Zankore

1.1 Identity, launch, and operating model

Zankore was launched in August 2026 as an Indonesian AI cloud and neocloud platform developed by Indosat Ooredoo Hutchison, Ooredoo Group, NVIDIA, and Nokia. The retained evidence supports Indonesia as the initial deployment market, but it does not identify a legal headquarters or fully describe the incorporated entity. The operating model is infrastructure-as-a-service: Zankore intends to provide on-demand NVIDIA GPU clusters for training, fine-tuning, inference, and agentic workloads to enterprises, developers, startups, and institutions. Its disclosed status is therefore early buildout rather than mature operation: an initial 100 MW deployment is being financed, approximately 200 MW is targeted for the first half of 2027, and 1 GW remains a longer-term target.[CO001, CO002, CO003, CO004, CO005, CO006]

1.2 Ownership, leadership, and governance visibility

Ooredoo Group is the founding shareholder and lead investor, with a disclosed 49% stake and an approximately US$800 million commitment for launch and initial development. Indosat contributes local infrastructure and execution, NVIDIA supplies the accelerated computing architecture and GPUs, and Nokia supplies AI-native networking. Public leadership visibility is much thinner than sponsor visibility. Vikram Sinha is identified as Zankore's director and chairman, while Aziz Aluthman Fakhroo speaks for the lead investor as Ooredoo Group CEO. The retained sources do not name a Zankore operating CEO, founders as individuals, a complete board, independent directors, or leadership changes. This leaves decision rights, sponsor reserved matters, and key-person dependence open for primary-document diligence.[CO002, CO007, CO014, CO015]

1.3 Capital structure and milestones

Zankore's largest disclosed financing is debt, not revenue, ARR, equity raised, or cash available. A five-bank consortium arranged a senior term loan facility of up to US$3.1 billion to acquire and deploy NVIDIA GPU infrastructure. Reporting also describes revenue-sharing and credit-support arrangements intended to pace deployment with demand. Natixis further disclosed that its tranche is secured by GPU equipment, material contracts, and take-or-pay offtake agreements with paying customers. Those protections indicate some contracted demand, but no retained source discloses customer names, count, utilization, pricing, draw status, maturity, interest cost, or covenant headroom. The financing headline therefore establishes available debt capacity, not enterprise value or proven operating scale.[CO009, CO010, CO011, CO012, CO013]

1.4 Diligence interpretation and unresolved cover metrics

The sponsor set brings telecom infrastructure, capital, networking, and GPU supply, but the platform's short operating history makes execution evidence more important than announced capacity. Ooredoo estimates approximately US$600 million of cumulative EBITDA contribution to itself over the investment's first five years; that is a sponsor forecast, not Zankore revenue or current EBITDA. An adverse analysis characterizes the disclosed debt as roughly four times founding equity and argues that utilization and rapidly repricing GPU collateral are the central risks. The retained pool cannot establish valuation, revenue or run-rate, customer count, headcount, deployed location count, or current utilization. Investment judgment should remain conditional on the cap table, loan documents, contracted backlog, deployment acceptance records, and independently reconciled operating metrics.[CO008, CO016, CO004, CO005, CO006, CO009, CO013]

1.5 Exhibits

2.1 Market boundary and status quo

Zankore's relevant market is regional GPU-as-a-Service and integrated AI infrastructure for Southeast Asian cloud, enterprise, and government workloads, initially deployed in Indonesia. The included spend is dedicated accelerated compute capacity, associated AI software and networking, and the power, cooling, and operating services needed to deliver that capacity. It is not the whole telecom market, all conventional data-centre colocation, or the full value of NVIDIA chip sales. Public descriptions place Zankore between a hyperscaler-like compute service and a carrier-managed infrastructure bundle: the platform is intended to sell GPU capacity for training, fine-tuning, inference, and agentic workloads, while combining telecom infrastructure with accelerated computing, networking, power, cooling, and operations (SM007; SM006). The practical status quo is a buyer assembling those layers from hyperscalers, independent neoclouds, data-centre operators, or an internal deployment. Zankore's proposed alternative is a regional provider that can bundle network and compute, which may simplify procurement but increases concentration and portability questions.[CM001, CM002, CM003]

2.2 Sizing lenses and evidence limits

The strongest public sizing lens is physical capacity rather than revenue. Ooredoo's announcement targets 1 GW of NVIDIA DSX AI Factory capacity across Southeast Asia, with approximately 200 MW expected in the first half of 2027 (SM005). Independent reports describe a staged plan beginning with a 100 MW Indonesian buildout and reaching approximately 200 MW in the first half of 2027, so these values are different deployment horizons rather than conflicting estimates (SM003; SM007). Ooredoo also cites industry analysis that Southeast Asian data-centre capacity demand could grow approximately 3.5x by 2030, but that is a regional demand indicator, not Zankore revenue or a serviceable market calculation (SM005). The capital lens is similarly not a revenue proxy: Ooredoo says it will commit approximately $800 million for a 49% stake, while independent reporting describes a senior term loan facility of up to $3.1 billion (SM005; SM003; SM007). These figures establish ambition and financing intensity, not customer demand, utilization, price, or cash returns. A quantitative SAM/SOM cannot therefore be isolated from the retained evidence without inventing utilization, GPU-hour pricing, contract coverage, or geographic allocation.[CM004, CM005, CM006, CM007, CM008, CM009, CM016]

2.3 Buyers, adoption drivers, and constraints

The disclosed buyer set has three primary segments. Cloud customers and AI developers need flexible GPU capacity for training, fine-tuning, inference, and agentic workloads; enterprises need a path from AI experimentation to mission-critical deployment; and governments need sovereign or regionally controlled infrastructure (SM007; SM006). In these segments, the economic buyer is mapped, rather than disclosed, as a cloud or technology organization, an enterprise infrastructure or AI budget owner, or a public-sector digital authority; the reviewed sources do not identify actual budget owners. The adoption case is strengthened by integrated power, cooling, networking, and compute, and by the company claim that DSX MaxLPS can enable up to 40% more compute within the same power budget (SM005; SM006). The main constraints are pre-commercial buildout, demand conversion, debt service, and vendor concentration. Reporting says the loan includes revenue-sharing and credit-support arrangements designed to link deployment pace to customer demand (SM007), while the first approximately 200 MW is not expected until the first half of 2027 (SM001; SM005). Buyers should therefore require workload portability, compute-specific service levels, pricing and capacity commitments, and evidence of contracted utilization before treating the capacity target as realized market penetration.[CM010, CM011, CM012, CM013, CM014, CM015, CM016]

2.4 Exhibits

3.1 Landscape and strategic position

Zankore is positioned as an AI neocloud rather than a general-purpose colocation operator. Ooredoo's release describes a target of 1 GW of NVIDIA DSX AI Factory capacity, while the initial deployment is described as approximately 200 MW in the first half of 2027; The Next Web separately reports a 100 MW starting build. Those figures indicate staged deployment, not one settled operating capacity. The public partner set combines Ooredoo's capital and regional reach, Indosat's Indonesian market and infrastructure position, Nokia networking, and NVIDIA hardware and software. The resulting proposition is a locally hosted, high-density compute platform for enterprises, developers, hyperscalers, governments, and AI companies. That is differentiated from ordinary cloud only if Zankore can translate local sovereignty, latency, and regulated-workload access into contracted demand. The current sources identify intended customer groups, but do not name anchor customers or disclose committed workload volumes.[CP001, CP002, CP003, CP004]

3.2 Buyer criteria, capability, and pricing

Buyers comparing Zankore with hyperscalers, incumbent Indonesian data centers, or another neocloud will care about usable GPU-hours, latency, data location, reliability, networking, support, security, and price after discounts. Zankore's disclosed capability case is specific on hardware and architecture: GB300 NVL72 systems, NVIDIA DSX, and DSX MaxLPS, with Ooredoo claiming that MaxLPS is designed to enable up to 40% more compute within the same power budget. That is a vendor-stated engineering proposition, not an independently measured customer outcome. Public materials do not provide a Zankore price card, utilization, service-level agreement, GPU-hour availability, or realized discount. The pricing comparison therefore separates observed list-price evidence from unknowns rather than inferring parity with global providers. A buyer may rationally pay for Indonesian location or sovereign handling, but that willingness remains an underwriting question until contracts or customer references are disclosed.[CP005, CP006, CP007, CP008]

3.3 Switching costs, distribution, and moat durability

Zankore's potential moat is infrastructural and relational, not a proprietary model moat demonstrated in the assigned evidence. Indosat contributes local execution and connectivity, Ooredoo supplies capital and regional scale, Nokia supplies AI-native networking, and NVIDIA supplies the accelerated-computing stack. This combination can shorten permitting, power, network, and deployment paths relative to a new entrant starting without a telecom partner. It can also create distribution into Indonesian enterprises and public-sector workloads. However, the same stack creates multi-homing and substitution risk: customers can compare Zankore with hyperscalers, existing regional facilities, and other NVIDIA-backed neoclouds. The Next Web reports that loan economics depend on customer demand and that GPU collateral could lose value as the market changes. That makes customer contracts, service reliability, and workload portability more important than headline capacity. No source establishes exclusivity, minimum-volume commitments, or a switching-cost schedule.[CP009, CP010, CP011, CP012]

3.4 Displacement and competitive risks

The adverse case is that Zankore builds an expensive supply position before regional demand is proven. The Next Web notes that the company launched only about one month before securing up to $3.1 billion of senior debt and says the structure leaves lenders exposed if customers do not sign enough contracts. AInvest makes the sharper analytical argument that the debt is roughly four times the approximately $800 million founding equity and that GPU value is periodically repriced; because it is an opinionated analysis, the leverage framing should not be treated as audited balance-sheet data. Remio's analysis adds a separate displacement constraint: the 1 GW ambition depends on cooling, water, and energy choices that remain undisclosed, while phased expansion should be conditioned on measured performance and demand. Together these sources imply three diligence tests: named anchor demand, financing and collateral terms, and a credible resource plan. Until those are available, Zankore's partner advantage may improve access to infrastructure without proving a durable economic moat.[CP013, CP014, CP015, CP016]

3.5 Exhibits

4.1 Monetization and public traction

Zankore's disclosed revenue mechanism is AI compute sold to cloud, enterprise, and government customers, with Indonesia as the first deployment market and Southeast Asia as the expansion footprint. The platform is described as a full-stack AI factory combining NVIDIA compute, Nokia networking, power, cooling, and operations; that architecture explains the intended service rather than proving a realized price or margin. Public materials identify a target of approximately 200MW in the first half of 2027 and a longer-term 1GW target, but they do not disclose live revenue, ARR, utilization, signed customer names, contract tenor, or realized price per GPU-hour. A third-party account says no anchor customers are named and the terms tied to deployment pace and customer demand are not disclosed. Accordingly, the current traction record supports capacity planning and sponsor intent, not underwriting of recurring revenue.[CI001, CI002, CI003, CI004]

4.2 Capital structure and financing dependency

The disclosed capital stack is unusually debt-heavy for a greenfield neocloud. Ooredoo says it will commit approximately US$800 million for a 49% founding stake, while later coverage reports a senior term loan facility of up to US$3.1 billion arranged by five banks to fund the GPU buildout. Those figures are not equivalent to cash available: the equity commitment is spread over the investment period, and the loan is a facility whose draw, covenants, collateral terms, pricing, amortization, and conditions precedent are not public in the assigned evidence. The resulting headline ratio is roughly US$3.1 billion of maximum debt facility to US$800 million of stated sponsor commitment, but that is not a verified leverage ratio for drawn debt or total capitalization. An adverse analysis characterizes the facility as collateralized GPU lending and highlights utilization, depreciation, and hardware repricing as the central risks. No cash balance, burn, runway, refinancing plan, or debt-service coverage is disclosed.[CI005, CI006, CI007, CI008, CI009]

4.3 Margin path and unit-economics evidence

Public unit economics are not sufficient for a conventional SaaS or marketplace underwriting model. An equity-research estimate assumes development capex of US$20-25 million per MW, annual revenue of US$10-13 million per MW, and a 69% EBITDA margin, but these are analyst assumptions based on management disclosure and peer ranges, not reported Zankore results. The same analysis estimates project IRR at 12-13% and equity payback at less than three years; these outputs should be treated as estimates contingent on utilization, pricing, power cost, GPU life, and financing terms. The company and parent materials instead disclose cumulative revenue/EBITDA or attributable EBITDA estimates without a customer-level bridge. There is no public realized price, GPU utilization, gross margin, power cost, maintenance cost, depreciation schedule, CAC, payback, or contribution margin. The key financial conclusion is therefore that margin potential is scenario-based and the evidence needed to test it is private.[CI010, CI011, CI012, CI013]

4.4 Underwriting verdict and diligence blockers

Zankore has a credible sponsor and infrastructure-partner base, but the financial case is not yet auditable from public information. Ooredoo's release describes approximately US$600 million of cumulative contracted EBITDA contribution to Ooredoo over the investment period, while other public estimates describe roughly US$13 billion of cumulative revenue and US$9 billion of EBITDA over five years; these are company or analyst estimates, not audited Zankore statements, and their differing scope and denominators prevent direct reconciliation. The initial build therefore appears capital-intensive and financing-dependent, with demand ramp and collateral value as the principal downside variables. Underwriting should require executed offtake contracts, customer concentration, pricing and utilization schedules, debt documentation, draw conditions, covenant headroom, power and cooling costs, capex invoices, and monthly cash-flow forecasts. Until those materials are supplied, the appropriate stance is conditional interest rather than a quantified valuation.[CI014, CI015, CI016]

4.5 Exhibits

5.1 Product and customer workflow

Zankore's product is GPU cloud capacity rather than a general-purpose application. The available descriptions identify training, fine-tuning, inference, and agentic workloads as intended uses, with customers ranging from AI-native companies and developers to enterprises, institutions, hyperscalers, and governments. The customer workflow is therefore to obtain dedicated or on-demand NVIDIA compute, run models close to regional data and users, and consume an integrated infrastructure service rather than assemble every layer independently. This is a coherent product definition, but it remains a launch and deployment plan: the public evidence does not disclose live tenants, service-level objectives, utilization, pricing, or measured workload outcomes. The commercial proposition should consequently be read as intended capacity and target segments, not verified adoption.[CE001, CE002, CE003, CE004]

5.2 Architecture and operating model

The proposed operating model combines NVIDIA DSX as the reference architecture with accelerated computing, networking, power, cooling, and operations in one AI-factory system. The disclosed hardware milestone is approximately 200 MW in the first half of 2027 using NVIDIA GB300 NVL72, within a longer-term target of up to 1 GW. NVIDIA DSX MaxLPS is described as dynamically optimizing power across the GPU fleet and as designed to enable up to 40% more compute within the same power budget. Nokia contributes AI-native networking, while Indosat contributes local infrastructure and execution. These are useful architecture and responsibility boundaries, but the sources do not publish rack topology, software interfaces, measured power-use effectiveness, redundancy design, or commissioning evidence.[CE005, CE006, CE007, CE008, CE009]

5.3 Maturity, dependencies, and differentiation

Zankore's differentiation is primarily ecosystem and infrastructure integration: Ooredoo supplies capital and regional scale, Indosat supplies Indonesian infrastructure and execution, NVIDIA supplies GPUs, software, and the DSX playbook, and Nokia supplies networking. That combination can reduce integration burden for regional buyers and support sovereign or data-resident deployments. It also creates concentrated dependencies on NVIDIA hardware and software, telecom-site execution, power availability, cooling, network delivery, and partner governance. Public descriptions establish the partners and planned milestones, but not proprietary software, patent protection, independent benchmark results, completed certifications, or a production roadmap with acceptance criteria. The strongest current moat is therefore execution access and partner coordination, not yet independently demonstrated technical superiority.[CE010, CE011, CE012]

5.4 Trust, reliability, and diligence blockers

The launch materials emphasize secure, scalable, sovereign, and trusted AI infrastructure, and identify governance through a board and chief executive. Those statements describe intended positioning rather than evidence of controls. No assigned source provides an uptime history, incident record, penetration-test result, SOC or ISO certification, data-processing terms, model or tenant isolation design, disaster-recovery targets, cooling telemetry, or customer support metrics. A financing report also notes that the loan's collateral was not clear and that GPU resale value could depend on the same demand expected to repay the loan; that is a capital and asset-risk consideration for the operating platform, not proof of a product defect. Before underwriting product maturity, diligence should obtain deployment acceptance records, security attestations, capacity reservations, customer contracts, and evidence that the first capacity milestone is operational.[CE013, CE014, CE015]

5.5 Exhibits

6.1 Customer universe and buying surface

Zankore's stated market spans enterprises, startups, developers, AI-native companies and institutions, initially in Indonesia and then across Southeast Asia. The intended workloads include model training, fine-tuning, inference and agentic AI, implying a mixed buying surface: infrastructure or technology leaders may procure capacity, technical teams use it, and regulated or institutional users may value local deployment. Independent descriptions also use broader "cloud and enterprise customers" language and an Asia-Pacific frame. These are target segments, not observed cohorts. The sources provide no segment-level customer count, pricing, revenue mix, sales channel attribution or evidence that any category is already using Zankore in production, so strategic value by segment remains unquantified.[CU001, CU002, CU003, CU010, CU011]

6.2 Adoption trajectory and named-customer proof

The observable trajectory is infrastructure-led. Reporting describes an initial 100 MW build, an official plan for approximately 200 MW in the first half of 2027, and a longer-term target of up to 1 GW. Zankore also describes revenue-sharing and credit-support arrangements intended to align the pace of deployment with demand. Those milestones may create capacity for customer adoption, but they do not establish active users, utilization, contracted backlog or realized revenue. Ooredoo's estimate of approximately US$600 million in cumulative EBITDA contribution over the first five years is an investor projection, not a disclosed customer metric. Across the nine assigned sources, no customer is named and no production deployment or pilot is documented at contract level.[CU004, CU005, CU006, CU007, CU008, CU009]

6.3 Durability, expansion and concentration risk

Public evidence does not yet support a retention or expansion thesis. None of the reviewed sources reports net or gross revenue retention, churn, renewals, contract length, cohort behavior, satisfaction, customer-level utilization, backlog, offtake or revenue concentration. The demand-linked financing structure makes those omissions material: TNW notes that inadequate contracting could leave lenders exposed to depreciating GPUs, while AInvest argues that anchor customers must arrive quickly enough to support the leveraged buildout. These are analytical warnings rather than evidence of actual customer loss. Diligence should therefore prioritize executed contracts, named references, utilization by cohort, renewal mechanics and top-customer exposure before interpreting planned megawatts as durable commercial traction.[CU012, CU013, CU014, CU015]

6.4 Exhibits

7.1 Leverage, demand, and hardware economics

Zankore is attempting a capital-intensive build before it has a demonstrated operating history. The announced senior term loan is up to US$3.1 billion, while Ooredoo's contemporaneous release describes an approximately US$800 million commitment for a 49% founding stake. The sources do not disclose Zankore's cash-flow profile, customer contract coverage, interest rate, covenants, amortization, or collateral package. The financing therefore creates a timing risk: the platform must bring capacity online and convert it into contracted, paying utilization while GPU values and rental prices can change. The Next Web report specifically notes that the loan's collateral remains unclear and that GPU collateral could lose value if demand disappoints. The practical monitoring test is signed, creditworthy offtake and realized utilization against debt service, not the headline gigawatt target. The public comparison between the maximum facility and Ooredoo's commitment is a warning signal, not a verified debt-to-equity ratio, because total paid-in equity and cash available to Zankore are not disclosed.[CR001, CR002, CR003, CR004, CR005, CR014, CR015]

7.2 Build, power, and operational execution

The rollout depends on delivering a technically integrated infrastructure program across Indonesia and later Southeast Asia. Public materials describe an initial 100 MW build, approximately 200 MW targeted in the first half of 2027, and an eventual 1 GW ambition, but these are plans rather than evidence of commissioned capacity. The stack also depends on NVIDIA GB300 NVL72 systems, NVIDIA DSX and DSX MaxLPS, network integration from Nokia, and local execution by Indosat Ooredoo Hutchison. That combination creates schedule, power, cooling, supply, commissioning, reliability, and service-quality dependencies. A missed phase could leave debt-funded equipment underutilized while financing exposure remains. The announced revenue-sharing and credit-support model is directionally aligned with demand, but its contracts and operating effect are not public. A prudent mitigation is stage-gated deployment tied to power-ready sites, acceptance testing, uptime and utilization thresholds, with an explicit pause trigger before the next capacity tranche.[CR006, CR007, CR008, CR009, CR010, CR016]

7.3 Partner, regulatory, and governance exposure

Zankore's partner model is a strength and a concentration risk at the same time. Ooredoo is described as lead investor with a 49% stake, Indosat contributes Indonesian market and infrastructure capabilities, NVIDIA supplies the accelerated-computing stack, and Nokia supplies networking. This creates dependency on counterparties that also have commercial interests in the build, while the public record reviewed here does not establish board controls, related-party protections, customer concentration, permits, data-governance approvals, incident history, or insurance. The reviewed adverse material raises questions about demand timing, GPU collateral, and the circularity of vendor financing, but it does not establish a default or misconduct. Accordingly, regulatory and legal risk should remain an evidence gap rather than an invented incident: diligence should obtain permits, financing documents, security and privacy controls, litigation searches, and a full dependency register before treating the risk as low.[CR004, CR011, CR012, CR013, CR014]

7.4 Exhibits

8.1 Investment call and price discipline

The recommended stance is research-more and track rather than buy. Ooredoo's approximately US$800 million commitment for a 49% stake and the up-to-US$3.1 billion senior term facility demonstrate sponsor and lender capacity, while the planned 100 MW-to-200 MW-to-1 GW expansion offers material operating leverage if utilization develops as planned. Those strengths do not establish an investable price. The public record does not disclose a current equity valuation, loan pricing, maturity, covenants, collateral package, preference stack, or a named anchor customer. Entry discipline should therefore be expressed as conditions rather than a point valuation: require a fully diluted capitalization table, financing waterfall, contracted revenue schedule, unit economics, and downside recovery analysis before underwriting a return.[CV001, CV003, CV007, CV008, CV009, CV014, CV016]

8.2 Thesis, anti-thesis, and financing context

The thesis rests on unusually strong strategic sponsorship, bank participation, and a regional capacity plan positioned against company-cited growth in Southeast Asian data-center demand. Ooredoo also estimates approximately US$600 million of cumulative EBITDA contribution over the first five years, but that is a sponsor forecast rather than audited Zankore performance. The anti-thesis is that lenders have committed substantial capital after only about four weeks of operating history, with deployment linked to customer demand and credit support whose terms remain undisclosed. NVIDIA is both a backer and the hardware supplier, so part of ecosystem demand is financed through a vehicle buying NVIDIA equipment. This alignment can accelerate deployment, but it also concentrates technology, residual-value, and utilization risk.[CV001, CV002, CV003, CV005, CV006, CV010, CV011, CV015]

8.3 Scenario framework and comparable limitations

A model-appropriate valuation should link enterprise value to contracted gross profit or EBITDA after power, hosting, hardware financing, and refresh costs, rather than apply a generic software revenue multiple. The bull case requires contracted demand, on-time 2027 delivery, high utilization, and evidence that power optimization improves saleable compute economics. The base case assumes phased deployment and delayed utilization, making value highly sensitive to financing cost and customer ramp. The bear case combines weak offtake, GPU obsolescence, covenant pressure, and lower collateral recovery. No defensible valuation range or probability weighting can be published from the reviewed evidence because current valuation, forecast cash flows, comparable transactions, and debt terms are absent.[CV005, CV007, CV009, CV010, CV012, CV013, CV014, CV015]

8.4 Exit readiness, diligence asks, and thesis-break triggers

Exit readiness is early: Zankore launched in August 2026, and the public record is dominated by sponsor announcements and financing coverage rather than operating results. A strategic exit could become plausible if the platform converts planned capacity into diversified, creditworthy contracts and proves cash yields after GPU refresh obligations; an infrastructure or public-market route would additionally require audited segment reporting and durable leverage metrics. Before investment, obtain signed offtake schedules, customer concentration, utilization and pricing cohorts, project-level cash flows, debt and security documents, GPU refresh commitments, cap-table rights, and downside recovery cases. Thesis-break triggers are material slippage in the 2027 capacity plan, utilization below debt-service needs, customer concentration without guarantees, covenant stress, or evidence that equipment values are falling faster than principal amortizes.[CV006, CV007, CV009, CV010, CV013, CV014, CV016]

8.5 Exhibits