As AI moves from experimentation to production, the conversation is shifting from model selection to inference strategy. Organizations are grappling with GPU scarcity, rising compute costs, fragmented ownership of production AI, and an operational layer that is proving harder to manage than the models themselves. At the same time, agentic AI is introducing new questions around authority, accountability, and workforce transformation that most organizations have not fully answered yet.

The result is a market defined by urgency and contradiction. Organizations are investing aggressively in AI inference while openly admitting that competitive fear is driving much of that spending. They are repatriating workloads from the cloud while adopting open standards like MCP at remarkable speed, and they believe autonomous agents are inevitable while simultaneously calling the technology overhyped. Navigating these tensions requires a clear view of how the market is actually behaving, not just where vendors say it is headed.

To gain further insights into these trends, Omdia surveyed 400 technical and business stakeholders in North America (US and Canada) involved in the strategy, decision making, selection, deployment, and management of AI initiatives and projects at their organization.