NTT Global Data Centers has opened a $1 billion capital raise for U.S.-based infrastructure projects, the first discrete tranche of a $10 billion global expansion aimed at AI-ready data center capacity. The Tokyo-headquartered unit of Nippon Telegraph and Telephone is running the book directly with U.S. institutional allocators rather than syndicating through traditional infrastructure funds. The move puts $1 billion of near-term deployment capital into competition for the same hyperscale tenant commitments that Digital Realty, Equinix, and a thickening field of private builds are chasing.
NTT Global operates 160-plus data centers across 20 countries and holds a Tier One carrier network advantage in Asia-Pacific routing that newer entrants lack. The company is positioning the U.S. raise as AI-specific infrastructure—higher power density per rack, liquid cooling optionality, and sub-10-millisecond latency to major cloud on-ramps. Management has not disclosed whether the $1 billion will fund greenfield construction, acquire distressed or stalled projects, or both. The global $10 billion program includes expansions in India, Europe, and additional U.S. metros, but the $1 billion U.S. book is being marketed as a standalone opportunity with independent return hurdles.
The raise matters because it signals a structural shift in data-center capital formation. NTT is not waiting for syndication or fund formation timelines—it is moving live capital into a market where AI inference workloads are driving 30%-40% annualized growth in rack demand but where actual deliverable capacity lags by 18-24 months. Allocators who move into this book are effectively pre-leasing infrastructure to hyperscalers who have publicly committed to $200 billion-plus in combined AI capex over the next two years. The tension is in tenant credit and lease duration. Microsoft, Google, and Amazon can command 10-15 year leases with escalators, but second-tier AI labs and model-training shops are offering 3-5 year commitments with higher per-kilowatt rates and structural rollover risk.
Operators should watch two follow-on events. First, whether NTT's U.S. book closes inside 90 days or stretches into Q2 2025—speed indicates how much of the $1 billion is already soft-circled by family offices and sovereign wealth allocators. Second, whether the company announces specific metro site acquisitions in Northern Virginia, Phoenix, or Dallas-Fort Worth within 120 days of the close. Those three metros account for 60%-plus of U.S. hyperscale expansion, and land parcels near existing fiber and power substations are being locked under option at a pace that suggests pre-negotiated tenant agreements are already in motion.
NTT's carrier-grade network and existing Asian customer base give it an edge in latency-sensitive AI inference workloads that require multi-region redundancy. If the $1 billion deploys into projects with signed hyperscale anchor tenants, the IRR will compress but the duration risk disappears. If it chases higher yields with shorter-term AI lab tenants, the return spreads widen but so does the rollover exposure when those labs consolidate or collapse.