Pavel Durov Launches Cocoon Network for Private AI Computing

The new project from the founder of Telegram, the Cocoon (Confidential Compute Open Network) network, which was announced earlier this fall, has launched. Its main objective is to provide an alternative to major cloud platforms like Amazon Web Services or Microsoft Azure in the field of artificial intelligence.

The network's operating principle is based on pooling computing power. Owners of powerful hardware, particularly graphics processing units (GPUs), can contribute their resources to process AI requests and receive real-time rewards in TON cryptocurrency.

However, unlike familiar "mining" models, this one comes with high technical requirements. Joining the network requires high-performance hardware, for example, a graphics processor at the level of an NVIDIA H100, more commonly used in data centers. This means that average owners of gaming video cards are unlikely to become part of the system, which alleviates potential concerns about its impact on the consumer electronics market.

A key focus of the project is privacy. According to Durov, the network ensures complete confidentiality of data processing, positioning itself as a response to growing user concerns when working with centralized AI services. For developers and small companies, this could become a way to reduce infrastructure costs. Theoretically, the decentralized model should offer more competitive pricing compared to traditional giants.

Some of Telegram's features related to automatic translation are already being processed through Cocoon, with near-term plans including scaling the network and implementing new AI functions in the messenger. All of the project's documentation and source code have been published openly.