Introduction

Rent NVIDIA H100s, H200s, B200s and RTX accelerators by the hour at outlet prices.

GPU Outlet is a clearance marketplace for high-end GPUs. We aggregate idle capacity from neoclouds, AI labs, and infrastructure providers, then let you rent it by the hour at prices the rest of the market can’t touch.

Quickstart

Sign up, top up, and launch your first H100 in under three minutes.

API reference

REST endpoints for auth, catalog, rentals, billing and SSH keys.

GPU catalog

RTX 4090 → B300 — current hourly rates and VRAM specs.

SSH access

Add your public key once. Every pod injects it automatically.

What you get

Per-second billing, prepaid wallet

Rentals are billed every hour but metered every second. Top up your wallet once via Stripe and let it draw down. No invoices, no contracts, no procurement calls.

Six flagship GPU families

RTX 4090 ($0.65/hr) · **RTX 5090** ($1.15/hr) · A100 80GB ($1.40/hr) · **H100 80GB** ($2.80/hr) · B200 ($6.00/hr) · **B300** ($7.50/hr) — cleared inventory from real providers, not synthetic spot pricing.

Instant SSH

Add your public key once in Settings → SSH keys. We inject it into every pod you launch — your private key never leaves your machine.

No lock-in

Stop an instance anytime; you’re only billed for seconds actually consumed. Reservations and contracts are opt-in, never the default.

How it fits together

The marketplace splits cleanly: the catalog surfaces what’s available right now, the rental orchestrator allocates pods from third-party providers, and the billing engine debits your prepaid wallet at the metered rate. Each layer has its own REST surface — see API reference.

Who this is for

GPU Outlet is built for developers and small teams that would rather ship than fight enterprise procurement:

• ML researchers running training jobs that don’t justify a long-term lease
• Inference operators spinning up nodes elastically as load fluctuates
• Hobbyists & students who want H100-class hardware for a weekend
• Agencies & studios rendering or fine-tuning on a per-project basis

If you need 18-month committed contracts with white-glove TAM coverage, you’re not our buyer — talk to one of the hyperscalers. If you want to pnpm dev against a real B200 by lunchtime, you’re in the right place.

Next steps

1. Create your account

   Email + 6-digit code, or one-tap Google / GitHub. No password. See Sign up.

2. Add a payment method

   Top up your prepaid wallet with Stripe. Card on file optional. See Topping up.

3. Add your SSH key

   Drop your id_ed25519.pub into Settings → SSH keys. See Connecting via SSH.

4. Launch an instance

   Pick a GPU family, pick a template, hit Continue. See Your first instance.

The dashboard at gpuoutlet.ai/app wraps every API call below — you don’t need to touch a terminal to use the product. The API exists for automation and CI pipelines.