Sell your idle GPUs and models
A marketplace to buy, sell, or lease pre-trained weights and GPU capacity from vetted ML teams. Recover value from assets you'd otherwise write off.
For ML teams who've retired models or hit GPU bottlenecks in the past year.
You're on the list. We'll be in touch soon.
Good assets, written off
You built a great model but have no product to sell it, so it feels like waste. Your GPUs sit idle between projects. Training costs have tripled, yet you keep retraining models that already exist somewhere. Right now there's no clean, trustworthy way to recover that value or reuse someone else's work.
Who it's for.
Teams with idle assets
ML Ops and infrastructure leads with underused GPU clusters or deprecated fine-tuned models who want to recover cost instead of eating depreciation.
Teams that need to ship faster
Smaller ML teams tired of training from scratch who'd rather buy or lease proven weights and capacity to move quicker.
Teams sitting on datasets
Groups with valuable datasets they'd share if there were a legal framework and a way to actually get paid for it.
Turn depreciating AI infrastructure into value
Recover sunk cost
Liquidate deprecated fine-tuned models and idle clusters instead of watching them depreciate.
Skip training from scratch
Buy or lease weights that already exist so your team ships faster with less redundant compute.
Vetted and traceable
Provenance and licensing tracked, so you can buy second-hand weights your compliance team will accept.
How it works
List your assets
Add idle GPU capacity, retired models, or datasets with licensing and provenance details.
Match with buyers
We connect your listings with vetted ML teams looking to buy or lease rather than build from scratch.
Get paid
Set your terms and recover value through a sale, lease, or revenue split.
Help shape the first version
We're picking a handful of pilot teams to build this around real inventory and real cost pressure. If you've trained or retired models this year, or fought GPU bottlenecks, we want to talk.
You're on the list. We'll be in touch soon.
Good to know.
What can I list? +
Deprecated or retired fine-tuned models, unused training datasets, and idle GPU capacity between projects.
How does pricing work? +
Pilot customers help us decide between a monthly fee or a revenue split. You keep control over what you charge for your assets.
Is this live yet? +
No. We're running a private beta and looking for 3-5 pilot teams to shape the roadmap before we open more broadly.
What about licensing and provenance? +
We're building the legal framework and provenance tracking as a core part of the platform, so buying and selling stays trustworthy.