DEPLOY Taxonomy · as of 2026-08-30 17:08 UTC

Neoclouds

A neocloud is a GPU-first cloud whose product is AI compute, rented to third parties, independent of the hyperscalers, running dedicated AI data-center capacity. This is DEPLOY’s definition, the operator classes it sits against, and the global roster we track. The physical campuses these operators run are in the data-center registry.

Neoclouds tracked

With verified record

Regions

Coverage targets

What counts as a neocloud

A neocloud IS

GPU / AI-accelerator-first: compute is the product, not incidental to another business.
Rents that compute to third parties (not solely for its own workloads).
Independent of the hyperscalers (not AWS / Azure / Google Cloud / Oracle).
Operates or contracts dedicated AI data-center capacity (a real physical footprint).

A neocloud is NOT

Hyperscalers (general cloud at scale; AI is one product of many).
Pure colocation / REITs (sell space + power, not compute).
Enterprise self-builders that do not rent (xAI, Meta, Tesla for own use).
AI labs that do not resell compute, and brokers/aggregators with no capacity of their own.

The operator classes

Where the neocloud sits in the AI-compute landscape. The borders are the taxonomy.

Neocloud

GPU-first cloud whose product IS AI compute, rented to third parties, independent of the hyperscalers. CoreWeave, Lambda, Crusoe.

Hyperscaler

General-purpose cloud at global scale where AI compute is one product among many. AWS, Azure, Google Cloud, Oracle.

Colocation

Sells space, power, and cooling, not compute. The building, not the GPUs. Equinix, Digital Realty.

Enterprise self-build

Builds AI compute for its own use, not to rent. xAI Colossus, Meta, Tesla Dojo.

Sovereign program

Nation-backed AI-compute programs; often neocloud-shaped but state-directed. G42/Core42, HUMAIN.

The roster

The neoclouds DEPLOY tracks, by region. Names in color link to a verified company record; the rest are coverage targets we are building out. Notes are positioning, not verified metrics.

North America

CoreWeaveUnited States
The reference neocloud; NVIDIA-anchored GPU cloud with a large owned/leased campus footprint.
LambdaUnited States
GPU cloud and workstation maker turned large-scale AI-training cloud.
CrusoeUnited States
Energy-first AI cloud; builds compute at stranded/low-carbon power sites.
Together AIUnited States
Training + inference cloud built around open models.
Voltage ParkUnited States
Nonprofit-backed H100 cloud offering on-demand and reserved capacity.
Vast.aiUnited States
GPU marketplace aggregating third-party supply; borderline on the capacity test.
FluidstackUnited States
AI-training cloud contracting large GPU clusters for labs and enterprises.
Denvr DataworksCanada
AI-focused cloud with owned data-center capacity in North America.
TensorWaveUnited States
AMD-Instinct-based AI cloud, a rare non-NVIDIA neocloud.
Cirrascale Cloud ServicesUnited States
Long-running dedicated GPU/AI cloud offering bare-metal accelerator infrastructure.
RunPodUnited States
Developer-focused GPU cloud renting NVIDIA and AMD accelerators by the second; a Secure Cloud in partner data centers plus a Community Cloud marketplace.
IRENAustralia / United States
NASDAQ-listed renewable-powered operator scaling an NVIDIA GPU cloud from a bitcoin-mining base; anchored by a large Microsoft AI-cloud contract.

Europe

NebiusNetherlands
Full-stack AI cloud (ex-Yandex lineage) with owned European GPU capacity.
NscaleUnited Kingdom
Vertically integrated AI cloud building sustainable GPU capacity in Europe.
Northern DataGermany
HPC/AI cloud operator pivoting from crypto infrastructure to GPU compute.
ScalewayFrance
Iliad-group cloud with a dedicated AI/GPU supercomputer offering.
SesterceFrance
European GPU cloud building sustainable large-scale AI-compute capacity.
NexGen Cloud (Hyperstack)United Kingdom
London GPU cloud renting NVIDIA H100/H200/Blackwell via its on-demand Hyperstack platform and reserved AI Supercloud; NVIDIA Elite Cloud Partner.
Verda (DataCrunch)Finland
Helsinki neocloud (rebranded from DataCrunch in 2025) renting NVIDIA GPUs from renewable-powered Nordic data centers; NVIDIA Preferred Partner.

Asia

Yotta Data ServicesIndia
Hiranandani-group AI cloud; among India's first large NVIDIA GPU deployments.
Sify TechnologiesIndia · conglomerate unit
Data-center and cloud operator building GPU-as-a-service capacity in India.
KrutrimIndia
Ola-backed AI company operating its own GPU cloud for Indian model training.
Jio (Reliance)India · conglomerate unit
Telco building nation-scale AI compute; neocloud-adjacent within a conglomerate.
GMI CloudTaiwan
NVIDIA-partner GPU cloud serving APAC AI workloads.
Sustainable Metal CloudSingapore
Immersion-cooled GPU cloud targeting efficient AI compute in SE Asia.
FPT Smart CloudVietnam · conglomerate unit
FPT-group AI factory / GPU cloud, an NVIDIA partner in Vietnam.
Sakura InternetJapan
Japanese cloud host scaling a government-supported GPU cloud.
E2E NetworksIndia
NSE/BSE-listed Indian GPU cloud running large NVIDIA H200/B200 clusters and the TIR platform; Larsen & Toubro is a strategic investor.
YTL AI CloudMalaysia · conglomerate unit
YTL Power's NVIDIA-built GPU cloud at the Kulai, Johor green data-center park; runs Blackwell for Malaysian sovereign AI.
Tata CommunicationsIndia · conglomerate unit
Tata Communications' Vayu NVIDIA GPU cloud; a GPU-first product line within a diversified telecom operator.

Middle East

Core42 (G42)United Arab Emirates · sovereign adjacent
G42's AI-cloud arm; neocloud-shaped but state-linked.
HUMAINSaudi Arabia · sovereign adjacent
PIF-backed national AI-compute champion; neocloud-shaped, sovereign-directed.

The campuses these operators run, ranked by capacity, are in the data-center registry. This taxonomy is DEPLOY’s; coverage status is stated honestly and fills in as records are verified.