AI chips run hotter than air cooling can handle. That's been true for a couple of years now, and it's why liquid cooling has been steadily working its way into every data centre roadmap. Rack densities have outpaced what fans and chilled air can dissipate, full stop.
What's new is the second pressure now sitting on top of that.
Forty cities have signed the Global Urban Data Centres Pact, coordinated by C40 Cities and launched at London Climate Action Week. London, Phoenix and Melbourne are among them. The pact sets shared standards on power, water and land: build on underused sites, run on clean energy, use less water, and stop treating local communities as an afterthought. The scale behind it is real: Melbourne's roughly 50 major data centres are on track to take up a fifth of the city's power demand by 2040 and tens of billions of litres of water a year. Phoenix has over 200 sites existing or proposed, enough to double the region's electricity demand.
London Mayor Sadiq Khan said in a statement that while AI and digital infrastructure would play "a major role in the future prosperity of cities around the world... residents are right to expect growth to be managed responsibly".
There's a reason cities moved together rather than waiting on national policy. Data centres already account for an estimated 2.5–3.7% of global emissions, more than aviation, and demand is climbing faster than power consumption overall. A shared standard across nearly 100 cities is harder for any single region to undercut by quietly relaxing the rules to win investment.
So operators now have two pressures stacked on each other, not one. The heat problem was already forcing the move to liquid cooling. The pact means that move now has to happen inside a shrinking power and water budget at the same time. It's no longer enough to solve the thermal problem, the cooling system has to do that and stay inside resource limits a city has the authority to enforce. Most existing infrastructure wasn't designed with both constraints in mind, because until recently, it didn't have to be.
That's a materially harder engineering problem than "install liquid cooling." It means systems validated before they go live, monitored for water chemistry and contamination over years of operation, and designed against power and water ceilings from day one. The industry's own language has shifted accordingly: this is lifecycle management now, not product selection.
And that's where this stops being a sustainability story and becomes a talent one.
Every operator now answering to a city pact, a planning authority, or ESG reporting needs people who can solve heat and resource limits as one problem, not two separate ones. Thermal engineers who can validate liquid cooling systems against both performance and water budgets before go-live. Facility engineers for whom energy load and water chemistry are core responsibilities, not someone else's. Compliance specialists who can turn planning requirements into design decisions before they become blockers.
These are specialist roles, and the pool is shallow, because the discipline itself is young, most of this sat at the margins of data centre engineering five years ago. It's now central to whether a project gets approved at all. Clients who wait for a planning authority to force the issue will be competing for the same small group of people as everyone else, at the same time. Clients who build this capability now are negotiating from strength when the regulation catches up with them.
If your background is in thermal management, liquid cooling, water systems or infrastructure compliance, this is the moment your specialism stops being a niche and becomes essential. The skills that used to sit at the edge of data centre engineering are moving to the centre of it, because heat made liquid cooling necessary, and now cities are making resource-efficient liquid cooling necessary.
This isn't a job-market blip. It's a structural shift, the kind that defines a career rather than a single move. Engineers who build deep expertise here will become the standard the rest of the industry is hired against.
We work inside mission-critical infrastructure every day, from data centres, to hyperscale builds and the engineering that keeps them running. We see talent gaps before they make headlines, because we're the ones trying to close them.
The pressure on data centres isn't one problem anymore, it's two, stacked on top of each other. We help clients build the capability to answer both ahead of the regulation, not in reaction to it. And we help the specialists doing this work find roles that match where the industry is actually heading.
