UtilityNet
High-performance computing System
공연
RD 인기지표
RD 성장지표
비교
업데이트 따라가기
팔로우 목록
인물
UtilityNet 소개
UtilityNet builds a decentralised compute network for AI. The idea is to pool scattered GPUs, TPUs and servers into a schedulable resource pool, where developers request capacity per task and providers are settled by measured contribution. The network reaches down to the chip layer, embedding a security engine so a device can attest its own compute and identity instead of competing for blocks through pointless hashing.
The consensus mechanism is PoCI, proof of computation integrity, which combines chip ownership, verifiable computation and VRF randomness so block production does not depend on power draw. Off-chain data is anchored through a Merkle state machine, and containerisation with graphics virtualisation allows TPUs to be rented in slices. A scheduling centre on top matches demand, packaging fragmented capacity into rentable instances that frameworks such as PyTorch can call directly. Rentals bill by the hour, instances are released when a job ends, charges follow actual occupancy, and node operators stake tokens as a service guarantee.