NTT Global Data Centers aims to double data center capacity to 4 GW amid the AI boom

NTT Global Data Centers plans to double the capacity of its global data center network to 4 GW to keep pace with the surge in demand for AI infrastructure. This is one of the most notable signals of spring 2026: the race for artificial intelligence is increasingly constrained not just by chips, but by electricity, facilities, and cooling.

Scale of Expansion

NTT's division, considered the third-largest data center operator outside China, wants to reach 4 GW of capacity within approximately two years. Currently, the company has more than 2 GW of critical IT load, over 160 facilities, and presence in more than 20 countries and regions.

To double its scale, NTT is running 34 projects simultaneously. In essence, this is not about launching a couple of new facilities, but rather about a major cycle of construction and expansion across multiple regions at once.

The level of demand readiness is particularly important. According to Doug Adams, the head of the business unit, contracts are already secured for more than 70% of these 34 projects. This is a strong market signal: NTT is not building speculatively, but for already confirmed client needs.

Moreover, the company expects that within a five-year horizon, it will be able to offer significantly more than 5 GW of capacity. If this plan succeeds, NTT will establish itself among those players who are winning from the AI boom not at the level of models, but at the level of physical infrastructure.

Who Is Taking the Capacity

Demand is coming not only from classical cloud companies. In March, NTT announced new commitments of nearly 115 MW across campuses in Virginia, Illinois, and California. Of this volume, more than 90 MW was reserved by a major hyperscaler, and nearly 20 MW was contracted by three enterprise customers.

This is a good indicator of market structure: the primary consumers are still cloud giants, but large enterprises are also beginning to reserve capacity in advance for their future AI workloads.

Clients need not just square meters for racks, but facilities that are designed from the start for dense computing and rapid growth. In NTT's materials, this looks like:

• support for high-density compute workloads for AI services
• ability to transition to liquid cooling as requirements grow
• proximity to cloud availability zones and low latency
• compliance with security and compliance requirements
• rapid capacity deployment and a clear path for further expansion

Over the past year, NTT has brought ten new facilities online in North America, EMEA, and the Asia-Pacific region, adding more than 370 MW of IT capacity. This shows that the company is no longer just announcing ambitious goals, but is able to quickly convert demand into real infrastructure.

Against the backdrop of the generative AI boom, this becomes a separate competitive advantage: large clients increasingly choose not the cheapest operator, but one who can deliver capacity on time and without compromises on cooling, networking, and resilience.

Money and Execution

The next question is how to finance such growth. NTT says it can use its own funds, partner capital, and instruments from its REIT — NTT DC REIT. This is an important detail because the data center market is currently growing faster than many developers can attract electricity, equipment, and contractors.

Having multiple financing channels reduces the risk of projects stalling halfway through, and pre-signed contracts make NTT's investment story significantly more compelling.

But money is only part of the task. In practice, the speed of deploying new AI-ready facilities is constrained by access to power grids, lead times for engineering equipment, design for high density, and preparation of cooling systems.

This is why NTT's announcement is important not just for the 4 GW figure. It shows that major operators have already learned to plan development for future demand years in advance and sell capacity before a facility is fully ready.

For customers, this means one thing: available capacity in key markets is shrinking, and reserving it needs to happen sooner rather than later.

What This Means

Winners of the AI race in 2026 are being determined not only in laboratories and not only in the GPU market. Increasingly significant are operators who can quickly turn electricity, land, and capital into ready-to-use data centers. NTT's plan for 4 GW shows that the infrastructure layer of the AI market is becoming as strategic as the models and cloud services themselves.