SKN | Telegram’s Pavel Durov Launches Cocoon, a Decentralized AI Network Built on TON

Telegram co-founder Pavel Durov has officially launched Cocoon, a privacy-focused decentralized AI compute network built on The Open Network (TON). The platform, designed to challenge centralized AI providers, is now live and already processing user requests and rewarding GPU contributors in Toncoin (TON).

Cocoon operates as a distributed computing system, allowing anyone with spare GPU capacity to rent out their hardware to the network. In return, contributors earn TON tokens while Cocoon handles AI queries and workloads in a privacy-preserving environment. According to Durov, the network’s early operations confirm that GPU owners are already earning rewards from supplying compute power.

A Direct Challenge to Centralized AI Infrastructure

Durov introduced Cocoon as an alternative to major cloud computing services—such as Amazon and Microsoft—which dominate the AI infrastructure market. He argued that these centralized providers inflate costs, limit transparency, and introduce privacy risks by controlling user data and compute supply.

By distributing computation across independent GPU operators, Cocoon aims to eliminate reliance on centralized intermediaries and reduce both the economic and security risks associated with legacy cloud models. The platform’s launch follows Durov’s announcement at Blockchain Life 2025 in Dubai, where he highlighted growing demand for AI systems that do not depend on traditional big-tech providers.

Cocoon leverages the TON blockchain for coordination, settlement and decentralized governance, aligning with Telegram’s broader ecosystem strategy and TON’s increasing integration into messaging-based applications.

Decentralized AI as a Counterbalance to Surveillance and Control

The launch of Cocoon arrives amid mounting concerns—expressed by technologists, cryptographers and Web3 developers—about the unchecked expansion of centralized AI systems. Critics warn that heavily centralized infrastructure grants governments and corporations unprecedented control over user data, behavioral patterns, and digital identity.

David Holtzman, chief strategy officer of decentralized cybersecurity protocol Naoris, recently told Cointelegraph that centralized AI platforms pose structural risks to privacy, cybersecurity, and societal autonomy. Holtzman argued that distributing AI across blockchain-based networks helps restore trust by offering tamper-proof data verification and enabling nodes to communicate without centralized oversight.

Researchers and executives from decentralized AI developer Onicai and the Dfinity Foundation have likewise proposed ethical guidelines for AI that hinge on permissionless blockchain infrastructure, open auditing and transparent computation.

Public sentiment appears to be shifting in favor of these decentralized alternatives. A 2024 survey by Digital Currency Group found that 77% of respondents believe decentralized AI will benefit society more than centralized systems, citing greater transparency and user sovereignty.

An Emerging Sector With Growing Demand

Cocoon enters a rapidly expanding competitive landscape for decentralized AI compute—including networks such as Bittensor, Akash and the Internet Computer—each seeking to offer scalable and censorship-resistant alternatives to traditional cloud providers.

TON’s large global user base, fueled by Telegram’s 900 million–plus monthly active users, may give Cocoon a unique runway for adoption. If compute supply scales, Cocoon could position TON as a major player in the decentralized AI industry while creating a new economic model for GPU owners and developers.

For now, the platform marks one of the most ambitious attempts to fuse AI workloads with blockchain-based autonomy—pushing forward a vision of privacy-preserving, user-controlled digital intelligence that stands apart from the corporate-dominated AI ecosystem.