Boundless Network Inc., a distributed computing company that built its graphics processing unit network to generate zero-knowledge proofs for the cryptocurrency industry, is expanding its network toward artificial intelligence inference, the process of providing responses to AI queries.

The company said today it will open its network of about 4,000 GPUs to AI workloads and run it as managed inference infrastructure. A full product launch is planned for later this summer.

Boundless assembled the network over the past four years to meet transaction-proving demand in crypto, work that required pooling GPU supply, scheduling compute-heavy jobs and getting scattered machines to act as a single reliable system. It is now tuning that same coordination layer for AI.

The pitch rests on where AI spending is heading. Gartner Inc. expects inference to eat up 55% of AI-optimized infrastructure-as-a-service spending this year and more than 65% by 2029. The more models a company puts into production, the more that inference bill decides how much it can actually run.

Much of the market is priced as though every workload needs scarce high-end data center chips, Boundless argues, when plenty of inference runs fine on cheaper hardware. That includes consumer-grade cards and GPUs originally bought for compute-heavy markets such as crypto mining and proving. The company says it finds that overlooked capacity and tunes the hardware and serving software around it.

“Four years ago, we set out to solve one hard compute problem,” said Chief Executive Shiv Shankar. “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. Open models are giving teams more control over what they build, but production inference is still constrained by cost, capacity and reliability.”

Shankar spent more than 15 years in engineering leadership roles at companies including Coinbase Global Inc., Ava Labs Inc. and Amazon.com Inc. before Boundless.

In early benchmarking, the company said inference ran as much as 50% cheaper than comparable hyperscaler cloud options. It expects the savings to matter most for asynchronous workloads, the kind that do not need an instant response.

Boundless also plans to give its native token, ZKC, a role in the AI network, mirroring how the token works in its proving business. Operators would stake ZKC to join, with the size of a stake tied to how much they can earn.

Boundless will keep running its zero-knowledge proving network alongside the AI expansion. The company is taking signups for the new service through a waitlist on its website.

Image: Boundless