The AI conversation is changing. The question is no longer "Which model should we use?" The question is: "How do we operate intelligence at enterprise scale?"
That shift was everywhere at HPE Discover 2026 in Las Vegas. Amazing energy, incredible conversations and deep discussions with customers, partners and leaders across the ecosystem. One shift stood out above everything else: AI at scale is no longer a model challenge. It's an infrastructure, operations and economics challenge.
Three leaders, one consistent message. Antonio Neri: you don't design a building around a single room, you design a structure that lets the whole system flow and develop over time. Rami Rahim: networking used to be behind the scenes; now it's centre stage, and the future is self-driving. Fidelma Russo: the real shift is people, systems and agents working together at scale to deliver entirely new outcomes. This is the agentic enterprise.
But here's what sharpened this most: the customer conversations off-stage.
The tokenomics moment
Agentic AI doesn't make one request to a model. It makes hundreds, back and forth, iteratively and continuously. Token costs compound fast, and every enterprise customer I spoke with that week had felt exactly this.
From cost per token to owning intelligence
The enterprises moving fastest aren't chasing cheaper tokens. They are asking a different question entirely: what does it cost to own this capability, and what do I unlock when my data never leaves my boundary?
From human in the loop to human in the lead
The agentic enterprise isn't about removing humans. It's about elevating them: not managing every decision, but setting direction, governing outcomes and leading with intent.
The customers I met aren't debating public versus private any more. They are designing for ownership: predictable economics, data they control and AI they can govern. AI is becoming a capital asset, not a consumption expense. The organisations that architect for that now will define the competitive gap.