What are neoclouds? Neoclouds vs hyperscalers vs GPU marketplaces

A neocloud is a cloud built around GPUs for AI workloads, rather than a general-purpose cloud that also rents GPUs. CoreWeave, Crusoe, Nebius and Lambda are the best-known examples. The term comes up constantly in AI infrastructure, but the line between a neocloud, a hyperscaler and a GPU marketplace isn't always clear. This guide explains the three kinds of GPU cloud, what each costs, and how to choose between them.

Three kinds of GPU cloud

There are a lot of GPU clouds. SemiAnalysis tracks 209 of them, up from 124 in its earlier count. They fall into three broad groups.

Type	Examples	What you get
Hyperscalers	Google Cloud, Oracle, Azure, AWS	GPUs inside a full general-purpose cloud
Neoclouds	CoreWeave, Crusoe, Nebius, Lambda	GPU-first clouds built for AI workloads
Developer clouds and marketplaces	RunPod, Together, Vast.ai, Lium	Low-cost GPUs, often self-serve, some from many hosts

Hyperscalers

The hyperscalers are the general-purpose clouds. GPUs are one product among hundreds, sitting next to storage, databases, networking and identity. If your company already runs on one of them, renting GPUs there keeps everything in one place, under contracts you already have.

Neoclouds

Neoclouds are clouds where the GPU is the product. They are built around AI training and inference, and they compete mostly on GPU price and availability rather than on a broad catalog of services. CoreWeave is the largest: it went public in March 2025 at about $23 billion.

The group is growing fast. Synergy Research Group puts neocloud revenue at over $25 billion in 2025, approaching $400 billion by 2031.

Developer clouds and marketplaces

At the low end of the price range sit developer clouds and marketplaces. Some run their own capacity with a self-serve, developer-first experience. Others are marketplaces that pool GPUs from many hosts. Prices are the lowest of the three groups, and how reliable and available the capacity is varies a lot from one provider to the next.

What each type costs: the H100 as an example

The clearest way to see the difference is the price of the same chip. Here is the on-demand NVIDIA H100 across all three groups:

Provider	Type	H100 on-demand, $ per GPU-hour
Google Cloud	Hyperscaler	$11.07
Oracle	Hyperscaler	$10.00
Azure	Hyperscaler	$6.98
AWS	Hyperscaler	$6.88
CoreWeave	Neocloud	$6.16
Crusoe	Neocloud	$3.90
Nebius	Neocloud	$3.85
Lambda	Neocloud	$3.29
RunPod	Developer cloud / marketplace	$2.89
Together	Developer cloud / marketplace	$1.99
Vast.ai	Developer cloud / marketplace	$1.79
Lium	Developer cloud / marketplace	$1.25

Source: getdeploying.com, 19 Sep 2026. On-demand list prices. SKUs vary by provider.

The bands are clear: $6.88 to $11.07 at the hyperscalers, $3.29 to $6.16 at the neoclouds, and $1.25 to $2.89 at developer clouds and marketplaces. Even inside one band the spread is wide. For the full breakdown, including contract and spot pricing, see our post on the H100 price per hour.

Price isn't the only difference: reliability

A low list price doesn't help if the job fails halfway or the capacity isn't there. Reliability varies a lot, and it isn't visible from a price page.

SemiAnalysis's ClusterMAX 2.0 rating (November 2025) grades GPU clouds on a scale from Platinum down to Underperforming. Only one cloud was rated Platinum. Most buyers never see ratings like this when they pick a provider. They learn which providers are reliable the hard way, one job at a time.

How to choose between them

• Choose a hyperscaler when your data, security setup and contracts already live there, and the GPU premium is worth keeping everything in one place.
• Choose a neocloud when GPUs are the main thing you buy, you want prices well below the hyperscalers, and you need clusters built for training and inference.
• Choose a developer cloud or marketplace when price matters most, the job can tolerate some variance in reliability, and you can test providers yourself.

Many teams end up using more than one. The trouble is that every extra provider means a new account and contract, a new API and tooling, and a new bill. We cover that cost in how to choose and switch GPU cloud providers.

Where Opengrid fits

Up front: Opengrid is pre-launch. The market view ships first and the routing API is in development.

Opengrid is built so you don't have to pick a type of cloud up front. The plan is a live view of pricing, availability and regions across hyperscalers, neoclouds and marketplaces, then one API key that routes each job to whichever provider wins on price, availability, region and reliability. To see the other tools that help you buy across clouds today, read our comparison of GPU cloud aggregators and marketplaces.