bring a workloadearn with your PC
workloads · YOM Containers

you keep the stack. we bring the GPU.

You keep your orchestrator, your protocol and your product. When it needs a GPU near a person, we place your container on a node in range and bill the hours it runs.

bring your container how a container runs

a rental sells you a machine. this sells you the hour, next to your user.

a GPU rental

  • you reserve capacity in advance and estimate your peaks
  • the region you booked, and the user is wherever they are
  • the booked hours are billed whether anything ran

this network

  • capacity on demand, shared across every project on the network
  • the machine is the nearest compatible one to the person
  • billed by the hour while your container runs; at Basic and Protected service levels nothing is reserved ahead
  • your own servers can join the pool and earn when they are idle

how a container runs

  1. ask You tell us what runs and where your users are; each container is started on a compatible node near the people using it.
  2. place We pick a compatible node in range of that person and start your container on it, in its own clean environment. Nothing is injected unless you ask for our streamer.
  3. run Your container runs on that machine. Your endpoint, your protocol, your users.
  4. bill Every hour it runs is billed at the hourly price of its compute class. The operator whose machine runs it is paid from that charge.

your stack runs in its own clean container

what runs
your stack in its own Docker container with GPU access through the NVIDIA Container Toolkit, destroyed when it stops so nothing carries to the next user; pre-release builds run only on telco or cloud capacity YOM controls
where
the nearest compatible node to the person using it
billing
by the hour while your container runs, at the price of its compute class; at Basic and Protected service levels nothing is reserved ahead
access
no shell into the host for you; your container is destroyed when it stops
your machines
your own servers join on the same node software that runs partner racks in production, and earn from other sessions when idle
overflow
when demand outruns operator supply, rented cloud GPUs take the overflow; that capacity can be part of your quote

a container is billed by the hour

Your container is billed for each hour it runs, at the price of its compute class. Tell us what it needs from a GPU and where your users are, and we come back with a quote.

an hour while it runs, by compute class

light
$0.12
standard
$0.24
premium
$0.48
ultra
$0.72
the same hour in a data centre

about $1.23

on a GPU reserved for the peak, against $0.24 for a Standard hour here

orchestration only

quoted with your fleet

when your own machines join the pool

how streaming and containers are priced

who is already asking

  • a streaming platform

    runs its own front end on commercial cloud today and wants a GPU nearer its players

  • a platform with its own orchestrator

    wants this network as one vendor beside the hyperscalers it already uses

  • a low-latency server host

    wants capacity next to its users

the ask

bring the container

Tell us what it runs, what it needs from a GPU, and where your users are. We come back with where the network can place it today and what it costs.

talk to us about capacity