Shadeform GPU Cloud Marketplace - Compare 30+ Providers

TL;DR

TL;DR: Shadeform aggregates GPU compute from 30+ cloud providers — including Lambda, Vultr, Crusoe, and Nebius — into a single marketplace with one API. Deploy H100, H200, B200, or GB200 instances on-demand, scale clusters to 64 GPUs, or reserve capacity at a discount.

Source and Accuracy Notes

⚠️ This section is MANDATORY. All links must be verified from actual source, not guessed.

• Project page: shadeform.com — verified via direct fetch
• Documentation: docs.shadeform.ai — verified via direct fetch
• HN launch thread: news.ycombinator.com/item?id=37164864 — 62 points, author edgoode
• Trust center: trust.shadeform.com — SOC compliance verified
• Platform: platform.shadeform.ai — verified via direct fetch
• License: Proprietary (no open-source repo)
• Source last checked: 2026-07-02

What Is Shadeform?

Shadeform is a GPU cloud marketplace that aggregates compute capacity from over 30 cloud providers into a single point of access. Rather than opening accounts with Lambda, Vultr, Crusoe, and Nebius separately, you browse GPU models, availability, and pricing through Shadeform’s platform and deploy via their API or web interface.

The product is aimed at ML engineers and AI teams who need to compare GPU pricing across providers without managing multiple vendor relationships. The platform is a Y Combinator W24 company.

From the site:

“Deploy high quality GPUs at the best prices across 30+ top clouds worldwide.”

Supported GPU models include B300 SXM6, GB300 NVLink, B200 SXM6, GB200 NVLink, H200 SXM5, H100 SXM5, H100 PCIe, RTX Pro 6000 PCIe, RTX 5090 PCIe, RTX 4090 PCIe, A100 SXM4, A100 PCIe, L40S PCIe, and A6000 PCIe.

Setup Workflow

Step 1: Create an account

Sign up at platform.shadeform.ai. The platform redirects to a managed onboarding flow. No credit card is required to browse the marketplace directory.

Step 2: Browse the GPU directory

The cloud directory lists all 30+ providers and their available GPU inventory. Each provider page shows instance types, interconnect options (InfiniBand, RoCE, Ethernet), and region availability.

Step 3: Deploy via API

Shadeform exposes a REST API documented at docs.shadeform.ai. Authentication uses an API key generated from the platform dashboard. The docs cover instance creation, cluster deployment, and reservation requests.

Step 4: Reserve capacity (optional)

For sustained workloads, submit a reservation request through the platform or API. The form collects GPU model, quantity, interconnect type, duration, preferred location, and target start date. Shadeform sends quotes from their provider network within 24 hours.

Deeper Analysis

Marketplace model

Shadeform does not operate its own GPU hardware. It acts as an aggregator, connecting buyers with providers who list capacity on the platform. This is similar to how cloud marketplace brokers work — the actual compute runs on the provider’s infrastructure (Lambda, Vultr, Crusoe, Nebius, etc.), while Shadeform handles the unified billing and orchestration layer.

GPU availability

The platform spans datacenter GPUs from NVIDIA’s full stack:

• Blackwell architecture: B300 SXM6, GB300 NVLink, B200 SXM6, GB200 NVLink
• Hopper architecture: H200 SXM5, H100 SXM5, H100 PCIe
• Ampere architecture: A100 SXM4, A100 PCIe, RTX Pro 6000 PCIe
• Ada Lovelace: RTX 5090 PCIe, RTX 4090 PCIe
• Ada generation: L40S PCIe, A6000 PCIe

Interconnect options

When deploying clusters, you can specify interconnect type:

• InfiniBand — for multi-node training workloads that require high bandwidth
• RoCE — RDMA over Converged Ethernet
• Ethernet — standard networking, lowest cost
• No preference — lets providers quote the best available option

Cluster deployments

On-demand clusters of up to 64 GPUs are available with no long-term commitment. This covers the “burst to multi-node” scenario for distributed training without pre-purchasing reserved capacity.

Provider network

Providers confirmed on the platform include Nebius, Lambda, Vultr, Crusoe, DigitalOcean, and Massed Compute. Shadeform’s trust page confirms SOC compliance, which is relevant for enterprise procurement.

No public pricing

Shadeform does not publish hourly GPU rates on the site. Pricing is discovered by submitting reservation requests or deploying through the API. This is a marketplace pricing model — the same GPU model may have different rates depending on the provider, region, and interconnect type. Budget-conscious teams should plan for the reservation request workflow to get firm quotes before committing workloads.

Practical Evaluation Checklist

• Browse shadeform.com/directory/clouds for available providers and GPU models
• Review docs.shadeform.ai for API authentication and endpoint reference
• Check trust.shadeform.com for SOC compliance documentation before enterprise procurement
• Submit a reservation request for workloads requiring 32+ GPUs to compare quotes across providers
• Test the marketplace directory to confirm your preferred GPU model is available in your target region before writing integration code

Security Notes

• SOC compliance documentation is available at trust.shadeform.com
• API authentication uses a platform-issued key — store it in environment variables, not in code
• The actual compute runs on third-party provider infrastructure (Lambda, Vultr, etc.), each with their own security posture — verify the provider’s compliance certifications separately for regulated workloads

FAQ

Q: Does Shadeform operate its own GPU hardware? A: No. Shadeform aggregates capacity from 30+ third-party cloud providers. The actual compute runs on the provider’s infrastructure.

Q: What GPU models are available? A: NVIDIA datacenter and consumer GPUs including B300, GB300, B200, GB200, H200, H100, A100, L40S, and consumer RTX models (5090, 4090). Availability varies by provider and region.

Q: Is there a minimum commitment? A: On-demand instances and clusters have no minimum commitment. Reserved commitments offer discounted rates for 1–12 month terms.

Q: Does Shadeform publish pricing? A: No. Hourly rates are not listed on the site. Prices are quoted through the reservation request flow or via API after deployment.

Q: Does Shadeform have an open-source SDK? A: The API documentation is at docs.shadeform.ai. No public GitHub repository was found for the Shadeform platform itself.

Conclusion

Shadeform solves the multi-cloud GPU procurement problem for ML teams. Instead of managing accounts with Lambda, Vultr, Crusoe, and Nebius separately, you access a unified marketplace with one API and one billing relationship. The platform covers 30+ providers, the full NVIDIA GPU stack from H100 to B300, and cluster sizes up to 64 GPUs on-demand.

The main trade-off is pricing opacity — you cannot browse hourly rates before creating an account, and firm pricing requires a reservation request. For teams already evaluating multiple GPU providers, Shadeform’s aggregation model likely saves more procurement time than it costs in discovery overhead.

For a hands-on evaluation, start at platform.shadeform.ai and browse the GPU directory to confirm your required model and region are available before writing integration code.