Compute resale gives an unused commitment another use.
A customer may reserve more GPU time than they ultimately need. Capacity resale is an arrangement that makes eligible unused capacity available to another customer, subject to the original agreement and a new commercial arrangement.
It is different from selling physical GPUs. What changes hands is access to capacity for some or all of the remaining reservation.
The original agreement determines what is possible.
A reservation cannot automatically be transferred just because the holder no longer needs it. The provider’s terms may permit resale, subleasing, assignment, or none of these. Approval may be required.
Check what can be transferred: the number of GPUs, the time window, the whole reservation, or only a defined portion. Storage, credentials, and data may have separate handover requirements.
A useful listing describes both the resource and the window.
A prospective buyer needs enough information to judge whether the capacity will work for their workload. That includes the GPU configuration, cluster topology, region, available dates, software environment, and access conditions.
The remaining duration matters. A powerful cluster with an unsuitable start date or too little time left may not meet the next buyer’s requirements.
Matching is followed by agreement and delivery.
Finding a buyer is only one step. The parties still need to agree on price, scope, payment, access, and responsibility for delivery. Credentials and customer data need an appropriate handover process.
Listing capacity does not guarantee it will sell. A resale also does not automatically release the original customer from its obligations. The actual agreement determines which responsibilities remain and which are transferred.
Questions to resolve before you list.
Gather the reservation details before seeking a buyer. Clear constraints make it easier to evaluate a possible match.
- Does the contract permit the proposed transfer or sublease?
- What quantity and dates are genuinely available?
- Whose approval is required?
- Who remains responsible for provider payments and delivery?
- How will access, storage, credentials, and data be handed over?
- When are proceeds due, and what happens if the resale is not completed?
Further reading
Ornn: resale documentationSF Compute: reselling computeThese sources provide technical or competitor examples. Their product terms and capabilities are not Corgi commitments.