Zankore Secures $3.1 Billion Loan to Build Nvidia AI Cloud Capacity in Indonesia

Zankore, the Nvidia-backed AI neocloud joint venture majority-fed by Ooredoo Group and Indosat Ooredoo Hutchison, signed a senior term loan facility of up to $3.1 billion on September 9, 2026 to fund its first phase of GPU infrastructure in Indonesia — an initial 100MW build that the company says will scale toward 1GW within three years.

What is known

Per Ooredoo’s own press release, Zankore launched in August 2026 as a joint venture between Indosat Ooredoo Hutchison, Nokia, and Nvidia, with Ooredoo as founding shareholder and lead investor holding a 49% stake and a commitment of roughly $800 million. A month later, the platform closed the debt side of its funding stack: a senior term loan of up to $3.1 billion, with Citi, ING, Natixis CIB, Qatar National Bank, and UOB acting as senior mandated lead arrangers, underwriters, and bookrunners, according to both the Ooredoo release and DCD’s reporting.

The stated targets, per Ooredoo and corroborated by The Next Web and FinanceAsia, are an initial 100MW of Nvidia AI infrastructure in Indonesia, roughly 200MW by the first half of 2027, and an ambition to reach 1GW of Nvidia DSX AI Factory capacity within three years. Nvidia is unusual here in sitting on both sides of the transaction — an investor in Zankore and the GPU supplier the loan proceeds will be spent with — a structure The Next Web notes reflects Nvidia’s broader push to mobilize third-party lender capital rather than fund AI buildouts from its own balance sheet.

What it means for a power buyer

Indonesia already has the cheapest commercial electricity of any market our price index tracks — around $0.063/kWh, roughly a quarter of Singapore’s $0.27/kWh — and our catalog lists 40 facilities split between Jakarta (data-sovereignty workloads) and Batam (an offshore extension of Singapore, roughly 16ms away by fibre per our market notes). A financing package aimed at deploying 100MW of GPU capacity in-country adds real, purpose-built AI load to that market rather than general-purpose colocation — a different buyer profile than the enterprise and hyperscale-storage demand that has driven most of Indonesia’s build-out so far.

For anyone shopping GPU capacity rather than shell-and-power, the more relevant number is on the compute side: our GPU price tracker currently puts on-demand H100 at a median of about $3.39/GPU-hr and H200 at about $4.40/GPU-hr globally. A new 100MW-scale Nvidia deployment coming online in Southeast Asia over the next year is a supply-side signal worth watching for regional GPU pricing — though Zankore’s capacity is aimed initially at enterprise and sovereign AI demand rather than the on-demand spot market our tracker follows, so it won’t move those numbers directly in the near term.

The financing structure is also a signal in its own right: a $3.1 billion syndicated loan from five banks, alongside revenue-sharing and credit-support terms tied to deployment pace, is debt underwritten against future GPU demand rather than a hyperscaler’s balance sheet — a funding model increasingly common for neocloud buildouts and one worth weighing against colocation capacity backed by long-term anchor tenant leases.

What we don’t know

Neither Ooredoo’s release nor the press coverage names the specific site(s) in Indonesia where the initial 100MW will be built, nor whether Zankore will build its own facilities or lease capacity from existing operators. The exact terms of the “revenue-sharing and credit-support arrangements” tied to deployment pace and customer demand haven’t been disclosed beyond that description, and no anchor customers for the GPU capacity have been named.