Supercomputers are not so super for the planet.
Supercomputers are a double-edged sword. While they are essential for complex tasks such as weather forecasting, scientific research, drug discovery and powering AI technologies like ChatGPT, they come with a hefty environmental price tag. To put it into perspective, running a single query to ChatGPT can consume as much electricity as lighting a bulb for 20 minutes, which is ten times more than a typical Google search.
Let’s dive deeper into the environmental footprint of these computational giants and explore the measures being taken to mitigate their impacts.
Understanding the carbon footprint of supercomputers—are they all bad?
Supercomputers are tasked with massive calculations at breakneck speeds, making them huge energy consumers and, consequently, hefty carbon emitters. For some perspective, research from the University of Edinburgh points out that the top 196 out of 500 supercomputers in the world use three billion kilowatt-hours of energy annually—roughly what about 277,777 U.S. households use in a year. Take Japan’s Fugaku, for example—one of the most powerful supercomputers in the world—it gobbles up over 28 megawatts, matching the energy footprint of tens of thousands of homes.
A 2022 report by the International Energy Agency adds more context: data centers, which house many supercomputers, consume 460 terawatt-hours of electricity—about 2% of the global total. This consumption is expected to double by 2026 as demands from AI technology grow.
However, it’s not all grim news. The environmental impact of supercomputers varies widely depending on how they’re powered. For instance, the Dutch national supercomputing facility, SURF, runs entirely on renewable energy. This slashes its carbon emissions dramatically.
Additionally, the location of these supercomputers significantly influences their carbon footprints. In countries like Australia, where electricity is mostly generated from fossil fuels, running supercomputing tasks can produce roughly 70 times more carbon emissions than in Switzerland, which primarily uses hydropower.
Considering the full lifecycle emission
When we talk about the environmental impact of supercomputers, it’s crucial to consider their full lifecycle emissions—not just the energy they consume while in operation. This broad view includes everything from manufacturing the hardware and maintaining infrastructure to cooling the components and eventually disposing of them.
AI startup Hugging Face found that its large language model BLOOM generated 50 metric tons of CO2 emissions throughout its lifecycle, with half stemming from training and the other half from manufacturing and maintaining the required computing infrastructure.
Gage Hills, Assistant Professor of Electrical Engineering at Harvard, notes that the energy required to manufacture a computer chip can exceed the energy it uses over its ten-year lifespan. This is due to the energy-intensive processes used to create advanced, nanometer-scale features on chips—a major factor in the carbon footprint during the manufacturing phase.
Disposal of supercomputers: An overlooked issue
When supercomputers reach the end of their useful life, they become part of a growing problem: electronic waste (e-waste), which carries its own environmental risks. Supercomputers are packed with metals, plastics and rare earth elements, some of which can be downright harmful if not disposed of correctly. Older computer components may even contain toxic substances like mercury and lead, known to cause serious health issues such as cancer and skin disorders. If not handled properly, these materials can contaminate soil and water.
You might think, “Can’t we just recycle supercomputer components since they’re mostly metals and plastics?” While that sounds like a solution, it’s not so straightforward. Globally, only about 17% of e-waste is recycled. This low rate is largely due to the complexity of recycling electronic components from large items like supercomputers. These components are often not designed to be easily taken apart, which complicates recycling efforts and ramps up costs. Consequently, a significant amount of e-waste ends up in landfills, where these hazardous materials can leach into the environment, posing a continuous threat.
What is being done to make supercomputers more eco-friendly?
Given their crucial role in science and technology, ditching supercomputers isn’t really an option—and who’d want to stop using ChatGPT anyway? Instead, scientists and engineers are hard at work reducing their environmental toll.
This effort includes designing more energy-efficient hardware. For example, newer supercomputers like Setonix at the Pawsey Supercomputing Research Centre are 30 times more energy-efficient than older models, thanks to advanced cooling techniques and renewable energy sources.
Research is also underway to develop software that demands less power. One interesting finding is that running simulations a bit slower can save a significant amount of energy without greatly affecting performance.
Innovations in cooling are also making a big difference. Traditional air cooling is being replaced with more efficient water-based systems, which can handle up to 95% of the heat generated by supercomputers. This reduces the energy needed for cooling, significantly lowering their power usage effectiveness (PUE).
Additionally, powering these facilities with renewable energy sources, like solar, wind or hydroelectric power, is becoming more common. Notable examples include NASA’s Modular Supercomputing Facility and the Oak Ridge National Laboratory (ORNL)’s Frontier supercomputer, which integrate renewable energy into their operations.
Companies like Hewlett Packard Enterprise (HPE) are at the forefront of developing tools for real-time monitoring and management of power usage in supercomputing environments. These tools help adjust operations based on the current workload and environmental conditions, optimizing energy use.
Last but not least, while recycling supercomputers presents challenges due to their complex structures, it’s not an impossible task. For instance, in 2019, the Oak Ridge National Laboratory successfully recycled materials like plastic and metals from its Cray XK7 Titan supercomputer. This progress suggests a promising future for reducing the environmental impact of these powerful machines.
As our reliance on these powerful machines grows, it’s vital that we champion innovations aimed at reducing their environmental footprint. The path forward for supercomputing must prioritize sustainability—not merely to satisfy our digital demands but for the well-being of our planet too. Hence, making supercomputers more eco-friendly is essential, ensuring they help rather than hinder our environmental efforts.
Also read:
- Tech Giants Racing to Unleash the Power of Quantum Supercomputers
- 5 Uses of Supercomputers for Startups and Entrepreneurs
- What Is Quantum Computing? Startups Working on Quantum Computing Tech
Header Image from Freepik
