NTT Global Data Centers weighs the price tag of not building fast enough

The next 200-megawatt data center permitted in a county will likely resemble buildings much different from those from a decade ago. Occupying the space of two or three football fields, these centers will consume more power than most nearby industrial parks. This intersection of AI-driven demand and physical constraints is the focus of NTT Global Data Centers' new AI Stress Test Report.

The report simulates three growth trajectories through 2030, testing each against constraints such as land availability, water consumption, and specialized labor. The middle scenario projects a $3.8 trillion investment increase in digital infrastructure annually, while the accelerated case sees this figure rise to $5.7 trillion.

Bruno Berti, NTT Global Data Centers' CTO, has observed the size envelope expanding significantly. Where once they built ten-megawatt buildings, they now construct hundreds of megawatts, comparable to the size of two or three football fields. Berti highlights permitting as the primary bottleneck in AI infrastructure development, noting that power availability and deployment timelines, rather than fiber capacity, are the decisive factors.

He asserts that focusing on rapid GPU procurement is misguided without securing permits for their deployment.

Berti advocates for a portfolio approach, treating capacity expansion as a system-wide coordination problem rather than a land race. Addressing community concerns, he clarifies that modern data centers use liquid cooling technology to minimize water waste and often partner with utilities for renewable energy sources to stabilize the grid, dispelling the notion that they drain resources.

Berti concludes that transparency is key to building trust with communities, emphasizing the need for education and partnership to facilitate the permitting process and enable AI infrastructure to benefit communities, policymakers, and businesses.