Skip to main content

​
Welcome to Galadriel

Galadriel is a GPU marketplace that connects ML engs and companies with GPU clusters through a transparent, market-driven orderbook system. Get instant access to 3000+ H100, H200, B200, and B300 GPU clusters with best prices on the market without long-term contract lock-ins.

​
Why Galadriel?

Best Price on the Market

Supply and demand determine prices through a transparent orderbook. Save 30-40% by using limit orders or off-peak scheduling.

Bare Metal Kubernetes

0-2% overhead vs 5-10% for VMs. Kubernetes pods run directly on bare metal nodes with no hypervisor layer.

DevExp easy as using Stripe

Install our CLI, run a single command, and get access to GPU clusters in under 60 seconds. No complex setup, no waiting lists.

Scalable Supply

Access to 3000+ GPUs with capacity to scale to thousands more. Whether you need 8 GPUs or 800, we have the infrastructure ready.

Resell Unused Time

Finished training early? Resell unused GPU time back to the marketplace and recover your costs.

​
Quick Example

# Install CLI
curl -sSL https://galadriel.com/install.sh | sh

# Configure
galadriel configure

# Buy 16 H100 GPUs for 24 hours (market order)
galadriel order buy --gpu-type h100 --gpus 16 --duration 24h --zone us-west-1

# Or use a limit order to save money
galadriel order buy --gpu-type h100 --gpus 16 --duration 24h --price 1.30 --zone us-west-1

# Deploy your workload
kubectl apply -f training-job.yaml

# Finished early? Resell remaining time
galadriel lease resell lse_xyz123 --price 1.20

​
How It Works

  1. Place an Order: Submit buy or sell orders through CLI, API, or SDK
  2. Match in Orderbook: Orders are matched using price-time priority
  3. Instant Provisioning: GPUs are provisioned in under 60 seconds
  4. Deploy Workloads: Use kubectl or SSH to run your workloads
  5. Resell if Needed: Finished early? Resell unused time to recover costs

​
Platform Economics

Galadriel operates as a two-sided marketplace:
  • 10% platform fee (transparent, no hidden costs)
  • Customers: Set maximum prices with limit orders
  • Providers: Set minimum prices for their GPUs
  • Orderbook matching: Continuous double auction, similar to stock exchanges

​
What’s Included

All instances include:
  • Full node exclusivity (8 GPUs minimum per customer)
  • InfiniBand 400 Gbps with RDMA support
  • At least 2TB of local NVMe storage per node
  • Unlimited network egress
  • SSH and kubectl access
  • Email/Slack support

​
Next Steps

Quickstart Guide

Get started in under 5 minutes

How the Marketplace Works

Learn about orderbooks and order types

API Reference

REST API documentation

Pricing Guide

Understand marketplace pricing