Pavel Durov unveils Cocoon — a decentralized AI compute network

Cocoon is a new platform designed to handle AI computations through a distributed network of GPUs, eliminating the need for large centralized servers.

Here’s how it works:
- GPU owners connect their machines to Cocoon and contribute computing power.
- AI developers send processing tasks to the network.
- The system automatically assigns these tasks to available GPUs.
- Contributors are rewarded in TON for every completed job.

The standout feature of Cocoon is data privacy, all information is encrypted, ensuring no participant can view the contents of a request. In essence, it’s private AI infrastructure secured by the TON blockchain. Telegram will be the first major adopter, planning to route its AI-related operations through Cocoon to promote both privacy and network adoption.

In short:
- GPU providers earn TON tokens.
- Developers gain affordable, secure compute power.
- Users benefit from fast, private AI performance.