The Great Power Rush: Are We Asking the Wrong Question?

The race for power dominated conversations at Datacloud Global Congress in Cannes. But what if the industry’s biggest challenge isn’t securing power at all?

As AI infrastructure scales at an unprecedented pace, many organisations are operating on the assumption that power can always be sourced, expanded, or engineered into existence. The reality is more complex. The companies best positioned to succeed won’t be those simply chasing power, but those that understand its constraints, risks, and wider implications ahead of the curve.

Cannes delivered its familiar mix of momentum and ambition, capital in motion, partnerships forming, and decisions being shaped that will define the next decade of infrastructure, all set against a Mediterranean backdrop that lends a sense of inevitability to the scale of investment being discussed.

AI dominated every agenda. Yet beneath the surface-level excitement, a consistent theme emerged, not as a threat, but as an opportunity to be better understood.

Power, and more specifically, the ability to understand and navigate its realities, is becoming a defining advantage.

Here are five things Datacloud confirmed about where the real opportunity in AI infrastructure truly lies:

1. The organisations moving fastest are asking better questions about power.

Everyone is chasing power. New connections, new locations, new generation strategies. That ambition is right. But the organisations gaining real ground are the ones going a step further, asking not just “where is the power?” but “what does our power situation actually expose us to?”

The grid has constraints. Viable power isn’t uniformly available. And self-generation, while increasingly part of the answer, carries its own technical and commercial complexity. The organisations navigating this well aren’t the ones with the most optimistic projections.

They’re the ones with the clearest understanding of their own position.

2. Clarity about power exposure is becoming a genuine competitive advantage.

The load profiles that AI infrastructure demands are categorically different from what the industry was designing for five years ago. That creates real questions about grid capacity, fault levels, protection systems, and the assumptions baked into existing infrastructure.

The good news is that these are answerable questions. Organisations that invest in getting independent, rigorous answers early, before disputes arise, before due diligence uncovers gaps, before planning applications stall are the ones making confident investment decisions. They’re the ones whose projects move.

Knowledge, applied at the right moment, is what separates confident progress from costly delay.

3. The question of what AI infrastructure actually needs is still wide open and that’s an opportunity.

Not every valuable conversation at Cannes happened under a spotlight. Poolside, with the agenda temporarily forgotten, a question came up that the industry hasn’t fully answered yet.

Does an AI factory actually need Tier 4?

It’s a genuinely interesting question. The downtime economics of AI inference workloads are categorically different from traditional enterprise IT. If a workload can go down for an hour, ramp down gracefully, and simply be rerun, the specification for the facility changes significantly. Tier 4 resilience is exactly right for some use cases: government disaster recovery, sovereign infrastructure backup, country-level facilities where full redundancy is precisely what’s needed.

But AI factories may represent an entirely new category, one that the industry has the opportunity to define properly, rather than defaulting to specifications written for a different world. That’s not a problem. It’s a chance to build something better.

4. The relationship between AI and the power system is about to get much more dynamic.

Follow the AI factory thread a little further and something genuinely exciting emerges. If these facilities can tolerate managed downtime, and if workloads can be programmed to run in alignment with electricity pricing and grid conditions, the relationship between AI infrastructure and the power system becomes something far more intelligent than anyone is currently designing for.

“We’re not there yet. But it’s not far away.”

The organisations building that understanding now, before the standards catch up, before the grid conditions shift, will be the ones with the clearest view when it matters most.

5. Knowing there’s a power challenge is the starting point. Understanding your own is the advantage.

The industry knows there’s a power challenge. The organisations pulling ahead are the ones that understand their own.

There’s a meaningful gap between the two and it’s where projects accelerate or stall, where investments are validated or challenged, where disputes are avoided or escalated. Closing that gap means knowing what your grid connection actually exposes you to. It means having protection studies that reflect how the network operates today. It means walking into any process, a due diligence, a planning application, a technical dispute with analysis that holds up because it was produced independently, by people with no stake in any particular outcome.

That’s not just good engineering. In the AI era, it’s good business.

The industry is reaching the same conclusion. The data backs it up.

What felt like a niche engineering conversation in Cannes is now a mainstream strategic priority. The data is catching up with what the most forward-thinking organisations have understood for some time.

The Uptime Institute’s 2026 predictions report is direct: the data centre industry faces a power challenge that technology alone will not resolve. Andy Lawrence, their executive director of research, notes that uncertainty about how AI will reshape demand is complicating both capacity planning and resiliency strategies, making independent, specialist expertise more valuable than ever.

The scale is significant. Of around 140 large-scale data centre projects representing approximately 12 gigawatts of power planned to go live in 2026, only a third are currently under construction, underscoring just how much depends on getting the power picture right from the outset.

As one leading industry analysis put it: the companies that navigate this most effectively, securing power, building understanding, and managing their exposure with confidence, will be best positioned to lead the AI era.

That’s the real quest. Not just finding power but truly understanding it.

The organisations that get there first will have an advantage that compounds. Knowledge, in this era, is the most valuable infrastructure of all.