# Asian neoclouds: the GPU-first AI clouds of Asia

DEPLOY tracks 50 verified GPU-first cloud operators across 14 Asian countries, each renting AI compute as its product, independent of the hyperscalers. Source: https://registry.deploy.report/asian-neoclouds

## Tracked Asian neoclouds

| Operator | Country | What it runs | NVIDIA partner |
|---|---|---|---|
| Yotta Data Services | India | - | Nvidia (H100 and L40S GPUs across Shakti Cloud AI Lab, Workspace VM, Bare Metal, and Cluster tiers) |
| Sify Technologies | India | - | - |
| Krutrim | India | - | - |
| Jio Platforms | India | - | - |
| GMI Cloud | Taiwan | - | - |
| Sustainable Metal Cloud | Singapore | - | - |
| FPT Smart Cloud | Vietnam | - | - |
| Sakura Internet | Japan | - | - |
| E2E Networks | India | NVIDIA B200 / H200 / H100 / A100; 2×1,024-GPU H200 clusters + 1,024-GPU B200 cluster; TIR platform | - |
| YTL AI Cloud | Malaysia | NVIDIA GB200 NVL72 Grace Blackwell (from H100); ~600 MW solar-supported data-center park | NVIDIA |
| Nxera | Singapore | - | Announced May 2024 with Nvidia to build a sovereign AI cloud on Nvidia GPUs in Singapore; specific GPU counts not publicly disclosed by Nxera |
| Tata Communications | India | NVIDIA H100 / H200 / L40S (Blackwell planned); HGX with InfiniBand | NVIDIA (full stack: NIM, NeMo, AI Enterprise) |
| Neysa | India | NVIDIA H100 / L40S on-demand; Velocis platform | - |
| NxtGen | India | NVIDIA A100 / H100 / L40S / T4; bare-metal + vGPU (SpeedCloud AI) | - |
| Cyfuture | India | NVIDIA H100 / H200 / A100 / L40S / V100; per-second billing | - |
| AceCloud | India | NVIDIA A30 / L4 / L40S / A100 / H100 / H200; INR billing | - |
| CtrlS | India | Bare-metal NVIDIA GPU clusters (GPU Private Cloud) | - |
| Pi Datacenters | India | NVIDIA A100 / H100; hourly pricing | - |
| ESDS | India | NVIDIA B200 / B300 / GB200 / GB300 + AMD Instinct MI300X SuperPODs | - |
| Ishan Technologies | India | 1,000+ GPUs; on-demand | - |
| NeevCloud | India | NVIDIA H100 / H200 / B200 / B300; HGX with 3,200 Gbps InfiniBand | - |
| Capitalonline | China | NVIDIA H20 / L20 (domestic) + overseas training clusters; bare-metal GPU | - |
| GMO GPU Cloud | Japan | NVIDIA H200 / HGX B300 (Blackwell Ultra); TOP500-ranked cluster | - |
| Highreso | Japan | NVIDIA HGX H200 / B200 (GPUSOROBAN) | - |
| RUTILEA | Japan | NVIDIA H100 on-demand | - |
| Ubitus | Japan | NVIDIA Blackwell GPU cloud (from H100 DGX) | - |
| KT Cloud | South Korea | GPU clusters (Moreh MoAI + AMD Instinct); Rebellions NPU-as-a-Service | - |
| SK Telecom GPUaaS | South Korea | 1,000+ NVIDIA B200 at Gasan AIDC; Petasus virtualization | Lambda (investor/tech), VAST Data |
| Siam AI | Thailand | NVIDIA H100 (1,024-GPU cluster) / H200 / GB200; on-demand pricing | Thailand's NVIDIA Cloud Partner |
| Lintasarta | Indonesia | NVIDIA H100 SXM (Deka GPU / GPU Merdeka) | First NVIDIA Cloud Partner in Indonesia |
| GreenNode | Vietnam | ~1,024 NVIDIA H100; bare-metal across 6 AZs | Preferred NVIDIA Cloud Partner |
| ePLDT | Philippines | NVIDIA H100 / H200; VITRO Sta. Rosa (50 MW AI-ready DC) | - |
| Viettel | Vietnam | NVIDIA H200 + DGX B200 (~1.5 exaFLOPS); vCGPU product | - |
| Freedom Holding | Kazakhstan | - | - |
| UzCloud | Uzbekistan | NVIDIA DGX B200 (8x B200/node); Run:ai orchestration | - |
| Bitdeer | Singapore | - | - |
| Lanyun | China | NVIDIA (RTX 4090, H100/H800) + domestic GPUs; own multi-province clusters | - |
| SenseCore | China | - | - |
| KDDI | Japan | NVIDIA GB200 NVL72; Osaka Sakai AI DC | - |
| Preferred Networks, Inc. | Japan | - | - |
| SoftBank AI Cloud | Japan | NVIDIA GB200 NVL72; Infrinia AI Cloud OS | - |
| NTT DATA | Japan | NVIDIA DGX B200 / DGX SuperPOD | - |
| Ubilink | Taiwan | NVIDIA H100 SuperPOD (~1,024 GPUs); 100% renewable | - |
| Taiwan AI Cloud | Taiwan | Taiwania + NVIDIA H100 / AFS PODs; pay-as-you-go | - |
| L&T Vyoma | India | NVIDIA H100 / H200 / B200 (B300 deploying); own AI DCs | - |
| Neutra Compute | Indonesia | NVIDIA + AMD accelerators; Cikarage Tier III/IV DC | - |
| Data Vault Pakistan | Pakistan | NVIDIA GPUs (H100-class); solar-powered Karachi AI factory | Telenor Pakistan |
| Alem.Cloud | Kazakhstan | 512 NVIDIA H200 (64 HGX servers, ~2 exaflops) | - |
| AutoDL | China | NVIDIA A100 / A800 / RTX 3090 / 2080Ti; own multi-node clusters | - |
| Lingjing Cloud | China | Own GPU/NPU capacity incl. ~5,000-card Wuxi Xinwei center | - |

## Notes

- Chinese AI compute is dominated by hyperscalers (Alibaba Cloud, Tencent, Huawei, Baidu), which fall outside the neocloud definition (independent, GPU-first, compute-as-the-product).
- Full definition, operator classes, and global roster: https://registry.deploy.report/neoclouds