Agentic AI infrastructure shifts enterprise focus from model choice to platform control
SiliconANGLE Victoria Gayton ● Covered by 4 sources
Enterprises are moving AI from demos to production and asking who controls the platform, data, and costs. That’s pushing them away from public-cloud-only setups and toward hybrid control.
Based on reporting by SiliconANGLE, Victoria Gayton — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Agentic AI is forcing a shift in enterprise thinking. The old debate was about which model to pick. The new one is about where the whole system runs, who can touch the data, and how much the bill balloons once “assistant” turns into something that never really sleeps.
Joe Fernandes, who leads Red Hat’s Artificial Intelligence Business Unit, said the jump from experimentation to production changes the math fast. Token costs rise sharply, he said, especially when companies move from simple chatbots to always-running enterprise agents. And cost is only part of it. Data exposure, compliance rules, and sovereignty requirements all start to bite once these systems get real work.
That is where platform teams come back to the center of the room. Fernandes argued that agents need the same basics every enterprise application needs — reliability, uptime, scale, security — but with a new problem layered on top: autonomy. If an agent can act on its own, someone has to decide what it can access inside the network or on a file system.
Red Hat is leaning into that problem with agent sandboxes, including Nvidia’s OpenShell, an open-source sandbox runtime for AI agents that Red Hat contributes to and maintains. Fernandes said the goal is to work with other vendors and users to push the space toward a standard, not to leave every company inventing its own guardrails from scratch.
The bigger picture is familiar if you’ve watched enterprise tech long enough. Public cloud is still part of the story, but not the whole story. Fernandes said agentic AI infrastructure will need to span public cloud, private environments, sovereign clouds and the edge. In other words: the AI era has arrived, and it still refuses to make life simple for platform teams.
My take — AI-written commentary, not fact-checked reporting
This is the part of AI that actually matters: control beats novelty once real systems go live. The hype crowd keeps selling smarter agents, but the grown-up question is who gets to decide where they run and what they can reach. Open tools and hybrid setups are not a purity test; they are the price of admission when autonomy meets enterprise reality.
Read more about this at: SiliconANGLE