Boundless expands 4,000-GPU network into AI inference while keeping ZK proving
The startup is extending a crypto-built GPU network into AI inference and plans to require operators to stake ZKC, with staking levels tied to revenue potential.
Boundless is broadening its distributed GPU network from zero-knowledge proving into AI inference, repurposing infrastructure first built for crypto compute demand. The company said its network of about 4,000 GPUs has been optimized for inference through hardware tuning, workload adaptation, routing and scheduling, while its zero-knowledge proof network connecting the Ethereum mainnet and Base to Bitcoin will continue to run in parallel. It also plans to give its native token, ZKC, a role in the AI network by requiring operators to stake ZKC to join, with the stake size tied to revenue potential. Boundless said early benchmarks showed that consumer GPUs and existing mining and proof hardware could cut inference costs by as much as 50% compared with major cloud services.
- zero-knowledge proof: A cryptographic proof that lets one party verify a computation or statement without revealing the underlying data.
- AI inference: The process of running a trained AI model on new inputs to generate outputs.
- stake: Lock up tokens as collateral to participate in a network or access certain roles.