Telegram’s Cocoon network goes live, challenges Amazon, Microsoft in AI compute
Major messaging platform Telegram has now officially entered the AI-focused compute services market, aiming to compete with Amazon and Microsoft with its crypto-powered compute network, Cocoon.
Major messaging platform Telegram has now officially entered the AI-focused compute services market, aiming to compete with Amazon and Microsoft with its crypto-powered compute network, Cocoon.
- Telegram launched Cocoon, a decentralized AI compute network on the TON blockchain, positioning itself as an alternative to Amazon and Microsoft.
- GPU providers earn TON tokens for contributing computing power, with Telegram claiming this reduces costs and improves privacy versus centralized providers.
- While promising confidentiality, the network faces criticism for working only with specific Intel processors, potentially hindering widespread adoption.
Key Takeaways by nexos.ai, reviewed by Cybernews staff.
According to Pavel Durov, the co-founder and CEO of Telegram, its "decentralized confidential compute network" went live this past Sunday, and the first AI requests from users are now being processed by Cocoon.
The network is built on the Telegram-supported TON blockchain and connects GPU owners, who provide computing power, with privacy-conscious applications that require running AI models.
"For GPU Providers, it defines how suitable hardware can become part of a confidential, attested compute layer – for developers, it is the backend that executes model requests and settles payments on-chain," according to the project's documents.
Meanwhile, to provide their services to this network, GPU owners earn the toncoin (TON) token. At the time of writing, TON is down almost 4% in a day and 78% in a year.
"Centralized compute providers such as Amazon and Microsoft act as expensive intermediaries that drive up prices and reduce privacy. Cocoon solves both the economic and confidentiality issues associated with legacy AI compute providers," Durov said.
According to him, over the next few weeks, the network will be onboarding more GPU supply and "bringing more developer demand." No information was provided about the current state of the network.
The CEO also promised Telegram users unspecified AI-related features built on "100% confidentiality," adding that the new network will bring control and privacy back to users.
Durov has also emphasized that Telegram is a significant consumer of AI compute itself, as it powers features such as translations, voice-to-text, and summaries. Additionally, Telegram Mini Apps are expected to add even more demand.
Meanwhile, some industry observers question whether this concept can take off at all.
"I see a huge ecosystem problem: The technology is tied to a narrow segment of Intel processors, ignoring the huge markets of AMD, ARM, and other companies' own developments (like Blackwell). It's like trying to create a worldwide standard for a product that only works on one specific phone model," Alex Petrov, the CIO of AI computing solutions company Hyperfusion, said.
However, others counterargue that this is not a problem with decentralized networks, as they usually don't depend on specialized hardware.
Unlock more exclusive Cybernews content on YouTube.