The Australian Energy Market Operator published its 2026 Electricity Statement of Opportunities late on Monday UTC. It forecasts data-centre consumption on the National Electricity Market rising from about 5 TWh in 2025-26 to 34 TWh by 2035-36 — from 3% of grid demand to 13%. Known projects have gone from 97 a year ago to 225; 165 data centres already operate on the NEM.

What the conventional framing gets wrong

The 34 TWh is not what the pipeline asked for. It is what AEMO believes will actually show up. Reported connection capacity in the queue runs to tens of gigawatts; 34 TWh a year is roughly 3.9 GW of average draw. AEMO gets from one to the other by applying two haircuts that almost nobody quotes alongside the headline. The first is attrition: over 40% of projects tracked since 2025 have dropped out or gone backwards in connection status. The second is utilisation: mature facilities average roughly 50% of their network connection capacity, and operators indicate sites "may take five-to-10 years to reach capacity." Convert "225 projects" or the headline gigawatt queue into energy without both haircuts and you overstate by close to an order of magnitude.

The second trap is the word "sevenfold"

A rise from 5 TWh to 34 TWh is a forecast revision, not measured growth. And the direction of the revision is the actual news: last year's ESOO put demand at this level "some time after 2050", with an accelerated-transition case reaching only about 24 TWh by mid-2035. The operator has pulled a 2050 number forward to 2036 in a single annual cycle.

Why a system operator's discount rate is worth having

Announced compute and connected compute are different quantities, and the gap between them is usually invisible — developers announce, utilities do not comment, and nobody publishes the attrition. This is one of the few places a system operator has put a number on its own scepticism. Every Australian project's power assumption now gets marked against it.

What is not asserted here

The exact gigawatt total in the connection queue. Outlets print 67 GW against a prior-year 38 GW; AEMO's own page was not reachable to confirm that pair, so it is left out. Around 40 GW of new generation and storage is committed or anticipated, with a further 33 GW supported by government programmes.