As artificial intelligence (AI) technology rapidly evolves, tech companies are grappling with the substantial energy and resource consumption it requires. According to Stanford's Artificial Intelligence Index Report 2023, a single training session of OpenAI's GPT-3 consumes 1,287 megawatt-hours, equivalent to the energy used by 3,000 Tesla Model Y vehicles driving 200,000 miles each.
At the recent World Economic Forum, OpenAI's CEO Sam Altman highlighted nuclear fusion as a potential future energy source, though it remains years away from practical use. In the meantime, major tech companies are considering nuclear power as a viable solution to AI's energy demands.
Google (GOOG, Financial) announced an agreement with Kairos Power to purchase power from six to seven small modular reactors (SMRs) totaling 500 megawatt-hours to meet its data centers and AI energy needs. The reactors are expected between 2030 and 2035, though the construction site is yet to be determined. Google's Energy and Climate Director emphasized the importance of clean, reliable energy for technological advancement.
This announcement led to a surge in nuclear energy stocks. Oklo, a U.S. nuclear startup backed by Sam Altman, saw its stock rise 9.73%, with an additional 7.97% increase in after-hours trading. Other companies, including Constellation Energy and Cameco, also experienced stock gains.
Prior to Google's move, other tech giants had already embraced nuclear power. In September, Microsoft (MSFT) entered a 20-year agreement with Constellation Energy to revamp and utilize the Three Mile Island nuclear plant. Oracle (ORCL) plans to power its data center with over 1 gigawatt of energy from three small reactors. Amazon (AMZN) aims to connect a new data center directly to a Pennsylvania nuclear plant.
The trend towards small modular reactors continues to grow, as they offer significant advantages in efficiency and cost. Kairos Power, supported by the U.S. Department of Energy, is developing these reactors to reduce costs and accelerate construction timelines. Their simplified design and safety features enable deployment in diverse locations, ensuring predictable project delivery.
