Cocoon is live! Telegram founder: Rent GPUs to earn TON against AWS giants.

Cocoon Decentralization AI network has officially launched. This is a privacy-protecting decentralized computing platform built on the open network (TON). TON is an independent layer 1 Blockchain associated with the Telegram instant messaging application. Telegram co-founder Pavel Durov stated that the Decentralization AI network has processed its first batch of user requests, and GPU owners have also begun to profit by renting out their hardware.

Durov declares war on Decentralization AI: Amazon and Microsoft are expensive intermediaries

(Source: 2025 Blockchain Life Conference)

According to Telegram co-founder Pavel Durov, the decentralized AI network has processed the first batch of requests from users, and GPU owners have started to profit by renting out their hardware. In a statement on Sunday, he said, “Centralized computing service providers like Amazon and Microsoft act as expensive intermediaries, driving up prices and reducing privacy. Cocoon addresses the economic and confidentiality issues brought by traditional AI computing service providers.”

This criticism of centralized AI service providers is not unfounded. Amazon AWS and Microsoft Azure dominate the global cloud computing market, charging exorbitant fees for AI training and inference services. According to industry data, the cost of running large language models on AWS can reach dozens to hundreds of dollars per hour, which excludes many developers and small businesses from AI innovation.

A more serious issue is privacy. When users upload data to Amazon or Microsoft's AI services, that data is actually exposed to the monitoring of centralized platforms. These companies can access user data to improve their own models, and they may even be required by government agencies to provide user information. For users and businesses that value privacy, this centralized architecture poses unacceptable risks.

Cocoon's solution is to decentralize AI computing across a network composed of independent GPU owners. This architecture has two core advantages. The first is cost advantages, as GPU owners provide services directly to users, eliminating the high premiums of intermediaries. Idle gaming graphics cards or mining rigs can become network nodes, and this diversity and competitiveness of supply will naturally drive down prices. The second is privacy protection, as users' AI requests are processed in a decentralized manner, with no single entity able to access all data comprehensively.

Durov chose to build Cocoon on the TON Blockchain, demonstrating his long-term commitment to this ecosystem. TON was initially developed by the Telegram team, and although it later became an independent project, it maintains a close connection with Telegram's 950 million user base. This user base provides a natural promotion channel for Cocoon, allowing any Telegram user to seamlessly access Cocoon services.

Renting GPUs to Earn TON: The Economic Model of Decentralized AI

Cocoon allows owners of Graphics Processing Units (GPUs) to rent out their computing power to the network, processing user queries and requests in exchange for Toncoin, the native token of the TON Blockchain. This “computing is mining” model creates a new source of income for GPU owners, particularly for gamers and former cryptocurrency miners who have idle graphics cards.

From an economic incentive perspective, Cocoon's model is highly attractive. GPU owners do not need to invest in expensive professional mining machines; ordinary gaming graphics cards like the NVIDIA RTX 4090 or AMD RX 7900 XTX can participate in the network. These hardware components are usually idle when not gaming and can now be used to process AI requests and earn TON tokens. Compared to traditional cryptocurrency mining (such as PoW), the energy efficiency of AI computing is higher and does not require specialized ASIC miners.

In terms of revenue model, GPU owners receive TON rewards based on the number of requests processed and the complexity of calculations. The network uses smart contracts to automatically distribute earnings, ensuring fairness and transparency. After deducting network transaction fees from the TON tokens paid by users, the remaining portion is allocated to the GPU nodes providing computational services. This peer-to-peer payment model eliminates intermediaries' commissions, allowing both supply and demand sides to obtain better value.

Cocoon Economic Model Three Major Advantages

Low Barrier to Entry: Ordinary gaming graphics cards can participate, no need for professional mining machines.

Idle Monetization: Convert idle GPUs into a continuous source of income

Transparent Distribution: Smart contracts automatically settle, with no intermediaries taking a cut.

For the TON ecosystem, the launch of Cocoon creates new token usage scenarios. Users need to pay TON to use AI services, and GPU providers earn TON and may hold or reinvest in the ecosystem for the long term. This practical application demand will provide intrinsic value support for the TON token, rather than relying solely on speculative demand.

From the perspective of market size, the global AI computing market is growing rapidly. According to market research, the computing demand for AI training and inference is growing by more than 50% annually, which provides a huge market space for decentralized AI platforms. If Cocoon can capture even just 1% of this market share, it will bring hundreds of millions of dollars in transaction volume to the TON ecosystem.

Privacy Protection and Decentralization: Countering AI Surveillance Dystopia

Durov announced the launch of Cocoon at the Blockchain Life 2025 conference held in Dubai, United Arab Emirates, in October, in response to user demand for an AI platform that can protect privacy and data from the encroachment of large centralized AI service providers. The blockchain community, privacy advocates, and crypto-punks have long warned about the negative social impacts of centralized AI and have advocated for decentralized AI networks as public goods.

David Holtzman, the Chief Strategy Officer of Naoris' decentralized security protocol, pointed out that centralized AI systems grant governments and corporations significant influence over individuals, which can undermine user privacy, threaten traditional cybersecurity measures, and lead organized actors to engage in social control. He added that by applying blockchain technology to AI, these threats can be mitigated by verifying information sources, ensuring records are immutable, and allowing nodes on a decentralized computing network to communicate in a trustless manner.

These concerns have already manifested in reality. Centralized AI service providers like OpenAI, Google, and Microsoft control vast amounts of user data and training data. There is a lack of transparency regarding how this data is used, its storage security, and whether it is accessed by government agencies. The multiple data leakage incidents from AI companies that surfaced in 2024 have further intensified public distrust in centralized AI.

Cocoon's Decentralization architecture provides a fundamental solution. User AI requests are distributed to multiple GPU nodes for processing, with no single entity able to access all data in its entirety. Additionally, the TON Blockchain's immutability ensures that all transactions and data flows are recorded and traceable while protecting user privacy. This “transparent yet anonymous” architecture is the core advantage of decentralized systems.

In 2024, researchers from the Dfinity Foundation (a non-profit organization guiding the Internet Computer Protocol) and executives from the decentralized AI developer Onicai outlined seven rules to ensure ethical AI. These measures include running AI on a permissionless blockchain network to ensure transparency and data integrity. The design philosophy of Cocoon is highly aligned with these ethical guidelines.

77% Public Support: Decentralization AI is More Beneficial for Society

A survey conducted by Digital Currency Group (DCG) in May showed that among the 2,036 respondents surveyed, 77% believe that decentralized AI is more beneficial to society than centralized systems. This overwhelming public support indicates that concerns about the monopoly of centralized AI have transformed into a demand for decentralized alternatives.

This trend in public opinion provides a strong social foundation for the Cocoon and TON ecosystem. When the vast majority of users recognize the value of decentralized AI, market demand will naturally tilt in that direction. Developers will choose to build applications on decentralized platforms, users will choose to use privacy-protecting AI services, and GPU owners will join in providing a more ethically aligned computing network.

The timing of the launch of Cocoon is just right. With the rapid development and widespread application of AI technology, discussions about AI regulation, privacy protection, and data sovereignty are heating up globally. The EU's AI Act, the US's AI Safety Framework, and various countries' antitrust investigations into large AI companies all indicate that regulators and the public are becoming more aware of the risks associated with centralized AI.

For the TON ecosystem, Cocoon represents a strategic breakthrough. Although TON is connected to Telegram's 950 million user base, it has long lacked a killer application to drive mass adoption. As a decentralized AI platform, Cocoon not only provides practical use cases, but more importantly, it taps into the hottest technology track globally: AI. If Cocoon can attract a large number of developers and users, TON will transform from “Telegram's blockchain” to the “infrastructure for decentralized AI.”