shadeform.com

What It Is

Shadeform is a GPU Cloud Marketplace. It is not positioned as an AI application that directly provides model capabilities, but rather as a platform that aggregates GPU cloud resources for training and inference teams. Its core selling points are that there are no sales calls or long-term commitments required: users can self-serve and deploy multi-node GPU clusters within minutes, while managing 30+ GPU clouds from a single platform.

Core Capabilities and Use Cases

Based on the main content, Shadeform supports GPU types such as H200, H100, and A100. On-demand clusters range from 16-64 GPUs, consisting of 2-8 nodes with 8 GPUs per node, and the platform emphasizes high-performance interconnect networking. This makes it better suited for distributed training, large-model fine-tuning/training, peak scaling for inference services, model testing, and teams that need temporary access to additional high-end GPUs. Customer feedback also focuses on “quick access to scarce GPUs,” “scaling across providers,” and “reducing procurement friction.”

Pricing and Trial

The pricing model is fairly clear, but lacks detail: on-demand clusters are billed by the second, have no minimum runtime, and can be stopped at any time, making them suitable for short-cycle workloads and elastic demand. If you need more than 64 GPUs, you must submit a reserved commitment, and reserved commitments may receive discounts depending on the term. The main content does not disclose specific GPU unit prices, free credits, trial policies, or payment methods, so it is not possible to assess the absolute cost level.

Pros and Cons

The advantages are that Shadeform covers mainstream high-end GPUs, supports interconnected 16-64 GPU clusters, offers a high degree of self-service, and provides unified access to multiple vetted cloud providers. Per-second billing also helps reduce idle costs. The limitations are that the minimum starting point is 16 GPUs, making it unsuitable for single-GPU use, small-scale experiments, or individual developers; clusters above 64 GPUs also cannot be launched directly through on-demand self-service. In addition, the main content does not explain SLA, regions, storage, bandwidth, image environments, API/CLI support, compliance, or data privacy mechanisms.

Who It’s For and Access from China

Shadeform is better suited for AI companies, data science teams, and infrastructure teams that already have clear training or inference workloads and need to scale GPU capacity quickly. Individual users and developers who only need low-cost single-GPU instances may be better served by RunPod, Vast.ai, Lambda Cloud, or GPU offerings from major public clouds. Access from mainland China, RMB payments, and Chinese-language support are not mentioned in the main content, so network availability should be considered unknown. If access or payment is limited, alternatives such as domestic cloud provider GPUs, AutoDL, or Lepton AI could be considered.