Global Neural Net Supercomputer

Everything Computes!

Most computing resources (workstations, servers, smart devices, and mobile devices) are idle 80% of time. Computes organizes and harnesses the power of these idle CPUs and GPUs to fulfill infinite compute requests. Best of all, our neural net supercomputing platform requires no infrastructure!

Computing Power to Change the World

Office Computes

Office Computes

Imagine if you could turn your idle office computers, servers, and IoT devices into a supercomputer...

Customer Computes

Customer Computes

Imagine if your customers' web browsers and mobile apps could process computations for you while simply surfing your website or interacting with your app...

Donate to
Greater Good

Donate to
Greater Good

Imagine if these computes could be used for the greater good to help cure cancer or cure Parkinson's disease or fight cyber terrorism etc...

Computes Runs Everywhere!

Any OS

Computes runs on Windows, Mac, Linux, iOS, and Android.

Any Browser

Computes runs on any web browser or any web page via tabs or extensions.


Computes runs on any microcontroller that supports JavaScript.


Computes runs on AWS Lambda or Azure Functions for scaling cores on-demand.

IoT Devices

Computes runs on smart meters and other industrial IoT devices.

BYO Supercomputer

Computes turns your datacenter and workstations into a supercomputer.

Computes Architecture

How Does It Work?
  • Network - Computes nodes run on any computing resource and automatically discover peers on edge networks or globally via the Internet. Computes nodes can connect as a private mesh networks using shared crypto keys or via the public Internet.
  • Storage - Computes nodes can read and write to a global, peer-to-peer (P2P) file system (private mesh network or Internet). In addition to sharing storage, all Computes nodes can also share memory and securely message each other like a single supercomputer.
  • Computes - Computes nodes can both initiate computations and process computations for itself and other P2P nodes. Computations are developed using Javascript or Node.JS. Each node is sandboxed so distributed computations cannot access local disk or memory.
Picture is worth 1,000 words (or cores)

Computes nodes are inter-connected to act like a single supercomputer

Computes nodes share memory, storage, and messaging as you would expect from a globaly distributed supercomputer!

Computes can also run as Lambda-like or Functions-like "serverless" platform using your own distributed machines rather than relying on a cloud infrastructure like AWS or Azure or Google. We leverage idle CPUs and GPUs in your existing infrastructure but you can also choose to leverage AWS Lambda or Azure Functions or Google Functions as your Computes cores.

One major advantage of using Computes over existing "severless" options is that applications are dynamic just like a traditional computer. Functions and algorithms do not need to be pre-deployed to an infrastructure. Kernels distribute dynamic Javascript scripts, gist URLs, or NPM-based applications along with datasets or IPFS links to your cores across the Internet.

End of Cloud Computing?

Peter Levine, a16z partner, explains how computes are moving from the cloud to the edge of the network. Watch video.

Neural Network

Computes nodes natively include machine learning and artificial intelligence tools. The more you train Computes, the smarter it becomes!

Build your own Computes powered RaspberryPi beowulf supercomputer!

Instructions and SD card image.
Checkout our blog for more Computes ideas, tips, & tricks.


Do you have a project in mind to help us test the scale of Computes?


Twitter: @computesio

Facebook: /computes