Distributed compute startup Boundless expands 4,000-GPU network from ZK to AI

Boundless, the zero-knowledge startup initially focused on settling zero-knowledge proofs on Bitcoin, is expanding its distributed GPU network to support artificial intelligence inference workloads, shifting capacity originally developed for crypto ZK proving to meet growing demand in the AI sector.

"Four years ago, we set out to solve one hard compute problem," Boundless CEO Shiv Shankar said in a statement on Tuesday. "In the process, we built something bigger: a network for coordinating distributed GPU capacity. AI now needs the same foundation at a much larger scale.”

According to the announcement, Boundless operates a network of approximately 4,000 GPUs that has now been optimized for AI inference, including hardware tuning, workload adaptation and managed operations such as routing and scheduling. This network initially served as a coordination and verification layer running a connection between the Ethereum mainnet and Base Layer 2 to Bitcoin. 

“ZK proving and AI inference are both resource-intensive and require the same operational layer, one that matches jobs to available capacity, keeps utilization high, and routes work efficiently,” the announcement read.

Boundless is far from the only crypto company shifting attention to the fast-growing AI sector. Bitcoin miners, which have built large-scale data center infrastructure and secured access to industrial power supplies, have increasingly sought to repurpose their facilities for high-performance computing (HPC) and AI workloads.

So far, it's been less common for protocol or Layer 2 teams to shift their focus to AI, though decentralized compute networks have increasingly begun directing capacity toward AI workloads and the agentic economy. Eigen Labs, initially focused on restaking, has shifted its infrastructure toward verifiable off-chain compute.

Of note, Boundless said it will continue running its zero-knowledge proving network in parallel, and plans to “give its native token, ZKC, a role in its AI network,” by requiring AI operators to stake ZKC to join the network, according to the announcement. The size of their stake is “tied to their earning potential.”

The company said early benchmarks show inference costs up to 50% lower than those from major cloud providers, particularly for asynchronous workloads, by using lower-cost capacity including consumer-grade GPUs and hardware previously purchased for crypto mining and proving. 

“Open models are giving teams more control over what they build, but production inference is still constrained by cost, capacity, and reliability,” Shankar said. “Boundless is built to change that."