SpareGPU
A marketplace for idle GPU capacity
SpareGPU connects machines with an idle graphics card to the workloads that need one. A worker application offers the GPU to the network; jobs — language-model inference, vision, image generation — are scheduled onto it and settled per job.
Clients call an OpenAI-compatible API and do not have to care where the job ran. Owners install the worker, decide when their card is available and earn on hardware that would otherwise sit idle.
What it does
OpenAI-compatible API
Point an existing integration at SpareGPU and keep the code you already wrote.
Worker app for GPU owners
Windows, macOS and Linux, with one switch for making the card available and taking it back.
Scheduling across the fleet
Each job goes to a worker that can actually run it, with a cloud fallback when no suitable card is free.
Accounting per job
Transparent usage and payout records for both sides of the marketplace.
Who it is for
For developers who need inference capacity without renting a whole cloud GPU, and for owners of hardware that idles most of the day.
Visit the website