Infrastructure at Gigawatt Scale: Power, Regulation, and the Limits of “Future Proofing”

The final Mainstage session at Tech Show London brought together senior US data centre leaders to examine how hyperscale AI demand is reshaping digital infrastructure. Moderated by Mark Gusakov (Digital Infrastructure Alliance), the panel featured Anahita Mouro (Google), Jeff Ivey (Crusoe), Buddy Rizer (Loudoun County), Adrian Mountstephens (Equinix) and Cliff Grossner (Open Compute Project Foundation).

The panel operated on the assumption that AI will drive continued infrastructure expansion. The discussion instead focused on what it means to build at gigawatt scale in real time, under power constraints, regulatory scrutiny and increasing public visibility.

From Data Centres to AI Factories

Early in the session, Anahita Mouro described the structural shift underway:

“We are moving from the era of just building their houses for servers into large scale, massive AI factories.”

AI workloads require densities, cooling strategies and integration approaches that diverge materially from earlier cloud deployments. Mouro emphasised the need to treat the data centre “as a one integrated massive system,” aligning silicon, hardware and operations within a unified architecture.

The change involves not only a greater scale but a different design philosophy. Power optimisation, supply chain resilience and system-level efficiency must be embedded from the outset.

Gigawatts, Not Megawatts

Jeff Ivey highlighted the magnitude of change:

“It’s gigawatts to scale and gig watch speed.”

Where previous growth cycles were measured in megawatts, hyperscale AI now requires gigawatt-scale thinking. Grid capacity, however, has not expanded at an equivalent speed.

“The power is just not there.”

Bridging strategies – including on-site generation and diversified energy portfolios – are becoming integral to deployment planning. Energy procurement has moved from an operational consideration to a structural constraint.

Infrastructure strategy now extends beyond site selection to active engagement with utilities and long-term power coordination.

Standardisation and Shared Engineering

Cliff Grossner described the Open Compute Project as:

“a collaborative environment where our members can come together and solve their shared problems.”

As AI deployment accelerates, rack design, interconnect standards, and power distribution approaches risk fragmenting across operators. Shared engineering frameworks can reduce inefficiencies that otherwise slow scale.

Experience from earlier cloud growth cycles suggests that excessive variation introduces friction. Standardisation, in this context, functions as an accelerator rather than a limitation.

The Limits of “Future Proofing”

When the discussion turned to long-term planning, one panellist responded candidly:

“I think the word future proof is total bullshit.”

Infrastructure requirements are evolving faster than traditional planning cycles anticipate. Chip generations shift, density assumptions change, and market signals move rapidly.

Adaptability offers a more realistic objective than certainty. Flexible design, modular deployment, and iterative expansion reduce exposure to stranded capacity and obsolete assumptions.

Regulation and Public Accountability

Regulatory oversight was framed as a function of sector maturity rather than resistance. Adrian Mountstephens observed:

“Ultimately, Globally. The data centre industry is going to end up being regulated completely in all jurisdictions.”

As data centres become central to national economies and digital life, regulatory expansion becomes likely.

Public perception also emerged as a strategic factor. Responding to questions around local opposition, Buddy Rizer summarised the approach succinctly:

“Communicate early, communicate often, be fact-based and be responsive.”

Infrastructure is increasingly visible and politically salient. Community engagement is no longer peripheral to expansion strategy.

Building Ahead of Certainty

The panel acknowledged a central tension: infrastructure is expanding in anticipation of AI demand that continues to evolve.

As one panellist remarked:

“AI isn’t here yet? This is all the hard work in AI is yet to come.”

Operators are building capacity in parallel with the development of usage models and monetisation maturity. Deployment timelines are advancing even as business cases continue to evolve.

The scale of expansion is tangible. The structural question concerns alignment between silicon development, power availability, regulatory readiness and market adoption.

Infrastructure as Strategic Capability

Across the session, recurring constraints became evident: accelerating silicon cycles, extended build timelines, tightening power availability and expanding regulatory oversight.

Scaling AI infrastructure depends increasingly on coordination across utilities, governments, operators and technology providers. Ambition alone is insufficient without delivery capacity.

The challenges discussed are not confined to the US market. They reflect structural dynamics shaping global AI infrastructure expansion.

In this environment, adaptability functions as a practical discipline. The ability to coordinate energy strategy, engineering standards, regulatory engagement and public communication determines how effectively hyperscale AI deployment can be sustained.

Join Data Centre World Frankfurt

6-7 May 2026, Messe Frankfurt

Power, cooling, capacity and sustainability — the operational realities shaping Europe’s digital backbone.

Get Your Free Ticket