Your Spark sleeps 20 hours a day.
Send it somewhere it works.
Ship us your DGX Spark and we run it. We erase it, rebuild it and rack it in our EU-Central facility next to our own nodes, and you keep 70% of every rented hour. Nothing below is live yet, and this page says plainly which parts are built and which are planned.
The math, upfront
Illustrative numbers at today's $2.90/hour on-demand rate. Your share is 70%; our 30% covers racking, power, bandwidth, billing, renter vetting and support, so you never talk to a renter and never see an electricity bill.
Quiet demand
~$370
per month at ~25% rented utilization — a machine that would otherwise have earned nothing in a cupboard.
Steady demand
~$740
per month at ~50% rented utilization. Your node is listed continuously, so this is fleet demand, not your calendar.
Fully rented
~$1,480
per month at 100% rented, which is a theoretical ceiling rather than anything we expect a node to hit.
These are arithmetic at the stated utilization levels, not forecasts. Utilization depends on demand we cannot promise you, and since the programme hasn't launched we have no owner earnings history to show you. You earn only for hours actually rented.
How it works
Apply, then ship it
Join the waitlist. At launch we agree the term in writing, you back up anything you want to keep, and you send us the machine. We do not run software on hardware sitting in your home or office — the Spark comes to us.
We erase it and rebuild it
On arrival the drive is wiped and DGX OS is reinstalled from scratch, so the node enters the fleet in a state we control end to end. Everything on the machine is destroyed at intake — that is the point, and it is why you back up first. It then runs as an ordinary EU-Central node, not a separate community pool: same facility, same wipe between rentals, same product a compliance customer is buying.
Wiped between every rental
Your node gets the same isolation as the rest of the fleet: the workspace is destroyed and the machine is handed back to the pool clean after every rental, and a readiness check refuses to re-let a node that has not come back clean. We publish the mechanism rather than asking you to take it on trust.
You get paid, and you can take it back
70% of every rented hour. Payout schedule, method, and the earnings view in the console are still being built; we'll agree those in writing with the first owners rather than announce them here — along with the notice period, who pays shipping each way, and the condition the machine comes back in. Withdraw it and we wipe it and send it home.
What we need from you
The machine, and its power brick
Shipped to us with the original power supply. Connectivity, power and cooling stop being your problem the moment it arrives.
A term you're happy with
Racking a machine for a week is not worth the shipping in either direction, so the agreement sets a minimum term and a notice period. Both are agreed before you ship, not after.
Stock hardware
An unmodified DGX Spark in working order. Whatever OS it arrives on does not matter — we reinstall it. Modified or damaged units we send back rather than rack.
Host FAQ
How much will I actually earn?
You keep 70% of every rented hour, about $2.03/hour at the current $2.90 on-demand rate. What that adds up to depends on fleet demand, not on your availability: your node is listed continuously and placement draws on our own machines and owner-supplied ones alike. Before you commit we will show you what fleet utilization actually looks like at that point, including if the answer is discouraging.
Can I still use my own machine?
No — it is in our racks, not your office, so for the length of the term it is not available to you. That is the trade: you give up access and stop paying for power, and in exchange the machine earns whenever anyone rents it instead of only when you remember to list it. If you want a Spark you can still touch, keep it and rent capacity from us instead.
What happens to the data already on the machine?
It is destroyed at intake, deliberately. We erase the drive and reinstall DGX OS before the node ever accepts a rental, because a node we did not build is a node we cannot vouch for. Back up anything you care about before you ship — we cannot recover it afterwards and will not try.
Who pays for power and bandwidth?
We do. The Spark peaks at 240 W, which is a few cents an hour, and once the machine is in our facility that cost sits on our side of the 70/30 split along with racking, cooling and connectivity. Your 70% is 70% of the rate, with nothing deducted for running the box.
Who is responsible if my hardware fails?
The machine stays yours, including the risk that it simply dies of age; custody does not extend the manufacturer's warranty. While it is with us we monitor thermals and pull a node from the pool rather than cook it, and we intend to cover damage caused in our facility or by a renter up to the replacement value of the unit. Transit risk in both directions, and the exact cover, belong in the host agreement — which we will write before anyone ships us anything, not after.
Is this available now?
Not yet. The programme is in development, and onboarding will open in stages, EU first. Join the waitlist and we'll tell you how deep the queue is and what current fleet utilization looks like.
Idle silicon is a subscription you pay.
Send us the Spark, and let the machine cover its own invoice.
Join the host waitlist