Global AI, a U.S. company that builds single-tenant, air-gapped data centers for governments and enterprises, has closed $441 million in financing, its first-ever debt raise. The senior secured credit facility was led and arranged by J.P. Morgan, with participation from a group of lenders the company did not name.
The announcement lands the same week another GPU infrastructure operator, Nscale, lined up a $900 million credit facility from a lender group that also included J.P. Morgan — two nine-figure debt packages for non-hyperscaler AI capacity within days of each other.
Co-Founder, Director, and CEO Sami Issa tied the raise to contracted demand. “We are significantly accelerating our capital-raising efforts to meet the exceptional demand we are seeing for dedicated, single-tenant, air-gapped AI infrastructure,” Issa said. “With $6.2 billion in contracted revenues, including $1 billion in contracted revenues already built and delivered to customers, we are scaling rapidly to meet this demand.”
What Global AI is actually building
Founded in 2024, the company describes itself as the world’s first sovereign AI hyperscaler: it designs, builds, powers, and operates data centers where one customer occupies the entire environment, physically isolated from other tenants and from public networks. Its model spans land, energy, construction, liquid cooling, and GPU-dense compute, aimed at nations and enterprises that want AI training and inference inside their own jurisdiction rather than on shared public cloud.
The hardware program is NVIDIA-based. The company, an NVIDIA Cloud Partner, owns and operates a data center in Endicott, New York, running GB200 NVL72 clusters. On December 3, 2025, it announced a partnership with HUMAIN, the AI company backed by Saudi Arabia’s Public Investment Fund, to develop large-scale AI data centers in the United States; that release also disclosed that NVIDIA had joined a recent Global AI fundraising round.
On March 16, 2026, the company said it had completed a GB300 NVL72 deployment it calls the largest such cluster in New York State, is actively deploying 7,000 GB300 systems, and plans to roll out NVIDIA’s next-generation Vera Rubin NVL72 racks across its U.S. footprint.
The capacity roadmap behind the raise
The debt comes against a buildout the company has already scheduled publicly: 16 megawatts of initial critical IT load at Endicott, ramping to 100 megawatts through 2026, with 250 megawatts planned for 2027 and 1 gigawatt of critical capacity targeted by the end of 2029.
For a company founded in 2024, the $441 million senior secured facility against $6.2 billion in contracted revenue is its first-ever debt raise, following earlier fundraising rounds that included NVIDIA. The stated next milestone is the ramp to 100 megawatts at Endicott by the end of 2026, followed by the Vera Rubin NVL72 deployments and the 250-megawatt mark in 2027.
