Jetpack
For people with devices
- Contribute idle hours of hardware you already own.
- You set the hours, the limits and the temperature cap.
- See an honest estimate before you sign up, including when the answer is zero.
Pre-launch · distributed computing network
SpareGrid is building a distributed computing network that connects unused processing power with businesses that need it. Contribute with Jetpack, or bring your workload to SpareGrid.
I have computing power
Explore JetpackI need computing power
Explore SpareGrid ComputeIllustrative simulation · not live network data
Start here
Jetpack
SpareGrid Compute
One device is a Jetpack. Millions of devices are SpareGrid.
How it would work
Nothing is running yet. This is the design we intend to build.
A business would submit a batch job.
The scheduler would cut it into small tasks and match them to idle devices.
Devices would run each task in a sandbox, and results would be checked before anyone is paid.
The business would get verified results, and contributors a share of what it paid.
Honest numbers
At these assumptions it is about a dollar a month for a mid-range gaming PC, a few dollars for a flagship GPU, and about nothing for a thin laptop or a phone. It only becomes meaningful if demand and rates grow.
| Device | Net a monthhypothetical, base | Rangeconservative to optimistic |
|---|---|---|
| Gaming desktopaMid-range GPU, 12 GB, 8 h a day | $1.14 | −$0.14 to $2.87 |
| Gaming desktopaFlagship GPU, 24 GB+, 8 h a day | $4.00 | $0.16 to $8.77 |
| Home server or NASbx86 server, 8+ cores, 24 h a day | $1.30 | $0.10 to $2.76 |
| LaptopbIntegrated graphics, 6 h a day | −$0.03 | −$0.05 to $0.04 |
| SmartphonebRecent smartphone, 7 h a day | −$0.01 | −$0.01 to $0.01 |
Workload fit
A network of home devices suits work that can be cut into small, independent, restartable tasks on data that is safe to distribute. It does not suit anything that must stay up or answer in real time.
Batch AI inference
Run a model over a large queue of inputs where no single answer is urgent.
Rendering
Frames or tiles rendered independently and assembled afterwards.
Media transcoding
Convert video or audio between formats and bitrates in segments.
Simulation and parameter sweeps
Thousands of independent runs with different seeds or parameters.
Data preparation
CPU-bound processing of non-sensitive data in chunks.
Real-time inference
Serving users who are waiting for an answer.
Large-model training
Multi-GPU training that needs fast interconnect.
Databases and long-running services
Anything that must stay up and keep state.
Status
Only the website and its forms exist. Nothing in the network itself has been built or tested.
Built and running in production.
A design we intend to build. No working code yet.
A direction with open questions. It may change or prove unworkable.
Vision
If a network like this grows, the people who supply its computing power could share in the value it creates. Whether that ever adds up to more than pocket money depends on demand, rates and decisions nobody has made yet.
Jetpack early access puts you on a waiting list. A compute request starts a conversation about a pilot.