The boutique build, sized · Working backwards from the premium-local demand
The smaller footprint
Twenty-two megawatts at energisation, expandable to twenty-eight.
Built against seven boutique segments. Filled by a demand trajectory
whose shape supports the build before the compounding does the rest.
·The demand, named
Seven segments populate the boutique premium-local tier. Each is small.
None individually justifies a facility. In aggregate they support a
standalone build sized for them and patient enough to fill it.
State agencies, federal civilian mid-sensitivity, universities
1.0 MW
PROTECTED-equivalent workloads not requiring Defence-classified facilities
Primary, year one
7.5 MW
Captured at modest take-rate against incumbents focused on scale.
Sizing assumes 30–60 per cent capture of the boutique-tier addressable within each segment, reflecting the service-tier gap incumbents leave unaddressed at scales below their target unit economics. At anchor pricing of AUD 5.3M per MW per year (the AUD 4.5M hyperscaler-Sydney baseline plus the eighteen per cent central-case premium) year-one revenue lands near AUD 40M against ramp opex.
·The compounding
Year-one demand fills less than half the proposed facility. The case
depends on what arrives behind it. Three forces, each on a different
time-frame.
Second order · 1–3 yrs
AI tooling and compliance
Copilot-class adoption inside each segment runs per-customer compute roughly two times over three years. APRA CPS 230 (in force July 2025), the Privacy Act reforms staged through 2025–26, and cyber-insurance underwriters distinguishing Australian residency from offshore each pull workload onshore. Royal Commission and class-action precedent compounds it: chain-of-custody is now a board-level question.
Third order · 3–5 yrs
Cluster and ecosystem
Cluster compounding is documented in every successful datacentre geography — Northern Virginia, Dublin, Singapore. One mining anchor brings software vendors and JV counterparties. One Big Four anchor brings audit-platform partners. The first two anchors are customers; the third anchor is the geography becoming the answer.
Sustaining
Why the compounding holds
Each effect is structural rather than narrative. AI tooling adoption is happening; compliance regimes are written; cyber-insurance pricing is set; cluster effects are mathematical. The risk is that growth arrives faster than expansion can be staged, not that it fails to arrive.
Figure · The demand, sized over five years
Primary today, second-order from AI tooling and compliance, third-order from cluster effects. Stacked against a capacity line that steps once. The numbers are illustrative; the shape is the argument.
Primary demand is the seven premium-local segments named above. The second-order overlay assumes AI tooling adoption runs roughly 2× per-customer compute over three years and that APRA CPS 230, the Privacy Act reforms, and cyber-insurance residency premia each pull workload onshore. The third-order overlay is cluster compounding. The numbers are illustrative; the shape supports a 22 MW initial build that expands once between years three and four.
·The footprint
The build
A standalone, twenty-eight megawatts at maturity
Sydney metro, REIT-model, specialty operator on triple-net.
Capex anchored at AUD 45 million per megawatt for the
powered shell — deliberately above the AUD 35 to 45M indicative range
for premium colocation at Sydney metro. Total deployed capital lands at
AUD 1.26 billion at twenty-eight megawatts, structured
as a long-tenor REIT-style hold with a specialty operator on triple-net
for the service tier. Unlevered yield 8.7 per cent at
steady state on capex; levered IRR 12.7 per cent at
fifty-five per cent LTV and 6.5 per cent senior coupon.
Macquarie Park, Eastern Creek, and Erskine Park each carry connection
capacity at this scale today without queueing into AEMO at the
gigawatt-class tier. Developer-owner role sits with Gowings; the
specialty operator runs the service tier; the residual asset value
— building, cooling, connection, land — sits with Gowings; the GPU
obsolescence cycle that breaks the merchant case sits with the tenants.
·The position
Build a Sydney-metro premium-tier facility
at twenty to thirty megawatts, sized to the boutique premium-local
demand named above, underwritten by a specialty operator on long-tenor
triple-net.