It's very easy to find articles (online of course) warning about how our Internet addiction is killing the planet through the phenomenal energy consumption of data centers. Whether it is watching streaming services, playing games online, or processing astronomical observations, the consumption of anything digital is described as a major contributor to global warming, aside from the minority of renewably powered centers. There's just one problem with these reports – they didn't check the data.
Northwestern University's Professor Eric Masanet, on the other hand, did. He found these reports are based on looking at old records of the amount of energy consumed, and factoring in the enormous growth in processing. Often this gets projected into the future, where our data thirst looks sure to continue. These reports ignore one crucial fact, however: data processing has become much more efficient.
In Science, Masanet and co-authors note: “Since 2010, electricity use per computation of a typical volume server – the workhorse of the data center – has dropped by a factor of four, largely owing to processor efficiency improvements and reductions in idle power.” Storage efficiency has improved even faster. Computations increased 550 percent between 2010 and 2018, but energy consumption grew by just 6 percent.
Data centers aren't off the hook entirely. Masanet calculates they are using 1 percent of the world's electricity – more than most countries. This makes it essential they switch to non-polluting sources as quickly as possible, something scientists whose work requires the use of supercomputers are pushing to make happen.
The other side of the equation is whether these efficiency gains can continue to roughly match increased demand. "While the historical efficiency progress made by data centers is remarkable, our findings do not mean that the IT industry and policymakers can rest on their laurels," Masanet said in a statement. "We think there is enough remaining efficiency potential to last several more years. But ever-growing demand for data means that everyone – including policy makers, data center operators, equipment manufacturers and data consumers – must intensify efforts to avoid a possible sharp rise in energy use later this decade."
Masanet and colleagues combined information not just on the amount of data processed and the computers involved but storage devices and cooling systems. It's the first “bottom-up” estimate allowing for regional differences and categories of center since 2010.
Masanat called for better monitoring of energy consumption so more accurate estimates can be made in future. The authors believe raising awareness of industry-leading efficiency practices will help improve the standards overall. They also call for more research into more efficient computing and technologies for heat removal, and identification of any centers that are lagging behind the rest of the field.
