Why Supercomputers Are the Unsung Heroes of the Green Energy Transition

The sound is the first thing you notice when you pass a supercomputing facility. It’s a steady, low roar that serves as a reminder that something serious is going on behind the walls, not the soft hum of an office server.

There are rows of black cabinets inside that blink softly, resembling library shelves. It is difficult to believe that devices like these, which are heavy, power-hungry, and nearly out of style in the era of thin phones, are performing some of the most delicate tasks in the worldwide movement toward cleaner energy. Nevertheless, they are.

Solar farms baking in the Andalusian sun or wind turbines on Scottish coastlines are typically used to tell the tale of the green transition. It is rarely communicated through code. However, there is a simulation running somewhere on a machine the size of a small warehouse behind nearly every significant renewable energy decision made today. The engineers creating the next generation of offshore wind farms would be speculating without those simulations. Additionally, it is very costly to guess in this industry.

Consider the Barcelona Supercomputing Center, which is home to MareNostrum 5, one of the world’s most potent systems. The head of the Earth System Services Group there, Albert Soret, has discussed how supercomputers enable scientists to create what are effectively digital twins of the Earth: replicas of the atmosphere, oceans, ice sheets, and biosphere that react, for the most part, in the same way as the actual planet. This type of modeling has the potential to influence climate policy more than any one international summit. Speaking with experts in the field gives me the impression that science is at last keeping up with the urgency.

The ambitious decarbonization goals set by the European Green Deal are essentially wishful thinking in the absence of modeling. Researchers must test how a new solid-state battery will deteriorate over a ten-year period, how a city’s grid will react when half of its rooftops suddenly become solar producers, and how a wind farm will react in turbulent coastal weather. A laptop cannot provide an answer to any of these questions. Thousands of processors must operate in parallel to break up the problem and reassemble the results in a manner that is nearly real-time.

The unequal distribution of this power is intriguing. The US Department of Energy’s Summit machine at Oak Ridge is operating at 200 quadrillion calculations per second, Europe is making significant investments, and initiatives like RISC2 are attempting to bring Latin American scientists into the same dialogue. Even excellent energy researchers, however, have never taken HPC seriously. There is a real expertise gap. Whether training programs will close it quickly enough is still up in the air.

Investors appear to have faith in the hardware aspects of clean energy, such as the EVs, turbines, and panels. Due in part to its poor photography, the software and simulation layer receives less attention. A magazine cover featuring a row of humming cabinets is not appealing. However, as this has developed over the past few years, it’s difficult to ignore the fact that some of the most important work in the energy story is being done by the people operating these machines in silence. There were once those who doubted Tesla. Wind power also did. In some respects, supercomputing in climate science seems to be in a similar stage of development—credible, accelerating, and still underappreciated.

These odd, noisy rooms will play a major role in the construction of the energy transition over the next ten years. They won’t be seen by most people. That seems to be about correct.