Zankore has signed a loan facility of up to $3.1 billion to acquire and deploy computing hardware for its planned AI cloud business. The September 9 announcement backs a service aimed at enterprises, developers and institutions in Indonesia and the wider Southeast Asian market.
A customer developing an AI model needs more than a collection of chips. It needs the machines connected, provisioned and kept running through demanding workloads. Zankore describes a system combining Nvidia GPU hardware, liquid-cooled facilities and software for managing the infrastructure. Its intended workloads include training models and inference—the computing needed to run those models after training.
The company presents itself as an infrastructure partner that designs around a customer's workload, jurisdiction and timeline. Data residency, or where information is held and processed, is part of that pitch. The managed software layer is intended to cover provisioning through day-to-day operational health, so customers do not have to build the complete computing environment themselves.
This is still a buildout, not a claim that the advertised capacity is already serving customers. Zankore says it is building an initial 100 megawatts of infrastructure; its launch plan targeted around 200 megawatts in the first half of 2027 and a larger eventual expansion. Those are measures of planned infrastructure capacity, not booked revenue or utilization.
The financing is a senior term loan facility, not an equity round, and the announced ceiling does not establish that the full amount has been drawn. Citi advised on the deal, with Citi, ING, Natixis, Qatar National Bank and UOB arranging it. The company also describes revenue-sharing and credit-support arrangements intended to connect deployment with customer demand.
The business question is how quickly that demand becomes paying use of the machines. The Next Web notes the relationship between Nvidia's backing and the Nvidia hardware the financing will purchase. For Zankore, delivering reliable computing to customers—and matching the pace of hardware deployment to their needs—will matter more than announcing another capacity target.