Zankore secures $3.1B AI loan for Southeast Asia compute build-out

A large loan for a new AI infrastructure platform

Zankore has secured a senior term loan facility of up to $3.1 billion to support the development and deployment of next-generation AI and GPU cloud infrastructure across Southeast Asia [1]. The financing is one of the day’s clearest signs that banks are beginning to treat AI compute not merely as a technology cycle, but as a bankable infrastructure category involving data centers, power capacity, GPUs, networking and long-term customer demand [3].

The company is starting with Indonesia as its first deployment market, where it says it is building an initial 100 MW of Nvidia AI infrastructure [1]. Its broader plan remains far larger: Zankore has linked the financing to a target of up to 1 GW of Nvidia DSX AI Factory capacity and roughly 200 MW of initial capacity in the first half of 2027 [1]. That combination of near-term deployment and long-term target makes the deal significant not just for Zankore, but for Southeast Asia’s effort to localize more of the compute capacity required for AI training, fine-tuning, inference and agentic applications [1].

The loan comes from a consortium of five banks, with Citi, ING, Natixis CIB, Qatar National Bank and United Overseas Bank acting as senior mandated lead arrangers, underwriters and bookrunners [1]. Citi also acted as exclusive debt adviser on the facility [1]. FinanceAsia described the transaction as a syndicated deal from five banks that will help the newly formed firm build GPU cloud infrastructure in Indonesia alongside Nvidia [2].

From launch announcement to financed build-out

Zankore was launched in August 2026 with Indosat Ooredoo Hutchison, Ooredoo Group, Nvidia and Nokia as central participants in the platform [1]. The new financing moves the project from announcement to a more concrete capital phase, because the proceeds are intended to support the acquisition and deployment of advanced Nvidia GPU infrastructure [1]. Reuters, carried by Boursorama, also reported that the debt will finance the deployment of cloud-computing and Nvidia GPU infrastructure as Zankore expands AI capacity in Indonesia and across Southeast Asia [5].

That matters because AI infrastructure is unusually capital intensive. A GPU cloud is not simply a room full of chips. It requires land, substations, resilient power supply, cooling systems, high-performance networking, software orchestration, maintenance teams and customers willing to reserve or consume capacity at scale. TechNode Global noted that the size and structure of the loan highlight AI infrastructure becoming a financing category in its own right, with GPU deployments requiring large upfront commitments beyond the processors themselves [3].

The project’s design also tries to match capital deployment with demand. Zankore said a revenue-sharing and credit-support model will align infrastructure deployment with customer demand and create a more capital-efficient path to scale [1]. The Next Web reported that the facility includes revenue-sharing and credit-support arrangements intended to link the pace of deployment to customer demand [4]. For lenders, that kind of structure can be important because the economics of AI infrastructure depend heavily on utilization: idle GPUs can become expensive very quickly, while contracted, power-ready compute can resemble infrastructure cash flow.

Why the banks are treating compute like infrastructure

The striking feature of the transaction is not only its size, but its form. Zankore is a recently launched platform, yet five major lenders have underwritten a multibillion-dollar facility tied to AI compute [2]. The Next Web highlighted that the company launched only a month before the loan announcement and is backed by Indosat Ooredoo Hutchison, Ooredoo Group, Nokia and Nvidia [4].

The bank group’s public comments frame AI compute as a new kind of infrastructure asset. Zankore’s announcement said the transaction reflects growing confidence in demand for AI computing and in the company’s ability to capture the opportunity [1]. Natixis CIB described AI compute as an emerging asset class requiring innovative financing solutions, while ING pointed to the substantial capital demands associated with AI compute and GPU deployments [1]. UOB framed the project as part of ASEAN’s digital future, while QNB linked the financing to Indonesia’s ambition to become a regional hub for AI infrastructure [1].

That language is important. It suggests lenders are increasingly willing to underwrite the physical layer of AI: power-hungry data-center capacity, accelerator purchases and cloud platforms designed for enterprise workloads. For years, venture capital and corporate balance sheets carried much of the risk of AI development. Zankore’s loan points to a different model, in which banks finance the compute stack because they see it as essential economic infrastructure.

Still, this is not risk-free infrastructure in the traditional sense. The Next Web noted that Nvidia is both a backer of Zankore and the supplier of the GPUs the financing will help buy [4]. That structure is disclosed, and it can align incentives, but it also means some of the demand for Nvidia hardware is being financed through the balance sheets of partner platforms and lenders [4]. If regional customers absorb the capacity, the model could accelerate Southeast Asia’s AI economy. If demand lags, lenders may be exposed to specialized hardware whose value can fall as chip cycles move quickly.

Indonesia as the first proving ground

Indonesia is central to the project because Zankore is beginning its build-out there [1]. Reuters reported that Zankore is initially building 100 MW of Nvidia AI infrastructure to serve AI-native entities in Indonesia and Southeast Asia [5]. The company says the platform will provide high-performance GPU clusters for training, fine-tuning, large-scale inference and agentic AI applications [1].

The Indonesian context gives the financing strategic weight. Southeast Asia’s AI users have often relied on cloud capacity concentrated in the United States, Europe and a few advanced Asian data-center hubs. Zankore’s pitch is that local and regional compute can allow enterprises, startups, developers and public institutions to build closer to their users, data and industries [1]. TechNode Global reported that the platform is intended to serve customers across Indonesia and the wider Southeast Asian market, with Indonesia positioned as the first deployment base [3].

Indosat’s role helps explain why banks might view the platform as more than a speculative start-up. The Next Web noted that Indosat Ooredoo Hutchison is one of Indonesia’s largest mobile operators and brings assets that matter to infrastructure finance, including land, power connections, regulatory position and balance-sheet support [4]. Reuters reported that Qatari telecom operator Ooredoo is Zankore’s founding shareholder and lead investor with a 49% stake [5].

Those relationships do not remove execution risk, but they give the project industrial sponsors, telecom infrastructure and a regional commercial channel. They also place the loan in a broader race among telecom operators, cloud providers and chip companies to control the physical foundation of AI adoption in fast-growing markets.

The bottlenecks: power, chips and utilization

The loan’s scale also sharpens the constraints that come with AI build-outs. Zankore can raise debt, but it must still turn financing into operating capacity: GPU procurement, data-center readiness, power availability, cooling, interconnection and skilled operations. The company says the financing will help accelerate deployment of large-scale Nvidia GPU computing and cloud-based AI services [1].

Power may become one of the most important tests. A 100 MW initial deployment is already a major energy commitment, while a 1 GW target would place Zankore among the most ambitious AI infrastructure platforms in the region [1]. Each phase will require not just equipment but dependable electricity, grid coordination and commercially viable operating costs. Zankore said its long-term ambition will be realized progressively, with each phase focused on strengthening infrastructure, commercial ecosystem and operational capabilities [1].

Utilization is the other test. The economics of a GPU cloud depend on whether enterprises, AI-native companies, developers and institutions actually buy the capacity. Zankore says its services will support model training, fine-tuning, inference and agentic AI workloads [1]. TechNode Global noted that how quickly proposed capacity is absorbed will depend on enterprise demand, regional AI adoption and the economics of operating power-intensive GPU infrastructure at scale [3].

The cloud now comes with a mortgage

Zankore’s $3.1 billion loan is best understood as a financing milestone for the AI era. The transaction does not merely fund servers; it attaches long-term debt to the physical systems that make AI possible. It shows that lenders are willing to finance GPUs, power and data-center capacity when the sponsors, technology partners and demand story are strong enough.

For Southeast Asia, the promise is a deeper local compute base and less dependence on remote capacity. For Zankore, the challenge is to convert a landmark financing package into operational infrastructure that customers use at scale. For the wider market, the message is simple: AI capacity is no longer being built only with venture capital or hyperscaler budgets. Increasingly, it is being financed like infrastructure — and the cloud now comes with a mortgage.

Sources from the last 72 hours

1. [1]Zankore Expands Southeast Asia’s AI Compute Infrastructure With NVIDIASep 8, 2026, 5:00 PM UTC
2. [2]Zankore secures $3.1bn loan for AI infrastructureSep 8, 2026, 4:00 PM UTC
3. [3]Zankore secures up to $3.1B financing for Southeast Asia AI infrastructure buildoutSep 8, 2026, 4:00 PM UTC
4. [4]Nvidia-backed Zankore raises a $3.1bn loan to buy Nvidia GPUs for IndonesiaSep 9, 2026, 11:20 AM UTC
5. [5]Zankore, soutenue par Ooredoo, lève 3,1 milliards de dollars pour la plateforme cloud d'IA de NvidiaSep 9, 2026, 9:57 AM UTC

AI-generated article based on recent web research, then preserved as a dated editorial snapshot.