Externalities, energy, and the internet

Data center technology processing passed the milestone of consuming 1% of world energy in 2010 and is projected to increase to 6% by 2030, according to a 2020 Science magazine study  by Eric Masanet, Arman Shehabi, Nuoa Lei, Sarah Smith, and Jonathan Koomey. The authors suggest that governments may need to take on a more considered approach to expanding data centers to meet the growing demand. State and local economic developers offering incentives to recruit data centers may want to take particular note. The non-captured costs, or externalities, of information processing hubs are mounting as the pace of human-induced climate change increases.    

According to The MIT Press Reader, the cloud now has a greater carbon footprint than the airline industry. At 200 terawatt hours (TWh) annually, all existing data centers combined  devour more energy than even some nation-states.

Electricity use and associated carbon emissions are just two external concerns. A Washington Post article reports a large data center can gobble up anywhere between one million and five million gallons of water daily—as much as a town of 10,000 to 50,000 people. The Post also reveals that, at least within America, these data centers can contribute to catastrophes like drought. Environmental pressure from data centers in water-stressed areas such as  Arizona might be expected, but water wars are also beginning in areas receiving above-average rainfall because of climate change, like central Ohio, which is anticipating multiple energy-water hits from multiple proposed data centers and the mega Intel manufacturing complex under development.

The Science magazine writers argue that many policymakers have not been able to make comprehensive adjustments because they’ve not had a suitable assessment of the full scope of data center energy and water use because of a lack of analysis from the bottom up. This lack of information causes a potential error that the researchers try to combat by taking data from both national studies and the latest fully replicable bottom-up study from a decade ago.

The article reports that between 2010 and 2018, global data center workloads and computing instances had increased more than sixfold, whereas data center internet protocol (IP) traffic had increased by more than tenfold, meaning that over the years, service demands had risen significantly. In contrast, since 2010, electricity use per computation in a server has decreased to one-fourth of what it was, largely owing to processor efficiency improvements and reductions in idle power. Simultaneously, the use of watts per terabyte of storage has been reduced by a factor of nine from gains in storage-drive density and efficiency.

In short, researchers found that data centers were able to keep up with growing demand and maintain their significant share of global electricity use by increasing their Power Usage Effectiveness (PUE): Total amount of energy used in data center/total energy used by IT equipment. According to the Science authors, more energy-efficient port technologies brought a 10-fold increase in data center traffic with only modest increases in network energy use.

However, the authors are still determining whether this rise in efficiency over the last decade will be able to last. They predicted that the next doubling of global data center compute instances may occur within 2023 or 2024. And though resources are present to absorb this doubling of compute instances, it’s noted that policymakers can still help by promoting already existing efficiency standards such as Energy Star for servers, storage, and network devices.

Second, the authors suggest investing in new technologies to manage future energy demand growth in the simplest manner possible once current efficiency trends reach their limits. Third, to be prepared for the increase in energy demand for data centers, much greater public information is required for understanding and monitoring data center energy use and its drivers and for designing and evaluating effective policies. The authors concluded that more public reporting from large data centers should be incentivized, but more research was needed to produce policy-relevant data for best practices.