What Is a Neocloud?
A neocloud is a cloud company that focuses almost entirely on renting high-end GPUs for AI work. Unlike hyperscalers that sell hundreds of services, neocloud providers keep their catalog small and center it on raw compute, bare-metal or thin-VM access, and fast networking.
According to SemiAnalysis, it's a new breed of cloud compute provider focused on offering GPU rental.
Neocloud Traits
- GPU-first with the latest NVIDIA hardware.
- Light virtualization for native-like speed.
- Simple pricing without complicated clauses.
- Easy to access and launch clusters in hours, not weeks.
Growth of Neocloud Providers
The rise of AI products and the demand for hardware caused a surge in neocloud providers. They are built around the GPU-as-a-Service (GPUaaS) model, where developers rent high-performance GPUs on demand instead of managing physical infrastructure.
Neocloud services provide flexibility for scaling AI workloads up or down instantly. That flexibility is especially valuable for training and inference with models that require significant compute.
Unstable Hardware Supply
The global computing market is no longer a predictable commodity cycle. Over the past several years, the industry has seen a series of boom-and-bust cycles that have made consistent GPU availability difficult for teams relying on hyperscalers.
There have been brief windows of price correction. Still, the overarching trend since 2020 has been defined by supply chain fragility and significant price spikes for core components.
| Year | GPU Market Status | RAM & Memory Trends | Sources |
|---|---|---|---|
| 2020/ 2021 |
Scarcity; prices reach 300% of MSRP due to high pandemic demand and crypto mining. | Prices rise steadily as global logistics fracture and remote work spikes. | Laptop Outlet (2025) |
| 2022 | Prices crash as supply surges due to the ETH "Merge," ending the mining boom. | Manufacturers overproduce to avoid shortages, leading to a market glut. | The Register (2022) |
| 2023 | Prices stabilize at retail. Availability of mid-range and high-end cards returns. | Record-low prices for DDR4/DDR5 as manufacturers clear excess inventory. | IntuitionLabs (2025) |
| 2024 | Focus shifts to AI silicon; consumer GPU supply becomes premium. | Prices rise as production shifts to High Bandwidth Memory (HBM) for AI. | JPR (2025) |
| 2025 | High-end GPU availability tightens; focus shifts to AI data centers. | RAMpocalypse: Consumer DDR5 prices surge by over 160% in several regions. | Digital Watch (2025) |
| 2026 | Structural shortage; enterprise lead times for GPUs stretch to 52 weeks. | RAM prices spike and account for roughly 23% of a standard PC's total cost. | Gartner (2026), Astute Group (2026) |
Cost Savings
Neocloud rates are 70-80% cheaper than hyperscalers for the same silicon. Thunder Compute rents an on-demand A100 80GB VM for $1.09/hr (source). By contrast, the same GPU on Oracle costs around $4/hr (source).
Focus and Speed
Because they run only GPU clusters, neoclouds ship new hardware first and tune their networks for AI collective-communication patterns. This lets builders train larger models sooner and at higher throughput.
Neocloud vs. Hyperscaler
| Comparison Point | Neocloud | Hyperscaler |
|---|---|---|
| Built Around | GPU compute | Breadth of managed services |
| Service Catalog | Narrow by design | Very broad, hundreds of services |
| Provisioning Speed | Seconds to minutes | Minutes, quota permitting |
| Typical A100 Price1 | $1.09-$2.79 per GPU-hr | $3.43-$5.07 per GPU-hr |
| Regional Footprint | Fewer regions | Global, dozens of regions |
| Typical Commitment | On-demand, often no minimum | On-demand, reserved, or committed spend |
The Three Types of Neocloud Providers
Not every neocloud is built for the same job, and the differences between them matter. The word covers at least three distinct business models, and picking the wrong tier wastes both money and time.
- Cluster and bare-metal specialists rent large, dedicated GPU clusters on multi-month or multi-year contracts. They target model builders running sustained, large-scale training, where guaranteed capacity is the product. CoreWeave and Crusoe operate here.
- On-demand platforms rent by the hour or the minute with no commitment. They are self-serve and built for developers who need a GPU right now for prototyping, fine-tuning, or a single-node run. Thunder Compute, Vast.ai, and Runpod operate here.
- Serverless inference platforms abstract the machine away entirely and bill per request or per second of execution. They suit teams deploying inference that scales to zero between calls. Modal and Baseten are common examples.
A provider focusing on 12-month cluster contracts and one that rents a single GPU for twenty minutes are both neoclouds, yet they are very different. Match the tier to the workload: reserved clusters for continuous training, on-demand instances for iterative development, serverless for spiky inference.
Thunder Compute sits in the on-demand tier, built for fast, self-serve access rather than long procurement cycles. You can connect straight from your editor with the VS Code or Cursor extension and launch an A100 in seconds, then shut it down when the run finishes. That model fits prototyping and single-node training, where paying for idle reserved capacity makes little sense.
How to Pick the Right Neocloud
- Check available GPUs: For training at scale, look for infrastructure with at least 400 Gbps InfiniBand or RoCE.
- Evaluate storage bandwidth: Look for at least 250GB/s aggregate.
- Compare pricing models: On-demand for tests, reserved or spot for long runs.
- Understand network infrastructure: Fat-tree or rail-optimized designs cut congestion.
- Run a benchmark: Fine-tune a familiar model to track tokens/second and total cost.
Neocloud Pricing
| Provider | A100 Price | H100 Price | Best For | Notes |
|---|---|---|---|---|
| Thunder Compute | $1.09 | $3.20 | On-demand prototyping and single-node training from VS Code or Cursor | US Central, on-demand VM |
| Vast.ai | $1.45 | $10.84 | Interruption-tolerant jobs at marketplace rates | Marketplace pricing |
| Hyperbolic | N/A | $3.19 | On-demand H100 inference | A100 not listed in current pricing reference |
| Runpod | $1.39 | $2.89 | On-demand pods and serverless inference | On-demand pricing |
| Nebius | N/A | $4.50 | European data residency and managed AI cloud | H100 NVLink on-demand |
| Crusoe Cloud | $2.00 | $4.50 | Sustainability-focused, energy-backed capacity | A100 PCIe on-demand |
| CoreWeave | $2.70 | $6.16 | Enterprise-scale, multi-node reserved training | 8-GPU node price, normalized per GPU |
| Lambda | $2.79 | $3.29 | Managed ML environments and self-serve clusters | US West, on-demand VM |
Hyperscaler Pricing
| Provider | A100 Price | H100 Price | Notes |
|---|---|---|---|
| AWS | $3.43 | $6.88 | US on-demand |
| Azure | $3.67 | $6.98 | Linux NC A100/H100 v4/v5, US pricing |
| Oracle Cloud | $4.00 | $10.00 | Bare-metal node price, normalized per GPU |
| Google Cloud | $5.03 | $10.98 | a2-ultragpu and a3-highgpu-1g VM pricing, US on-demand |
A Five-Step Action Plan
- Define the job: Model size, training days, budget cap.
- Short-list three neocloud companies with GPUs in stock.
- Spin up a 4-GPU node and run your workflow end-to-end.
- Track dollars per thousand training tokens as the metric.
- Reserve capacity once you hit the target price-performance.
When to Stay on Your Current Cloud
If you need dozens of managed services, strict FedRAMP or HIPAA compliance across many regions, or deep integration with existing enterprise IAM, the big clouds may still be smoother. Many teams blend approaches: train on a neocloud, then deploy inference on AWS, Azure, or GCP.
Last Thoughts on Neoclouds
Testing a neocloud is easy. Marketplace A100 listings start near $1.09/GPU-hr, and Thunder Compute offers fixed managed A100 access at $1.09/GPU-hr with instant H100 virtual machines at $3.20/GPU-hr. Spin up a VM, move your data, and see if it beats your current bill.
FAQ
What is a neocloud?
A neocloud is a cloud provider that specializes almost exclusively in renting high-end GPUs for AI workloads. Unlike hyperscalers that offer hundreds of services, neoclouds focus on raw GPU compute, bare-metal or thin-VM access, and fast networking for AI training and inference.
What's the best neocloud for renting H100s?
Thunder Compute offers available managed H100s at $3.20/hr. Other options include marketplace-style providers for lower rates, CoreWeave for enterprise multi-node clusters, and Lambda for managed ML environments.
What is a neocloud provider?
A neocloud provider is a specialized cloud platform focused on GPU-as-a-Service for AI workloads. Unlike hyperscalers, neoclouds offer faster provisioning, simpler infrastructure, and significantly lower costs for GPU compute.
What are neocloud companies?
Neocloud companies are a new category of cloud providers focused on delivering GPU infrastructure for AI. Leading neocloud companies include Thunder Compute, CoreWeave, Lambda, Crusoe Cloud, Nebius, Vultr, and Runpod.
How much cheaper are neoclouds compared to AWS or Azure?
Neoclouds typically cost 70-80% less than hyperscalers for equivalent GPU hardware. Thunder Compute charges $1.09/hr for an A100 80GB versus AWS at $3.43/hr and Azure at $3.63/hr for comparable configurations.
What are the main advantages of using a neocloud?
The primary advantages of neoclouds include lower costs per training hour, direct GPU access, elastic capacity, faster access to new NVIDIA hardware, and simpler pricing with less vendor lock-in than hyperscalers.
What are the disadvantages of neoclouds?
Neocloud trade-offs include: fewer geographic regions and compliance certifications, limited managed services like databases and event streaming, and the need to manage more of your own infrastructure stack compared to hyperscalers.
Who should use a neocloud instead of AWS or Google Cloud?
Neoclouds are ideal for AI researchers, ML engineers, and companies primarily training or fine-tuning large language models. Neoclouds work best for teams that need cost-effective GPU access and can manage their own environment without extensive managed services.
How do I choose the right neocloud provider?
Verify GPU type and interconnect speed, check storage bandwidth, compare on-demand versus reserved pricing, confirm support availability, and run a benchmark test before committing.