ShareMyGPU is an online marketplace for renting GPU compute by the hour. According to information on its official page, the platform is currently in the construction and security verification stage, rentals and new listings are paused, and no payments are accepted. The following content is compiled from the official page.
Purpose ShareMyGPU aims to provide sharing and rental of GPU compute: on the one hand, allowing users who need compute to rent GPUs by the hour; on the other hand, allowing GPU owners to earn income by sharing idle compute. The official description is "Share compute. Protect the host.", that is, protecting the host while sharing compute.
Main capabilities and plans The platform is building "managed GPU rentals" capabilities, the core of which includes automatic expiration and host isolation. The official statement clearly says that it is building a marketplace so that rentals technically terminate at the end of paid time. The planned features currently listed on the page include: before renting, being able to view GPU model, VRAM, location, and hourly price; users pay by the hour and can choose 1, 4, 8, or 24 hours, with Stripe handling payments; after payment, a managed workspace is configured for the paid duration and automatically shut down at expiration.
Security controls The official statement emphasizes that a paid timer alone cannot protect the owner's machine, so multiple pre-checks are in place that must be passed before rentals can be opened. These checks include: isolating workloads so that tenants run in a restricted environment and cannot access host files, administrator privileges, or the Docker socket, and GPU hosts require dedicated setup; setting CPU, memory, process, disk, and GPU limits to prevent tenants from occupying the entire machine; automatically terminating access by issuing short-term leases and shutting down the workspace at expiration, including when the marketplace cannot contact the host.
How to use (planned) The officially described future rental process is: view the GPU model, VRAM, location, and hourly price; pay for the selected number of hours (1, 4, 8, or 24 hours), with Stripe handling payment; configure a managed workspace for the paid time and automatically shut it down at expiration. Connection information will be displayed on the rental page and sent by email after payment. However, the official statement also explains that because managed access is still being built, rentals are paused, and booking steps will be announced after security checks are passed.
Owner participation and pilot The platform plans a revenue share for owners of 75% (before deducting fees, refunds, and adjustments). However, the official statement clearly says that during the rental pause, no payments will be made to owners. The platform is looking for a small number of owners to test the managed host agent on dedicated machines in order to verify isolation, resource limits, and automatic shutdown. The pilot requires a dedicated Linux machine and a supported NVIDIA GPU. Interested owners can send GPU model and operating system information, but the official notice warns not to include credentials or connection details. Owners will be invited only after the agent can automatically stop access and pass isolation tests.
Use cases and limitations The use cases are renting GPU compute by the hour and owners sharing GPU compute to earn income. Limitations include: rentals and new listings are currently paused, no payments are accepted, and GPUs cannot be rented immediately; the pilot requires a dedicated Linux machine and a supported NVIDIA GPU; the revenue share for owners is planned, and there are no payments during the rental pause; the platform has not yet announced complete system requirements, the list of supported GPU models, pricing details, and usage steps. The official statement says that booking steps and exact setup requirements will be announced after security checks are passed.
