ZK computing power networks are moving into the AI sector—are the old hands in the staking mining model seeing “cut costs in half” as genuinely great, or just a sales pitch?

The zero-knowledge proof and distributed computing project Boundless announced that it will expand its compute network, which is made up of about 4,000 GPUs, into AI inference services. The network previously mainly handled Ethereum mainnet and Base-related zero-knowledge proof tasks; it has now completed hardware tuning, workload adaptation, and routing and scheduling upgrades for AI inference, and the existing ZK proving network will continue running in parallel. In the future, AI compute operators will need to stake ZKC to join the network, and the staking amount will be linked to potential rewards. The company said that the early testing costs for some asynchronous inference tasks are up to about 50% lower than those of major cloud service providers. (The Block)