Boundless Expands Distributed GPU Compute Network To Support AI Inference Workloads

Boundless has introduced its enlargement into the bogus intelligence sector, with plans to deploy its distributed GPU community to assist the rising demand for AI inference. The transfer extends the corporate’s present infrastructure, which was initially developed to energy zero-knowledge (ZK) proving workloads for the cryptocurrency trade.

“Four years in the past, we got down to resolve one arduous compute drawback,” mentioned Shiv Shankar, CEO of Boundless. “In the method, we constructed one thing greater: a community for coordinating distributed GPU capability. AI now wants the identical basis at a a lot bigger scale. Open fashions are giving groups extra management over what they construct, however manufacturing inference remains to be constrained by price, capability, and reliability. Boundless is constructed to vary that.”

The firm’s ZK proving community was designed to coordinate distributed GPU sources, schedule compute-intensive duties, and keep environment friendly operations throughout a number of machines. Boundless said that the identical infrastructure is now being tailored to handle growing demand for AI inference as organizations develop the deployment of synthetic intelligence functions.

According to the corporate, AI inference is rising as a major driver of computing demand, with infrastructure prices turning into an more and more vital consideration. Citing Gartner, Boundless famous that 55% of spending on AI-optimized infrastructure-as-a-service is predicted to be allotted to inference in 2026, growing to greater than 65% by 2029.

Boundless Repurposes GPU Network for AI Inference

Boundless said that many AI inference workloads don’t require premium knowledge middle GPUs and might as a substitute function effectively on lower-cost {hardware}, together with consumer-grade graphics processors and GPUs initially acquired for compute-intensive functions similar to cryptocurrency mining and ZK proving. The firm mentioned it goals to determine and optimize this underutilized capability by integrating {hardware}, software program, and inference-serving techniques.

The firm at present operates a distributed community of roughly 4,000 GPUs that was initially developed for ZK proving. As a part of its AI technique, Boundless is optimizing the community to supply managed inference infrastructure, together with {hardware} optimization, workload adaptation, routing, batching, scheduling, monitoring, and operational assist.

Boundless reported that early benchmarking signifies inference prices might be diminished by as a lot as 50% in contrast with hyperscale cloud suppliers, with the best potential financial savings anticipated for asynchronous workloads. The firm additionally plans to combine its native ZKC token into the AI community, with operators anticipated to stake the token to take part and with staking ranges linked to incomes potential.

Boundless said that the enlargement is meant to allow organizations to deploy AI inference workloads with decrease computing prices. The transition to AI infrastructure was introduced immediately, whereas the total product launch is scheduled for later this summer time. The firm added that it’ll proceed working its zero-knowledge proving community alongside its AI inference providers.

The put up Boundless Expands Distributed GPU Compute Network To Support AI Inference Workloads appeared first on Metaverse Post.