As capital floods into data centres on the promise of AI, the gap between investor expectations and delivery reality is widening fast. Shifting hyperscaler strategies are reducing demand visibility and driving mispriced return assumptions. At the same time, the economics of AI are evolving, with real compute costs increasingly colliding with unclear or uneven monetisation models. As boards and capital providers sharpen their focus on unit economics, growth narratives alone will no longer justify infrastructure build-out.

Private equity backed platforms are under pressure as development risk, power delivery delays and front-loaded capex collide with optimistic models. Rapid compute evolution is raising the risk of mismatched products and stranded assets. Power, not buildings, is emerging as the true source of defensible value, reshaping capital structures, portfolios and returns. This talk cuts through growth narratives to examine where risk is really sitting today, which parts of the stack retain pricing power as AI margins compress elsewhere, and how you can best position your capital as the market moves from expansion to economic discipline.