What data centres can teach project managers about sustainability

The Cloud isn鈥檛 weightless. A new article in PM World Journal argues that project managers have a unique role in making AI infrastructure genuinely sustainable 鈥� starting with how we write business cases.

We talk about technology as though it floats. 鈥淭he Cloud,鈥� 鈥渨ireless,鈥� 鈥渧irtual鈥� 鈥� the language suggests something weightless. But every cloud is bolted to a building site, and every AI query draws power from a data centre consuming electricity and evaporating water.

In my latest article for PM World Journal, I examine whether the environmental benefits of AI 鈥� discovering new battery materials, cutting building energy use by over 20%, optimising its own cooling systems 鈥� outweigh the costs of the data centres that make it possible. The answer isn鈥檛 straightforward, and that鈥檚 precisely why project managers matter.

Here are two ideas from the article that deserve wider discussion.

The synergy hiding in your business case

Small Modular Reactors (SMRs) could provide zero-carbon baseload power for data centres, but their deployment is stalled by one thing: the cost of nuclear-grade security. Perimeter fencing, 24/7 guards, biometric access, surveillance 鈥� it all adds up.

Here鈥檚 the thing: Tier 4 data centres already have all of that. They鈥檙e among the most physically hardened commercial buildings in the world. If you co-locate an SMR within an existing data centre鈥檚 secure perimeter, you share the security cost across two projects that would otherwise bear it separately. The SMR gets affordable security; the data centre gets zero-carbon, islanded power with no transmission losses.

This isn鈥檛 an engineering idea 鈥� it鈥檚 a project management idea. It requires someone in the initiation phase to look beyond the immediate project scope and ask: what else could share this infrastructure? That鈥檚 cross-portfolio thinking, and it鈥檚 exactly the kind of synergy that Total Value Contribution analysis is designed to surface.

The IEA鈥檚 April 2026 report 鈥淜ey Questions on Energy and AI鈥� confirms that data centre operators have already signed conditional offtake agreements for 45 GW of SMR capacity 鈥� up from 25 GW just eighteen months earlier. Data centre demand may be the catalyst that brings SMR technology to commercial scale, benefiting the wider energy system far beyond the technology sector.

The efficiency trap

Most of us assume that making something more efficient makes it more sustainable. Often, the opposite happens. The economist William Stanley Jevons observed in 1865 that as steam engines became more fuel-efficient, total coal consumption increased 鈥� because cheaper energy made more activities economical.

The same pattern is playing out in AI. Energy efficiency per AI task improves by roughly an order of magnitude each year, yet . Efficiency makes compute cheaper; cheaper compute generates more demand; more demand consumes more energy. Project managers who report efficiency gains without tracking total resource consumption may be measuring the wrong thing.

This is why I argue that every data centre business case needs Quintuple Bottom Line stage-gate controls 鈥� assessing People, Planet, Prosperity, Peace and Partnership at every decision point, with increasing specificity as the project progresses. Efficiency gains are welcome, but only if governance ensures they deliver actual environmental savings rather than simply enabling more work.

What you can do

If you鈥檙e involved in technology projects 鈥� and increasingly, who isn鈥檛鈥� three questions are worth asking at your next stage gate:

What is the whole-life carbon and water footprint of this project, including decommissioning?
Could the infrastructure we鈥檙e building serve a second purpose 鈥� energy generation, district heating, material recovery?
Are we measuring total environmental impact, or just efficiency per unit?

The full article, , is available in PM World Journal (Volume XV, Issue VI, June 2026). You can find it and the on sustainability and regeneration in projects.