初创公司尽调
尽调报告 AI infrastructure and cloud computing early buildout 2026-10-22

Zankore

AI 新云基础设施平台

Zankore 获得了可信的战略支持,并制定了雄心勃勃的 AI 算力计划;但目前的融资规模已经领先于公开披露的运营和商业验证。

覆盖范围与披露说明

公开证据集中在 2026 年 8 月至 9 月的发布和融资窗口。产能目标、发起方承诺、债务能力和预测均有披露,但现有留存证据仍无法确认当前估值、收入、利用率、客户数、员工数及完整融资条款。

封面要素

领投方持股 01
49 % [CO002]
Ooredoo 承诺 02
800 US$ million [CO002]
高级债务融资 03
3100 US$ million [CO009]
长期产能目标 04
1 GW [CO005]

公司概况

Zankore 是一家新成立、聚焦印度尼西亚的 AI 新云,由 Ooredoo Group、Indosat Ooredoo Hutchison、NVIDIA 和 Nokia 支持。平台仍处于早期建设阶段,公开披露主要集中在发起方资本、大额高级债务融资和计划中的 GPU 产能,而非经营表现。

官网
zankore.com
创立地点
Indonesia
产品
按需提供 NVIDIA GPU 基础设施和云端 AI 算力,用于训练、微调、推理和智能体工作负载,并集成网络、电力、制冷和运营。
客户
面向印度尼西亚及更广泛东南亚市场的云客户、AI 开发者、企业、初创公司,以及政府或受监管用户。
商业模式
计划提供基础设施即服务和 AI 新云产能;公开来源未披露定价、利用率或签约收入。
阶段
Early buildout
融资情况
Ooredoo 披露将以约 8 亿美元承诺换取 49% 股权;报道还提到最高 31 亿美元的高级定期贷款额度。
[CO001, CO002, CO003, CO009, CO016]

执行摘要

主要优势

  • 战略支持汇集了电信基础设施、GPU 供应、网络能力和发起方资本。
  • 如果需求兑现,公开披露的融资和产能计划可以支撑区域 AI 算力规模化。
  • 首个印度尼西亚部署项目设定了明确的近期产能目标,也明确了对应的买方群体。

主要风险

  • 利用率、签约需求、定价、收入和客户集中度均未公开披露。
  • 大规模债务融资建设带来了偿债、GPU 残值和再融资风险敞口。
  • 对供应商和发起方的集中依赖可能限制可迁移性、治理透明度和下行保护。

未决问题

  • 获取当前估值、资本结构表、稀释条款和所有权分配顺序。
  • 核实已签署的承购合同、客户名称、签约 MW、定价、利用率和订单积压。
  • 审阅贷款文件,确认提款状态、定价、到期日、契约、抵押品和追索权。
  • 对照已投运产能、项目现金流、GPU 更新义务和经审计的运营指标。

目录

Chapter 01

01公司概览

1.1 身份、发布与运营模式

Zankore 于 2026 年 8 月发布,是由 Indosat Ooredoo Hutchison、Ooredoo Group、NVIDIA 和 Nokia 开发的印度尼西亚 AI 云和新云平台。留存证据支持印度尼西亚是首个部署市场,但没有确认法定总部,也未完整说明注册实体。其运营模式是基础设施即服务:Zankore 计划向企业、开发者、初创公司和机构按需提供 NVIDIA GPU 集群,用于训练、微调、推理和智能体工作负载。因此,公司披露的状态仍是早期建设,而非成熟运营:首个 100 MW 部署正在融资,2027 年上半年目标约 200 MW,长期目标为 1 GW。[CO001, CO002, CO003, CO004, CO005, CO006]

快照 KPI 与证据状态
指标数值或状态日期或期间置信度证据缺口
发布于 2026 年 8 月发布2026-08中未披露法定注册日期
初始市场印度尼西亚;计划扩展至更广泛东南亚截至 2026-10-22中未披露法定总部
商业模式面向 AI 工作负载的按需 NVIDIA GPU 产能截至 2026-10-22中未披露定价和利用率
初始建设100 MW NVIDIA AI 基础设施宣布于 2026-09中未披露已投运产能
近期目标约 200 MW2027 年上半年中交付时间表未独立验证
长期目标1 GW NVIDIA DSX AI Factory 产能长期目标中未披露分期和已承诺资本
领投方承诺约 8 亿美元换取 49%宣布于 2026-08-06中未披露出资时间表和已缴金额
债务额度最高 31 亿美元高级定期贷款宣布于 2026-09-09中未披露已提款余额、定价、到期日和契约
收入或年化收入留存来源未披露截至 2026-10-22低取得经审计管理账目和签约订单积压
估值、客户和员工数留存来源未披露截至 2026-10-22低取得股权结构表、客户名单和工资核对

数值区分承诺、债务能力、预测和产能目标;不可用指标不视为零,且任何显示金额都不代表手头现金。

[CO001, CO003, CO004, CO005, CO006, CO007, CO009]

1.2 所有权、领导层与治理透明度

Ooredoo Group 是创始股东和领投方,披露持股 49%,并承诺约 8 亿美元用于发布和初期开发。Indosat 提供本地基础设施和执行能力,NVIDIA 提供加速计算架构与 GPU,Nokia 提供原生 AI 网络。公开领导层信息远少于发起方信息。Vikram Sinha 被确认是 Zankore 董事兼主席,领投方则由 Ooredoo Group CEO Aziz Aluthman Fakhroo 代表发声。留存来源未披露 Zankore 的运营 CEO、个人创始人、完整董事会、独立董事或领导层变动。因此,决策权、发起方保留事项和关键人依赖仍需通过一手文件核查。[CO002, CO007, CO014, CO015]

领导层与创始人表
人员或席位公开描述的角色背景或覆盖依赖评估
Vikram SinhaZankore 董事兼主席Indosat 总裁董事兼 CEO;本地电信执行可见的发起方—高管依赖度高;运营授权未披露
Aziz Aluthman FakhrooOoredoo Group CEO 兼领投方发言人资本发起方和区域基础设施组合发起方影响力明显;Zankore 正式董事会权利未披露
运营 CEO 及更广泛董事会留存来源未披露无法测试职能覆盖重大治理和关键人缺口

本表仅列出指定证据中点名的领导层和发起方高管;未列入不代表不存在其他高管或董事。

[CO002, CO007, CO014, CO015]
利益相关方或投资者地图
利益相关方已披露角色控制或经济重要性优先尽调要求
Ooredoo Group创始股东和领投方持股 49%,承诺约 8 亿美元已执行认购文件、出资时间表、保留事项
Indosat Ooredoo Hutchison本地发起方及基础设施运营商印尼市场地位、土地、电力、监管和执行能力资产投入、服务协议、关联方定价
NVIDIA技术支持方及 GPU 供应商架构、软件生态、GPU 获取渠道及融资采购的硬件供应配额条款、集中度、更新义务、利益冲突
NokiaAI 原生网络合作伙伴规划中 AI 基础设施的网络底座商业范围、质保、验收里程碑
Citi、ING、Natixis CIB、QNB 与 UOB牵头安排行、承销商及簿记行最高 US$3.1 billion 的高级定期贷款融资贷款协议、提款条件、担保方案、契约
已签约客户或承购方Natixis 提及的保底付费及重大合同为有担保融资提供潜在需求支撑客户名称、已签约 MW、期限、信用质量、终止权

根据公开发布和融资披露整理的部分利益相关方清单;除 Ooredoo 的 49% 股权及贷款方承诺外,其他经济利益仍未量化。

[CO002, CO007, CO010, CO011, CO012, CO013]

1.3 资本结构与里程碑

Zankore 披露的最大融资是债务,而不是收入、ARR、已募股权或可用现金。五家银行组成的银团安排了最高 31 亿美元的高级定期贷款额度,用于采购和部署 NVIDIA GPU 基础设施。报道还提到收入分成和信用支持安排,意在让部署节奏跟随需求。Natixis 进一步披露,其贷款份额以 GPU 设备、重大合同及与付费客户签订的照付不议承购协议作担保。这些保护措施说明可能存在部分已签约需求,但留存来源没有披露客户名称、数量、利用率、定价、提款状态、到期日、利息成本或契约余量。因此,融资主标题 只能证明可用债务额度,不能证明企业价值或已验证的经营规模。[CO009, CO010, CO011, CO012, CO013]

FO001: Zankore 发布、融资与产能里程碑

公开证据集中在一段较短的发布和融资窗口内,大部分运营产能仍以未来目标的形式存在。

日期反映保留来源中的公告或目标时期;该时间线不把未来产能目标视为已投产基础设施。

[CO001, CO005, CO006, CO007, CO009, CO013]

1.4 尽调解读与封面指标缺口

发起方组合带来了电信基础设施、资本、网络和 GPU 供应,但平台经营历史很短,执行证据比宣布的产能更重要。Ooredoo 预计投资前五年累计为自身贡献约 6 亿美元 EBITDA;这是发起方预测,不是 Zankore 的收入或当前 EBITDA。逆向分析称,披露债务约为创始股权的四倍,并认为利用率和快速重定价的 GPU 抵押品是核心风险。留存材料无法确认估值、收入或 年化收入、客户数、员工数、已部署地点数或当前利用率。投资判断应以股权结构表、贷款文件、签约订单积压、部署验收记录和独立核对的经营指标为前提。[CO008, CO016, CO004, CO005, CO006, CO009, CO013]

1.5 展示项

Chapter 02

02市场分析

2.1 市场边界与现状

Zankore 所处市场是东南亚云、企业和政府工作负载使用的区域 GPU 即服务及集成 AI 基础设施,首个部署市场为印度尼西亚。纳入范围包括专用加速计算产能、相关 AI 软件和网络,以及交付这些产能所需的电力、制冷和运营服务;不包括整个电信市场、所有传统数据中心托管业务,也不包括 NVIDIA 芯片销售的全部价值。公开描述将 Zankore 定位在类似超大规模云的算力服务与运营商管理的基础设施组合之间:平台计划销售训练、微调、推理和智能体工作负载的 GPU 产能,同时整合电信基础设施、加速计算、网络、电力、制冷和运营(SM007;SM006)。现实中的替代方案是买方从超大规模云、独立新云、数据中心运营商或内部部署中自行拼装这些层。Zankore 提出的方案是区域供应商打包网络和算力,可能简化采购,但也增加集中度和可迁移性问题。[CM001, CM002, CM003]

市场定义表
细分市场 / 类别纳入的支出排除的支出买方 / 付款方相关性
区域 GPU 即服务用于训练、微调、推理和智能体工作负载的 GPU 容量不含计算能力的一般电信连接云及开发者基础设施预算核心服务类别
一体化 AI 基础设施加速计算、网络、电力、制冷和运营无关的企业软件及消费者电信企业基础设施或 AI 预算体现捆绑式方案的差异化
主权或区域 AI 容量以印尼为首、服务东南亚云、企业及政府工作负载的容量全球超大规模云厂商收入政府及受监管行业预算支持本地化和控制要求
现状替代方案超大规模云厂商、独立新云、托管机房或自建方案零支出或非 AI 项目选择其他交付模式的同一批买方划定竞争边界

边界根据公开披露的目标工作负载和平台组件综合整理;未披露 Zankore 特有的定价和客户合同。

[CM001, CM002, CM003]

2.2 规模化视角与证据限制

公开证据中最强的规模化视角是物理产能,而非收入。Ooredoo 宣布目标是在东南亚建设 1 GW 的 NVIDIA DSX AI Factory 产能,预计 2027 年上半年约 200 MW(SM005)。独立报道描述了分阶段计划:从印度尼西亚 100 MW 建设开始,2027 年上半年达到约 200 MW;因此这些数值代表不同部署阶段,并非相互冲突的估计(SM003;SM007)。Ooredoo 还引用行业分析称,东南亚数据中心产能需求到 2030 年可能增长约 3.5 倍,但这只是区域需求指标,不是 Zankore 收入或可服务市场计算(SM005)。资本视角同样不能替代收入:Ooredoo 称将以约 8 亿美元换取 49% 股权,而独立报道提到最高 31 亿美元的高级定期贷款额度(SM005;SM003;SM007)。这些数字证明的是雄心和融资强度,而非客户需求、利用率、价格或现金回报。基于留存证据,无法在不臆造利用率、GPU 小时价格、合同覆盖率或地域分配的情况下拆出定量 SAM/SOM。[CM004, CM005, CM006, CM007, CM008, CM009, CM016]

规模测算与融资视角
发布方年份地理范围数值CAGR 或增长信号方法信心度局限
Ooredoo Group2026东南亚1 GW 目标容量未披露公司关于规划中的 NVIDIA DSX AI Factory 容量的公告中容量目标不等于收入或已签约需求
Ooredoo Group2026印度尼西亚约 200 MW2027 年上半年交付公司关于规划中第一阶段的公告中规划容量和利用率未经独立验证
Data Center Dynamics2026印度尼西亚,继而扩展至东南亚初始建设 100 MW;最终目标 1 GW三年内扩张关于融资和部署计划的新闻报道中初始建设与后续 200 MW 里程碑属于不同的扩张周期
Ooredoo Group2026东南亚到 2030 年,数据中心容量需求约为 3.5 倍2030 年达到 3.5 倍公司公告引用的行业分析低未披露底层研究、分母或 Zankore 的份额
Ooredoo Group2026Zankore 投资约 $800 million;49% 股权五年承诺期公司投资公告中融资不等于平台可用现金或市场收入
Data Center Dynamics 与 The Next Web2026印尼首期部署最高 $3.1 billion 的高级定期贷款部署与客户需求挂钩关于 GPU 采购和部署的债务融资报道中债务不等于收入、市场规模、募集的股权资金或已证实可用的现金

这些视角不可相加,且使用的单位不同。表格区分了容量、区域需求增长、投资承诺和高级债务融资;市场份额及定价数据为 null,明确表示存在信息缺口。

[CM004, CM005, CM006, CM007, CM008, CM009, CM016]
FM001: 受证据约束的市场规模视角

已公开披露的产能和需求信号从区域背景逐步收窄到 Zankore 的计划部署,但商业变现规模仍无法量化。

各层宽度仅表示层级,不代表数值比例。

各项是单位不同的独立证据层,并包含两个明确未知项;该金字塔不按比例绘制,也不推导 TAM、SAM、SOM、转化率或市场份额。

[CM004, CM005, CM006, CM007, CM013, CM016]

2.3 买方、采用驱动因素与约束

已披露买方有三个主要群体。云客户和 AI 开发者需要灵活的 GPU 产能来训练、微调、推理和运行智能体工作负载;企业需要从 AI 试验走向关键任务部署;政府需要主权或区域可控的基础设施(SM007;SM006)。在这些细分市场中,经济买方只是被映射出来,并非已披露:可能是云或技术组织、企业基础设施或 AI 预算负责人,或公共部门数字主管;已审阅来源没有确认实际预算负责人。集成电力、制冷、网络和算力,以及公司声称 DSX MaxLPS 可在同一电力预算内实现最高 40% 更多算力,强化了采用逻辑(SM005;SM006)。主要约束是商业化前建设、需求转化、债务偿付和供应商集中度。报道称,贷款包含收入分成和信用支持安排,旨在把部署速度与客户需求挂钩(SM007);首批约 200 MW 预计要到 2027 年上半年才交付(SM001;SM005)。因此,在把产能目标视为已实现市场渗透前,买方应先要求确认工作负载可迁移性、算力专属服务水平、价格和产能承诺,以及已签约利用率证据。[CM010, CM011, CM012, CM013, CM014, CM015, CM016]

细分市场与买方图谱
细分市场买方用户付款方工作流预算负责人采用触发因素
云和 AI 开发者云平台或开发者服务负责人模型和平台团队云或计算预算训练、微调、推理、智能体工作负载云基础设施获得区域 GPU 容量
大型企业CIO 或 CTO 部门数据科学和应用团队企业 AI 或基础设施预算从实验转向关键任务部署CIO 或 CTO一体化计算与网络性能
政府及受监管用户数字事务部门或公共部门技术主管机构政府 AI 和数据团队公共数字基础设施预算主权或受控 AI 工作负载公共部门技术区域控制和国家数字竞争力目标
电信及工业部署网络或工业技术买方运营和 AI 工程团队网络或工业资本 / 运营支出预算专用网络或工业 AI 部署网络或运营负责人网络与计算捆绑采购

买方角色和工作流依据已披露的客户细分及工作负载描述作出分析映射;未保留具名的 Zankore 客户或已签署合同。

[CM010, CM011, CM012, CM014, CM015]

2.4 展示项

Chapter 03

03竞争者

3.1 格局与战略定位

Zankore 的定位是 AI 新云,而非通用型托管运营商。Ooredoo 的公告目标是 1 GW NVIDIA DSX AI Factory 产能,初始部署则描述为 2027 年上半年约 200 MW;The Next Web 另称起步建设为 100 MW。这些数字说明部署分阶段推进,而不是已有一个确定的运营产能。公开合作方组合了 Ooredoo 的资本和区域覆盖、Indosat 的印度尼西亚市场及基础设施地位、Nokia 的网络能力,以及 NVIDIA 的硬件和软件。最终方案是面向企业、开发者、超大规模云、政府和 AI 公司的本地托管高密度算力平台。只有当 Zankore 能把本地主权、低延迟和受监管工作负载接入转化为签约需求时,它才真正区别于普通云。当前来源只确认了目标客户群,没有确认锚定客户或签约工作负载量。[CP001, CP002, CP003, CP004]

竞争对手与替代方案画像
替代方案类别规模或融资证据目标客户差异化局限
Zankore直接的 AI 新云进入者目标 1 GW;2027 年首期约 200 MW;最高 $3.1B 高级定期贷款企业、开发者、超大规模云厂商、政府、AI 公司印尼选址,加上电信、网络、GPU 和资本合作伙伴没有具名锚定客户、公开价目表或已披露利用率
超大规模云厂商在位云厂商指定来源未提供可比的容量或融资数字广泛的企业和开发者工作负载成熟的服务广度、软件生态和多区域冗余印尼主权属性或本地延迟带来的差异可能较小
现有印尼或区域数据中心在位或相邻基础设施指定证据未量化其 AI-GPU 容量企业及托管工作负载现有设施、运营记录及可能更低的执行风险AI 专用 GPU 密度、软件和网络集成情况未知
其他 NVIDIA 支持的新云直接 / 相邻 AI 计算The Next Web 将 Firmus 描述为另一个 NVIDIA 支持的印尼项目;没有可比的 Zankore 合同数据AI 开发者和企业工作负载专用加速计算及 NVIDIA 生态获取能力共享 GPU 供应和融资模式可能加剧价格及抵押品风险
自建替代方案没有公开的 Zankore 对比或基准工作负载可预测的大型企业和政府掌控数据、架构和进度需要资本、运营能力、电力和专业人才

规模和融资字段仅限于指定公开来源;空白的比较指标不是估算值。竞争对手行代表市场类别,未保留具名公司的证据。

[CP001, CP003, CP009, CP013]

3.2 买方标准、能力与定价

买方在 Zankore、超大规模云、印度尼西亚现有数据中心或另一家新云之间比较时,会关注可用 GPU 小时、延迟、数据位置、可靠性、网络、支持、安全,以及折扣后的价格。Zankore 披露的能力论证具体涉及硬件和架构:GB300 NVL72 系统、NVIDIA DSX 和 DSX MaxLPS;Ooredoo 声称 MaxLPS 的设计目标是在同一电力预算内实现最高 40% 更多算力。这是供应商声称的工程命题,不是独立测量的客户结果。公开材料没有提供 Zankore 价目表、利用率、服务水平协议、GPU 小时可用量或实际折扣。因此,定价比较应区分已观察到的 标价证据和未知项,而不能推断与全球供应商价格相当。买方可能愿意为印度尼西亚所在地或主权数据处理付费,但在合同或客户参考资料披露前,这仍是投资判断问题。[CP005, CP006, CP007, CP008]

能力与采购标准对比
采购标准Zankore 证据超大规模云厂商或在位厂商对比对买方的含义
加速计算已点名 GB300 NVL72 和 NVIDIA DSX本证据池未提供对比硬件和配额信息核实交付的 GPU 类型、可用性和更新计划
电力与工作负载效率DSX MaxLPS 据称设计为在相同功耗范围内最多多提供 40% 的计算能力没有独立基准或对比结果在获得工作负载级测试前,将其视为公司口径或合作伙伴口径
网络与本地接入已描述 Nokia AI 原生网络以及 Indosat/Ooredoo 基础设施区域在位网络性能未量化测试延迟、交叉连接、出口流量和故障域
主权与受监管工作负载已说明以印尼为首的部署及政府 / 企业目标细分市场指定来源未覆盖超大规模云厂商的司法管辖区和服务控制获取数据驻留、审计和事件响应条款

单元格总结指定的公开描述,并非同口径产品测试。缺失的对比值仍属未知。

[CP004, CP005, CP006, CP009]
定价与打包证据
供应商或选项价格或合同模式包含的能力折扣或未知项竞争含义
Zankore据报道可提供用于训练、微调、推理和智能体工作负载的 GPU 容量没有公开标价、实际价格、SLA 或最低承诺目前无法评估定价能力和毛利潜力
超大规模云厂商广泛的云服务可作为类别级比较对象指定来源没有可比报价或 GPU 价格获取同口径的区域 GPU 小时报价
现有数据中心托管机房或基础设施替代具有可行性AI-GPU 打包方式及电力条款未知比较实际交付的计算能力,而不是宣传口径的 MW 数字
自建客户拥有整套技术栈并自行做运营决策资本、人员和利用率要求未知仅在工作负载规模足以支撑固定成本控制时采用

价格为 null,表示指定证据没有公开价格或合同报价;不表示服务免费或成本为零。

[CP007, CP008, CP010, CP012]
FP001: 有证据支持的竞争定位

序数定位突出了 Zankore 的本地基础设施准入优势,以及相对于广泛的在位者和替代类别仍未解决的商业验证问题。

条目本地基础设施与主权适配度(序数,1 为低,5 为高) (2–5)有公开证据的运营与商业成熟度(序数,1 为低,5 为高) (1–5)说明
1. Zankore 4 3 已披露的集成能力和本地准入较高;价格和已签约需求的公开证据较少
2. 超大规模云 2 5 生态广泛且运营成熟;本文未确立其本地主权适配性
3. 现有区域数据中心 3 4 可能拥有运营历史;AI 专用产能和集成技术栈未知
4. 内部建设 5 1 客户控制力最大;假设的执行负担最高
5. 其他 AI 新云 3 2 专用 GPU 产能;区域合同和融资的公开证据各不相同
6. 区域电信云 4 3 可能具备本地网络准入;AI 专用产能和软件集成未知

坐标是分析师根据引用证据给出的序数评分,不代表市场份额、性能或价格测量。

[CP003, CP007, CP009, CP013]

3.3 切换成本、分销与护城河持久性

Zankore 潜在的护城河来自基础设施和关系,而不是指定证据中已证明的专有模型。Indosat 提供本地执行和连接,Ooredoo 提供资本和区域规模,Nokia 提供原生 AI 网络,NVIDIA 提供加速计算栈。这一组合相较于没有电信伙伴的新进入者,可能缩短许可、电力、网络和部署路径,也可能把分销带入印度尼西亚企业和公共部门工作负载。但同一套栈也带来多重托管和替代风险:客户可以将 Zankore 与超大规模云、现有区域设施及其他 NVIDIA 支持的新云比较。The Next Web 报道称,贷款经济性取决于客户需求,而 GPU 抵押品可能随市场变化贬值。因此,客户合同、服务可靠性和工作负载可迁移性比 融资标题中的产能更重要。没有来源确认排他性、最低量承诺或切换成本表。[CP009, CP010, CP011, CP012]

3.4 替代与竞争风险

逆向情景是,Zankore 在区域需求得到验证前就建成昂贵的供应能力。The Next Web 指出,公司在取得最高 31 亿美元高级债务前仅成立约一个月,并称若客户没有签署足够合同,贷款方将承担风险。AInvest 给出更尖锐的分析:债务约为约 8 亿美元创始股权的四倍,且 GPU 价值会周期性重估;由于这是带有观点的分析,杠杆表述不应视为经审计的资产负债表数据。Remio 的分析提出另一项替代约束:1 GW 雄心取决于仍未披露的制冷、用水和能源选择,分阶段扩张应以实测表现和需求为条件。综合来看,需要做三项尽调测试:已命名的锚定需求、融资和抵押品条款,以及可信的资源计划。在这些材料到位前,Zankore 的伙伴优势或许能改善基础设施接入,但尚未证明持久的经济护城河。[CP013, CP014, CP015, CP016]

3.5 展示项

Chapter 04

04财务

4.1 变现与公开商业进展

Zankore 披露的收入机制是向云、企业和政府客户销售 AI 算力,首个部署市场为印度尼西亚,扩张范围为东南亚。平台被描述为整合 NVIDIA 算力、Nokia 网络、电力、制冷和运营的全栈 AI 工厂;这一架构说明了计划提供的服务,却不能证明实际价格或利润率。公开材料给出 2027 年上半年约 200 MW 的目标和长期 1 GW 目标,但没有披露实际收入、ARR、利用率、已签客户名称、合同期限或每 GPU 小时的实际价格。第三方报道称没有公开锚定客户,部署速度和客户需求相关条款也未披露。因此,当前商业进展记录支持的是产能规划和发起方意图,而不是经得起投资判断的经常性收入。[CI001, CI002, CI003, CI004]

收入来源与证据质量
收入来源机制单位或驱动因素当前公开状态质量尽调问题
AI 计算容量向云、企业和政府用户销售 GPU 基础设施MW / GPU 容量已设定目标;未披露实际收入公司目标已签署客户合同和已承诺 MW
全栈 AI 工厂一体化计算、网络、电力、制冷和运营容量及每单位功耗范围的有效计算能力已描述架构;未披露价格和利润率公司 / 合作伙伴描述成本和服务水平计划
区域扩张在东南亚新增 AI 基础设施长期容量最高 1GW战略目标;未披露站点和融资计划公司目标按站点拆分的资本开支和融资计划
AI 软件与生态模型和网络能力与计算能力并列定位挂靠基础设施的软件 / 服务在发布报道中被点名;未披露变现方式仅作描述SKU、附加率和合同经济性
Ooredoo 可归属经济利益发起方体现的 Zankore 经济利益份额约 US$600m 累计 EBITDA 估算投资期内母公司的估算公司声称的估算并表、所有权和现金分配路径

公开材料描述的是计划中的机制和目标,而非经审计收入。null 或未披露字段需要私下尽调;MW 是容量,不是收入。

[CI001, CI002, CI003, CI014]
定价与变现证据
定价项目标价 vs 实际价格公开数值信心度含义
GPU 计算价格实际价格不可得低无法计算收入或毛利率
合同模式合同条款不可得低需要期限、最低量、保底付费和调价机制
利用率运营利用率不可得低容量目标不能证明需求覆盖
客户集中度客户身份不可得低无法评估交易对手或续约风险
折扣与抵扣不可得低无法区分标价经济性和实际经济性

指定来源没有披露 Zankore 的标价、实际价格、利用率或合同条款。null 表示证据缺口,不是零值。

[CI003, CI004]
FI001: 公开披露的收入到融资桥接

股东承诺和债务能力拟用于资助 GPU 基础设施,而基础设施必须转化为已签约客户需求;运营转化步骤仍未验证。

节点 / 连接节点 / 起点终点说明
节点 1 Ooredoo 承诺 [sponsor-equity]US$800m stated commitment 49% 创始股东持股;时间和现金提款未完全披露
节点 2 高级定期贷款融资 [senior-debt]Up to US$3.1bn maximum facility 提款金额和条款未披露
节点 3 初始 AI 产能 [gpu-build]100MW / ~200MW stated deployment targets 对第一阶段的披露不一致;目标是在 2027 年上半年达到约 200MW
节点 4 已签约客户需求 [customer-demand]not disclosed 指定证据中没有具名锚定客户或承购条款
节点 5 收入与 EBITDA [revenue-ebitda]not reported 存在股东和分析师估计;实际结果不可得
节点 6 电力、冷却与服务成本 [operating-costs]not disclosed 需要这些数据来检验利润率和债务偿付假设
连接 1 Ooredoo 承诺 [sponsor-equity]初始 AI 产能 [gpu-build]提供资金
连接 2 高级定期贷款融资 [senior-debt]初始 AI 产能 [gpu-build]提供资金
连接 3 初始 AI 产能 [gpu-build]已签约客户需求 [customer-demand]需要
连接 4 已签约客户需求 [customer-demand]收入与 EBITDA [revenue-ebitda]转化为

这是证据图谱,而非现金流预测;其中保留不可得的节点,不推断收入。

[CI005, CI006, CI007, CI008, CI009]

4.2 资本结构与融资依赖

对一家从零建设的新云而言,已披露资本结构的债务占比异常高。Ooredoo 称将以约 8 亿美元取得 49% 创始股权,后续报道则称五家银行安排了最高 31 亿美元的高级定期贷款额度,为 GPU 建设提供资金。这些数字不等于可用现金:股权承诺分布在投资期内,而贷款额度的提款、契约、抵押品条款、定价、摊还和先决条件,在指定证据中都未公开。由此得到的核心比率是最高 31 亿美元债务额度对 8 亿美元已声明发起方承诺,但这不是已提款债务或总资本的经验证杠杆率。逆向分析将该额度描述为以 GPU 抵押的贷款,并把利用率、折旧和硬件重定价列为核心风险。现金余额、现金消耗、资金续航、再融资计划或偿债覆盖率均未披露。[CI005, CI006, CI007, CI008, CI009]

4.3 利润率路径与单位经济证据

公开单位经济还不足以支撑传统 SaaS 或平台型市场投资模型。某份股票研究估计假设每 MW 开发资本支出为 US$20-25 million、每 MW 年收入为 US$10-13 million、EBITDA 利润率为 69%,但这些是基于管理层披露和同行区间的分析师假设,不是 Zankore 已报告的结果。同一分析估计项目 IRR 为 12-13%,股权回本期少于三年;这些结果应视为取决于利用率、定价、电力成本、GPU 寿命和融资条款的估计。公司和母公司材料披露的是累计收入/EBITDA 或可归属 EBITDA 估计,没有客户层面的桥接。公开信息中没有实际价格、GPU 利用率、毛利率、电力成本、维护成本、折旧表、CAC、回本期或贡献利润率。因此,财务结论是利润率潜力仍基于情景,验证所需证据掌握在私有材料中。[CI010, CI011, CI012, CI013]

单位经济与资本需求
指标数值或 null信心度重要性尽调问题
每 MW 开发资本开支每 MW:US$20m–US$25m/MW中决定首期建设所需资本发票、范围、预备费和调试计划
每 MW 年收入US$10m-US$13m/MW 估算中在假设利用率和定价情景下,将容量与收入连接起来已签署的承购协议及价格—数量计划
EBITDA 利润率69% 估算低决定偿债能力和股权回报实际成本结构及利润率桥接
项目 IRR12%-13% 估算低概括情景回报,但对假设敏感含敏感性分析和融资条款的模型
股权回收期估算 <3 年低检验资本回收速度现金瀑布及债务摊销
手头现金低评估流动性和资金续航所必需最新资产负债表及受限现金
运营现金消耗低评估融资依赖所必需月度现金流量表
偿债覆盖率低检验 US$3.1bn 融资额度所必需债务计划、契约及贷款方模型

标注为“估算”的数值是分析师假设或发起方估算,并非 Zankore 已报告的结果。null 值代表重大的私人证据缺口。

[CI005, CI006, CI009, CI010, CI011, CI012, CI013, CI016]

4.4 投资判断与尽调阻碍

Zankore 拥有可信的发起方和基础设施伙伴,但仅凭公开信息,财务案例尚未能够审计。Ooredoo 的公告描述了投资期内累计约 6 亿美元、已签约的对 Ooredoo EBITDA 贡献,其他公开估计则给出五年累计约 130 亿美元收入和 90 亿美元 EBITDA;这些是公司或分析师估计,不是经审计的 Zankore 报表,且范围和分母不同,无法直接核对。初始建设因此显得资本密集且依赖融资,需求爬坡和抵押品价值是主要下行变量。投资判断应要求已执行的承购合同、客户集中度、价格和利用率计划、债务文件、提款条件、契约余量、电力和制冷成本、资本支出发票及月度现金流预测。在这些材料交付前,合适的态度是有条件关注,而不是给出量化估值。[CI014, CI015, CI016]

4.5 展示项

Chapter 05

05产品与技术

5.1 产品与客户工作流

Zankore 的产品是 GPU 云产能,而非通用应用。现有描述把训练、微调、推理和智能体工作负载列为目标用途,客户则从 AI 原生公司和开发者延伸至企业、机构、超大规模云和政府。因此,客户工作流是取得专用或按需 NVIDIA 算力,让模型靠近区域数据和用户运行,并消费集成式基础设施服务,而不是自行拼装所有层。这一定义逻辑完整,但仍是发布和部署计划:公开证据没有披露实际租户、服务水平目标、利用率、定价或实测工作负载结果。因此,商业命题应理解为目标产能和目标细分市场,而不是已验证的采用。[CE001, CE002, CE003, CE004]

产品模块与资产矩阵
模块或资产用户或买方状态或成熟度差异化尽调缺口
GPU 云容量AI 原生公司和开发者计划部署NVIDIA 加速计算与区域位置没有公开的实时容量或利用率证据
AI 训练与微调企业和模型构建者预期工作负载用于模型开发的专用 GPU 集群没有公开基准或客户结果
推理与智能体工作负载企业和机构预期工作负载将容量置于区域数据与用户附近没有公开延迟、SLA 或隔离数据
AI 工厂平台超大规模云厂商和政府已宣布初始建设NVIDIA DSX 将计算、网络、电力、冷却和运营整合起来没有投产或验收记录

状态反映公开发布和融资描述,并非对商业可用性的独立验证;运营字段为 null,意味着仍需向客户和部署方尽调。

[CE001, CE002, CE003, CE004, CE005]
工作流与用例表
用户任务当前工作流或需求公司方案可量化收益局限
训练模型获取区域 GPU 容量用于训练的 NVIDIA GPU 集群计划提供按需访问没有披露可用性或价格
微调模型将基础模型适配到本地数据GPU 云与集成式 AI 技术栈据称区域数据距离更近没有披露数据治理条款
提供推理服务运行生产模型请求用于推理的 AI 计算计划提供安全且可扩展的容量没有延迟或正常运行时间证据
运行智能体 AI从试验走向关键任务部署全栈计算与网络平台计划用集成基础设施减少碎片化未披露生产环境参考客户

本表区分预期用例与已量化收益。公开材料没有提供实际转化率、利用率、延迟或客户留存指标。

[CE001, CE002, CE003, CE013]

5.2 架构与运营模式

拟议运营模式把 NVIDIA DSX 作为参考架构,在一套 AI 工厂系统中整合加速计算、网络、电力、制冷和运营。已披露的硬件里程碑是 2027 年上半年使用 NVIDIA GB300 NVL72 实现约 200 MW,长期目标最高 1 GW。NVIDIA DSX MaxLPS 被描述为可动态优化 GPU 集群的电力,并设计为在同一电力预算内实现最高 40% 更多算力。Nokia 提供原生 AI 网络,Indosat 提供本地基础设施和执行能力。这些信息划定了有用的架构和责任边界,但来源没有公布机架拓扑、软件接口、实测电力使用效率、冗余设计或调试投运证据。[CE005, CE006, CE007, CE008, CE009]

技术与运营架构
层或流程角色依赖风险或尽调问题
加速计算NVIDIA GB300 NVL72 GPU 基础设施NVIDIA 供应与平台支持确认交付计划和硬件验收
AI 工厂编排NVIDIA DSX 参考架构与生态系统作战手册NVIDIA 软件与经验证设计获取软件物料清单和运营手册
电力优化NVIDIA DSX MaxLPS设施电力容量和集群遥测验证部署配置中的最高百分之四十设计宣称
网络Nokia AI 原生网络与连接架构Nokia 设备与 Indosat 网络执行验证拓扑、延迟、冗余和安全控制
设施运营电力、冷却和场站运营印度尼西亚场站与公用事业验证 MW 通电、冷却余量和韧性
商业交付收入分成与信用支持模式客户需求和融资交易对手获取合同、承购安排和与需求挂钩的契约
安全与主权平台被描述为安全、可信且具主权性仅属营销或表述的意图获取安全架构和独立鉴证
可靠性与可用性资料描述了规模化和高性能基础设施在指定来源中未找到正常运行时间或事故历史获取 SLA、冗余和停运记录
隐私与数据处理区域上靠近数据和用户被定位为一项优势条款和隔离控制未披露审阅 DPA、数据驻留、租户隔离和删除控制
质量与认证资料描述了合作伙伴的专业能力指定来源中没有认证或验收证据获取 ISO/SOC、投产和质量记录

架构层是根据披露的合作伙伴角色和平台描述重构的;来源没有提供详细工程图或独立技术验证。

[CE005, CE006, CE007, CE008, CE009, CE011, CE013, CE014, CE015]
FE001: Zankore 产品架构图

公开描述的各层把区域执行和资本连接到 NVIDIA AI 工厂基础设施及客户工作负载。

客户工作负载
平台拟服务的区域 AI 工作负载。
模块
  • 训练与微调
  • 推理
  • 智能体 AI 应用
云服务
算力接入与商业交付层。
模块
  • GPU 集群
  • 专用或按需 AI 算力
  • 收入分成与信用支持模式
AI 工厂平台
NVIDIA 参考架构与 GPU 部署层。
模块
  • NVIDIA DSX 参考架构
  • GB300 NVL72
  • DSX MaxLPS 电力优化
设施与网络
物理站点、电力、冷却和连接层。
模块
  • 电力与冷却
  • Nokia 原生 AI 网络
  • Indosat 数字基础设施与执行
区域平台
本地执行与区域扩张层。
模块
  • Ooredoo 资本与区域规模
  • 印尼发布市场
  • 东南亚扩张目标
治理与生态
支持规模化的合作伙伴与治理结构。
模块
  • Ooredoo、Indosat、Nokia 与 NVIDIA 的合作
  • 董事会与 CEO 监督
  • 超大规模云厂商、企业、政府和开发者重点

这是基于证据的架构摘要,而非完整工程图;各层边界和依赖关系仍需尽调。

[CE005, CE008, CE010, CE011]

5.3 成熟度、依赖与差异化

Zankore 的差异化主要来自生态和基础设施整合:Ooredoo 提供资本和区域规模,Indosat 提供印度尼西亚基础设施和执行,NVIDIA 提供 GPU、软件和 DSX 方法,Nokia 提供网络。这一组合可以减少区域买方的集成负担,也支持主权或数据本地化部署;但它同时集中依赖 NVIDIA 硬件和软件、电信站点执行、电力可用性、制冷、网络交付和伙伴治理。公开描述确认了伙伴和计划里程碑,却没有确认专有软件、专利保护、独立 基准、已完成认证或带验收标准的生产路线图。因此,当前最强的护城河是执行接入和伙伴协同,而非尚未独立证明的技术优势。[CE010, CE011, CE012]

5.4 信任、可靠性与尽调阻碍

发布材料强调安全、可扩展、主权和可信的 AI 基础设施,并提到通过董事会和首席执行官进行治理。这些表述描述的是定位意图,而不是控制措施证据。指定来源没有提供正常运行时间历史、事故记录、渗透测试结果、SOC 或 ISO 认证、数据处理条款、模型或租户隔离设计、灾备目标、制冷遥测或客户支持指标。一份融资报道还指出贷款抵押品并不明确,GPU 转售价值可能取决于同样需要偿还贷款的需求;这是运营平台的资本和资产风险考量,不是产品缺陷证据。在评估产品成熟度前,尽调应取得部署验收记录、安全鉴证、产能预留、客户合同,以及首个产能里程碑已投入运营的证据。[CE013, CE014, CE015]

5.5 展示项

Chapter 06

06客户

6.1 客户范围与采购场景

Zankore 声称的市场覆盖企业、初创公司、开发者、AI 原生公司和机构,先在印度尼西亚,随后扩展到东南亚。目标工作负载包括模型训练、微调、推理和智能体 AI,意味着采购场景多元:基础设施或技术负责人可能采购产能,技术团队使用产能,受监管或机构用户可能重视本地部署。独立描述还使用更宽泛的“云和企业客户”表述,并采用亚太视角。这些是目标细分市场,不是已观察到的客户群。来源没有提供按细分市场划分的客户数、定价、收入组合、销售渠道归属,也没有证据表明任何类别已在生产环境使用 Zankore,因此各细分市场的战略价值仍无法量化。[CU001, CU002, CU003, CU010, CU011]

客户分层表
客户群买方 / 付款方主要用户预期用例公开规模证据战略价值证据缺口
企业基础设施或技术负责人AI 工程和业务团队企业应用的训练与推理未披露客户数量或具名部署潜在的大规模经常性计算需求价格和生产利用率未披露
AI 原生公司和初创企业创始人或技术负责人模型开发者和平台工程师训练、微调和智能体工作负载未披露具名账户或客户群产品驱动的区域计算需求合同规模、获客渠道和留存未披露
开发者个人开发者或开发者团队预算负责人软件和机器学习开发者用于开发和推理的按需 GPU 容量未披露活跃开发者指标生态采用与工作负载试验接入模式、价格和转化率未披露
机构机构采购或公共部门预算负责人研究、技术或运营团队本地 AI 开发和生产工作负载未披露具名机构或采购记录数据本地化和主权计算需求采购周期、部署状态和结果未披露

客户群和用例取自 Zankore 所述目标市场;买方角色、战略价值和缺口属于分析分类,任何一行都不代表经过验证的客户群。

[CU001, CU002, CU003, CU010, CU011]
FU001: 拟议客户旅程与证据检查点

已披露的商业路径从目标客户群走向与需求挂钩的产能扩张,但部署和续约检查点仍缺少客户验证。

这是根据目标客户群、工作负载和产能里程碑推断出的尽调旅程,并非已观察到的转化漏斗,也不暗示任何转化率。

[CU001, CU003, CU004, CU005, CU006, CU007, CU012]

6.2 采用轨迹与命名客户证明

可观察到的轨迹由基础设施带动。报道描述了最初 100 MW 建设、2027 年上半年约 200 MW 的官方计划,以及长期最高 1 GW 目标。Zankore 还描述了收入分成和信用支持安排,意在让部署速度与需求一致。这些里程碑可能创造客户采用所需的产能,但不能证明活跃用户、利用率、签约订单积压或实际收入。Ooredoo 预计前五年累计贡献约 6 亿美元 EBITDA,这是投资人预测,不是已披露的客户指标。在九个指定来源中,没有客户被点名,也没有合同层面的生产部署或试点记录。[CU004, CU005, CU006, CU007, CU008, CU009]

客户增长与采用轨迹表
指标或里程碑数值时期证据置信度客户含义缺失分母
初始基础设施建设100 MW据报道为 2026 年 9 月中计划在近期为客户工作负载提供供给没有已投产容量或利用率百分比
首期计划容量约 200 MW2027 年上半年对该计划而言高可服务容量可能跃升没有签约需求或客户分配
长期容量目标最高 1 GW多年期目标对该目标而言高,但不代表交付区域扩张抱负没有带日期的客户爬坡或利用率目标
投资者经济预期约 US$600 million 的累计 EBITDA 贡献Ooredoo 投资的前五年中,因为这是公司估算对未来变现的间接预期没有客户收入、ARR、积压订单或利润率桥接

这些是供给侧计划和一项投资者估算,并非活跃采用度量;客户分母为 null,不得解读为零。

[CU004, CU005, CU006, CU007, CU009]
具名客户证明表
客户证据类别客户群部署 / 用例生产环境 vs 试点结果局限
未识别出具名生产客户未披露未披露合同层面的工作负载尚未建立未报告客户结果审阅来源描述的是目标客户群,而非具名生产部署
未识别出具名试点客户未披露未披露试点范围尚未建立未报告试点结果审阅来源没有提供客户背书或实施证据
未识别出具名锚定或承购客户资料泛泛讨论云和企业需求未披露承诺的容量分配尚未建立未报告签约需求指标一项不利来源重复了蓝筹客户宣称,但没有名称或合同细节

截至 2026-10-22,本章分配到的九个成功预取来源所覆盖的部分公开证据清单;在这一限定语料库中未出现,并不能证明不存在私人客户关系。

[CU007, CU008, CU014, CU015]

6.3 持久性、扩张与集中度风险

公开证据尚不足以支持留存或扩张逻辑。已审阅来源没有一家报告净收入留存、毛收入留存、流失、续约、合同期限、客户队列 行为、满意度、客户层面利用率、订单积压、承购或收入集中度。需求挂钩的融资结构使这些缺口具有实质影响:TNW 指出,签约不足可能让贷款方暴露于 GPU 贬值风险;AInvest 则认为锚定客户必须足够快地到来,才能支撑高杠杆建设。这些是分析警告,不是实际客户流失证据。因此,在把计划中的兆瓦数解读为持久商业进展前,尽调应优先确认已执行合同、命名客户参考、按客户队列 划分的利用率、续约机制和头部客户暴露。[CU012, CU013, CU014, CU015]

6.4 展示项

Chapter 07

07风险

7.1 杠杆、需求与硬件经济性

Zankore 在尚无已证明经营历史前就尝试建设资本密集型平台。公告的高级定期贷款最高为 31 亿美元,而 Ooredoo 同期公告描述了约 8 亿美元、对应 49% 创始股权的承诺。来源没有披露 Zankore 的现金流状况、客户合同覆盖率、利率、契约、摊还或抵押品包。因此融资带来时间风险:平台必须上线产能,并把它转化为已签约、付费的利用率,同时 GPU 价值和租赁价格可能变化。The Next Web 特别指出贷款抵押品仍不明确,若需求失望,GPU 抵押品可能贬值。实际监控测试应看已签署且资信可靠的承购以及相对于偿债的实际利用率,而不是融资主标题中的千兆瓦目标。最高额度与 Ooredoo 承诺之间的公开比较是警示信号,不是经验证的债资比,因为 Zankore 的实际缴入股权和可用现金均未披露。[CR001, CR002, CR003, CR004, CR005, CR014, CR015]

监管 / 法律风险登记表
排名规则 / 牌照 / 案件司法辖区状态可能性证据实现后的严重性已披露缓解措施残余风险尽调路径
1优先贷款担保、契约和追索权未披露已宣布最高 US$3.1bn 融资额度;已执行条款不可得未量化若追索权或契约消除下行灵活性,则对论点至关重要已宣布收入分成和信用支持文件审阅前未知获取融资协议、担保、担保物清单、契约和下行情景模型
2融资、用电、建设和环境审批印度尼西亚指定证据中没有许可登记表未量化若审批延误投产,则对论点至关重要已描述 Indosat 的执行角色;许可控制未披露未知获取逐场站许可矩阵、电力承诺、环境审批和投产证书
3隐私、数据驻留和安全义务印度尼西亚及计划进入的东南亚市场指定证据中尚未建立合规资料包未量化若不合规阻断受监管工作负载,风险较高指定证据中没有认证或控制资料包未知获取法律顾问分析、数据流图、安全认证、事故日志和客户 DPA 条款
4诉讼、执法、制裁和出口管制风险相关运营及供应商司法辖区尚未建立案件或执法记录;缺失并不证明不存在未量化若供应或运营受限,则影响重大未识别出公开缓解措施未知通过法律顾问开展司法辖区法院、监管机构、制裁和出口管制检索

截至 2026-10-22 的部分公开证据登记表;当指定证据未建立事故、许可状态或法律结果时,不对可能性评分。

[CR001, CR003, CR010, CR012, CR015]
FR001: 风险热力图

该矩阵有条件地排列影响程度,同时显示公开证据不足以评估主要风险发生的可能性。

[CR003, CR004, CR005, CR007, CR012, CR014, CR015, CR016]

7.2 建设、电力与运营执行

推进计划依赖于在印度尼西亚、随后在东南亚交付技术集成的基础设施项目。公开材料描述了最初 100 MW 建设、2027 年上半年约 200 MW 目标和最终 1 GW 雄心,但这些是计划,不是已投运产能的证据。该栈还依赖 NVIDIA GB300 NVL72 系统、NVIDIA DSX 与 DSX MaxLPS、Nokia 的网络集成,以及 Indosat Ooredoo Hutchison 的本地执行。这一组合带来进度、电力、制冷、供应、调试、可靠性和服务质量依赖。任何阶段延误,都可能让债务融资设备闲置而融资敞口继续存在。已宣布的收入分成和信用支持模式在方向上与需求一致,但其合同和运营效果不公开。审慎的缓解措施是分阶段放量,并以电力就绪站点、验收测试、正常运行时间 和利用率阈值为条件,在进入下一产能批次前设置明确暂停触发条件。[CR006, CR007, CR008, CR009, CR010, CR016]

运营 / 质量 / 安全风险登记表
失效模式可能性严重性缓解成熟度残余风险未解决缺口
电力、冷却或投产延误未量化若首期 2027 年阶段出现重大延期,则对论点至关重要已宣布合作伙伴和集成架构;运营证明不可得尽调关口场站电力、冷却、许可、验收测试和已投产 MW 证据
GPU 采购或技术迭代未量化若可用性、价格或过时风险损害经济性,则风险高已披露 NVIDIA 技术和融资;尚未建立多元化重大且未量化采购订单、交付计划、保证、升级权和残值压力测试
集群利用不足未量化对论点至关重要,因为收入必须支撑债务融资部署已宣布收入分成和信用支持;合同细节不可得尽调关口具约束力的客户管线、签约 MW、价格、信用质量和月度利用率
可靠性或安全故障未量化若服务中断或控制薄弱阻断生产工作负载,则风险高指定证据中没有公开 SLA、认证或事故记录未知SOC 报告、事故历史、正常运行时间 SLA、保险和响应计划

影响标签是有条件的投资后果,而非事故概率;公开证据不支持数值或分类的可能性估计。

[CR004, CR006, CR007, CR008, CR009, CR010, CR012, CR016]

7.3 伙伴、监管与治理暴露

Zankore 的伙伴模式同时是优势和集中度风险。Ooredoo 被描述为持股 49% 的领投方,Indosat 提供印度尼西亚市场和基础设施能力,NVIDIA 提供加速计算栈,Nokia 提供网络。平台因此依赖同时拥有商业利益的交易对手,而本次审阅的公开记录没有确认董事会控制、关联方保护、客户集中度、许可、数据治理批准、事故历史或保险。已审阅的逆向材料对需求时点、GPU 抵押品和供应商融资的循环性提出疑问,但没有证明违约或不当行为。因此,监管和法律风险应保持为证据缺口,而不是臆造事故:在把风险视为低位前,尽调应取得许可、融资文件、安全与隐私控制、诉讼检索结果及完整依赖关系登记。[CR004, CR011, CR012, CR013, CR014]

合作伙伴 / 依赖风险登记表
依赖项交易对手角色失效情景严重性缓释措施与剩余敞口
资本与治理Ooredoo Group49% 领投方持股,并承诺约 US$800m承诺时间或治理支持与建设计划不一致若资金或控制权安排打断部署,将直接影响投资论点审查实缴时间表和保留事项;剩余敞口仍未知
印尼执行Indosat Ooredoo Hutchison本地基础设施、市场准入与执行能力站点、电力、许可或客户准入延迟若初始部署无法投产,将直接影响投资论点采用合同阶段门、服务水平和备用运营商;条款未公开
GPU 与 AI 技术栈NVIDIAGB300、DSX、DSX MaxLPS、硬件供应与生态供应、定价或技术迭代削弱经济性风险高,因为融资资产与核心技术栈共用一家供应商核查质保、升级权、多代采购和转售假设
原生 AI 网络Nokia平台的网络层网络集成或支持延迟若生产就绪度延后,风险较高要求验收测试、支持 SLA 和可互操作的备用设计

合作伙伴角色基于已披露的发布材料;集中度、排他性和备用权利仍需审查合同。

[CR006, CR011, CR012, CR013]

7.4 展示项

Chapter 08

08估值

8.1 投资判断与价格纪律

建议的态度是继续研究并跟踪,而不是买入。Ooredoo 以约 US$800 million 换取 49% 股权的承诺,以及最高 US$3.1 billion 的高级定期贷款额度,证明了发起方和贷款方的资金能力;若利用率按计划形成,计划中的 100 MW→200 MW→1 GW 扩张也会带来显著经营杠杆。但这些优势不能确定可投资价格。公开记录没有披露当前股权估值、贷款定价、到期日、契约、抵押品包、优先股堆叠或已命名锚定客户。因此,入场纪律应表达为条件,而不是点估值:在评估回报前,要求完全稀释后的股权结构表、融资 清偿瀑布、签约收入计划、单位经济和下行回收分析。[CV001, CV003, CV007, CV008, CV009, CV014, CV016]

建议摘要
决策事项当前判断对决策的影响
建议继续研究 / 跟踪在价格和保护条款披露前,不要对其进入市场进行投资判断。
信心中融资和产能证据较强;估值和商业证据不完整。
风险评级高大规模债务融资部署早于已披露的运营历史和锚定需求。
估值立场公开支持不足将 US$800 million 承诺视为融资背景,而不是公司估值。
进入纪律有条件要求提供股权结构表、债务清偿瀑布、已签约需求、单位经济性和回收情景。
改变判断的证据经审计的经济性与具有约束力的承购协议只有在利用率、利润率、债务偿付和稀释完成建模后,才确定价格。

分析师判断基于截至 2026-10-22 审阅的公开证据;在指定来源集合中,没有公开可得的当前股权估值或完整融资清偿瀑布。

[CV001, CV003, CV008, CV009, CV014, CV016]
FV001: 建议逻辑

可见的股东和贷款方融资能力支持有条件地继续跟踪,因为商业验证、债务保护和有依据的进入估值仍然缺失。

节点 / 连接节点 / 起点终点说明
节点 1 股东资本 [sponsor-capital]承诺投入约 US$800 million,换取 49% 持股。
节点 2 债务能力 [debt-capacity]五家银行提供最高 US$3.1 billion 的高级定期贷款融资。
节点 3 规模计划 [scale-plan]初始建设 100 MW,2027 年上半年约 200 MW,目标达到 1 GW。
节点 4 商业验证缺口 [proof-gap]没有具名锚定客户或已披露的签约收入时间表。
节点 5 下行保护缺口 [downside-gap]贷款抵押品、契约、追索权和优先权条款未披露。
节点 6 继续研究 / 跟踪 [decision]只有在已签约需求、单位经济性和回收保护得到验证后,才确定价格。

该顺序是分析性的,而非按时间排列;它不量化转化率、概率、企业价值或投资回报。

[CV002, CV004, CV011, CV012, CV013, CV015, CV016]

8.2 投资论点、反论点与融资背景

投资论点建立在异常强的战略发起方、银行参与和区域产能计划之上,并对应公司引用的东南亚数据中心需求增长。Ooredoo 还估计前五年累计贡献约 6 亿美元 EBITDA,但这是发起方预测,而非经审计的 Zankore 表现。反论点是,贷款方在仅约四周经营历史后就承诺了大额资本,而部署与客户需求及条款未披露的信用支持挂钩。NVIDIA 既是支持方又是硬件供应商,因此部分生态需求是通过购买 NVIDIA 设备的载体融资的。这种一致性可以加速部署,但也集中技术、残值和利用率风险。[CV001, CV002, CV003, CV005, CV006, CV010, CV011, CV015]

投资论点与反论点
立场论点当前证据哪些信息会改变判断
投资论点战略股东可以提供资本和执行支持。Ooredoo 承诺约 US$800 million,持股 49%。股东支持不足,或进行大幅稀释的资本重组。
投资论点银行融资可以加快区域规模化 AI 算力建设。五家银行安排或承销了最高 US$3.1 billion 的融资。融资提款条件或契约严重限制部署。
投资论点如果需求已签约,产能可以创造经营杠杆。已公布路径是初期 100 MW、2027 年上半年约 200 MW,以及三年内达到 1 GW。已签约需求无法跟上已投产产能。
反论点资本形成领先于运营验证。融资在发布后约四周完成。经审计的利用率、留存率和现金利润率队列数据可得。
反论点GPU 剩余价值与客户需求构成相互关联的下行风险。信用支持和部署都与需求挂钩,而抵押品细节仍不清楚。长期照付不议合同和保守的抵押品折扣得到记录。

股东预测和产能目标不是经审计结果;判断变化需要交易原始文件和运营证据,而不是更多公告。

[CV001, CV003, CV005, CV006, CV007, CV009, CV010]

8.3 情景框架与可比限制

合适的估值模型应在扣除电力、托管、硬件融资和更新成本后,把企业价值与签约毛利或 EBITDA 挂钩,而不是套用通用软件收入倍数。乐观情景要求有签约需求、2027 年按时交付、高利用率,以及证明电力优化改善可销售算力经济性。基准情景假设分阶段部署和利用率延迟,价值会高度受融资成本和客户爬坡影响。悲观情景则结合承购疲弱、GPU 过时、契约压力和较低抵押品回收。由于当前估值、预测现金流、可比交易和债务条款均缺失,基于已审阅证据无法发布可辩护的估值区间或概率权重。[CV005, CV007, CV009, CV010, CV012, CV013, CV014, CV015]

乐观、基准与悲观情景
情景假设估值与回报逻辑关键风险概率信号
乐观具有约束力且多元化的承购需求;2027 年产能按时交付;高利用率;电力优化转化为可销售算力。按融资后的已签约 EBITDA 估值,只有在可重复的现金收益得到证明后才给予溢价。执行、集中度、技术更新和杠杆。已签合同、经审计的利润率和契约余量。
基准分阶段交付;利用率落后于产能爬坡;股东支持仍可获得。使用保守的基础设施倍数,并充分计入 GPU 更新、利息和稀释。闲置产能和融资成本带来较低回报。季度利用率队列,以及高于下行情景阈值的债务偿付覆盖率。
悲观承购需求令人失望;GPU 价值压缩;建设延迟;契约或再融资压力出现。以高级债权索偿主张受偿顺序之后的回收价值为锚,而不是已宣布的融资或重置成本。减值、资本重组、贷款方控制和股权稀释。客户取消、部署延迟、契约豁免或二手 GPU 价格下跌。

这些是定性情景,而非按概率加权的估值区间;收入、现金流、债务经济性、资本表条款和可辩护的可比交易均未披露。

[CV005, CV007, CV009, CV010, CV012, CV013, CV014, CV015]

8.4 退出准备度、尽调要求与论点失效触发条件

退出准备度仍早期:Zankore 于 2026 年 8 月发布,公开记录主要是发起方公告和融资报道,而非经营结果。如果平台把计划产能转化为多元且资信可靠的合同,并证明扣除 GPU 更新义务后的现金收益,战略退出才可能变得可行;基础设施或公开市场路径还需要经审计的分部报告和持久的杠杆指标。在投资前,应取得签署的承购计划、客户集中度、利用率和价格 按客户队列拆分的利用率和定价、项目层面现金流、债务和担保文件、GPU 更新承诺、资本结构权利及下行回收情景。论点失效触发条件包括 2027 年产能计划明显延误、利用率低于偿债需要、没有保障的客户集中度、契约承压,或设备价值下降速度快于本金摊还。[CV006, CV007, CV009, CV010, CV013, CV014, CV016]

8.5 展示项

免责声明

本报告是对截至 2026-10-22 保留公开证据的尽调分析,不构成投资、法律、会计或财务建议。

证据索引

结论
编号陈述可信度来源
CO001 Zankore was launched in August 2026. 高 SO003, SO004
CO002 Indosat Ooredoo Hutchison, Ooredoo Group, NVIDIA, and Nokia developed and backed Zankore. 高 SO002, SO003, SO004, SO006, SO007
CO003 Zankore sells on-demand NVIDIA GPU capacity for AI training, fine-tuning, inference, and agentic workloads. 中 SO001, SO003, SO007
CO004 Zankore is building an initial 100 MW of NVIDIA AI infrastructure in Indonesia. 中 SO001, SO003, SO004, SO007
CO005 Zankore targets approximately 200 MW of AI capacity in the first half of 2027. 中 SO001, SO002, SO003, SO004, SO006, SO007
CO006 Zankore's long-term announced capacity target is 1 GW. 中 SO002, SO003, SO004, SO005, SO006, SO007
CO007 Ooredoo Group disclosed a 49% stake and an approximately US$800 million commitment to Zankore. 高 SO005, SO006, SO008
CO008 Ooredoo estimates approximately US$600 million of cumulative EBITDA contribution to itself over the first five years of the investment. 中 SO006
CO009 Zankore secured a senior term loan facility of up to US$3.1 billion. 高 SO001, SO003, SO004, SO007, SO008
CO010 Citi, ING, Natixis CIB, Qatar National Bank, and UOB acted as senior arrangers, underwriters, and bookrunners. 高 SO001, SO003, SO007
CO011 The loan proceeds are designated for acquiring and deploying NVIDIA GPU infrastructure. 高 SO001, SO003, SO007
CO012 Zankore describes revenue-sharing and credit-support arrangements intended to align deployment with customer demand. 中 SO003, SO007
CO013 Natixis says its tranche is secured by GPU equipment, material contracts, and take-or-pay agreements with paying customers. 中 SO001
CO014 Vikram Sinha is publicly identified as Zankore's director and chairman. 高 SO004, SO007
CO015 Aziz Aluthman Fakhroo is Ooredoo Group CEO and the lead investor's public executive sponsor for Zankore. 中 SO002, SO006
CO016 AInvest characterizes Zankore's disclosed debt as roughly four times its approximately US$800 million of founding equity. 中 SO008
CM001 Zankore is described as a GPU-as-a-Service or neocloud platform for regional AI workloads. 中 SM001, SM007
CM002 The disclosed workload scope includes AI training, fine-tuning, inference, and agentic workloads. 中 SM007
CM003 Public descriptions position Zankore as an integrated compute, networking, power, cooling, and operations platform rather than a standalone GPU rental listing. 中 SM006
CM004 Ooredoo's public announcement targets 1 GW of NVIDIA DSX AI Factory capacity across Southeast Asia. 高 SM004, SM005
CM005 Ooredoo says approximately 200 MW of Zankore AI capacity is expected in the first half of 2027. 高 SM003, SM005, SM006, SM009
CM006 Data Center Dynamics reports an initial 100 MW Indonesian buildout before an ultimate 1 GW target. 中 SM003, SM007
CM007 Ooredoo cites industry analysis projecting approximately 3.5x growth in Southeast Asian data-centre capacity demand by 2030 without identifying the underlying study or denominator. 中 SM005
CM008 Ooredoo says it will commit approximately $800 million for a 49% stake in Zankore. 高 SM004, SM005, SM009
CM009 Zankore reportedly secured a senior term loan facility of up to $3.1 billion for GPU infrastructure deployment. 中 SM003, SM007, SM008
CM010 Indosat says Zankore can serve hyperscalers, enterprises, and governments across Asia-Pacific. 中 SM006
CM011 The platform is described as selling GPU capacity to enterprises and developers across the region. 中 SM007
CM012 Ooredoo says NVIDIA DSX MaxLPS is designed to enable up to 40% more compute within the same power budget. 中 SM005, SM006
CM013 Zankore remains in buildout, with its first approximately 200 MW not expected until the first half of 2027. 中 SM001, SM005
CM014 Bundling network and compute may simplify procurement while increasing concentration and workload-portability risk. 中 SM001
CM015 The reported loan includes revenue-sharing and credit-support arrangements designed to link deployment pace to customer demand. 中 SM007
CM016 The retained evidence does not quantify revenue TAM, SAM, SOM, pricing, utilization, contracted GPU volumes, or named customers. 低
CP001 Zankore targets up to 1 GW of NVIDIA DSX AI Factory capacity. 中 SP005, SP004
CP002 Public sources describe a staged deployment with approximately 200 MW planned for the first half of 2027 and a separate 100 MW initial build description. 中 SP006, SP007
CP003 Zankore combines Ooredoo, Indosat Ooredoo Hutchison, Nokia, and NVIDIA in its disclosed partner stack. 中 SP005, SP006
CP004 The stated target customer groups include hyperscalers, enterprises, governments, developers, and AI companies. 中 SP006, SP004
CP005 The disclosed technical stack includes NVIDIA GB300 NVL72 systems, NVIDIA DSX, and DSX MaxLPS. 高 SP005, SP006
CP006 Ooredoo's release says DSX MaxLPS is designed to enable up to 40% more compute within the same power budget. 中 SP005
CP007 The assigned sources do not disclose a Zankore list price, realized discount, GPU-hour price, or customer-specific contract price. 中 SP007, SP009
CP008 The assigned sources do not provide a public Zankore service-level agreement, utilization rate, or named anchor customer. 中 SP007, SP009
CP009 Indosat's local infrastructure and execution, Ooredoo's capital and regional scale, Nokia's networking, and NVIDIA's accelerated-computing ecosystem are presented as complementary sources of distribution and supply access. 中 SP005, SP006
CP010 Zankore's local deployment could differentiate it for Indonesian or regulated workloads, but the assigned sources do not establish a customer switching-cost schedule. 低 SP004, SP006
CP011 The Next Web reports that Zankore's financing includes revenue-sharing and credit-support arrangements tied to deployment pace and customer demand. 中 SP007
CP012 The assigned sources do not establish exclusivity, minimum-volume commitments, or long-term anchor contracts for Zankore's capacity. 中 SP009
CP013 The Next Web reports that GPU demand shortfalls could leave lenders with hardware whose value declines as the market changes. 中 SP007
CP014 AInvest characterizes the up-to-$3.1 billion senior loan as roughly four times the approximately $800 million of founding equity. 中 SP008
CP015 Remio's analysis identifies undisclosed cooling design, water demand, water sources, and drought procedures as constraints on evaluating the full-buildout proposition. 中 SP001
CP016 The assigned evidence supports requiring named anchor demand, realized pricing, financing terms, and resource-performance disclosures before treating the 1 GW ambition as a durable moat. 中 SP001, SP007, SP009
CI001 Zankore is described as an AI compute and neocloud platform serving cloud, enterprise, and government customers. 中 SI006, SI008
CI002 Zankore targets approximately 200MW of AI capacity in the first half of 2027 and up to 1GW of NVIDIA DSX AI Factory capacity over the longer term. 高 SI007, SI008
CI003 The assigned evidence does not disclose realized Zankore revenue, ARR, utilization, GPU-hour pricing, or named anchor customers. 中 SI004
CI004 Public launch coverage describes the intended integrated stack but does not provide customer-level revenue contracts or realized margin data. 高 SI007, SI008
CI005 Ooredoo Group announced an approximately US$800 million commitment for a 49% founding stake in Zankore. 中 SI007
CI006 Coverage reports a senior term loan facility of up to US$3.1 billion for Zankore's AI capacity buildout. 高 SI004, SI005
CI007 The disclosed US$3.1 billion facility is a maximum facility amount and does not establish the amount drawn or total Zankore capitalization. 中 SI004, SI005
CI008 An adverse analysis describes the headline financing as debt roughly four times the approximately US$800 million of founding equity. 中 SI003
CI009 The assigned evidence does not disclose Zankore cash on hand, burn, runway, debt pricing, covenants, amortization, or debt-service coverage. 中 SI004, SI005
CI010 An equity-research analysis estimates development capex of US$20-25 million per MW for the initial build. 中 SI001
CI011 The same equity-research analysis estimates annual revenue of US$10-13 million per MW and a 69% EBITDA margin under its assumptions. 中 SI001
CI012 The equity-research analysis estimates project IRR at 12-13%. 中 SI001
CI013 The equity-research analysis estimates an equity payback period of less than three years. 中 SI001
CI014 Ooredoo's launch release estimates approximately US$600 million of cumulative EBITDA contribution to Ooredoo over the investment period. 中 SI007
CI015 An analyst account describes cumulative Zankore revenue and EBITDA estimates of approximately US$13 billion and US$9 billion over five years. 中 SI001
CI016 The public estimates in the assigned evidence use different scopes and denominators and cannot be reconciled into audited Zankore operating results. 中 SI001, SI007
CE001 Zankore is described as an AI compute and neocloud platform serving cloud, enterprise, hyperscaler, government, developer, and AI-native users. 高 SE001, SE003, SE005
CE002 Public descriptions identify model training, fine-tuning, inference, and agentic workloads as intended Zankore use cases. 中 SE003, SE008
CE003 Zankore is intended to provide dedicated or on-demand GPU compute through NVIDIA-powered cloud infrastructure. 中 SE003, SE008
CE004 Assigned sources do not disclose live customer utilization, production workload outcomes, pricing, service-level objectives, or realized performance for Zankore. 低
CE005 Zankore targets up to 1 GW of NVIDIA DSX AI Factory capacity and approximately 200 MW of AI capacity in the first half of 2027. 高 SE001, SE004, SE006
CE006 NVIDIA DSX is described as integrating accelerated computing, networking, power, cooling, and operations into a unified AI-factory system. 高 SE001, SE007
CE007 Zankore plans to use NVIDIA GB300 NVL72 systems for the approximately 200 MW first-half-2027 capacity milestone. 高 SE001, SE004, SE006
CE008 NVIDIA DSX MaxLPS is described as dynamically optimizing GPU-fleet power and as designed to enable up to 40% more compute within the same power budget. 高 SE001, SE006, SE007
CE009 Nokia is described as supplying AI-native networking while Indosat supplies Indonesian digital infrastructure and execution capabilities. 中 SE001, SE002, SE007
CE010 Zankore's current differentiation is primarily the combination of Ooredoo capital, Indosat local execution, NVIDIA compute and software, and Nokia networking. 中 SE001, SE002, SE006
CE011 The disclosed architecture depends on NVIDIA hardware and software, partner delivery, Indonesian power and facilities, cooling, and network execution. 中 SE001, SE003, SE006, SE008
CE012 Assigned sources do not provide independent benchmark results, completed technical certifications, proprietary-software evidence, or a production roadmap with acceptance criteria. 低
CE013 Launch materials describe Zankore as secure, trusted, scalable, and sovereign, but assigned sources do not provide independent security attestations, tenant-isolation details, or data-processing terms. 中 SE001, SE002
CE014 Assigned sources do not provide an uptime history, incident record, service-level target, disaster-recovery target, or cooling telemetry for Zankore. 低
CE015 The Next Web reported that the loan's collateral was unclear and that GPU resale value could depend on the same demand expected to repay the loan. 中 SE008
CU001 Zankore publicly targets enterprises, startups, developers, AI-native companies and institutions as customers. 中 SU001, SU004
CU002 Zankore's initial customer market is Indonesia, with a broader Southeast Asian and Asia-Pacific ambition. 中 SU001, SU002, SU004
CU003 The proposed service is designed for model training, fine-tuning, inference and agentic AI workloads. 中 SU001, SU004, SU008
CU004 Reporting in September 2026 described an initial 100 MW Zankore infrastructure build in Indonesia. 中 SU001, SU004, SU005, SU008
CU005 Zankore's official plan called for approximately 200 MW of AI capacity in the first half of 2027. 高 SU004, SU005, SU007, SU008
CU006 Zankore's disclosed long-term capacity target is up to 1 GW of NVIDIA AI Factory infrastructure. 高 SU002, SU004, SU005, SU006, SU007, SU008
CU007 Zankore describes revenue-sharing and credit-support arrangements intended to align infrastructure deployment with customer demand. 中 SU001, SU004, SU008
CU008 None of the nine reviewed sources identifies a Zankore customer by name or provides contract-level evidence of a production deployment or pilot. 中 SU001, SU002, SU003, SU004, SU005, SU006, SU007, SU008, SU009
CU009 Ooredoo estimates that its Zankore investment will contribute approximately US$600 million in cumulative EBITDA over the first five years. 中 SU007
CU010 Dealroom describes Zankore's intended buyers broadly as cloud and enterprise customers. 中 SU006
CU011 Telecompaper reports an Asia-Pacific customer scope with an initial focus on Southeast Asia. 中 SU002
CU012 TNW warns that insufficient customer contracting could leave lenders exposed to GPUs whose value may decline. 中 SU008
CU013 AInvest argues that Zankore's leveraged economics depend on anchor customers appearing quickly enough to support the debt-funded buildout. 中 SU009
CU014 None of the reviewed sources reports Zankore retention, churn, renewal, contract-length, cohort or customer-satisfaction metrics. 中 SU001, SU002, SU003, SU004, SU005, SU006, SU007, SU008, SU009
CU015 None of the reviewed sources discloses customer-level utilization, backlog, offtake or revenue concentration. 中 SU001, SU002, SU003, SU004, SU005, SU006, SU007, SU008, SU009
CR001 Zankore announced a senior term loan facility of up to US$3.1 billion for its GPU and cloud infrastructure rollout. 高 SR001, SR003, SR007
CR002 Ooredoo Group disclosed an approximately US$800 million commitment for a 49% founding stake in Zankore. 高 SR004, SR005
CR003 The assigned sources do not disclose Zankore's interest rate, covenants, amortization schedule, or collateral package for the senior term loan. 中 SR001, SR003, SR007
CR004 The public evidence reviewed does not establish binding customer contracts or a realized utilization rate sufficient to assess debt-service coverage. 低
CR005 The Next Web reports that the value of GPU collateral could decline if customer demand falls short. 中 SR007
CR006 Zankore is building an initial 100 MW of NVIDIA AI infrastructure in Indonesia. 高 SR001, SR002
CR007 Zankore targets approximately 200 MW of AI capacity in the first half of 2027 and an eventual 1 GW platform. 高 SR005, SR006
CR008 The disclosed technical plan depends on NVIDIA GB300 NVL72 systems and NVIDIA DSX or DSX MaxLPS technologies. 高 SR005, SR006
CR009 The launch materials assign partner roles across Indonesian execution, accelerated computing, and AI-native networking. 高 SR005, SR006
CR010 Zankore's announced revenue-sharing and credit-support model is intended to align deployment with customer demand, but no operating outcome is disclosed. 中 SR002
CR011 Ooredoo is described as lead investor with a 49% stake, while Indosat Ooredoo Hutchison, Nokia, and NVIDIA are described as launch partners. 高 SR004, SR005, SR006
CR012 The reviewed public sources do not establish Zankore's permits, privacy approvals, security certifications, insurance, litigation posture, or incident history. 低
CR013 The Next Web describes NVIDIA as both a Zankore backer and the supplier of the hardware the financing will purchase. 中 SR007
CR014 AInvest characterizes the structure as roughly four dollars of senior debt for each dollar of founding equity and warns that demand must arrive quickly to service the leverage. 中 SR008
CR015 Ooredoo describes US$800 million as its commitment to support launch and initial development, while the assigned evidence does not establish Zankore's total paid-in equity or cash available. 中 SR004, SR005
CR016 Failure to evidence demand-linked staging or progress toward the first-half 2027 capacity target would be an observable warning that deployment and debt exposure are diverging. 中 SR002, SR006, SR007
CV001 Ooredoo announced an approximately US$800 million commitment as lead investor for a 49% stake in Zankore. 高 SV004, SV005
CV002 Ooredoo estimates its Zankore investment will contribute approximately US$600 million of cumulative EBITDA to Ooredoo over the first five years. 中 SV005
CV003 Zankore secured a senior term loan facility of up to US$3.1 billion involving Citi, ING, Natixis CIB, Qatar National Bank, and UOB. 高 SV001, SV003, SV006, SV007
CV004 The up-to-US$3.1 billion facility is intended to fund the purchase and deployment of NVIDIA GPU infrastructure. 中 SV001, SV003, SV007
CV005 The financing includes revenue-sharing and credit-support arrangements designed to link deployment pace to customer demand. 中 SV001, SV007
CV006 Zankore launched in August 2026, approximately one month before the September 2026 loan announcement. 高 SV007, SV008
CV007 Zankore's stated capacity path is 100 MW initially, approximately 200 MW in the first half of 2027, and 1 GW within three years. 高 SV001, SV003, SV007, SV008
CV008 The US$800 million commitment for a 49% stake does not by itself establish a current equity or enterprise valuation because investment staging, dilution, rights, and liability assumptions are undisclosed. 中 SV004, SV005
CV009 Reviewed public sources do not disclose loan pricing, maturity, covenants, collateral, recourse, or the full preference and security waterfall. 中 SV001, SV007
CV010 The Next Web identifies downside risk if repayment depends on demand for GPUs whose collateral value can decline as the market changes. 中 SV007
CV011 Ooredoo cites industry analysis projecting Southeast Asian data-center capacity demand to grow approximately 3.5 times by 2030, predominantly from AI workloads. 中 SV005
CV012 Zankore and Ooredoo state that NVIDIA DSX MaxLPS is designed to enable up to 40% more compute within the same power budget. 中 SV005, SV008
CV013 Zankore targets GPU capacity for AI training, fine-tuning, inference, and agentic workloads across enterprise, developer, government, and hyperscaler segments. 高 SV007, SV008
CV014 The reviewed evidence names no anchor customer and does not disclose whether the initial facilities will be owned or leased. 中 SV001
CV015 KuCoin reports that Asian bank-led GPU financings have been scarce because of chip residual-value uncertainty and geopolitical risk. 中 SV002
CV016 Available public evidence supports a research-more and track recommendation, not a price-approved investment, until contracted demand, financing protections, and fully burdened economics are verified. 中 SV001, SV005, SV007
来源
编号出版方标题引文
SO001 Global Trade Review Natixis applies commodity trade finance structure to US$3.1bn Zankore AI loan Natixis’ tranche finances the purchase, installation and operation of the equipment and is secured against the GPU equipment itself, as well as material contracts and take-or-pay offtake agreements with paying customers.
SO002 TechNode Global Indonesia's Indosat jointly launches 1GW AI factory platform Zankore Zankore will deliver approximately 200 MW of AI capacity in the first half of 2027, powered by NVIDIA GB300 NVL72 processors.
SO003 TNW Nvidia-backed Zankore raises a $3.1bn loan to buy Nvidia GPUs for Indonesia One important detail about Zankore’s loan remains unclear: what exactly secures it.
SO004 Data Center Dynamics Zankore secures $3.1bn senior term loan facility for AI cloud buildout Zankore was founded in August 2026 when Indonesian telecoms operator Indosat Ooredoo Hutchison paired with the wider Ooredoo Group, Nokia, and Nvidia to launch an Indonesian AI cloud provider.
SO005 Dealroom News Ooredoo takes 49% of Zankore in $800M AI compute bet Zankore will supply AI computing capacity to cloud and enterprise customers, with plans to scale to as much as 1 gigawatt of NVIDIA DSX AI Factory capacity.
SO006 Ooredoo Group Ooredoo Group Scales its AI Compute Ambitions through Investment in Zankore As founding shareholder and lead investor with a 49% stake, Ooredoo Group will commit approximately US$800 million to support the platform's launch and initial development.
SO007 Developing Telecoms Zankore lands US$3.1b loan facility to finance Nvidia-powered AI platform Proceeds from the debt financing will support the acquisition and deployment of Nvidia GPU infrastructure.
SO008 AInvest Zankore's $3.1 Billion "Financing" Is Debt: Reading the Leverage Behind the AI Cloud Hype That $3.1 billion is debt, and it is roughly four times the approximately $800 million of founding equity behind the venture.
SM001 TeckNexus When Your Connectivity Provider Becomes Your Compute Provider Zankore’s initial capacity isn’t due until the first half of 2027, and the platform, in both its Ooredoo Group and Indosat-branded forms, is still in build-out rather than full commercial operation.
SM003 Data Center Dynamics Zankore secures $3.1bn senior term loan facility for AI cloud buildout A joint venture between Indosat Ooredoo Hutchison, Nokia, and Nvidia, Zankore is targeting an initial buildout of 100MW of Nvidia AI infrastructure in Indonesia, ultimately aiming to expand to 1GW.
SM004 Dealroom News Ooredoo takes 49% of Zankore in $800M AI compute bet Zankore will supply AI computing capacity to cloud and enterprise customers, with plans to scale to as much as 1 gigawatt of NVIDIA DSX AI Factory capacity.
SM005 Ooredoo Group Ooredoo Group Scales its AI Compute Ambitions through Investment in Zankore alongside Indosat Ooredoo Hutchison, Nokia and NVIDIA Zankore will deliver approximately 200 megawatts (MW) of AI capacity in the first half of 2027, powered by NVIDIA GB300 NVL72
SM006 TechNode Global Indonesia's Indosat jointly launches 1GW AI factory platform Zankore with Qatar's Ooredoo, Finland's Nokia, US' NVIDIA In a statement on Thursday, Indosat said the platform can serve hyperscalers, enterprises, and governments across Asia-Pacific, with Indonesia as the initial deployment market.
SM007 The Next Web Nvidia-backed Zankore raises a $3.1bn loan to buy Nvidia GPUs for Indonesia The company sells GPU capacity for AI training, fine-tuning, inference and agentic workloads to enterprises and developers across the region.
SM008 AInvest Zankore's $3.1 Billion "Financing" Is Debt: Reading the Leverage Behind the AI Cloud Hype None of that headline capacity means anything until it is filled.
SM009 Reuters via Zawya Qatar's Ooredoo leads $800mln investment in AI platform Zankore The initial investment is expected to contribute approximately $600 million in cumulative EBITDA over its first five years, according to current estimates
SP001 Remio ZanKore AI Factory Targets 1 GW, but Central Java Faces a Water Test The ZanKore AI factory has disclosed its computing targets, hardware direction, partners, and prospective customers. It has not publicly provided the same level of detail about cooling technology, expected water consumption, water sources, or drought procedures.
SP002 TechXplore Zankore's AI factory: Can Central Java afford the water cost? The question is therefore not simply how much investment enters Batang, but how much value stays there—and who bears the environmental costs.
SP003 Data Center Dynamics Zankore secures $3.1bn senior term loan facility for AI cloud buildout Zankore has secured $3.1 billion in financing for the first phase of its AI cloud buildout.
SP004 Dealroom News Ooredoo takes 49% of Zankore in $800M AI compute bet Zankore will supply AI computing capacity to cloud and enterprise customers, with plans to scale to as much as 1 gigawatt of NVIDIA DSX AI Factory capacity.
SP005 Ooredoo Group Ooredoo Group Scales its AI Compute Ambitions through Investment in Zankore alongside Indosat Ooredoo Hutchison, Nokia and NVIDIA The platform also provides a foundation for future capacity expansion.
SP006 TechNode Global Indonesia's Indosat jointly launches 1GW AI factory platform Zankore with Qatar's Ooredoo, Finland's Nokia, US' NVIDIA Zankore will deliver approximately 200 MW of AI capacity in the first half of 2027, powered by NVIDIA GB300 NVL72 processors.
SP007 The Next Web Nvidia-backed Zankore raises a $3.1bn loan to buy Nvidia GPUs for Indonesia If customers sign enough contracts, the model works for everyone. If demand falls short, the lenders are left with a large pool of GPUs whose value could decline as the market changes.
SP008 AInvest Zankore's $3.1 Billion Financing Is Debt: Reading the Leverage Behind the AI Cloud Hype That $3.1 billion is debt, and it is roughly four times the approximately $800 million of founding equity behind the venture.
SP009 Coloprice Zankore Secures $3.1 Billion Loan to Build Nvidia AI Cloud Capacity in Indonesia 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.
SI001 Indo Premier ISAT - Zankore: entering the AI factory era We estimate a project IRR of 12-13% and an equity payback period of <3 years.
SI002 MegaDeal Indonesia ZanKore, Growth $800M – Global VC Megadeal Briefing There is no individual founder narrative here; the entity’s basis for existence is the combined capital and technology of four blue-chip sponsors — Ooredoo Group (Qatari telecom group), Indosat Ooredoo Hutchison (Indonesia’s second-largest mobile operator, “IOH”), NVIDIA, and Nokia.
SI003 AInvest Zankore's $3.1 Billion "Financing" Is Debt: Reading the Leverage Behind the AI Cloud Hype That $3.1 billion is debt, and it is roughly four times the approximately $800 million of founding equity behind the venture.
SI004 Coloprice Zankore Secures $3.1 Billion Loan to Build Nvidia AI Cloud Capacity in Indonesia 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.
SI005 Data Center Dynamics Zankore secures $3.1bn senior term loan facility for AI cloud buildout The senior term loan facility has been provided by a consortium of regional and international financial institutions and will be used to support the development and deployment of Zankore's AI capacity.
SI006 Dealroom Ooredoo takes 49% of Zankore in $800M AI compute bet Zankore will supply AI computing capacity to cloud and enterprise customers, with plans to scale to as much as 1 gigawatt of NVIDIA DSX AI Factory capacity.
SI007 Ooredoo Group Ooredoo Group Scales its AI Compute Ambitions through Investment in Zankore alongside Indosat Ooredoo Hutchison, Nokia and NVIDIA As founding shareholder and lead investor with a 49% stake, Ooredoo Group will commit approximately US$800 million to support the platform’s launch and initial development.
SI008 TNGlobal Indonesia's Indosat Ooredoo Hutchison has launched Zankore by Indosat In a statement on Thursday, Indosat said the platform can serve hyperscalers, enterprises, and governments across Asia-Pacific, with Indonesia as the initial deployment market.
SE001 Indosat Ooredoo Hutchison Indosat, Ooredoo Group, Nokia, and NVIDIA Launch Zankore by Indosat, Ready to Serve the Region with Full Stack AI Infra Platform Zankore by Indosat is building one of Southeast Asia’s largest AI factory platforms for the nation, using NVIDIA DSX, the reference architecture, platform and ecosystem playbook for designing and operating AI factories.
SE002 HyperFRAME Research Nokia Powers Sovereign APAC AI Infrastructure through Zankore Alliance By delivering an ultra-low-latency networking stack, Nokia ensures that high-throughput agentic workloads scale predictably and securely.
SE003 Data Centre Solutions Zankore secures $3.1B loan to build NVIDIA-powered AI infrastructure in Indonesia The platform, built on NVIDIA's full-stack AI computing technologies, will provide GPU clusters for AI training and inference, giving customers access to dedicated AI compute on demand.
SE004 Data Center Dynamics Zankore secures $3.1bn senior term loan facility for AI cloud buildout Ultimately planning a 1GW build-out, the company is targeting approximately 200MW of AI capacity in the first half of 2027, which will be powered by Nvidia GB300 NVL72.
SE005 Dealroom News Ooredoo takes 49% of Zankore in $800M AI compute bet Zankore will supply AI computing capacity to cloud and enterprise customers, with plans to scale to as much as 1 gigawatt of NVIDIA DSX AI Factory capacity.
SE006 Ooredoo Group Ooredoo Group Scales its AI Compute Ambitions through Investment in Zankore alongside Indosat Ooredoo Hutchison, Nokia and NVIDIA Zankore will deliver approximately 200 megawatts (MW) of AI capacity in the first half of 2027, powered by NVIDIA GB300 NVL72.
SE007 TNGlobal Indonesia's Indosat jointly launches 1GW AI factory platform Zankore with Qatar's Ooredoo, Finland's Nokia, US NVIDIA The platform uses NVIDIA DSX, a reference architecture that integrates accelerated computing, networking, power, cooling, and operations into a unified system.
SE008 The Next Web Nvidia-backed Zankore raises a $3.1bn loan to buy Nvidia GPUs for Indonesia The risk of the AI infrastructure buildout is therefore shared with a growing group of lenders.
SU001 TechNode Global Zankore secures up to $3.1B for Southeast Asia AI infrastructure Zankore said it plans to offer cloud services to enterprises, startups, developers, AI-native companies and institutions across Indonesia and the wider Southeast Asian market.
SU002 Telecompaper IOH, Nokia, Nvidia launch Zankore AI joint venture backed by USD 800 million Ooredoo investment The platform will serve customers across the Asia-Pacific region, with an initial focus on Southeast Asia.
SU003 Subhadra AI Indonesia's AI Surge: Zankore AI Factory, CoreWeave Expansion, and 40% Enterprise Adoption
SU004 Data Centre Solutions Zankore secures $3.1B loan to build NVIDIA-powered AI infrastructure in Indonesia Through the collaboration, Zankore plans to offer NVIDIA-powered cloud services to AI-native companies, enterprises, startups, developers and institutions across Indonesia and Southeast Asia.
SU005 Data Center Dynamics Zankore secures $3.1bn senior term loan facility for AI cloud buildout Zankore is targeting an initial buildout of 100MW of Nvidia AI infrastructure in Indonesia, ultimately aiming to expand to 1GW.
SU006 Dealroom Ooredoo takes 49% of Zankore in $800M AI compute bet Zankore will supply AI computing capacity to cloud and enterprise customers, with plans to scale to as much as 1 gigawatt of NVIDIA DSX AI Factory capacity.
SU007 Ooredoo Group Ooredoo Group scales its AI compute ambitions through investment in Zankore Zankore will deliver approximately 200 megawatts (MW) of AI capacity in the first half of 2027, powered by NVIDIA GB300 NVL72.
SU008 TNW Nvidia-backed Zankore raises a $3.1bn loan to buy Nvidia GPUs for Indonesia If customers sign enough contracts, the model works for everyone. If demand falls short, the lenders are left with a large pool of GPUs whose value could decline as the market changes.
SU009 AInvest Zankore's $3.1 Billion "Financing" Is Debt: Reading the Leverage Behind the AI Cloud Hype None of that headline capacity means anything until it is filled.
SR001 The Economic Times Nvidia: Ooredoo-backed Zankore secures $3.1 billion financing for Nvidia AI cloud platform Zankore is a dedicated AI compute and neocloud platform being developed with Indonesian telecom operator Indosat Ooredoo Hutchison, Nokia and Nvidia.
SR002 The Asian Banker Zankore expands Southeast Asia’s AI compute infrastructure with NVIDIA A revenue-sharing and credit-support model will align infrastructure deployment with customer demand, helping broaden access to large-scale accelerated computing while providing Zankore with a capital-efficient path to scale.
SR003 Data Center Dynamics Zankore secures $3.1bn senior term loan facility for AI cloud buildout Zankore has secured $3.1 billion in financing for the first phase of its AI cloud buildout.
SR004 Dealroom News Ooredoo takes 49% of Zankore in $800M AI compute bet For Ooredoo, a minority stake in Zankore is a bet on owning compute as well as the networks that connect it.
SR005 Ooredoo Group Ooredoo Group Scales its AI Compute Ambitions through Investment in Zankore alongside Indosat Ooredoo Hutchison, Nokia and NVIDIA Ooredoo will lead this landmark transaction with approximately US$800 million commitment to develop Zankore
SR006 TNGlobal Indonesia's Indosat jointly launches 1GW AI factory platform Zankore with Qatar's Ooredoo, Finland's Nokia, US' NVIDIA Zankore will deliver approximately 200 MW of AI capacity in the first half of 2027, powered by NVIDIA GB300 NVL72 processors.
SR007 The Next Web Nvidia-backed Zankore raises a $3.1bn loan to buy Nvidia GPUs for Indonesia One important detail about Zankore’s loan remains unclear: what exactly secures it.
SR008 AInvest Zankore's $3.1 Billion "Financing" Is Debt: Reading the Leverage Behind the AI Cloud Hype Whatever the cheerleading says, the binding constraint is not the spec sheet — it is whether anchor customers appear fast enough to service four-to-one leverage on hardware that is worth less with every shipping cycle.
SV001 ColoPrice Zankore Secures $3.1 Billion Loan to Build Nvidia AI Cloud Capacity in Indonesia 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.
SV002 KuCoin Zankore Secures $3.1 Billion in GPU Financing Amid Rising AI Demand in Asia In Asia, bank-led deals have been scarce due to uncertainty over chip residual values and geopolitical risks stemming from U.S.-China technological competition.
SV003 Data Center Dynamics Zankore secures $3.1bn senior term loan facility for AI cloud buildout Zankore has secured $3.1 billion in financing for the first phase of its AI cloud buildout.
SV004 Dealroom Ooredoo takes 49% of Zankore in $800M AI compute bet is investing about $800 million for a 49% stake in
SV005 Ooredoo Group Ooredoo Group Scales its AI Compute Ambitions through Investment in Zankore As founding shareholder and lead investor with a 49% stake, Ooredoo Group will commit approximately US$800 million to support the platform's launch and initial development.
SV006 FinanceAsia Zankore lands $3.1bn syndicated loan for AI infrastructure buildout As Asia's AI buildout continues apace, the large deal from Citi, ING, Natixis, UOB and QNB, will help the new firm build GPU cloud infrastructure in Indonesia with Nvidia.
SV007 The Next Web Nvidia-backed Zankore raises a $3.1bn loan to buy Nvidia GPUs for Indonesia Five international banks have now underwritten up to $3.1bn for a business with only about four weeks of operating history and a telecoms parent.
SV008 Zankore ZanKore by Indosat launch press release ZanKore will deliver approximately 200 megawatts (MW) of AI capacity in the first half of 2027, powered by NVIDIA GB300 NVL72.