Amazon, Google, and Microsoft Are Investing in Nuclear Power to Secure the Future of AI. It’s Not Going to Be Easy

  • Amazon, Google, and Microsoft have announced several projects to create or reactivate nuclear power plants.

  • This will help the companies meet the energy demands of their AI data centers.

  • However, uncertainty about future energy requirements and the need to balance these with renewable sources has created some tension.

Nuclear
No comments Twitter Flipboard E-mail

If there’s one thing AI needs, apart from endless data for training, it’s energy. A significant amount of energy, to be precise. Big tech companies are learning this lesson the hard way. Their massive investments in new data centers require a colossal energy supply, and many of them are already finding solutions. Investing in nuclear power is one of them.

Amazon. On Wednesday, Amazon announced an agreement with energy company Dominion Energy to develop a small modular reactor (SMR) in Virginia. The project is just one of several. In total, the company plans to invest $500 million in three projects across different states. Each of these projects could generate 320 MW of power, meaning that if all three come to fruition, their combined capacity could reach 960 MW.

Google. For its part, Google announced on Monday that it had reached an agreement with startup Kairos Power to build seven SMRs by 2030. This project is expected to generate up to 500 MW of greenhouse gas emission-free electricity, which will power Google’s AI data centers. Plans are also in place to potentially commission additional SMRs by 2035.

Microsoft. Meanwhile, Microsoft is making moves of its own. The company has reached an agreement with energy company Constellation Energy to reopen the Three Mile Island nuclear power plant, which has been closed since 2019. Constellation Energy will invest $1.6 billion to restore and modernize the plant. Once operational, the plant will supply its 835 MW of power to Microsoft’s data centers, which require significant energy for their AI projects.

Additionally, in 2023, the tech company made investments in Helion Energy, a company backed by OpenAI CEO Sam Altman and focused on the future of nuclear fusion.

Energy consumption

Will SMRs be the solution? The energy consumption of data centers operated by companies like Google and Microsoft is enormous. A recent study indicates that their energy usage amounts to 24 TWh, nearly equal to the annual consumption of countries such as Slovakia or Ecuador. The projected output of these projects is between 500 and 1,000 MW, which will undoubtedly help meet future energy needs in the short and medium term.

However, another issue looms.

There’s a sense of urgency. While these projects offer promising solutions, the plants won’t be ready immediately. For instance, Microsoft’s Three Mile Island nuclear plant isn’t expected to be operational for at least four years. In China and South Korea, constructing and commissioning a new nuclear power plant typically takes about five years, making them the “fastest” countries for these kinds of projects.

Concerns about the future. It’s unclear if current efforts will be sufficient. A recent Goldman Sachs study revealed that energy demand from data centers in the U.S. is expected to triple between 2023 and 2030. This will require an additional 47 GW of energy. While companies will meet some of this demand with renewable sources, they’ll likely need significant public and private investment to address the shortfall.

A resurgence of interest in nuclear power. After years of negative perceptions and a bleak outlook for nuclear energy, many countries are currently revitalizing projects to develop more nuclear power plants. Notably, the U.S. aims to activate 300 new next-generation reactors by 2050. The increasing energy demands driven by the growth of generative artificial intelligence have sparked a newfound interest in nuclear energy among major tech companies.

Energy demand changes Changes in energy demand by region, 2022-2026. | Image: International Energy Agency

Global electricity consumption predictions. According to the International Energy Agency, global electricity consumption is expected to double from an estimated 460 TWh in 2022 to more than 1,000 TWh by 2026. This amount of electricity is roughly equivalent to Japan’s annual consumption.

It could get worse. However, this prediction could be conservative. If generative AI becomes as popular as some analysts suggest, demand may increase even further. In August, Reuters reported that OpenAI’s ChatGPT had 200 million weekly active users, a significant rise compared to the 100 million users in November 2022. If user engagement continues to soar, the energy requirements of the data centers supporting these platforms will also significantly rise.

Criticism still surrounds nuclear power. The debate continues, primarily fueled by activists emphasizing the environmental and safety risks of nuclear power plants. On its website, Greenpeace says, “Nuclear power is incredibly expensive, hazardous, and slow to build.”

Concerns about competition with renewable energy sources. Critics of nuclear energy don’t view it as a valid alternative to renewables. In reality, the competition between these two energy sources is evident. The support that renewables have received from the government in recent years contrasts sharply with the current push for nuclear energy, particularly from the private sector and public opinion.

This could lead to a potential imbalance: If new nuclear power plants replace cheaper renewable sources, electricity prices may increase.

Image | Jason Mavrommatis

Related | Google Is Following in Microsoft’s Footsteps: It’s Exploring the Use of Modular Nuclear Reactors for Its AI Data Centers

Home o Index