In recent weeks we’ve covered the hyperscalers multi-billion-dollar investment in AI infrastructure and the commensurate demand for power. Progress accelerated this week with Amazon announcing a partnership with Energy Northwest to build modular nuclear reactors (SMRs), whilst Google signed a 500MW nuclear deal with Kairos Power.
Small modular nuclear reactors could play a crucial role in powering AI systems. Nuclear energy offers low-carbon, continuous power – qualities that make it an attractive alternative to traditional energy sources, especially as wind and solar struggle with intermittency issues. The focus on AI’s energy consumption marks a shift from previous concerns about blockchain technology. While crypto-related energy use has begun to stabilise, AI has emerged as the next significant computational technology to watch in terms of energy demands.
According to the International Energy Agency’s 2024 report, AI’s overall energy consumption remains relatively modest compared to sectors like industrial manufacturing (2026 share of total global demand: 1.30% standard data centres, 0.33% crypto, and 0.26% AI). But the density of AI power demand inside a data centre and adding capacity supply to regions with multiple existing datacentres, means the challenges are still immense.
Beyond nuclear power, other innovative approaches are being explored. Crusoe Energy and Blue Owl Capital recently launched a $3.4 billion joint venture to develop a 100,000 GPU AI data centre for Oracle and OpenAI powered by on and off-site renewables. Crusoe have previously led the way in using waste gas for crypto mining facilities. Waste gas is a byproduct of oil extraction, and if oil field operators have no economical use case for the gas or are unable to transfer it elsewhere, it’s often simply burned. According to a recent Crusoe report, for every ton of CO2 that the company produced in 2022, it reduced over 1.6 tons through avoided methane emissions. Founder Cully Cavness stated: “There is a huge amount of flared gas around the world, if you captured it all, it would power like two thirds of all of Europe’s electricity and it would power the entire datacentre industry many times over.”
Takeaways: The tech world is preparing for a future where AI drives unprecedented demand for sustainable power. The industry is actively seeking sustainable energy solutions to power its growing infrastructure. Next generation nuclear energy and innovative approaches like using flared gas are emerging as potential answers to the energy demand challenge.