Your neighborhood is the data center.
Nostram places small compute nodes in ordinary homes and links them into one network. Heavy jobs are split into pieces, spread across the neighborhood, and put back together. The result is a data center built with the people who live in it.
How a neighborhood does the work of a data center
A traditional data center is one building holding thousands of machines. Nostram is thousands of small machines with no building at all. Each piece is simple. The coordination between them is what we build.
-
A node goes in a home
A compact, quiet unit plugs into power and internet. It runs on its own and never touches the household's devices or data.
-
A point of presence connects the neighborhood
Nearby homes route through a local point of presence, the same way an ISP aggregates a neighborhood. Each point of presence connects to the network operations center, which schedules the work: it breaks each job into pieces, sends every piece to a node that is ready, and tracks all of them until the job is done.
-
Results come back together
Finished pieces are checked and reassembled. If a node goes offline mid-job, its piece is reassigned to another node. Nothing depends on any single machine.
Three ways into the network
The network only exists because people choose to take part in it. Whether you need compute, can host it, or want to help build it, there is a place for you.
Buy compute
For teams with heavy processing jobs that don't want to build or rent a data center to run them: model training, inference, simulations, and large batch work.
Talk to us about a workloadHost a node
For homeowners with a spare outlet and a stable connection. The node takes care of itself, and you are paid for the capacity it contributes. Your home becomes part of the infrastructure, and you become a partner in the network.
Ask about hostingBuild with us
For investors, hardware partners, and utilities who see where distributed compute is going and want a seat at the table early.
Start a conversationThe compute we offer
The network is built for work that splits cleanly into many pieces. Four kinds of work run on it today.
-
Medical simulations
Molecular modeling, drug-interaction studies, and other simulations that would otherwise wait in line for a research cluster.
-
AI training
Training and fine-tuning runs distributed across many nodes, with the scheduler managing every piece of the job.
-
Inference
Running trained models at scale, close to where the requests come from, without routing everything through one central facility.
-
Data processing
Large image and data sets, from drone inspection photos of utility lines to satellite imagery, analyzed thousands of files at a time.
Compute by the people, for the people
Today's compute is concentrated in enormous buildings (data centers) owned by large corporations. Nostram hands control back to the communities that use its compute. The people who host the network are partners in it, and the people who need compute get it without paying for a big corporation's empire.
Think of what rooftop solar did to the power industry. Generation used to happen only in a few giant plants; now it happens on thousands of rooftops, feeding the same grid, and the grid is stronger for it. Decentralized hardware does the same thing for the internet. Compute stops living in a handful of buildings and starts living everywhere, on the same backbone, with no single place that can take it all down. That is the future we are building toward: redundant, connected, and shared.
- One building, one site, one landlord
- Dedicated power feed the size of a small town
- Industrial cooling running around the clock
- One outage takes everything down
- Owned by one company
- Millions of homes on existing fiber, no building to build
- Runs on power connections that already exist
- No cooling plant; a home handles the heat easily
- Lose a node, the work reroutes
- Hosted by partners in the community
More sustainable
A hyperscale data center needs new land, new buildings, industrial cooling, and a grid connection the size of a small city. Nostram uses homes that already exist, on power connections that are already there, with no cooling plant and nothing new built. The heat a node gives off is not significant, and the home's existing climate control handles it easily.
Redundant by design
A single data center is a single point of failure. One fiber cut, one power event, one fire, and everything in it goes dark. Nostram has no center to take out. Work is spread across hundreds of homes on different streets, different circuits, and different providers, and the scheduler reroutes around any node that drops.
Questions people ask first
The short answers. We are happy to give the long ones too.
Does a node use my home's data or devices?
No. The node is a separate machine on your connection. It cannot access anything else in your home.
What happens if my internet drops?
The scheduler notices, reassigns the piece of work your node was doing, and picks you back up when you are online again. Nothing is lost.
Is this the same as cloud computing?
It is the same idea, renting compute, but the machines are spread across homes instead of stacked in one building. That changes the cost, the resilience, who owns it, and where the power comes from.
Is it running today?
Yes. A working network with a live operations center and scheduler is running today, and we are growing it one neighborhood at a time.