MS Utilizes Nuclear Power for AI Operation... AI Used in SMR Licensing Process

Microsoft (MS) is reportedly sourcing the electricity required to operate artificial intelligence (AI) from carbon-zero nuclear power plants and using AI to obtain approval for small modular reactors (SMRs).

According to the Wall Street Journal (WSJ) on the 12th (local time), MS decided in June to purchase nuclear power from Constellation Energy, the largest nuclear power owner in the U.S., to operate its data center in Virginia.

AI and supercomputing operations require enormous amounts of electricity. The power needed to run a single data center is equivalent to the electricity consumption of hundreds of thousands of households. Especially, AI operation demands even more computing power.

Constellation Energy projected that "the power demand for data centers due to the emergence of AI will be 5 to 6 times greater than the demand required for electric vehicle charging."

Nuclear power generates no carbon emissions and, unlike other renewable energy sources, can supply electricity 24 hours a day. However, safety concerns remain a drawback.

As a result, SMRs are gaining attention as an eco-friendly future energy source to replace existing nuclear power plants. SMRs do not produce carbon during power generation and have a lower risk of radioactive leakage or explosion accidents. Bill Gates, MS founder, has viewed SMRs as the next-generation energy source since 2008 and has made substantial investments.

However, SMRs require significant time and cost to obtain permits. Currently, there is only one SMR developer approved in the U.S. The approval process reportedly cost $500 million, with the application consisting of 12,000 pages and a total of 2 million pages of documentation.

MS is training large language models (LLMs) using AI to expedite the SMR approval process. They are collaborating with Terra Praxis, a UK-based nonprofit organization supporting nuclear power usage. MS provides coding expertise, while Terra Praxis supplies knowledge related to nuclear regulations.

Eric Ingersoll, co-CEO of Terra Praxis, said, "AI can reduce the time required to approve new nuclear reactors by up to 90%," adding, "It can also help accelerate the permitting process for renewable energy developers."