This is a direct challenge to the AI world dominated by tech giants. Telegram founder Pavel Durov officially unveiled the Cocoon project at the 2025 Blockchain Life conference in Dubai.
"Centralized computing providers like Amazon and Microsoft act as expensive middlemen, driving up costs and reducing privacy." Durov pinpointed the core issue in today’s AI computing market.
01 Project Overview: The Birth and Core Vision of Cocoon
Cocoon stands for Confidential Computing Open Network. Durov described this decentralized AI network, built on the TON blockchain, as a "confidential computing open network." Its goal is to create an ecosystem that connects users with powerful computers to developers who need computing resources, all while protecting user data privacy.
Within this system, GPU owners can share their computing power with the network and, in return, earn TON tokens as rewards. Developers and companies can leverage this computing power to run AI programs, eliminating the need to rely solely on tech giants like Amazon or Google.
Announcing the project, Durov expressed concern about current tech trends: "The world is heading in a strange direction. Over the past 20 years, we’ve been gradually losing our digital freedom."
02 Technical Architecture: A Privacy-First Decentralized Computing Model
Cocoon’s technical architecture stands apart from traditional AI services through its uncompromising focus on privacy. All data processed across the network is encrypted, meaning even GPU owners who provide computing power cannot access the information they’re handling.
This encrypted data processing model directly addresses growing concerns over how centralized AI companies manage user data.
The network essentially operates as a marketplace for computational power. GPU owners connect their hardware to the network and earn TON tokens as their computers process AI tasks. Developers pay for these services using the same TON tokens, creating a direct transaction between both parties.
Telegram will be Cocoon’s first major client, planning to integrate the network into its mini-app and bot ecosystem starting in November 2025.
03 Market Positioning: Differentiating from Traditional Cloud Services
Cocoon positions itself as a challenger to the dominance of infrastructure services provided by tech giants like Amazon Web Services and Microsoft Azure. By creating a market-driven infrastructure, the network promotes transparent pricing, enabling it to offer more competitive rates than traditional cloud providers.
According to a survey conducted by Digital Currency Group in May 2025, 77% of the 2,036 respondents believe that decentralized AI will benefit society more than centralized systems.
Cocoon represents a paradigm shift, uniquely combining blockchain technology, data protection, and artificial intelligence.
Traditional centralized AI systems pose significant risks, such as privacy breaches, data vulnerability, and potential algorithm manipulation. Blockchain technology can help verify the accuracy and integrity of AI-generated data by creating permanent records on a decentralized ledger.
| Comparison Dimension | Traditional Centralized Cloud Services | Cocoon Decentralized Network |
|---|---|---|
| Data Privacy Protection | Providers can access raw data, posing privacy risks | End-to-end encrypted processing, hardware providers cannot view data |
| Pricing Mechanism | Typically set unilaterally by vendors, lacks transparency | Market-driven, supply and demand determine pricing, fully transparent |
| Censorship Resistance | May be subject to vendor policies or government mandates | Decentralized architecture offers stronger resistance to censorship |
| Resource Utilization | Relies on centralized data centers, may leave resources idle | Harnesses global idle GPU resources, improving overall efficiency |
04 Ecosystem: The Combined Power of TON and Telegram
Cocoon’s success is closely tied to the overall development of the TON ecosystem. Max Crown, CEO of the TON Foundation, described Cocoon as "a shift toward an open, user-driven computing economy, giving control of AI infrastructure back to users rather than the companies exploiting it."
Telegram, with over one billion users worldwide, is expected to rapidly bring decentralized AI to hundreds of millions. The company will use Cocoon to enable AI features such as message summaries and draft writing. These functions will run on the decentralized network rather than sending user data to centralized AI providers, ensuring private conversations remain confidential.
Prior to announcing Cocoon, Durov had already begun establishing an AI lab in Kazakhstan, leveraging the country’s supercomputing infrastructure. This partnership underscores Durov’s commitment to building AI technologies free from the control of Western tech giants.
05 Investment Outlook: Institutional Funding and Market Response
The Cocoon project has received strong backing from institutional investors. AlphaTON Capital announced plans for significant investment in GPU infrastructure to support Cocoon. The firm will deploy high-performance GPUs in data centers worldwide to ensure the network has ample computing power from the outset.
AlphaTON Capital CEO Brittany Kaiser called the launch "a pivotal moment for the convergence of blockchain technology, data protection, and artificial intelligence." This investment demonstrates robust industry support even before the project’s official release.
Financially, AlphaTON has already invested $82.5 million to deploy over 1,000 NVIDIA B200 GPUs. These investments are expected to yield an internal rate of return of 59.7% and a return on investment of 615%, leveraging Telegram’s massive user base to accelerate AI adoption.
Currently, the latest price of TON tokens on Gate is around $2.20, roughly unchanged from when the project was announced. While short-term price fluctuations have been limited, trading volume has increased, indicating growing investor interest in Cocoon’s potential.
06 Challenges and Outlook: The Road Ahead for Decentralized AI
Despite its promising outlook, Cocoon still faces multiple challenges. Decentralized AI networks must navigate complex data protection regulations such as GDPR, which may conflict with anonymous data processing. Success also depends on attracting enough GPU providers to meet demand and ensuring the network can compete with centralized cloud services in speed and reliability.
The official launch in November 2025 will test whether Cocoon can deliver on its promises. In addition to convincing developers to migrate from existing cloud services, Cocoon faces competition from other decentralized networks like Akash Network and Render Network.
Cocoon is more than just another decentralized computing network. It embodies a philosophical stance—a powerful challenge to the current AI service model dominated by a handful of tech giants. By building an architecture that protects data privacy and gives users direct access to computing resources, Telegram is exploring an alternative path for AI development.
Future Outlook
Within a short time of Cocoon’s official launch, users have already submitted initial requests, and GPU owners are starting to earn income by renting out hardware. David Holtzman, Chief Strategy Officer at decentralized security protocol Naoris, warns that centralized AI systems give governments and corporations tremendous influence over individuals, potentially harming user privacy and threatening traditional cybersecurity defenses.
As the TON ecosystem continues to expand—with offerings like tokenized US stock trading and institutional-grade financial products—Cocoon is gradually becoming a vital bridge between Web3 and traditional finance. With the AI server market expected to exceed $100 billion by 2029, the rise of privacy-first decentralized computing networks is well underway.
