A working investor pack on AI data centres in Australia, organised around a single recommended position. The synthesis names the build. Six further pages size it, cost it, test it, and place it inside the global picture.
Build a Sydney-metro boutique standalone at twenty-two to twenty-eight megawatts. The frontier refusal stands; this recommendation is what was inside it.
Seven premium-local segments populate the boutique tier. A demand stack across five years against a capacity line that steps once. The shape supports the build.
The line-by-line accounting beneath the underwriting. AUD 1.26B of one-time capex, AUD 33M of annual opex, AUD 180M of thirty-year refresh, with the Australian premia surfaced at the lines they bite.
Sydney boutique against hyperscaler-Sydney: flex four numbers and pick a customer mix. The deployed capital, year-three NOI, unlevered yield, and levered IRR update on the right with a position read.
A demand cascade and a four-thesis scorecard, responsive to three diagnostic inputs. Locates where money originates and which thesis the boutique sits inside.
An AI data centre, from chip to hall, at four scales of magnification. The chip does the work; everything else exists to keep it cool, fed, and connected.
What the customer is actually buying when they buy compute. The middle of the data flow is the only thing billed; what goes in and what comes back is the customer's own.
Start with the synthesis. It names the recommended position — a Sydney-metro boutique standalone at twenty-two to twenty-eight megawatts — and shows live what the build pays at conservative inputs. From there, the footprint sizes the seven premium-local segments and the five-year demand trajectory; the cost stack decomposes every line on which the underwriting rests; the underwriting calculator stress-tests the position at the corner cases.
The thesis dashboard places the capital inside the global four-thesis frame. The two visual notes — Compute and Demand — are scale-agnostic and orient the reader to the asset itself. They are best read after the synthesis if the supply-side or demand-side mechanics need grounding, or first if the asset itself is unfamiliar.