AI data centers' power needs might require built-in nuclear reactors
As artificial intelligence (AI) data centers face increasing power demands, experts are considering alternative energy sources like small nuclear reactors to meet their needs. Tech giants like Microsoft and OpenAI are already investing in small modular reactor (SMR) technology, which could potentially power individual data centers and reduce reliance on the grid. Digital Realty CTO Chris Sharp told the BBC, "Our industry has to find another source of power."
What are Microsoft and OpenAI up to?
In 2023, Microsoft put out job listings that suggested it was planning to roll out its SMRs. The goal is to use them to power its AI data centers eventually. Meanwhile, OpenAI CEO Sam Altman has invested in a nuclear start-up known as Oklo, which is working on self-regulating SMRs. Altman also claimed that AI is so power-hungry that it will require a "breakthrough energy source."
SMRs are potential solution
Small modular reactors (SMRs) are mini power plants designed to provide on-site power. Michael Bluck, the director of the Centre for Nuclear Engineering at Imperial College London, said, "Data centers are power-hungry things, but with AI we're moving into a new level of power requirements." Although around 50 SMR designs exist, the challenge is building them in repeatable units and standardizing production lines. Currently, no commercial SMRs are in operation despite a growing industry focus.
Regulatory hurdles and future prospects
One major hurdle for SMRs is regulation, as nuclear energy carries inherent risks. The United States (US) Nuclear Regulatory Commission has been working with several SMR designers on pre-application activities in recent years. In 2020, it approved the first SMR design by Oregon-based start-up NuScale Power. However, it's still uncertain if SMRs will be the ultimate solution for AI data centers' increasing energy needs. Companies are urgently looking for techniques to scale up operations now rather than waiting for years.