Skip to content
Compute

GPU compute · Coming soon

GPU instances are on their way

We're bringing GPU virtual machines to Edge for inference, training, rendering and transcoding — with hardware from NVIDIA, AMD and Intel. They aren't available to order yet. Join the waitlist for early access.
Join the waitlistNeed GPUs now?

No commitment. We'll email you when GPU compute launches.

What to expect

Built the way Edge Compute already works

GPU instances will join the same console, CLI and API you use for CPU VMs today.

  • More than one vendor

    NVIDIA, AMD and Intel GPUs, so you can pick hardware for the workload rather than for the provider.

  • Hourly billing

    Pay for the hours you use. No reserved-capacity contract, no long-term commitment.

  • At the edge

    Run inference close to your users for lower latency, on the same network as your CDN and storage.

Use cases

What GPU instances are for

Accelerated workloads that CPU VMs handle slowly, or not at all.

  • LLM inference

    Serve large language models with low latency, close to where requests start.

  • Machine learning

    Train, fine-tune and deploy models at scale.

  • AI at the edge

    Low-latency inference next to users, not a continent away.

  • 3D rendering

    GPU-accelerated rendering for creative and product pipelines.

  • Video transcoding

    Hardware-accelerated encoding for streaming and media libraries.

  • Scientific computing

    CUDA-accelerated simulation and analysis.

Need GPU compute now?

Talk to us before you wait

Tell us about the model, the throughput and the deadline. We may be able to arrange early access, or help you find an interim setup that moves cleanly to Edge later.
Talk to the compute teamAI & ML on Edge

FAQ

GPU questions

Can I deploy a GPU VM today?
No. GPU instances aren't generally available yet. Everything on this page describes what's coming; CPU virtual machines are available now.
When will GPUs launch?
We haven't announced a date. Join the waitlist and you'll hear first, with an offer of early access.
Which GPUs will be available?
We're planning support for NVIDIA, AMD and Intel hardware. Specific models and prices will be published at launch.
How will GPU instances be billed?
Hourly, with no long-term commitment. Rates aren't published yet.

Get first access to GPUs on Edge

Join the waitlist and we'll be in touch at launch. In the meantime, CPU VMs deploy in under a minute.

Join the waitlistDeploy a CPU VM

Joining the waitlist is free and commits you to nothing.