Telegram taps Cocoon network for decentralized AI compute and privacy on TON

Telegram is rolling out a new way to power AI features, with the Cocoon network embedded in TON ecosystem to preserve user privacy while rewarding GPU owners.

What is the Cocoon network launching on TON?

On Nov. 30, 2025, Telegram founder Pavel Durov announced that Cocoon, a decentralized confidential compute network built on the TON blockchain, is now live and operational.

The project, formally called the Confidential Compute Open Network, is designed to process AI requests while fully protecting user privacy and data confidentiality.

According to Durov, Telegram will be the first major customer of the network and also its largest promoter. Moreover, this tight integration positions TON as a core infrastructure layer for confidential AI workloads inside Telegram.

How does Cocoon enable crypto earnings for GPU owners?

Cocoon allows anyone with a GPU to earn crypto by running AI models for applications that require strict privacy guarantees. Durov noted that some GPU owners have already started contributing their computing power to AI tasks and are earning TON tokens in return. However, the system is built so that these contributors never see the underlying user data they are processing.

The model creates a decentralized AI compute marketplace where GPU providers are matched with developers and applications needing confidential processing. That said, the network aims to scale as more GPU operators, including small miners and data centers, connect to its infrastructure for decentralized ai compute rewards.

How does Cocoon protect Telegram AI privacy?

Cocoon processes AI requests from Telegram users with full confidentiality, positioning itself as an alternative to centralized AI providers that cannot guarantee data privacy. The network routes AI workloads to independent GPU providers while enforcing strict isolation between the model execution layer and user inputs.

To achieve this, Cocoon uses Trusted Execution Environments (TEEs), such as Intel TDX, to deliver private, verifiable and attested model execution. Moreover, this architecture underpins confidential ai execution by ensuring that even node operators cannot inspect or leak the data they process.

What role does Telegram play in the new confidential compute network?

Telegram serves as Cocoon’s first major customer, integrating the network’s confidential AI capabilities directly into its messaging platform. This means AI-powered features inside Telegram can tap Cocoon for processing while maintaining end-user privacy and data secrecy.

Durov previously said Telegram would not only rely on the system but also heavily promote the confidential compute network. Furthermore, the app will act as the initial demand engine as Cocoon onboards more GPU providers and application developers building across the TON ecosystem.

How does TON integration support Telegram’s in-app economy?

The cocoon network fits into a broader strategy in which TON powers Telegram’s in-app economy. The blockchain already supports features such as creator payouts and ad payments, and now also underpins the incentivization layer for AI compute.

As Cocoon expands, this ton blockchain integration could tie confidential AI services, GPU rewards, and user-facing Telegram features into a single economic loop. However, the long-term impact will depend on how quickly developers and GPU operators adopt the new infrastructure.

In summary, Cocoon’s launch on TON creates a decentralized confidential compute layer for AI, giving Telegram a private alternative to centralized providers while opening a new crypto revenue stream for GPU owners.