Meta and Amazon Web Services continue to search for ways to meet their data centers’ growing power demand, requesting proposals for nuclear reactor construction and announcing new efficiency measures.
Meta said Dec. 3 it wants to add 1 GW to 4 GW of new U.S. nuclear generation capacity by the early 2030s to help meet its AI innovation goals and sustainability objectives. It said it is taking an open approach with its RFP so it can partner with others in the industry to bring new nuclear generation online.
AWS said Dec. 2 it has designed new data center components to support innovation with artificial intelligence and boost the energy efficiency of its facilities. It said this simultaneously will support the next wave of generative AI, increase computing power 12% and improve the availability and efficiency of the data centers.
Meta’s announcement is Big Tech’s latest embrace of nuclear power, which holds the potential to supply large amounts of baseload emissions-free electricity — if new reactors can be built quickly, affordably and in large numbers.
Microsoft, Google and Amazon earlier in 2024 announced deals to run their facilities on nuclear power. In November, media outlets were abuzz about a report that Meta’s plan to build an AI data center next to an existing nuclear plant was thwarted by the presence on-site of a population of rare bees that could be disrupted by the construction.
So Meta is looking elsewhere to meet its parallel goals of reducing its carbon footprint and increasing its computing power, an effort that already has yielded more than 12 GW of renewable energy contracts for its operations.
“Supporting the development of clean energy must continue to be a priority as electric grids expand to accommodate growing energy needs,” it said in its announcement. “At Meta, we believe nuclear energy will play a pivotal role in the transition to a cleaner, more reliable and diversified electric grid.”
Meta explained it is engaging projects earlier in the process because nuclear generation is more expensive, takes longer to build, faces more regulatory oversight and has a longer operating lifespan than other generation technologies.
It said: “We are looking to identify developers that can help accelerate the availability of new nuclear generators and create sufficient scale to achieve material cost reductions by deploying multiple units, both to provide for Meta’s future energy needs and to advance broader industry decarbonization.”
The growth of power-intensive AI and the data centers in which it exists has been presented as a seismic change, and one the U.S. power industry is not prepared to meet.
In the past several months, for example, Goldman Sachs predicted a 160% increase in data center demand by 2030. EPRI predicted data center demand could more than double to as much as 9% of U.S. electricity generation by 2030. The U.S. Department of Energy predicted total U.S. demand could grow 15 to 20% in the next decade. S&P Global predicted a need for 50 GW of new generation capacity by 2030, with accompanying upgrades in transmission — total cost $75 billion.
Not everyone is convinced the increase in electric demand from data centers will be so steep, however — the sector may not grow as expected, or technology improvements could reduce the power consumption of the hardware.
This latter scenario is the focus of the AWS initiative.
The new data center components announced Dec. 2 incorporate improvements in power, cooling and hardware design. They will be used in new U.S. data centers starting in early 2025; some existing facilities already have been retrofitted.
The upgrades include:
-
- Simplified electrical and mechanical designs reduce the required number of conversion and distribution processes, each of which is a point of inefficiency, energy loss and potential failure.
-
- Backup power is moved closer to the server racks, reducing the number of cooling fans needed.
-
- Novel liquid-to-chip mechanical cooling solutions are integrated with air cooling systems to maximize performance and efficiency while minimizing cost.
-
- AI is used to predict the most efficient way to position racks, reducing the amount of power that is stranded, unused or underused.
-
- In-house innovations in power delivery are expected to yield a 6X increase in rack power density within two years and an additional 3X increase further in the future.
-
- Telemetry tools provide real-time diagnostics and troubleshooting to optimize operating conditions.
Prasad Kalyanaraman, vice president of infrastructure services at AWS, said in the news release: “These data center capabilities represent an important step forward with increased energy efficiency and flexible support for emerging workloads. But what is even more exciting is that they are designed to be modular, so that we are able to retrofit our existing infrastructure for liquid cooling and energy efficiency to power generative AI applications and lower our carbon footprint.”