GPU as a Service
Offer EU-sovereign GPUs to your customers in real time. Live availability and your partner pricing, over one simple API.
$
The fleet
Your available GPU fleet
Dedicated NVIDIA cards in Amsterdam. Public list prices per GPU-hour; partners get a flat discount on every one.
Start hereNVIDIA A30
24 GB
- Dedicated card, 24 GB VRAM
- Inference and small fine-tuning
- Our most affordable GPU
Scale upNVIDIA A100
80 GB HBM2e
- 80 GB HBM2e per card
- Large fine-tuning and multi-GPU training
- The production workhorse
Latest BlackwellRTX 6000 Blackwell
96 GB GDDR7
- 96 GB GDDR7 per card
- 5th-gen Tensor Cores, FP8 and FP4
- The newest cards in the fleet
EUR, excluding VAT. On demand, billed per hour. Full list at leaf.cloud/pricing.
How it works
Your brand. Our racks.
- Step 1
Check what's free
Query live availability and your partner prices. Every number reflects capacity a new instance can actually land on.
- Step 2
Provision for your customer
Create the project and attach the GPU through our OpenStack APIs, in your own flow, under your own brand.
- Step 3
Invoice on your terms
You set end-customer prices and invoice them directly. We bill you monthly with a per-customer usage breakdown.
The API
Five endpoints. Zero ceremony.
Read-only JSON with a bearer token. Pretty in your terminal, compact in your code, with an OpenAPI spec for everything else.
What you can launch right now
Free GPUs per model and, for every public flavor, how many instances fit right now. Counts only include hosts that accept new instances, and each instance must fit on a single host (GPUs and memory).
One line per GPU model
A compact summary per GPU model: free GPUs, the largest instance you can launch now, which flavors fit, and the cheapest price per GPU-hour.
One flavor
Specs, your price and current availability for a single flavor. Flavor names are case-insensitive.
Price list
Every public flavor with its hourly price, after your discount. Doesn't depend on live capacity, so it's safe to cache for a day.
Availability over time
Hourly average of free GPUs per model, for spotting quiet hours or planning launches.
Bring your own tooling
Import into Postman, Insomnia or a client generator.
Partners
Built for infrastructure intermediaries
Brokers and advisors
Match buyers to Amsterdam capacity with live availability and operator pricing.
AI compute platforms
Lease capacity under your own platform. Your customers stay yours; invoicing runs through you.
EU cloud providers
Add a GPU tier without buying cards. OpenStack on both sides keeps the integration short.
Consultancies and MSPs
Provision sovereign GPUs for client projects in regulated sectors, under one agreement.
EU jurisdiction by design
Sovereign compute your customers can sign off on.
- Dutch-owned and Amsterdam-based, with no US parent and no CLOUD Act path.
- Servers run at Leaf Sites, where their heat is reused.
- GDPR Article 28 DPA for your customers' data.
Start offering GPUs this week.
Tell us about your expected volume and we'll send a personalized rate sheet within one business day.