Sign up for daily news updates from CleanTechnica on email. Or follow us on Google News!
Elon Musk has announced that he is building the world’s largest supercomputer. The problem is that it will require the same amount of electricity as would power 100,000 homes. Musk’s potential supercomputer power usage has many regional community members in the southwest Memphis area concerned, who say the artificial intelligence data center is compounding their pollution burden and adding stress on the local electrical grid.
AI has brought impressive results and multiple practical uses which have already changed society. Silva and researchers from Portugal’s Center for Power and Energy Systems explain that such high-performance computing relies on performance-oriented infrastructures with access to powerful computing resources. Otherwise, they aren’t able to complete tasks that contribute to solving complex problems in society.
So, what’s the answer? Clearly, decarbonization of supercomputing centers is key to improving their environmental performance — and it would also accentuate companies’ financial pictures, too.
Why does a supercomputer demand so much electricity? Supercomputing clearly has an intensive use of resources alongside the increase in service demand. That demand is due to emerging fields of science, combined with climate change concerns, rising energy costs, and an exascale paradigm. (Exascale computing is a type of ultra-powerful supercomputing, with systems performing billions of computations per second utilizing an infrastructure of CPUs and GPUs to process and analyze data.)
What’s Musk’s background with supercomputers? Musk started XAI as a competitor to ChatGPT, the chatbot powered by OpenAI, which he helped found but walked away from in 2018 after disagreements with other co-founders.
Why are these Memphis-area residents complaining? Residents of the heavily industrial community already have to deal with an oil refinery, a steel mill, and chemical plants. They explain that their area is already a cloud of smog, and it’s been made worse by Musk’s xAI data center.
What’s the background of Musk’s xAI data center? It sits in an old manufacturing plant on more than 550 acres. Before beginning operations there in July, xAI rolled in flatbed trucks loaded with almost 20 mobile power plants, fueled by natural gas, to help meet its electricity demands. As described by the New York Times, xAI is using the Memphis facility to develop its artificial intelligence models on a network of thousands of high-powered computer servers. Some of its models are trained on data from Musk’s social media platform, X. “We’re getting more and more days a year where it is unhealthy for us to go outside,” KeShaun Pearson, president of Memphis Community Against Pollution and a lifelong resident of the area near the xAI site, told the Times.
How does xAI define its mission? The company’s website describes xAI as “a company working on building artificial intelligence to accelerate human scientific discovery. We are guided by our mission to advance our collective understanding of the universe.”
What methods has the xAI team generated so far? The team has contributed numerous methods in the field — the Adam optimizer, Batch Normalization, Layer Normalization, and the discovery of adversarial examples. They also list “innovative techniques and analyses” such as Transformer-XL, Autoformalization, the Memorizing Transformer, Batch Size Scaling, μTransfer, and SimCLR. They say they “have worked on and led the development of some of the largest breakthroughs in the field,” including AlphaStar, AlphaCode, Inception, Minerva, GPT-3.5, and GPT-4. They partner with X Corp to bring the technology to more than 500 million users of the X app.
Is all supercomputer power usage the same? A recent study shows that supercomputer power usage varies considerably and “is always significantly below the peak provisioned power.” Several factors cause this — the machine may not be fully utilized, applications’ computational characteristics are not those which maximize power usage, and/or applications can be waiting on resources external to the node.
What happens if a supercomputer center uses less power than anticipated? Coming in below peak power translates into significant one-time fixed costs, such as the installation of a greater amount of power distribution or cooling infrastructure, as well as ongoing costs, to purchase electricity. In addition, the rise of energy consumption by the HPC community and large machine learning models in industry has environmental impacts, e.g. carbon emission, which need to be taken into consideration as well. In short, power demands in HPC have become a key limiting factor in both current exascale system operations and the design of future supercomputers.
Is xAI adding additional power features to the existing infrastructure? The scale of potential power consumption from the xAI supercomputer center means that the plan required approval from the Tennessee Valley Authority, which operates the electric grid for much of the region. xAI is requesting an additional 100MW of power, for a total of 150MW of power, which will require several actions to be completed. Memphis Light, Gas, and Water assured the community in an online fact sheet that the company “is paying for all upgrades” involved in supplying electricity and that “there will be no impact to the reliability of availability of power to other customers from this electric load.” With the followings conditions met, MLGW insists “there will be no impact to other electric customers:”
Approval by the TVA Board of Directors to serve the total 150MW demand;
Enrollment in a Demand Response (DR) program in which the company agrees to reduce its electric consumption when conditions warrant it to preserve power for existing customers;
$1.7M of improvements to a 161kV transmission line (per the MLGW electric system impact study); and,
Construction of a new substation adjacent to the facility.
How is MLGW pitching the extra power needed to its current customers? Calling it “value for the MLGW customers, they’re saying that increased electric sales will result in:
~$500K of additional PILOT payments annually to the City of Memphis;
~300+ new, high paying jobs and will reduce demand on the aquifer;
AI joining in the design and proposed construction of a greywater/recycled wastewater facility that will take in treated wastewater and produce water suitable for cooling processes;
MLGW and XAI coordinating procurement and early installation of more than 50MW of utility scale battery storage in the MLGW system starting in 2024.
While these are nice additions to the area, the environmental cost of supercomputing can depend hugely from where the energy to power the devices comes. That means Musk, xAI, and MLGW must turn to renewables for their power supplies. It’s imperative. The locals and the planet deserve it.
Chip in a few dollars a month to help support independent cleantech coverage that helps to accelerate the cleantech revolution!
Have a tip for CleanTechnica? Want to advertise? Want to suggest a guest for our CleanTech Talk podcast? Contact us here.
Sign up for our daily newsletter for 15 new cleantech stories a day. Or sign up for our weekly one if daily is too frequent.
Advertisement
CleanTechnica uses affiliate links. See our policy here.
CleanTechnica’s Comment Policy
Share this story!