Why Software Developers Should Be Thinking about the Climate

Why should we as software developers and engineers care about our impact on climate change? That was the core question in this episode of The New Stack Makers, a conversation between Alex Williams, founder and publisher of The New Stack and myself, Charles Humble, an IT industry veteran who writes regularly for us and acts as an adviser to a number of start-ups.

The starting point for building software in a more sustainable way, I suggested, is through better operations, and this has some tangible business benefits. For one thing, it saves you money.

Getting better at cloud Ops is also better from a security point of view, since if you have machines running in your own data centers that aren’t being actively used anymore, they’re unlikely to be patched and represent a security risk.

We have already made progress with combating climate change, I said in this episode, and there are grounds to be cautiously optimistic. As an example, I cited acid rain, which was mostly caused by sulfur dioxide, a gas that is produced when we burn coal. It had severe effects on ecosystems and we solved it. What happened was that government officials signed international agreements, placed emissions limits on power plants and started to reduce coal burning.

Interventions were incredibly effective. In Europe, sulfur dioxide emissions fell by 84% and in the U.S. by 90%. Some countries have reduced them by more than 98%.

We did something similar with the ozone hole — after countries signed the Montreal Protocol in 1987, emissions of ozone-depleting substances fell by more than 99%.

So there is reason to think we can solve the issues we currently face with greenhouse gasses: primarily for the IT industry, carbon. And while we may get less attention than an industry like aviation, we are, via our hardware and data centers, a major contributor to the problem.

Data Centers and AI

While it’s true that since 2010, emissions from data centers and network traffic have grown only modestly, data centers remain significant consumers of electricity. Global data center electricity use in 2021 was 220 to 320 terawatt-hours, and if we factor in data transmission networks, we can add a further 260 terawatt-hours.

For context, this makes the upper estimate of our industry’s electricity consumption similar to a country like Brazil, and much larger than a country like the U.K., which used 331.3TWh in the same period.

What’s more, while real numbers are hard to come by, training and running AI models is extremely power intensive so likely to exacerbate the issues we already have. It will also, Williams suggested, accelerate other scarcity problems such as the availability of GPUs and water.

I talked about the Maturity Matrix from the Green Software Foundation, which has two stated goals:

• 24/7 carbon-free electricity in data centers.
• 10-year life expectancy on all end-user devices.

In this episode of Makers, we spent some time discussing the validity of these goals, how we might be able to achieve them, and the role of regulation.

Listen to the whole episode for more, and check out these resources recommended in it:

• “The Cloud Native Attitude,” which I co-wrote with Anne Currie.
• “Building Green Software” forthcoming on O’Reilly by Currie, Sarah Hsu and Sara Bergman
• “How to Avoid a Climate Disaster” by Bill Gates
• “Not the End of the World” by Hannah Ritchie
• “Building Green Software,” Currie’s public training course.