The Nuclear AI Cloud: Nuclear Power Could Handle AI Energy Consumption Requirements
With all of the AI talk and the need for compute power and therefore electricity, here is a bit of information about the sources of electric power generation in the United States. My view is that nuclear energy is the best solution for generating clean, long-term, low-cost electricity.
Nuclear energy accounted for 20% of electricity production in the United States in the month of December, behind only natural gas at 42%.
The Waynesboro, Georgia-based Alvin W. Vogtle nuclear power plant’s two new reactors - reactors 3 and 4 - will produce 17.2 million megawatt hours annually.
For comparison, Google consumed approximately 22 million megawatt hours of energy globally in 2022.
The energy output of reactors 3 and 4 will power 500,000 homes and businesses.
As it relates to nuclear plant construction, production, maintenance and management, if I led the Department of Energy (DoE), I would recruit Google, Microsoft, Amazon, Apple, Facebook, Oracle, OpenAI and other large energy consumers to contribute significant amounts of capital to fund nuclear plant construction, production, maintenance and management. I would want the DoE to only have a support role.
Depending upon the plant in question and which firm would be its largest consumer (outside of American citizens), I would want that firm (Google for example) to run point on negotiations with the various contractors as Federal and State governments have proved time and again that they overpay for everything, including energy. For example, the Vogtle reactors 3 and 4 build was originally budgeted at $14 billion and the total cost is likely to be around $35 billion. Vendors love when they can convince Government clients to engage in cost-plus contracts.
