The Applied Layer

The Applied Layer / Pillars

Economics & Platform Choice

The real cost of enterprise AI and where the vendors actually land, beyond the marketing.

Pillar 4 of 5

The economics of enterprise AI are misrepresented in two directions: vendors understate steady-state cost, and sceptics overstate it. The actual shape is more interesting. Token cost falls; surrounding cost (retrieval, evaluation, integration, governance) does not. This pillar covers the unit economics that hold up under audit, total cost of ownership over realistic time horizons, and vendor-independent analysis of platform decisions. We file the work that makes a procurement decision defensible.

Key findings

  • Token cost is a small fraction of total cost of ownership in most production deployments.
  • Platform lock-in is real but is not where the cost asymmetry lives.
  • The cheapest production system is rarely the cheapest demo.
  • Hosted, open, and self-hosted economics converge faster than the vendor narratives suggest.
  • Cost forecasts that exclude evaluation and observability are not forecasts.

First piece coming soon

We are publishing one anchor research report per pillar through 2026, plus weekly briefings drawn from primary research.

Become a Member to be notified when the first Economics & Platform Choice piece publishes.

How this pillar connects