What is Telegram’s Cocoon network and how does it work?

Pavel Dorf, CEO of Telegram, recently announced the decentralized network Cocoon Officially launched on the TON blockchain platform. The project aims to end the dominance of centralized giants like Amazon and Microsoft over AI infrastructure. Dorf believes that these companies have become expensive intermediaries that both drive up prices and violate users’ privacy. But what exactly is Cocoon and how does it work? In the following, you will get to know more about this network and its technical architecture.

What is Cocoon telegram network?

Simply put, Cocoon, an acronym Confidential Compute Open Network or “Confidential computing open network“, is a decentralized platform for running artificial intelligence models built on the TON blockchain. This network acts like a free market where two groups are connected:

1. Hardware Owners: Those who have powerful servers and want to rent their processing power.
2. Developers: Those who need processing power to run their AI applications but don’t want to pay the exorbitant costs of traditional cloud services.

The main difference between Cocoon and the competition is the use of “trusted execution environments” (TEE). This technology ensures that the user’s data (for example, the question he asks the chatbot) is not visible even to the owner of the server and is processed completely encrypted.

Technical architecture of Cocoon Telegram network

To understand the security and efficiency of this network, we need to know its components in more detail. Cocoon consists of three main components (client, proxy, and worker), all of which run in TEE protected environments:

1. the client

This part is a software library that runs on the developer side (for example, in Telegram’s backend infrastructure). Its task is to send artificial intelligence inference requests to the network and pay fees. Before sending any sensitive data, the client checks the proxy’s security certificate to make sure it is connected to a secure node.

2. proxy

Proxies play the role of an intelligent interface and themselves run in a secure environment (TEE). The proxy takes the request from the client and sends it to the most appropriate Worker based on the type of artificial intelligence model and network traffic. Currently, proxies are managed by the Cocoon team, but in the future, anyone can take on the role of proxy.

3. Worker

Workers are the main part of this network; In fact, these are the servers equipped with graphics cards (GPU) that do the actual processing. Workers run AI models inside a protected virtual machine (TEE). Due to the use of Intel TDX technology, even the physical owner of the server (Host) cannot access the model’s requests or responses.

In general, in the Cocoon network, when you ask an artificial intelligence (for example, “draw a picture of a cat” or “summarize this text”), your application (the client) packages the question or request and sends it to a proxy to deliver it to the appropriate server.

Because there is no central bank or company to guarantee that you will pay later, you have to send the fee to do it right away (in TON cryptocurrency). The proxy keeps this money and only gives it to the server owner (Worker) when the work is done.

Security in Cocoon also goes beyond simple encryption; All communication between client, proxy and worker is done through RA-TLS protocol. In this method, before establishing a connection, each party must provide a digital certificate to the other party to prove that it is running in a secure hardware environment (TEE) and that its software has not been tampered with.

Applications and advantages of the Cocoon network

If you own a data center or powerful servers, Cocoon is a golden opportunity to earn passive income. By becoming a Worker, you will receive TON cryptocurrency for every request you process. Of course, you need professional equipment to join this network:

• Linux operating system (Kernel 6.16 or higher for full TDX support).
• Processor with Intel TDX capability to create a reliable processing environment (5th generation Intel Xeon Scalable or Intel Xeon 6 processor is required to enable TDX).
• Nvidia graphics card with confidential processing capability such as H100 series and above.

However, why should AI application developers use Cocoon instead of Amazon and Microsoft? This platform has three advantages:

1. If you’re building an app that deals with sensitive data (medical, financial, or personal), Cocoon is the only place you can guarantee users that their data won’t be seen by anyone (not even you or the cloud infrastructure).
2. By removing intermediaries such as AWS and Azure and creating a competitive market between GPU suppliers, processing costs are greatly reduced.
3. Payments are made through the fast TON blockchain, which allows for the processing of millions of micro-requests.