> ## Content Index > Fetch the complete content index at: https://bytevyte.com/llms.txt > Use this file to discover other available public pages before exploring further. # Zankore $3.1 Billion GPU Loan Shifts AI Compute Risk to Lenders - URL: https://bytevyte.com/zankore-3-1-billion-gpu-loan-shifts-ai-compute-risk-to-lenders/ - Published: 2026-09-11T14:21:04.000Z - Updated: 2026-09-11T14:21:04.000Z - Description: Zankore $3.1 billion GPU loan funds Nvidia-powered AI capacity in Indonesia, shifting capex and credit risk onto five banks. - Author: Bytevyte Editorial - Tags: ai-beats Indonesia's **Zankore** has secured a senior term loan facility of up to $3.1 billion to acquire and deploy **Nvidia** graphics processing units, a Zankore $3.1 billion GPU loan that ranks among the largest AI infrastructure financings ever assembled in Asia. The borrower is an AI infrastructure platform backed by Nvidia, Ooredoo Group, Indosat Ooredoo Hutchison and Nokia, and the facility was signed this week. Five banks provide the money: Citi, ING, Natixis, QNB and UOB. Zankore will use the proceeds to buy and deploy GPU capacity across Indonesia and Southeast Asia for model training, inference and agentic AI workloads. The company calls it the first syndicated transaction of this scale in the region. What sets the deal apart from a routine data centre loan is who carries the downside. Nvidia holds equity in Zankore and is its primary hardware supplier, and the structure pairs revenue-sharing arrangements with credit-support mechanisms. The GPUs themselves partially collateralise the debt. That design places part of the capital expenditure and the credit risk on the banks rather than the chip vendor. If rental economics for AI compute weaken, the lenders holding the facility absorb losses before Nvidia does. ## Inside the Zankore $3.1 Billion GPU Loan Zankore launched in August 2026, roughly a month before the signing. An entity with about one month of operating history has raised more debt than the annual capital budgets of many established regional telecom and data centre operators, leaning on hardware collateral and the credit support of its own supplier. The backer group mixes four different kinds of balance sheet: a chip designer in Nvidia, two telecom operators in Ooredoo Group and Indosat Ooredoo Hutchison, and a network equipment vendor in Nokia. Each brings something the others lack, from silicon supply to connectivity to regional operating licences. Zankore's intended customers are startups, enterprises and institutions in Indonesia and the wider region. The company frames AI compute as a new strategic digital asset class, and the loan documentation follows that logic by treating GPU fleets as financeable, revenue-producing infrastructure rather than depreciating equipment. The credit-support component is the piece that makes the pricing work. By sharing part of the downside with the lenders, Nvidia lowers the risk premium the consortium charges, which trims Zankore's cost of capital and raises how much hardware the loan can buy. The lending group spans American, French, British, Qatari and Singaporean banks. Syndication spreads a single operator's exposure across five institutions, which lets each participant hold a smaller slice of GPU-backed credit than any of them would take alone. Because the transaction is described as the first of its size in the region, the banks are also building underwriting models for an asset class with little historical loss data. For those lenders the appeal is a contracted, dollar-denominated cash flow tied to a technology whose demand has outstripped supply for three straight years. The exposure is an asset with a short economic life and a resale value that depends on whether newer silicon leaves it behind. The workload mix matters for the credit. Training runs are power-intensive and booked in long, discrete blocks, while inference and agentic AI workloads arrive as steady, latency-sensitive traffic. A fleet sized for one profile does not monetise equally well against the other, and this facility covers all three. ## From 100MW to 1GW of Capacity Zankore is building an initial 100MW of Nvidia-powered capacity, with a longer runway laid out in stages. | Stage | Capacity | Target | | -------------- | -------- | ------------------ | | Initial build | 100MW | Under way | | Second phase | 200MW | First half of 2027 | | Long-term goal | 1GW | Within three years | Each stage multiplies the power, cooling and grid requirements, and each needs fresh capital. The $3.1 billion facility is sized against the first phase. Reaching 200MW and then 1GW would require additional financings, from the same consortium or from new lenders. Power tends to be the binding constraint on builds of this size, and Indonesia's grid and land availability will determine whether the 200MW and 1GW targets are reachable on schedule. Any slippage pushes revenue further out against a loan that starts charging from the day it is drawn. The 1GW target also sets an implicit benchmark. Few single operators in Southeast Asia have announced comparable capacity, and the Nvidia DSX AI Factory label ties the buildout to a specific reference architecture rather than generic accelerated computing. That tie has commercial value for Nvidia and a concentration cost for Zankore, which is buying into one vendor's stack. ## Where the Risk Ends Up Nvidia's role extends past supplying chips. It holds equity in Zankore, supports the credit, and stands to collect the loan proceeds, which are earmarked for its own GPUs. Money lent by the consortium becomes revenue for the chipmaker before a single customer workload runs. The structure answers a specific gap. AI compute demand has concentrated among a small group of US hyperscalers able to fund data centres from their own balance sheets. Markets such as Indonesia have large digital populations and limited domestic capital for AI infrastructure, so operators there cannot self-fund comparable buildouts. Vendor-supported debt lets them borrow against the hardware they are buying. Compare the funding path for a US hyperscaler. A company of that scale can order accelerators, build campuses and depreciate them against cloud revenue it already collects. Zankore has to borrow first, deploy second and earn third, with interest accruing from the start. The trade-off is concentration. When a supplier underwrites demand for its own product, lenders' credit quality tracks that supplier's product cycle. A price cut, a faster generational refresh, or a shift in customer preference toward a rival accelerator all feed into the collateral value behind the loan. Zankore sits in a growing group of neocloud operators financed this way. These firms rent accelerated compute to AI developers without owning the hyperscale balance sheets that made the first wave of data centre construction self-financing. They depend on debt, on utilisation rates, and on GPU rental prices holding up long enough to service it. Two mismatches sit underneath the numbers. The Zankore $3.1 billion GPU loan carries dollar-denominated debt against customer revenue that is largely rupiah-based, which adds currency exposure on top of technology risk. And a facility sized for 100MW assumes utilisation levels that only materialise if the workloads arrive, the same assumption every neocloud makes and none fully controls. What the deal shows is that Nvidia's exposure to its own customers' credit is now part of how it sells hardware. The company is a shareholder in the borrower, a supplier of the collateral, and a beneficiary of the proceeds, three positions that normally sit with separate parties in a financing. ## Why this matters The Zankore $3.1 billion GPU loan shows how far Nvidia is willing to go to seed demand in markets that cannot fund AI infrastructure on their own. Indonesian startups and enterprises gain access to domestic GPU capacity, while the five banks and any neocloud operator that copies the template carry the residual risk. If AI compute rental economics hold, the model opens a financing channel for Nvidia's growth beyond the US hyperscaler market. If they soften, losses land first on lenders whose collateral is a rapidly depreciating GPU fleet. ## Related Articles - [ByteDance Unsecured AI Loan: Banks Bet $29.6B on Cash Flow, Not Collateral](https://bytevyte.com/bytedance-unsecured-ai-loan-banks-bet-29-6b-on-cash-flow-not-collateral/) - [Nvidia OpenAI Circular Financing: $250 Billion Guarantee Reshapes AI Infrastructure Risk](https://bytevyte.com/nvidia-openai-circular-financing-250-billion-guarantee-reshapes-ai-infrastructure-risk/) - [General Compute's Inference Chip Loan Opens Financing Path](https://bytevyte.com/general-computes-inference-chip-loan-opens-financing-path/) ✔Human Verified --- *Researched and cross-referenced against primary sources by the Bytevyte editorial team. This article was generated with the assistance of artificial intelligence and reviewed by the Bytevyte editorial team.*