NTT Global Data Centers' new AI Stress Test Report quantifies the collision between AI-driven demand and physical reality. It models three growth trajectories through 2030, with a baseline scenario of about 23% annual demand growth that pencils out to roughly $3.8 trillion of additional digital infrastructure investment. The accelerated case climbs to $5.7 trillion. The facilities involved are no longer the ten-megawatt buildings of a decade ago; they are now hundreds of megawatts, spanning the footprint of two or three football fields.
The report's central finding is that the choke point is not GPUs, transformers, or switchgear. It is paperwork, planning, and process. Power availability and deployment timelines, not fiber capacity, are the gating factors in the scenarios modeled. NTT CTO Bruno Berti puts it plainly: it does not matter how quickly you can get GPUs if you cannot get a permit to put them anywhere. That reframes strategy away from a "build, baby, build" reflex and toward treating capacity expansion as a system-wide coordination problem.
Berti also addresses two objections that come up in almost every community consultation: that data centers waste water and steal power from the grid. He says both are misconceptions rooted in older facilities. Modern designs use liquid cooling, which removes heat more effectively and eliminates excess water waste. Operators also plan the new class of hundred-megawatt-plus sites in partnership with utilities years in advance, often with on-site solar, wind, or other renewable generation that can help stabilize the grid.
His conclusion is blunt: transparency builds trust. The industry must stop treating what it does as a secret and instead educate and partner with communities, policymakers, and businesses if it wants the permits needed to turn capacity into reliable AI infrastructure through the decade. The report is presented in a sponsored post by NTT Global Data Centers.