Neoclouds by operating region
40 of 95 providers publish an operating footprint. Some specify the location of a GPU cluster. Others list every market their cloud or partner network serves. Check the provider’s regional catalog for availability of a particular accelerator.
| Operating footprint | GPU models | Deployment and orchestration | |||
|---|---|---|---|---|---|
| GMI Cloud | US, Europe and Asia-Pacific (company regional footprint; GPU availability varies). | NVIDIA H100, H200 and Blackwell (Blackwell shown as pre-order on the home page). | Kubernetes-based platform with automated scaling, load balancing and multi-region deployment. | 85 | 2026-11-18 |
| SHARON AI | NEXTDC Tier IV facility, Melbourne, Australia. | NVIDIA H200, H100, L40S and A40; H200/B200 supercluster described separately. | Bare metal, virtualization, Docker and Kubernetes clusters. | 106 | 2026-11-18 |
| WhiteFiber | Madison, North Carolina, US (NC-1). | — | — | 143 | 2026-11-18 |
| Latitude.sh | 26 metro regions worldwide (cloud footprint, not a GPU availability guarantee). | — | Bare metal and Latitude Kubernetes Service (LKS). | 204 | 2026-11-18 |
| Global AI | Endicott, New York, US | NVIDIA GB300 NVL72 | — | 214 | 2026-11-18 |
| GMO GPU Cloud | Japan (domestic data centers). | NVIDIA HGX B300 and H200. | Managed HPC clusters with Slurm; bare-metal service. | 233 | 2026-11-18 |
| Hyperstack | Vestland, Norway · Quebec, Canada · Texas, United States | — | — | 234 | 2026-11-18 |
| Scaleway | France, Poland, Italy and the Netherlands; Sweden and Germany are upcoming, excluded from this footprint. | — | — | 295 | 2026-11-18 |
| Firebird | Armenia | NVIDIA B200; VR200 planned | API for bare-metal provisioning and remediation | 323 | 2026-11-18 |
| Neysa | India; platform positioned for Indian data residency. | — | GPU-as-a-Service and managed AI PaaS via Velocis. | 323 | 2026-11-18 |
| Shadeform | North America, Europe, Asia and Australia through partner clouds. | — | — | 344 | 2026-11-18 |
| BUZZ HPC | Canada and the Nordics. | NVIDIA GB200 NVL72, B200, H200 and H100. | Bare metal, managed Slurm and managed Kubernetes. | 393 | 2026-11-18 |
| FPT AI Factory | Vietnam and Japan | NVIDIA H100 and H200 | Managed Kubernetes clusters on bare metal or GPU VMs | 394 | 2026-11-18 |
| Taiga Cloud | Chester. | — | — | 502 | 2026-11-18 |
| SaladCloud | Nodes in more than 190 countries | — | Salad Container Engine for distributed GPU workloads | 522 | 2026-11-18 |
| E2E Networks | India, including GPU instances offered by Jarvislabs. | NVIDIA B200, H200 and H100 | — | 534 | 2026-11-18 |
| Hetzner | European Union | NVIDIA RTX PRO 6000 Blackwell Max-Q | Dedicated GPU servers | 543 | 2026-11-18 |
| PaleBlueDot AI | 60+ regions through its compute network. | — | Reserved GPU clusters and custom token-factory deployments. | 543 | 2026-11-18 |
| Dārya AI | Zev village, Tajikistan | NVIDIA H200 and Blackwell | — | 593 | 2026-11-18 |
| Exabits | Partner sites in North America, Europe and Asia | H200, H100, A100 and L40 | On-demand GPU provisioning through an API | 603 | 2026-11-18 |
| emma | Luxembourg (own infrastructure) | L40, A100, H100, H200, B200 and B300 | Bare metal for training and optimized inference instances | 614 | 2026-11-18 |
| Sakura Internet | Ishikari first zone, Japan | NVIDIA V100, scheduled to retire March 31, 2027. | Hourly GPU virtual machines | 613 | 2026-11-18 |
| TensorDock | 100+ partner locations in more than 20 countries. | NVIDIA H100 SXM5, A100 SXM4 and RTX 4090 (among the models listed). | KVM virtual machines with dedicated GPU passthrough and root access. | 612 | 2026-11-18 |
| SK Telecom | South Korea | NVIDIA Blackwell | Petasus AI Cloud virtualization and AI Cloud Manager | 653 | 2026-11-18 |
| Civo | India (Civo cloud region) | NVIDIA H100, H200 and B200 | Bare-metal instances and managed Kubernetes | 663 | 2026-11-18 |
| Cassava AI | South Africa | NVIDIA H200 | NVIDIA Multi-Instance GPU (MIG) partitions | 694 | 2026-11-18 |
| Cirrascale | US-West, US-East and US-South. | NVIDIA GB300, B300, B200, RTX PRO 6000; AMD MI355X and MI350; Qualcomm AI 100 Ultra | Dedicated multi-accelerator cloud servers; monthly billing and term blocks. | 694 | 2026-11-18 |
| Claro GPUaaS | Brazil | — | Elastic GPU capacity scaled on demand. | 693 | 2026-11-18 |
| Clore.ai | 10 bare-metal regions | B300, B200, H200, H100 and A100 | Dedicated servers with root SSH access | 692 | 2026-11-18 |
| CUDO Compute | UK, EU, North America, Asia-Pacific and Middle East. | NVIDIA H100 and H200 deployed in the company case study; newer architectures advertised separately. | — | 693 | 2026-11-18 |
| Denvr Dataworks | Canada and the USA (owned and operated data centers). | NVIDIA H200, H100, A100; Intel Gaudi2 | Bare metal, virtual machines and containers; self-service cluster provisioning. | 693 | 2026-11-18 |
| evroc | Europe | L40S and B200 | B200 bare-metal clusters | 694 | 2026-11-18 |
| Fluence | Asia (Q3 2026 auction capacity) | NVIDIA H100 (auctioned capacity) | Containers, virtual machines and bare metal | 693 | 2026-11-18 |
| GPUNET | Asia-Pacific, Europe and North America | NVIDIA H100 | On-demand virtual machines through a marketplace | 693 | 2026-11-18 |
| Hot Aisle | Switch Pyramid, Grand Rapids, Michigan. | AMD Instinct MI300X (eight GPUs per Dell PowerEdge XE9680 server). | Isolated on-demand compute, provisioned through an automated platform; customer Kubernetes runs above the platform. | 694 | 2026-11-18 |
| Lintasarta | Indonesia | NVIDIA H100 SXM | — | 693 | 2026-11-18 |
| Nebul | Europe | — | Cloud GPUs and isolated private AI infrastructure | 693 | 2026-11-18 |
| SoftBank | Japan | NVIDIA GB200 NVL72 | Multi-tenant Kubernetes and managed inference APIs | 692 | 2026-11-18 |
| Telefónica | Spain; edge services in Madrid, Valencia, Seville, Bilbao and A Coruña | NVIDIA GPUs in Spain edge pilot | GPU virtual machines at edge locations | 693 | 2026-11-18 |
| YTL AI Cloud | Johor, Malaysia | NVIDIA GB200 | YTL Cloud Orchestration Platform | 693 | 2026-11-18 |
Every neocloud on this list
Full rankingHow to read the columns
Methodology- Operating footprint
- Company or service geography as sourced; a general cloud footprint does not guarantee GPU stock in each region.
- GPU models
- Hardware advertised or tested, with reservations and future availability explicitly noted. Context only; not a fleet count or scoring input.
- Deployment and orchestration
- Available provisioning and cluster management services; future features are labelled as such.
- Rank
- Position by composite score, computed at build time from the signals below with the weights on /methodology/. Ties share the higher position. Never set by hand and never for sale.