For the sake of the environment, the electricity consumption driven by the internet should not continue along its current trajectory. But it’s unlikely we all suddenly lose interest in accessing the world’s information easier and faster than ever before. Both modern economies and personal convenience dictate the availability and seamless flow of data. Therefore, humanity has to innovate its way to a better ratio of carbon emissions to computing requirements. We need new solutions in hardware, software, processes, and regulations that align with our environmental goals without sacrificing computing capacity. This is even more true as all companies, even technology laggards, begin to view technology as a horizonal shared service that enables their business.
Right now, we have data centers sprawled across six of seven continents consuming 7.4 gigawatts (GW) of power, a 55% increase from the 4.9 GW in 2022. This growth is fueled entirely by our exponentially increasing slope of “technology advancement,” such as increasingly data-reliant applications and enterprises, the adoption of live streaming services, a preponderance of IoT devices, and the much-loved youngest sibling on the scene, AI, with its voracious appetite for cloud-hosted services. Thanks to AI, analysts predict a 15% compound annual growth rate (CAGR) in data center power demand from 2023-2030 in the United States, while global data center power consumption could more than double by 2026, representing similar electricity consumption as the country of Japan.
Can an already struggling climate bear this surge in demand?
Fortunately, there are actionable paths forward to mitigate environmental impact without sacrificing growth:
Better data governance: As companies hoard an increasing volume of data, they are becoming savvier by adopting more stringent data governance policies. This shift means that not all data is stored indefinitely, which translates to lower energy demands.
Sustainable design: Data centers themselves are pursuing more innovative design with the use of renewable energy sources, with the five US-based hyperscalers now accounting for over half of corporate renewable purchases worldwide. Advanced cooling technology explores immersion cooling and automated or even AI-managed thermal management systems. Companies that store their data in these centers can reduce computing requirements by optimizing server utilization, improving coding practices, and adopting edge computing.
Advances in technology: Although more nascent in the technology maturity curve, advances in quantum computing will lead to performing the same computations with fewer resources, making computing more sustainable going forward. Advances in quantum will not only use less energy, but will unlock more advanced performance. The constant cost reduction imperative driving more efficient technology development often works against ecological interest, especially where new technology is involved; however, in this case, they work towards the same goal with an urgency buoyed by promises of lower operating costs.
Updated regulation: What private industry is not willing to do itself, regulation can tackle. The European Union’s new energy efficiency directive requires emissions reports filed for any data center usage larger than 500 kilowatts. Robust policy frameworks and regulatory measures can enforce sustainable practices in data centers and technology firms broadly, ensuring that progress does not continue to come at the expense of environmental health.
The shift to a more sustainable environment will not emerge from a single group. It will require repeated interest in making changes from independent companies, data centers and service providers who support their business, private and post-secondary researchers revealing the cutting-edge, and governments at every level creating the rules and regulations that guide or govern us all. While we may not be able to reduce reliance on the data that runs our modern world, together, we can keep at bay the storm gathering on the horizon. The clouds may be darkening, but our actions will determine whether they unleash the rain.
Stefani Ruzic | Senior Consultant, Deloitte Consulting LLP
Behavior/Demand will only change when carbon is priced into the product/service. Today we are learning to engineer AI prompts to improve output. Tomorrow we will learn to engineer AI prompts to minimize cost. I predict that in the long run, this optimization will impact language!