Microsoft's ambitious plan to power AI with small nuclear reactors

Microsoft is reportedly considering using small modular reactors (SMRs) to power its data centers for training large AI models.

The move is driven by the energy-intensive nature of AI training, which has a significant carbon footprint. SMRs are viewed as a more cost-effective and flexible alternative to traditional nuclear power stations.

In theory, SMRs offer several advantages, including lower construction costs and greater flexibility in deployment. However, Microsoft faces several challenges before implementing this plan. It must source a working SMR design, secure a supply of highly enriched uranium fuel required for small reactors, and address the long-term storage of nuclear waste.

While Microsoft founder Bill Gates is involved in nuclear innovation through TerraPower, the company currently has no agreements to supply reactors to Microsoft. Additionally, Microsoft is exploring nuclear fusion as a potential energy source, signing a power purchase agreement with fusion startup Helion earlier this year.

With the growing popularity of artificial intelligence and increasing energy consumption requirements, Microsoft is actively exploring various green power options for its data centers.