What’s Wrong With Cocoon

At the Blockchain Life 2025 conference, Pavel Durov announced that in November a new decentralized network for confidential AI computations will launch. The network, called Cocoon, will allow GPU owners to rent out their computing power in exchange for Toncoin rewards.

Sounds familiar? It’s quite close to what I described a few months ago — a cryptocurrency where mining tasks would actually help AI rather than perform meaningless computations. But there’s one big difference: in Durov’s version, the cryptocurrency itself seems unnecessary. It would make much more sense to pay network participants in stablecoins instead.

When early crypto enthusiasts were asked to perform computations and were rewarded with Bitcoin, they did it to support the functioning of a financial system where Bitcoin itself was the core value. Without miners, that system couldn’t exist. They mined because they wanted to contribute to a system they believed in — and the rewards mattered because they were paid in something that had meaning within that same system.

In Cocoon’s case, participants are offered rewards in a cryptocurrency that the network doesn’t actually need. And these computations aren’t required for that cryptocurrency to function either.

So essentially, participants will be giving up real resources — computing power and the electricity it consumes — in exchange for something that, even within the network itself, has no clear purpose.

Now, if access to AI services required paying in TON, that would be a completely different story. Then the earnings would truly make sense.

But as things stand, I expect most network participants will just sell their rewards — putting downward pressure on the TON price.

By the way, you can always swap TON for BTC or stablecoins at the best rate on rabbit.io.