Anyone in Europe with a GPU can join. You install our agent with one command, your machine delivers compute for training and benchmarking Loes, and you earn per second your GPU computes for us. Like mining, but the work is AI.
Early access: we onboard the first participants personally.
Rates track the current market price of your GPU class and refresh a few times a day. The monthly figure is the maximum at full load all month; how much work there really is depends on our training demand.
A gaming card in the attic or a workstation idling at night: connect it and it earns whenever we have training work. Only our agent runs on your machine, no ports need to open, and you can stop at any moment.
Do you run a rack or a pool of professional cards? We connect you personally to the shared Router and agree a fee per consumed GPU hour. We handle the workloads, the billing and the customers.
The agent is open and small, your machine stays yours, and we measure the work on our side.
Create a pairing code in your dashboard and paste the command into your terminal. The agent calls out to us; no ports need to open and it works fine behind your home router.
The agent reports your hardware and we automatically set the rate that fits your GPU class. Hardware that would cost you more in power than it earns is rejected.
Your machine receives encrypted work from our training pipeline and benchmarks: generating training data, running models. No customer data, no third-party code, only our agent.
We measure the work time on our side and count per second. Once a month we pay out: as credit or to your bank account, with an invoice we prepare for you.
Not sure your card is worth it? You see your device's rate right after pairing, before anything runs.
You earn per second your GPU actually computes for us, at an hourly rate that tracks the current market price of your GPU class. Idle time does not pay. The rate is shown next to your device in the dashboard, together with an honest estimate of the monthly maximum.
Only our agent, with one AI model from our catalog. Work arrives encrypted and consists of training and benchmark work for our own models: no customer data, no third-party code, no access to your system. Removal is always one command.
Monthly, in euros. You choose: as credit on your HostYourAI balance (instant, no threshold) or to your bank account via SEPA from 25 euros. We prepare the invoice for you (self-billing); you do not have to draw up anything.
Modern NVIDIA cards from the 30 series do well; Apple Silicon works for light jobs. More important than the card is the arithmetic: the rate has to sit well above your power costs. Machines where that cannot work are rejected, so nobody loses money without knowing it.
Yes. Removing the agent is one command on your own machine, and you delete the device in the dashboard. Accrued earnings are simply paid out at the next monthly run.
Customer data does not belong on hardware we do not manage. That is why participant machines only receive synthetic work: training and testing our own Loes models. This is how Loes gets trained with compute from people across Europe.
Then you are a capacity partner: we connect you personally, benchmark your machines and agree a fee per consumed GPU hour. Email your specs to info@hostyourai.com and you will hear back the same week.
Create an account, connect your device and see right away what your card is worth. Questions? One email is enough.