Partner program
Put your GPUs to work.
We handle the rest.
Plug your GPUs into Runcrate, from one machine to a datacenter, and set your own rate. We find the customers, bill them and support them.
Runcrate Hosts
Plug in your GPUs.
Here is the deal.
Paste one command into the machine, then the token from your dashboard. It checks every GPU, then waits for you to list them.
curl -fsSL https://get.runcrate.ai/host | sudo shOne GPU-hour
You get $1.95 per GPU-hour. Customers pay $2.79; Runcrate keeps $0.84 (30% of what the customer pays).
Paid every Monday through Stripe, for earnings more than 7 days old, once you're owed $10. Runcrate pays Stripe's fees.
Install one command
Ubuntu 22.04, 24.04 or Debian 12 on bare metal, with virtualization on in the BIOS. Datacenter cards and RTX 30, 40 and 50 series.
Set your own rates
Price each GPU per hour for spot, on-demand and reserved. You see what the customer pays beside every rate.
Set up payouts
Tell us where to pay you, once. Every rental, downtime refund and payout shows in your dashboard.
On your machine
What runs on your machine.
And what never does.
Customers run in VMs only
Each rental is a virtual machine with your GPUs passed through. Customer code never runs on your host itself.
Outbound HTTPS only
The agent calls Runcrate over HTTPS and never opens an inbound port on your network. NAT is fine.
Walled off from your network
A VM reaches the internet and nothing else: not your host, your LAN or other VMs. The rules are re-applied every 5 seconds, and VM networks go down whenever the agent isn't running.
Wiped after every rental
The VM's disk is overwritten with zeros, and the GPUs' memory is cleared before they can be rented again.
Signed software only
The installer checks the agent's signature against Runcrate's release key and refuses anything unsigned.
The honest part
Root on your machine could read a VM's memory, as on any server. Customers are protected from that by your agreement with Runcrate.
Why partner
Your GPUs, your terms.
Help when you need it.
Flexible terms
Spot asks for no commitment: take your GPUs back after the notice you set. On‑demand and reserved bring steadier income.
Support
Write to support any time. We reply within 1 business day.
Large fleets
Bringing a whole fleet?
Datacenters, GPU clouds and HPC sites: tell us about your infrastructure and we plan the onboarding with you.
- A named partner manager and priority engineering support
- A provider adapter for your existing infrastructure
- No exclusivity required
