Zankore has closed a $3.1 billion GPU financing facility to fund NVIDIA processor purchases for a new AI compute factory in Indonesia, according to reporting from Bloomberg. The facility will launch at 100 megawatts and is designed to scale to 1 gigawatt.
The financing structure includes a revenue-sharing and credit-support arrangement from NVIDIA, which reduces lender risk and allows GPU-backed infrastructure to be financed more readily in emerging markets. NVIDIA effectively backstops portions of the debt, making the deal more attractive to institutional lenders unfamiliar with the economics of large-scale AI compute capacity.
Citigroup, ING, Natixis, Qatar National Bank and United Overseas Bank underwrote the syndicated loan. The underwriting group includes both global and regional capital: ING and Natixis bring European institutional depth, Citi provides global markets reach, while UOB and QNB anchor exposure to Southeast Asian and Middle Eastern investors.
Indonesia's infrastructure base and labor costs position it competitively for large compute facilities. Regional jurisdictions have become destinations for AI compute deployment as North American data center capacity tightens and power availability in established markets constrains expansion. The facility's staged build, 100MW initially, then scaling to 1GW, allows Zankore to validate demand and optimize operations before committing full capital.
The deal structures GPU financing through collateral and revenue commitments rather than direct equity raises, a model gaining traction as demand for inference compute clusters accelerates. NVIDIA's participation shows confidence in Zankore's ability to monetize the asset and service the debt through long-term compute contracts, likely with cloud platforms or AI service providers.
The transaction adds $3.1 billion to global AI infrastructure funding, which has exceeded $50 billion across major deals and announced projects in 2026. Zankore's access to syndicated bank financing at scale, backed by NVIDIA credit support, demonstrates institutional lenders' appetite for GPU infrastructure loans when borrower revenue is hedged or collateralized.
The facility's path to 1GW hinges on sustained demand for inference-class compute and Zankore's ability to secure long-term contract commitments from customers to support debt service. If customer off-take agreements do not materialize, the staged rollout may stall before reaching full capacity.
