Artificial intelligence and nuclear energy.. A promising partnership or an existential threat

 

Over the past years, we have witnessed a radical shift in the relationship of technology giants with traditional energy sources. With the increasing demand for energy to power huge data centers that support artificial intelligence, major technology companies such as Google, Amazon, and Microsoft have turned to exploring more sustainable and efficient alternatives.

After decades of abandoning nuclear power due to safety concerns and high costs, this technology has come to the fore again, driven by new innovations in the field of small nuclear reactors (SMRs), which are safer and more flexible than conventional reactors.

But what makes nuclear energy attractive to these giant companies and not renewable energy such as wind or solar energy, and what small nuclear reactors, and what challenges the United States faces in developing them, and will these reactors push innovation and sustainability in the field of artificial intelligence or are we on the verge of a new era of disasters such as Chernobyl, Fukushima and(Three Mile Island), which are still etched in memory

First, which companies have started investing in nuclear energy projects

Google, Amazon, and Microsoft have announced huge investments in nuclear energy projects, and these investments are aimed at building new nuclear power plants or restarting old ones. These competing companies are the three largest companies providing cloud computing solutions in the world, and they were also among the largest companies that adopted artificial intelligence models and technologies and presented them to customers, both other companies and users, so it is no coincidence that these companies are among the largest investors in this field.

Here's a glimpse at the most important nuclear AI projects announced so far:

1. Google:

Google has signed an agreement with Kiros Power, a startup specializing in nuclear energy, to build 7 small nuclear reactors to power its data centers. These reactors are a promising solution for providing clean and reliable energy, as they are characterized by their simple design and advanced safety systems.

The partnership is expected to provide about 500 megawatts of carbon-free energy by 2035, supporting Google's goal of achieving carbon neutrality.

2. Microsoft:

Microsoft has announced plans to restart the closed Three Mile Island nuclear power plant in Pennsylvania – the site of the worst nuclear accident in the United States – in cooperation with Constellation Energy by 2028, and this project aims to provide about 835 megawatts of energy to meet the needs of Microsoft data centers. Never before has an American nuclear plant returned to service after being decommissioned, and not every production of one commercial nuclear power plant has been dedicated to one customer before either!

Microsoft did not stop there, but signed an agreement with Helion Energy to explore the possibilities of nuclear fusion energy as a future source of energy, and fusion energy is one of the most promising technologies in the field of nuclear energy, as it provides huge amounts of clean energy without producing dangerous nuclear waste.

In a parallel step, some reports earlier this year indicated that Microsoft and OpenAI are secretly developing an unprecedented supercomputer that uses millions of server chips to run artificial intelligence, and it may cost up to 100 billion dollars, codenamed (Stargate), but it will require up to 5 gigawatts of power to run it, equivalent to about what a big city like New York consumes, and all for just one computer!

3. Amazon:

Amazon announced last October the signing of 3 agreements to develop small nuclear reactors in the United States, as Amazon is working with Energy Northwest to develop four small nuclear reactors, which are expected to generate 320 MW in the first stage, with the possibility of increasing to 960 MW.

In addition, Amazon is exploring the possibility of developing small nuclear reactors with Dominion Energy in Virginia, adding at least 300 megawatts to meet growing energy demand. Amazon's current deal with Talen Energy also includes a USD 650 million investment in a data center in Pennsylvania powered directly by nuclear energy.

And then all these huge investments make us wonder about small nuclear reactors and how they differ from conventional reactors

Secondly, what are small nuclear reactors

These reactors are also flexible and adaptable, as they can be manufactured at the factory and transported as modules that can be assembled at the reactor site, so they are an ideal alternative to conventional large nuclear reactors, which must be designed for certain sites, which sometimes leads to delays in construction.

Moreover, small nuclear reactors rely on a variety of cooling liquids, including light water, liquid metals, and molten salt, which provides high design flexibility. These reactors are characterized by their strong Inherent Safety systems, which rely on the natural cooling cycle of the core, which reduces dependence on human intervention and significantly improves safety.

The US Nuclear Energy Agency estimates that by 2035, the global market for small nuclear reactors could reach 21 GW of power.

Third, why specifically nuclear energy and not wind or solar energy

The data centers used to train AI models require huge amounts of electrical energy. As these models grow in size and complexity, energy demand increases dramatically. Currently, most data centers rely on electricity produced from fossil fuels, which contributes to an increase in carbon emissions.

An important question arises here: how much energy do large linguistic models consume

Training large artificial intelligence models consumes a lot of energy, for example, the GPT-3 model with 175 billion Teachers uses 1287 MW per hour, while the DeepMind model with 280 billion Teachers uses up to 1066 MW per hour, which is about 100 times the average energy used by an American household in one year, which confirms the significant environmental impact of this industry.

Companies are striving to achieve carbon neutrality by 2030 or 2035, but total dependence on renewable energy sources such as the sun and wind faces significant challenges, such as uneven production due to weather conditions, solar energy is not available at night, and wind energy is affected by weather fluctuations, which requires complex storage and management solutions to ensure continuity of operation. Nuclear energy is emerging as a promising solution to meet this growing demand, for several reasons:

High power: nuclear reactors can generate huge amounts of electrical energy in a relatively small space, making them ideal for meeting the needs of huge data centers. Reliability: Nuclear power has the advantage of being a stable and reliable source of energy, unlike volatile renewable energy sources such as the sun and wind. Low carbon emissions: Nuclear energy is a clean energy source, as it does not produce greenhouse gases or other pollutants, which contribute to achieving the goals of combating climate change. Fourth, what are the challenges facing the United States in the development of small nuclear reactors? The main challenge facing the United States in this industry is the construction of the first small nuclear reactor. According to the IAEA, there are only three small nuclear reactors operating in the world, two in China and Russia, the central geopolitical rivals of the United States, and there is also a test reactor operating in Japan. So far, American companies are working on developing the design of small nuclear reactors, and the United States now has about 70 designs for small reactors, but only one of them has received approval to start construction, the Kairos Power Project, which will be in Tennessee. The reason for this delay is that design approvals usually take several years due to the complexity of reactor designs and the risks of radioactive materials. Rafael Grossi, Director General of the International Atomic Energy Agency (IAEA), stressed the need to develop flexible regulatory systems capable of keeping pace with the rapid developments in the nuclear energy industry. He pointed out that lengthy regulatory procedures that take many years hinder the growth of this industry, and suggested that the nuclear industry should be inspired by the aviation sector, which is based on unified global standards for aircraft design and operation, which contribute to improving quality and reducing costs. In addition, small nuclear reactors need highly enriched nuclear fuel (HALEU), which has a higher enrichment ratio than the fuel currently used in conventional reactors, making it more efficient and capable of producing energy. HALEU fuel has a higher level of enrichment than 5% enriched uranium-235, the main fissile isotope that produces energy during the chain reaction, which is used by conventional nuclear reactors, according to the US Department of Energy. HALEU fuel requires 5 to 20% enrichment, which makes it more efficient and able to produce the most energy in smaller reactors, and also contributes to improving the performance of small reactors and increasing their operational life, as this fuel helps the sodium element used in the cooling process to last longer, and produces more energy using less fuel. But the problem here is that the production of this fuel on a commercial scale in the United States is still very limited, and the globally dominant exporter is Russia. The US Department of Energy estimates that the United States will need more than 40 metric tons of this fuel by the end of the current decade. To meet this growing demand, the US Department of Energy is developing innovative solutions to produce HALEU fuel by recycling used nuclear fuel from research reactors. These solutions will save very small amounts of this fuel in the near term, but still not enough to meet the future needs of the industry. Fifth, artificial intelligence and nuclear energy.. A promising partnership or an existential threat Artificial intelligence permeates various aspects of our daily lives, but nuclear power systems do not need it, as these systems rely on known and predictable physical principles and do not need the complexity of artificial intelligence at all. Regulatory Commission (NRC), treat this with extreme caution and prevent its integration into these systems.