Can We Afford to Keep Up with AI’s Growing Energy Demands?

This week is Climate Week NYC, and many businesses and leaders are participating to discuss how we can advance clean energy and net-zero technologies to meet our planetary climate goals.

It’s imperative that businesses and organizations work to reduce our collective climate impact. Initiatives such as Emissions First, ZEROGrid and organizations such as the Ceres, the International Organization for Standardization, and the Clean Energy Buyers Association have been pivotal in helping us take important steps forward. And I’m particularly proud that Akamai continues to invest in renewable energy and other projects to push us closer to our goal of achieving net-zero emissions across our platform by 2030.?

Despite all this, there is much more that needs to be done.?And of course, one of the big goals of Climate Week is to have these discussions.

Specifically, for those of us in the technology industry, I think we need to have a serious conversation about AI and its sustainability implications. Tools driven by generative AI are particularly energy-intensive, with the data centers powering these technologies already accounting for 1% of global electricity use. AI’s demand is expected to grow tenfold by 2026, and The World Economic Forum is projecting a 26-36% annual growth rate in AI power demand.

Are the benefits of AI — generative AI in particular — worth the energy trade-offs? What do we really gain by being able to ask a chat bot to draw us pictures of cats? Are other tools we have at our disposal better able to solve these problems??

These are some of the questions we should be asking. And the answer isn’t always straightforward. It’s clear large language models (LLMs) are power-hungry. And there are many cases where more energy efficient algorithms solve our problems just as well (although they may take a little more work to build than simply leveraging a pre-built LLM).

On the flip side, AI and machine learning algorithms in some ways have also been a boon for energy efficiency. For example, Google’s DeepMind has leveraged AI to achieve a 40% reduction in energy used for cooling its data centers. By implementing similar AI-driven efficiency initiatives, the industry can significantly enhance its contribution to a sustainable energy future. Moreover, generative AI offers a promising avenue for helping researchers and scientists find new ways to address the growing climate crisis.?

So, is the use of AI going to increase our carbon footprint, or will it help us meet our climate objectives??

The answer likely lies somewhere in between. The trick will be sorting through the AI hype and ensuring we don’t start using AI for the sake of using AI. Implementing AI where it is the best tool for the job will give us the strongest shot at balancing innovation and efficiency.

Which is why I think it’s imperative that companies developing and deploying AI systems consider how such use cases may impact the environment and slow down efforts to combat the climate crisis. Raising stakeholder awareness of AI's environmental impact and your company's responsibility in addressing it is a crucial first step to driving action.?

So, as we celebrate Climate Week NYC, I hope we continue to think seriously about the impact of AI and how we can balance innovation with our climate future.