Microsoft is delving into the possibility of harnessing nuclear energy to fuel its AI data centers, addressing the colossal energy demands posed by AI models such as ChatGPT.
In a recently posted job listing, the company is actively seeking a “principal program manager for nuclear technology” to evaluate the feasibility of utilizing nuclear energy for the data centers housing AI models.
As initially reported by CNBC, Microsoft’s focus lies in the application of microreactors and Small Modular Reactors, which are notably more cost-effective to construct and operate compared to their larger nuclear counterparts. These innovations are intended to “power the data centers that the Microsoft Cloud and AI rely upon.”
AI models consume a substantial amount of computational power, with estimates suggesting that ChatGPT could accrue daily operating costs of up to $700,000 due to extensive server expenses. The job posting implies that Microsoft perceives nuclear energy as the solution to meet this escalating energy demand.
Microsoft previously entered into a $10 billion agreement with OpenAI, the creators of ChatGPT, which included Microsoft providing cloud services necessary for the startup’s model operations. Additionally, Microsoft integrated OpenAI’s technology into its Bing search engine, unveiling an AI-driven search tool earlier this year.
Growing apprehension surrounds the environmental impact of generative AI due to its immense energy consumption. Academic studies estimate that training GPT-3 resulted in over 550 tons of carbon dioxide emissions and a demand for 3.5 million liters of water.
In recent developments, Microsoft has made commitments to expedite the transition of its data centers to renewable energy sources, reduce emissions, and aspire to attain carbon negativity, water positivity, and zero waste status by 2030.