Introduction

Thunder Compute is a cloud platform optimized for prototyping with GPUs. On a Thunder Compute instance, you can do each of the following instantly with one command:

• Add or remove GPUs from an instance
• Switch which GPU type is attached to an instance
• Scale to a cluster of multiple GPUs

Additionally, while developing on Thunder Compute:

• Develop exactly the same as with a local GPU
• Use the same instance from prototyping through production
• Only pay when a process actively uses the GPU

Thunder Compute instances with network-attached GPUs behave exactly like linux machines with physically attached GPUs. One instance can be used throughout the lifecycle of a project without config and idle GPU time.

Here's a video of this in action: