The growth of AI, particularly generative AI has spurred a serious boom in data centre construction.

However, there is a huge downside. Environmental, governmental, and energy experts are concerned that the needs of generative AI far outstrip the ability of current power generation to power these new data centres.

To make matters worse, generative AI data centres are far more power hungry than the data centres of the past. With the need to accommodate other power-intensive phenomena including growth in electric automobiles and rapidly evolving man-made climate change, new data centres for generative AI are problematic.

Generative AI data centre problems

The problem spreads far beyond just power consumption. Data centres require tremendous amounts of water to keep them cool. As CPUs and GPUs (plus all the other supporting silicon) have grown more complex, the amount of power they use has risen dramatically. More power use means the amount of heat that must be conducted out of the data centre has increased accordingly.

Government regulation has already descended, with cities restricting data centre construction and governments eyeing even more regulation to address environmental concerns. Companies and investors that have strong commitments to green principles are less inclined to invest in generative AI or data centre construction. Further, concerns about supplies and environmental effects of rare mineral resources also weigh on the minds of governments, investors, and customers.

A hot topic

There isn’t much that can be done about the status of power generation and distribution – renewable power solutions and grid expansion are projects that take years to implement. New and innovative ways to dissipate heat in data centres and repurpose that heat to other uses are always being developed.

In theory, generative AI itself should grow more efficient over time. None of that will arrive soon enough or be efficient enough for the projected proliferation of data centres that are needed to support the growth projections.

Enterprises should be assessing their vendors when it comes to generative AI and insist on renewable energy and a reduction in water use at data centres. There will be a backlash against AI over environmental issues and enterprises should insulate themselves by requiring suppliers to employ green practices, even if it slows down implementation of generative AI.