of which have ambitious net-zero emissions commitments. This has materialized primarily through PPAs signed with energy providers, providing long-term security for both nuclear project developers and technology companies. Regarding the United States, these analysts note that the country is entering a new era of nuclear energy growth, driven by rising electricity demand from data centers, manufacturing offshoring, and electrification. In terms of the sector itself, data center expansions have increased over the past 18 months, primarily as a result of pandemic-related projects. New construction is accelerating, driven by the U.S., and much of the ChatGPT-driven construction surge has yet to come online. Growth has been strong over the past 18 months, with a 31% increase in IT load, which measures the energy consumption of servers in data centers. The total data center energy load, which includes server power consumption and other facility needs such as cooling, has more than doubled since 2020, surpassing 70 gigawatts, according to Bloomberg NEF asset data. The launch of ChatGPT in November 2022 prompted new investment and project announcements as developers sought to meet the growing demand for AI. However, most of this capability is still under construction. Construction starts have remained steady at around six gigawatts per semester, with the first half of 2025 setting a record of 6.4 gigawatts. These data centers are likely to come online within one to three years, indicating that expansions will accelerate further in the near future. The U.S. currently accounts for 49.3% of global data center IT capacity, its largest share on record, reflecting a reversal of previous declines. After gradually losing share in the early 2010s, the U.S. market is once again extending its lead as AI-driven projects become concentrated in the region. By early 2024, the U.S. accounted for more than 70% of new capacity under construction. The next five largest markets in terms of capacity under construction are China, Malaysia, Canada, India, and Brazil, although they represent less than a fifth of the capacity under construction in the U.S.