Nvidia-Backed Indonesian AI Platform Zankore Launches $3.1 Billion GPU Loan
nashnova research
Indonesian AI platform Zankore is syndicating a $3.1 billion loan to buy GPUs for its new AI factory, backed by a six-year Nvidia contract — a sign that GPUs themselves are becoming bankable assets as Asia's AI infrastructure financing wave accelerates.
What does this loan look like?
Zankore is raising $3.1 billion over five years at SOFR + 330 basis points to buy GPUs for a new AI factory.
UOB and Citigroup are among five original underwriters. They are now inviting other banks to join at the senior level as underwriters or bookrunners.
This means → the deal is being syndicated, spreading risk across multiple lenders — a sign the banking market is warming to GPU-backed debt.
What role does Nvidia play?
The loan is supported by a six-year Nvidia contract that includes revenue-sharing and credit support.
In plain terms = if Zankore cannot sell its compute capacity, Nvidia absorbs part of the revenue risk — so lenders are underwriting Nvidia's credit, not just Zankore's.
The arrangement covers unsold capacity within an initial 100 MW of compute. A U.S. fintech company is the designated buyer of Zankore's GPU capacity.
This reflects Nvidia's evolution from chip seller to financing enabler — using its own balance sheet to unlock downstream demand.
How big is Asia's AI infrastructure financing wave?
Blackstone's AirTrunk has secured A$4.25 billion (≈US$3 billion) in green loans for a Sydney data-center campus. Blue Owl Capital's Stack Infrastructure is seeking A$8.6 billion for a Melbourne project.
Moody's estimates global data-center investment will exceed $3 trillion over the next five years.
This means → regional banks are moving beyond funding data-center construction to funding the GPUs inside them — GPUs are becoming a collateralizable, priceable financial asset.
Where is the risk in this model?
The central question is whether Nvidia's revenue-sharing and residual-value guarantees can cover risk across the full five-year loan term.
In plain terms = GPUs are not real estate. They depreciate fast and get replaced by next-generation chips — no one can guarantee what these GPUs will be worth in five years.
This reflects a broader truth: AI infrastructure financing is in a "model-proving" phase — capital is flooding in, but the tools and track record for pricing the risk are still immature.
市场有风险，内容仅供研究参考，不构成投资建议。