Flexera report finds 78% of enterprises adopting agentic AI cloud consumption

The shift toward autonomous agents necessitates a transition from reactive IT management to proactive FinOps-centric governance. For the streaming industry, this means infrastructure costs may become increasingly volatile as AI agents autonomously scale compute resources for encoding or personalization tasks. As organizations compete for limited GPU capacity to train large language models, the resulting supply constraints could delay the deployment of next-tier generative video features. The broader ecosystem must now balance the efficiency gains of automation against the financial risks of unmonitored resource spikes. Watch for the adoption of granular monitoring tools designed to track the specific spend of individual AI agents.

Flexera's research on autonomous AI agents aligns with a broader wave of enterprise cloud spending pressure driven by agentic workloads. In its June 2025 Mobility Report, Ericsson found that generative AI traffic already accounts for 0.06% of total mobile network data but carries a 26% uplink ratio compared to the typical 10%, signaling that AI-native applications are reshaping infrastructure demand patterns in ways that mirror what Flexera observes at the cloud layer. The report projected that immersive AI experiences, including augmented reality agents and multi-user streaming, will flood uplinks with video and sensor data, creating the same unpredictable consumption spikes that Flexera's FinOps framework is designed to address. On the vendor and deployment side, Ericsson has moved aggressively to commercialize agentic AI for network operations. In a July 2025 blog post, Ericsson outlined its agentic AI pathway to autonomous network level 5, claiming an 80% reduction in time spent on analysis and decision-making when agents handle routine troubleshooting and optimization. At MWC Barcelona in March 2026, Ericsson networks chief Per Narvinger detailed how AI models integrated into RAN link adaptation algorithms deliver 10% additional spectrum efficiency on hardware already optimized for 30 years, a gain he valued at billions of dollars given current spectrum prices. Ericsson also announced plans to ship 10 AI-ready radio models with embedded neural accelerators by end of 2026, expanding the hardware base on which agentic software can operate. Competitive activity in the AI-RAN and agentic network space is intensifying, providing benchmarks against which Flexera's cloud-side findings can be measured. In August 2026, Samsung and NTT Docomo demonstrated per-user AI prediction that reduced throughput degradation events by 44% in simulation using real network data, a granularity level that goes beyond cell-wide optimization. The same report noted that Nokia and NVIDIA launched the first GPU-based commercial AI-RAN platform in July 2026, targeting double spectrum capacity at the cell level, while South Korea selected SK Telecom to lead its first industrial AI-RAN pilot testing equipment from Nokia, Ericsson, Samsung, and a domestic vendor simultaneously. These developments confirm that agentic AI traffic scaling breaks traditional serverless and on-demand models, reinforcing Flexera's finding that governance models must adapt to autonomous resource decisions at every layer. As these systems scale, are becoming a primary concern for enterprise budget planning. To mitigate these risks, enterprises are increasingly adopting to standardize deployment patterns, while also addressing to prevent machine-speed data breaches. Because of current operational expenses, organizations are also looking to to optimize efficiency at the edge, while to further harden these deployments. With by 2028, these infrastructure challenges are only expected to intensify, even as .