Computer Science expert calls for stemming rising energy costs in supercomputing

Prof. Jack Dongarra, a leading U.S. Computer Scientist, has called for innovation to stem the rising energy costs in supercomputing for sustainability.

He also cautioned that the growing demand for faster supercomputers is driving energy consumption to unsustainable levels.

Dongarra, a Turing Award winner and professor at the University of Tennessee, Knoxville, stated this on Thursday at the 12th Heidelberg Laureate forum in Germany.

He said that the world’s most advanced machines, while indispensable for science, consume electricity at the scale of small cities.

“The fastest computer today consumes about 30 megawatts, which translates to around 30 million U.S. dollars annually in energy costs.

“Another system dedicated to artificial intelligence in Tennessee uses 300 megawatts; ten times more,” Dongarra said.

According to him, the machines are vital for forecasting weather, simulating climate change, aiding drug discovery and advancing artificial intelligence.

“Yet their cost and power demands are raising questions about sustainability.

“These are sophisticated scientific instruments, like telescopes in astronomy. But to continue advancing, we must design systems that are not only faster but also more energy-efficient,” Dongarra said.

He explained that supercomputers are typically built with millions of processors, but efficiency is low, with many applications achieving less than 10 per cent of theoretical peak performance.

This gap, he noted, adds pressure on researchers and engineers to innovate.

Dongarra said companies such as Google, Amazon and Microsoft are already investing in custom-designed chips to improve efficiency, while China has shifted to developing indigenous processors following U.S. technology export restrictions.

Despite these challenges, he said supercomputing remains indispensable, “It allows us to predict, to simulate, and to prepare for the future.

“Energy is the constraint, but the science it enables is too important to ignore.”