Cocoon, the first privacy-focused decentralized AI platform based on The Open Network (TON), is now live. The launch is one of the most significant developments in making AI computational power accessible to humans via a blockchain protocol.
The new platform is powered by TON, a layer-1 blockchain heavily integrated into the Telegram ecosystem. It’s also designed to provide inexpensive, quick, and private AI computing — not from huge tech companies, but through a distributed network of graphics cards owned by regular people.
Cocoon had already received its first live user requests, said Telegram co-founder Pavel Durov, confirming the launch. He also mentioned that GPU holders are currently minting Toncoin (TON) by leasing their mining facilities to the network.
GPU providers earn as Cocoon begins processing live AI tasks
Cocoon offers anyone with a graphics card the opportunity to rent out their processing power. AI workloads are computed on a decentralized, global GPU resource cloud whose capacity surpasses that of even the top commercial providers.
The hardware provider earns Toncoin every time a task is processed, making the service more attractive to more participants. The system is designed to scale rapidly as additional GPUs join, meaning that Cocoon was built with the potential to evolve into as large and far-reaching an AI compute marketplace as possible.
The new platform helps solve two of the biggest problems in the current AI ecosystem, Durov said: the high costs and loss of privacy.
Among these are the advantages of centralized providers, such as Amazon and Microsoft. They own the infrastructure, set prices, and have access to user data, which many users and privacy advocates say is a dangerous prospect.
Cocoon flips that model. It maintains data encrypted everywhere, including during processing, so that GPU providers can’t see what they are computing. This ensures that users’ questions are kept secure, along with training data and final outputs, making Cocoon one of the most privacy-oriented AI networks available to the public.
Cocoon strengthens privacy-first AI with blockchain-powered compute
The launch is coming as governments, technologists, and civil society organizations become increasingly concerned about the strength of centralized AI. Critics argue that when a small number of companies manage AI models and compute resources, they have the power to manipulate social behavior, compromise cybersecurity, and violate human rights.
Decentralized AI projects are similar to Cocoon in that they aim to help mitigate those risks. They shard toward tens of thousands of untrusted nodes and log activity on an open blockchain. It does offer tamper-evident records, mutable data sources, and a trustless relay of information between nodes.
Cocoon is also being funded with a substantial amount of money for development. Tech funds backing the TON project have pledged massive GPU farms to boost the network’s throughput. This is how Cocoon can tackle heavier AI workloads, such as large language model queries, video processing, and ML tasks more generally.
Worldwide, there’s a growing desire for AI to be less centralized. Most people these days want to believe that decentralized AI will be more just than the centralized internet’s algorithms. Though most folks already feel that way, the first single from Cocoon’s debut album further normalizes that sentiment.
Cocoon aims to reimagine how AI infrastructure is built. Rather than giving the giants control, it gives back to users (who provide the hardware) and their profits.
If the adoption continues, Cocoon could become one of the initial instances of what it calls “community-owned AI,” where regular people control computing power, privacy, and rewards.
Join a premium crypto trading community free for 30 days - normally $100/mo.
