Good, Better, Best: How Sustainable Should Computing Be?

It has been quite a year. Increasing numbers of uncontrolled wildfires and extreme weather events have inspired new awareness and activism around climate change. This awareness has reached a broad range of computing communities: software and system developers, cloud operators, researchers and academics, policymakers, and increasingly business executives. The awakening of the community is evident in common questions: What is the problem? What can I do? Why can’t we do that? And there have been especially vibrant debates around AI/ML’s growing environmental impact.

As financial pressures on climate risk and reputation grow, computing industry executives have made new commitments to reduce carbon emissions and impact.

As financial pressures on climate risk and reputation grow, computing industry executives have made new commitments to reduce carbon emissions and impact. They are responding to pressure and expectations of investors, peers, and customers. How much should we demand?

In Western capitalism corporations do “what they must” to secure their current and future profits. It falls to our awareness, activism, and engagement to drive the corporate profit calculus. I believe we should expect them to do heroic things—far more than the corporations “think” they can do, or “know how or can price” to do. This is essential to overcome business conservatism. And we can drive that corporate behavior with who we choose to buy from, work for, and respect. And frankly, who we consider to be a climate “greenwasher” or worse “a climate destroyer.” So, my answer is that we should set the bar high and expect computing and particularly tech companies reaping billions in annual profits to not just mitigate computing’s own damage, but to drive progress on climate change broadly and aggressively. Consider the spectrum depicted in the figure here:

Figure. Models for computing’s sustainability

GOOD. Carbon Neutral—companies buy renewable via long-term PPAs. This is perhaps not a sacrifice (many PPAs will save cloud companies money). This is offsetting (like buying trees). But yes, a Good commitment.

BETTER. 24×7 hourly matching—a more ambitious goal.^a Only one cloud provider has made this commitment. This goal is match generation of renewables to datacenter power consumption synchronously. 24×7 can still be negative for grid renewable absorption and renewable fraction. So, it reflects a reduction of the environmental damage of datacenter operation. A Better commitment.

BEST. Datacenters as flexible loads. Creating a negative grid carbon footprint. By flexing loads at large-scale both temporally and geographically at dynamic ranges of 50% or even 100%, in cooperation with the grid would enable increased renewable absorption. Power grids need this help badly to achieve the radical renewable energy goals proposed for the coming decade(s). This dispatchable datacenter load approach has been studied extensively and is of growing commercial interest.^b While promising, this approach presents significant research challenges for flexible workloads, computing systems and hardware, that we should embrace.^c

The grid faces major challenges to decarbonize, and we can and should help accelerate the process.

I propose a call to action—let’s help the grid decarbonize, shaping computing activity to accelerate the addition of renewables and more important, the effective absorption of such renewable generation! The grid faces major challenges to decarbonize, and we can and should help accelerate the process. This means research. And change in practice and goals of computing companies—because they can (have technology and resources), they should win business (greener services), and they have a vested interest to ensure their economic future.

In 2022 and beyond, let’s challenge and hold ourselves and computing companies to the highest standard of “best,” driving to support the accelerated decarbonization of the power grid.

Andrew A. Chien, EDITOR-IN-CHIEF