Insights
Notes from the substation to the token.
Short, technical pieces on what actually decides whether AI capacity gets built in Australia: transmission, rack density, hardware generations and where the model runs.
The Qinari model: why every AI infrastructure partnership should be its own SPV.
Qinari is QORINAI's SPV platform — one ring-fenced vehicle per project. Why infrastructure partners, equipment lenders and overseas co-investors prefer a single-purpose company to a share of the whole business.
READ →Design · 2026-09-18Why the next AI hall is designed at 250 kW per rack.
Vera Rubin racks draw 190–230 kW and Rubin Ultra is heading for 600 kW. A data hall designed at 150 kW per rack in 2026 is obsolete on the day it energises. What a 250 kW design point changes.
READ →Power · 2026-09-18Australia's data-centre gap is power, not demand.
Victoria carries 9 GW+ of planned data-centre capacity and expects ~0.7 GW online by 2030. The gap is transmission connection, contingency limits and terminal-station capacity — not customers.
READ →Enterprise AI · 2026-09-18Why enterprises are bringing LLMs inside their own network.
Data sovereignty, IP retention, predictable cost and latency are pushing Australian enterprises from public AI APIs to on-premises and private-cloud LLMs. What changes, and what it takes to run.
READ →