Home Artificial Intelligence Microsoft Looks to Go Nuclear to Support Energy-Hungry AI

Microsoft Looks to Go Nuclear to Support Energy-Hungry AI

by Joey De Leon

Artificial Intelligence (AI) has become an integral part of many big tech companies’ products and services. However, the extensive energy and water consumption required to operate AI at scale has raised concerns about its environmental impact. Despite these concerns, companies such as Google and Microsoft continue to invest heavily in AI development.

In an effort to maintain its lead in the AI race, Microsoft has recently advertised a position for a program manager in “Nuclear Technology.” The job description suggests that Microsoft intends to use “microreactors” and “Small Modular Reactors” to power its data centers, which are crucial for the operation of its Cloud and AI services. While the exact plans for nuclear-powered AI remain unclear, Microsoft’s global presence makes the scope of its ambitions vast.

Small Modular Reactors (SMRs) are a proposed class of reactors that are smaller in size and power capacity compared to traditional nuclear plants. The advantage of SMRs is their potential to be built in one location and then transported to another. However, prototypes of SMRs are limited, with only a few implemented in countries like Russia and China. The approval of the first SMR design by the U.S. Office of Nuclear Energy was granted just this year.

Microsoft’s interest in nuclear energy is not entirely new. In May, the company signed a power purchase agreement with nuclear fusion startup Helion, which is set to begin in 2028. Fusion, a different process than fission used in SMRs, has shown promising developments in recent years. However, a significant energy pivot towards fusion or any other form of nuclear power is likely to take some time.

The need for immense power to sustain AI operations has driven companies like Microsoft to explore alternative energy sources. The company has made significant investments in generative AI and has integrated it into various software products. However, the sustainability report published by Microsoft revealed a 30% increase in water consumption year over year to cool its AI supercomputers. In fact, training OpenAI’s ChatGPT model, which Microsoft heavily collaborates on, consumed enough water to fill a nuclear reactor’s cooling tower.

Studies have shown that AI is responsible for substantial carbon emissions, and OpenAI’s GPT-3 model, in particular, has been found to emit more CO2 compared to other large language models. The introduction of OpenAI’s GPT-4, which is even more powerful and trained on significantly more data, suggests that the energy demands of AI will only continue to grow.

Despite the environmental concerns and high costs associated with AI, big tech companies like Microsoft are forging ahead with their AI ambitions. As the demand for AI continues to rise, finding sustainable and efficient energy sources will be crucial to mitigate its environmental impact. The integration of nuclear technology could potentially offer a solution, but questions regarding feasibility, safety, and long-term sustainability remain.

You may also like