en.Wedoany.com Reported - On June 1, Aible, a US-based enterprise agentic artificial intelligence company, announced that its enterprise solution for governed, long-running AI agents, AibleClaw, has integrated with NVIDIA Cloud Functions (NVCF), bringing a serverless GPU economic model to scheduled enterprise AI workloads. Aible stated that this integration extends the up to 200x total cost of ownership advantage for end-to-end generative AI demonstrated in its previous benchmarks to long-running enterprise agent tasks.

AibleClaw targets long-running agents that enterprises are accelerating deployment of—AI workloads that need to perform tasks continuously or periodically in the background. Compared to one-shot Q&A or short-duration inference, such tasks often exhibit more pronounced peak-and-valley characteristics, such as scheduled analysis of meeting agendas, generating work briefs, scanning business data, processing customer operations leads, tracking supply chain changes, or performing compliance checks. Aible refers to these tasks as "Claws," characterized by potential durations of several minutes, relatively controllable task trigger times, and lower sensitivity to instantaneous cold-start latency than real-time interactive applications. This makes them more suitable for on-demand scheduling, elastic inference, and cost optimization via NVCF. By combining AibleClaw with NVCF, enterprises no longer need to maintain always-on GPU clusters for all long-running agent tasks. Instead, they can execute planned tasks during periods of lower GPU demand or when resources are more suitable, thereby improving compute utilization and reducing idle costs in private AI deployments.

This solution is built on the NVIDIA DSX OS software stack. NVCF serves as a unified API layer for running and scaling inference, fine-tuning, batch processing, and simulation workloads across Kubernetes clusters, supporting auto-scaling, multi-tenant isolation, and higher GPU utilization. For Aible, the value of NVCF lies in the orchestration layer for enterprise AI agents: enterprises can integrate private servers, edge servers, desktop supercomputers, major cloud platforms, and NVIDIA cloud partner resources into the same scheduling system, prioritizing local execution when conditions are suitable, and distributing tasks to other locations or shared data center resources when necessary.

AibleClaw also incorporates the NVIDIA OpenShell runtime and the NemoClaw blueprint to support governed, long-running agents. The company previously launched AibleClaw with NVIDIA Nemotron 3 Super for governed long-running enterprise agents, and AibleClaw with NVIDIA Nemotron 3 Nano Omni for edge multimodal inference. This integration with NVCF extends AibleClaw's focus from model capabilities to enterprise AI cost structures, private deployment, and resource scheduling methods. For enterprises that have already integrated AI agents into their business processes, the variables truly impacting large-scale deployment are not limited to model response quality, but also include GPU resource utilization, token cost fluctuations, data residency requirements, auditability, and predictable costs for long-running tasks.

Aible emphasizes that its platform can run in environments such as major clouds, private servers, NVIDIA cloud partners, desktop supercomputers, and edge servers, supporting enterprises in privatizing the execution of generative AI and agentic AI workloads on their own servers. This means enterprises can gradually integrate AI agents around their existing IT architecture without needing to build large-scale centralized data centers all at once. As enterprises transition from AI demonstrations to production-grade deployments, the cost model for long-running agents is becoming a critical factor in infrastructure selection. The integration of AibleClaw with NVCF also reflects that enterprise AI competition is shifting from "whether agents can be built" to "whether agents can be run long-term at low cost, with governance and auditability."

This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com