The Boundless team, initially focused on zero-knowledge proof (ZKP) computations for Bitcoin, is radically changing its development vector. As I found out, the company is repurposing its distributed network of approximately 4,000 graphics processing units (GPUs) for AI model inference tasks. This is a strategic move that reflects market reality: demand for AI computing power is growing exponentially, while the ZKP niche is not yet generating comparable monetization.
According to Boundless CEO Shiva Shankar, the network, which previously served as a coordinator and verifier between Ethereum, Base, and Bitcoin, is now optimized for AI workloads. The key advantage is cost. Early benchmarks show that inference on Boundless is almost 50% cheaper than with major cloud providers, especially for asynchronous tasks. Savings are achieved through the use of consumer-grade graphics cards and equipment previously purchased for cryptocurrency mining and ZKP proof generation.
However, the company is not completely abandoning ZKP — details of future work in this direction are not yet disclosed. But the key signal is the ZKC token. Boundless plans to implement a staking mechanism for the native token for AI operators wishing to join the network. The stake size will be "tied to potential revenue," creating a direct economic link between computing power and the token.
This decision is a clear indicator of a general trend: crypto infrastructure created for mining and proofs is naturally migrating to the AI sector. Recall at least TeraWulf, which raised $3.5 billion in July to build a data center for Anthropic. Boundless demonstrates a more flexible approach — not building from scratch, but repurposing existing capacity.
Expert commentary: This case is an excellent example of "evolution through adaptation." Networks that cannot compete in performance with cloud giants win through price and access to "gray" equipment. But the main question is whether Boundless can provide reliability and speed comparable to AWS or Google Cloud for critical AI tasks. If so, we will see a new model of decentralized computing where ZKP startups become AI providers.