Cloud computing with a GPU

vGPUInstances

Some workloads a CPU will never solve, no matter how large you make it: training a model, serving inference, rendering. We build you a virtual server in the cloud with a GPU the size your work actually calls for, no more and no less.

What is included

What it is for, and what it is not

A GPU is expensive and only pays off on the right workload. We would rather tell you that up front than sell you a machine you will not use.

  • Training and inference

    Model training and inference in production, with the usual frameworks of the ecosystem.

  • Rendering and video

    3D rendering, transcoding and video processing, where the GPU turns hours of CPU into minutes.

  • Advised sizing

    You tell us what you will run and at what volume, and we propose the instance. No fixed catalogue: every workload asks for something different.

  • CPU and memory to match

    A GPU starved of CPU spends its day waiting for data. We balance the rest of the machine so the accelerator is not left idle.

  • NVMe storage

    Fast disk, so reading the dataset is not the bottleneck. Expandable with our cloud storage.

  • Your environment, not ours

    Full access to the machine: install the drivers, the libraries and the containers your work needs.

  • Scale with the moment

    Adjust the capacity when the project asks for it. There is no sense paying for training-grade power in a month that only serves inference.

  • Guided start

    We hand the machine over ready and help with the first deployment. A GPU being available does not mean your workload is using it.

How we work

From the first message to day-to-day operation

A few clear steps.

You tell us what you needStep 1 of 3
  • Project
  • Workload
  • Budget

Before you decide

Frequently asked questions

What people usually ask before paying for a GPU.

Not here? Write to us

You may not. If your workload is intermittent, small, or solved by somebody else's API, a dedicated GPU is money standing still. Tell us what you plan to run and we will say so before selling you anything.

Yes. You have full access to the machine: install the drivers, the frameworks and the containers your work needs.

Yes. There is no sense paying for training-grade power in a month that only serves inference. We adjust the machine when the project asks for it.

Tell us what you are going to run

Which model or which process, at what data volume and how often. We will propose the instance and the cost. If your workload does not need a GPU, we will tell you that too.

Ask about availability
  • 24/7 support
  • Local engineers
  • Cúcuta, Colombia
  • Training and inference
  • Rendering and video
  • Advised sizing
  • CPU and memory to match
  • NVMe storage
  • Your environment, not ours
  • Scale with the moment
  • Guided start