AI hype meets infrastructure reality: observations from MWC

Mobile World Congress has changed. It used to be about telecom infrastructure and the next wave of connectivity. Now the entire conversation seems to orbit around AI.

After a few days walking the floor and talking with vendors, operators, and infrastructure teams, a few patterns start to emerge. Clarifying how these emerging AI infrastructure trends influence strategic planning can help industry professionals understand the practical implications and prepare accordingly. And interestingly, many of the more thoughtful conversations are moving away from the hype and toward something more practical.

Three themes kept surfacing.

First, the growing gap between AI ambition and physical reality

Walk the floor and it quickly feels like every conversation promises massive AI capability. But the infrastructure required to support that scale is another story. Power, cooling, and grid capacity are quickly becoming the real constraints, raising questions about infrastructure feasibility. At some point, the conversation has to move beyond the marketing numbers and ask whether the world actually has the energy to support everything currently being proposed.

That reality is starting to creep into the discussions.

Second, the conversation is moving beyond GPUs alone

For a while, the narrative around AI infrastructure has focused heavily on GPU density. But more operators are beginning to talk about the broader system equation: power availability, CPU balance, edge processing, network architecture and how workloads are distributed across those layers.

In other words, AI infrastructure is increasingly a systems challenge rather than simply a hardware race.

Third, sovereignty is no longer a theoretical discussion

In many conversations, there is a noticeable shift in how organisations talk about control. Concerns about hyperscaler dependence and cloud lock-in are surfacing more openly. In response, there is growing interest in architectures that allow companies and countries to retain greater control over where data lives and how infrastructure is operated.

That includes renewed attention to open platforms and sovereign infrastructure models.

For infrastructure providers, these signals matter. The future of AI will not be defined only by the most powerful chips. It will be defined by the ability to balance power, compute, control and operational resilience across an increasingly complex system.

That balance is where the next phase of the industry conversation is likely to move.

For a deeper look at how AI scale is already reshaping infrastructure operations, read our perspective on the sovereign cloud shift.

• AI