Telegram starts Cocoon to connect people with spare GPUs to AI builders who need compute. The network pays in cryptocurrency and turns idle video cards into earners. The goal is straightforward: link sellers and buyers with Telegram’s software and let distributed hardware handle AI work.
Sellers list their graphics cards, buyers submit AI tasks, and the system manages the exchange. Each job is split into pieces, sent to many cards, recorded for verification, and followed by coin payouts to the owners. Instead of a single company operating a giant server hall, thousands of small machines share the load.
Card owners gain extra cash while model builders get cheaper processing without long cloud contracts. A GPU is the same chip that powers video games, and it also crunches the numbers needed to train and run AI programs, letting everyday hardware participate in AI workflows.
How Cocoon works and who it serves
If the idea scales, work may drift from big data centers to basements and bedrooms. More buyers could chase limited cards, pushing new GPU prices up. New tokens may appear and trade, creating fresh pockets of supply and demand across the network.
Payments arrive as coins, not dollars, so earnings swing with token price. The coins paid today might drop in value tomorrow, cutting the real wage of card owners. Uptime and price must match or beat the big clouds, or buyers will remain with Amazon, Google, or Microsoft.
Crypto income is taxable upon receipt in many places, so each seller must track price, date, and local rules. Side products like token staking, lending, and futures are likely, drawing speculators and adding volatility to pricing.
Cocoon is live. Whether it shifts the AI market, lifts card prices, or stalls under rule books will become clear within weeks, as performance, pricing, and regulation test the model in real conditions.
