With a growing demand for data centers to power AI technology, tech giants such as Microsoft, Google, and Amazon have struck deals with nuclear power plant operators and developers. Through this, the companies are also looking to fulfill their carbon-negative commitment.
Recently, Google revealed that it had signed “the world’s first corporate agreement to purchase nuclear energy” from SMRs developed by California-based Kairos Power.
Google CEO Sundar Pichai announced on X, “It’s the latest step in our history of accelerating clean energy sources and will help support AI investments.”
Besides, Microsoft has also agreed to pay an energy company to revive Constellation Energy’s shuttered Three Mile Island nuclear power plant in Pennsylvania. “This agreement is a major milestone in Microsoft’s efforts to help decarbonise the grid in support of our commitment to become carbon negative,” Bobby Hollis, vice president of energy at Microsoft, said in a statement.
Even Amazon has joined the race for nuclear power by signing three agreements to develop nuclear energy projects. This includes deals with public utility conglomerate Energy Northwest, developer X Energy to build multiple SMRs in the region and an agreement with Dominion Energy in Virginia.
“Nuclear is a safe source of carbon-free energy that can help power our operations and meet the growing demands of our customers while helping us progress toward our Climate Pledge commitment to be net-zero carbon across our operations by 2040,” Amazon Web Services CEO Matt Garman said in a statement announcing the deals.
#Nuclear energy stocks are soaring, fueled by rising demand for AI data centers!
With #Amazon and #Google signing major power deals, developers like #Oklo and #NuScale are seeing share prices skyrocket. pic.twitter.com/uvKHBvF8ye
— Radioactive Friends (@RadioactiveFrnd) October 21, 2024
Goldman Sachs Research estimates that data center power demand will grow 160% by 2030. Additionally, the International Energy Agency (IEA) projects global electricity consumption from data centers and AI will double from 460TWh in 2022 to over 1,000TWh by 2026.
However, analysts also point to the supply chain constraints that might come up with Nuclear energy sources. The U.S. government has banned uranium imports following Russia’s 2022 full-scale invasion of Ukraine. It is also probing whether China is reported buying Russia’s nuclear power and exporting its own production to the U.S.
As the energy demands of data centres, which power every critical digital infrastructure and technologies like generative AI, 5G, Io, etc., these reactors or SMRs are a critical move towards clean and sustainable data centres powered by safe nuclear energy.
“A normal data centre needs 32 megawatts of power flowing into the building. For an AI data centre, it’s 80 megawatts,” says Chris Sharp, CTO at Digital Realty, a US data centre giant.
Oklo’s reactors powered by nuclear fission energy stand as a viable option for data centres as they can generate 15MW each and can function for a minimum of 10 years before requiring refuelling.
Equinix intends to purchase power from Oklo’s upcoming SMR installations to fuel its US data centres. It will possess the first option for 36 months to acquire between 100MW and 500MW of cumulative capacity from specific Oklo powerhouses.
Additionally, smaller microreactors, with capacities ranging from 1 to 20 MW, are being developed specifically to power data centres and industrial sites. Startups like Oklo aim to deploy factory-built microreactors by 2028 to meet these energy demands.
Besides Oklo, several other US-based firms, including NuScale Power, Kairos Power, and X-energy, are actively developing small modular reactors (SMRs). Additionally, UK-based Rolls-Royce is also pursuing SMR technology.
Owing to the latest advancements in GPUs such as NVIDIA’s Blackwell and H100 Hopper series, innovations in AI and high-performance computing (HPC) are reaching new heights. Big tech companies like Oracle, AWS, and Google are readily adopting these GPUs to empower their AI infrastructure by making it faster and more efficient.
For instance, Oracle has announced plans to deploy a supercluster featuring 130,000 NVIDIA Blackwell GPUs to power advanced AI workloads. This far exceeds the capacity of traditional power grids and renewable energy sources alone.
Oracle is building new data centers that will house such colossal infrastructure, which will include power from three SMRs. These reactors, with a capacity of up to 1 gigawatt, are designed to meet the immense energy demands of the data centers.
How SMRs Could Save Money and Energy
Small modular reactors (SMRs) are a type of nuclear reactor that can produce up to 300 MWs of electricity. They are smaller than conventional reactors and can be built in factories and transported to the site. They also have a subcategory called microreactors, which can produce up to 10 MWs of electricity. These reactors could be ideal for Microsoft, as their data centres have similar power requirements.
SMRs have several advantages over conventional reactors. They are modular, which means they can be assembled and installed quickly and easily. They are said to be cost-effective, as they require less capital investment and maintenance. A 300 MW SMR costs about $900 million to $1 billion to build, while Microsoft pays about $7-8 million in electricity bills for each data centre per year. By switching to nuclear power, Microsoft can save money and reduce its dependence on fossil fuels.
According to Oracle founder Larry Ellison, the company plans to build a gigawatt-scale data center powered by three small nuclear reactors (SMRs).
It is fulfilling its strategy of obtaining unprecedented processing power, supporting zettascale computing with NVIDIA’s latest GPUs, while reducing its carbon footprint. On October 8, OpenAI announced via X that its team received the first engineering builds of Nvidia’s DGX B200. These new builds promise three times faster training speeds and fifteen times greater inference performance than previous models.
While GPUs are fuelling massive growth for NVIDIA thanks to the burgeoning demand from AI companies, the costs incurred by these companies don’t end with just GPUs. The energy cost of running these GPUs in data centres is enormous. Recently, a study showed that data centres consumed approximately 1,000 kWh per square metre, which is about 10x the power consumption of a typical American household. BLOOM, an LLM, utilised 914kWh over an 18-day period while running on 16 NVIDIA A100 40GB GPUs, managing an average of 558 requests/hour.
Facebook co-founder Dustin Moskovitz, OpenAI CEO Sam Altman, and Peter Thiel’s Mithril Capital have invested in Helion Energy, a Washington-based nuclear research company.
After the funding round, Sam Altman enthusiastically spoke about Helion in his blog. “Helion has a clear path to net electricity by 2024 and has a long-term goal of delivering electricity for 1 cent per kilowatt-hour,” he noted.
Way Out of SMRs
Recently, Vinod Khosla, in his article titled AI: Dystopia or Utopia? Summary mentioned, “My bet is on fusion boilers to retrofit and replace coal and natural gas boilers rather than building whole new fusion or nuclear plants. There are additionally promising efforts using geothermal, solar and advanced battery systems for clean, dispatchable electric power. Multiple vectors are driving down the environmental cost of compute.”
He further adds that advancements in algorithmic efficiency and hardware innovation enable AI systems to deliver greater performance while using considerably less power.
“New techniques and the integration of web search functions are helping AI scale more effectively without drastically increasing energy consumption. This push for optimized compute not only supports the growing energy demands of AI but also ensures that this technology can expand sustainably without straining global infrastructure,” he further added.
“When you look at a data center, in my view, you would still think about the mix of renewable and potentially gas,” says Som Shantanu, Asia Engineering Leader, Gas Power, GE Vernova.
He goes on to say that the main challenge with nuclear energy lies in its turnaround time. If a nuclear project is initiated, the process is long-term and may take several years.Additionally, nuclear plants often lack the flexibility to adjust their output as rapidly as gas or renewable energy sources. As a result, gas and renewables look more advantageous to nuclear energy.
The post Tech Giants Turn to Nuclear Energy to Power AI Technology appeared first on AIM.