Red Hat & NVIDIA expand AI Factory with agent controls - IT Brief UK

The Red Hat–NVIDIA "AI Factory" expansion is not a compute upgrade; it is a power grab over the orchestration layer. Agent control mechanisms signal the industry's pivot from "AI-as-a-service" to "AI-as-autonomous-workforce." Whoever defines agent lifecycle, permission boundaries, and resource scheduling owns the enterprise AI operating system. Technical ripple: GPU scheduling shifts from tensor parallelism to agent-level resource isolation. OpenShift must grow agent lifecycle APIs—echoing how Kubernetes re-encapsulated Docker in 2014. Application developers move from prompt engineering to designing agent autonomy boundaries. Compliance: The EU AI Act's audit mandates for autonomous systems are converting agent controls from an engineering feature into compliance infrastructure. Locking the orchestration layer into a single alliance stack recreates the supply-chain concentration risk that defined the 2022 chip shortage. Market dynamics: Microsoft AutoGen, AWS Bedrock, and Google Agent Engine are all racing for the same entry point. Red Hat's open-source DNA layered on NVIDIA's hardware lock-in carves a "bring-your-own-cloud + agent governance" lane—using Linux openness to counter Microsoft's closed ecosystem. 12–24 month outlook: Agent control protocols will ignite a standards war, replaying the container-image format battles of 2014–2016. "AI Factory" will morph into "AI Workforce Platform." The moat is not FLOPS; it is whoever writes the agent's constitution.