Power approvals now set the clock on NextDC's 565 MW forward order book

Divide underlying EBITDA by the megawatts actually billing and you get A$1.42 million per megawatt per year ^[18], which is the figure that makes the rest of the result legible. Apply it to the 565.1 MW of forward order book, which NextDC says are binding contracts rather than options, reservations or pipeline ^[3], and you get roughly A$803 million of incremental annual EBITDA ^[24], about 8.2% on the A$9.75 billion of new capital raised since August 2025 ^[25]. For infrastructure that is a respectable yield. The catch is the verb tense.

Billed capacity grew 58% to 175.0 MW last year ^[11]; compound that and the 740.1 MW already contracted ^[1] takes about 3.2 years to reach the meter ^[27]. The book itself is arithmetic rather than ambition: 740.1 contracted less 175.0 billing leaves exactly the 565.1 unbilled ^[17]. The build cost, meanwhile, is banked now. One year of capex divided by the forward book is A$6.01 million per megawatt ^[19], which at A$1.42 million of EBITDA per megawatt is 4.2 years of EBITDA to repay the gross build before a dollar of interest ^[26], and capex ran 13.7 times underlying EBITDA ^[22]. Craig Scroggie called FY26 "the largest contracting year in NEXTDC's history" ^[5]; the A$397 million overshoot of a top guide already lifted three times, 13.2% over ^[10][21], is what that sentence costs.

Then the grid. The New South Wales consultation on data centre connections and consumption closes on 14 September 2026, with national recommendations expected to take effect from 2027 ^[14], and work by Oxford Economics Australia in association with AEMO has National Electricity Market data centre usage going from 5.1 TWh in FY26 to 33.7 TWh by FY36, a factor of 6.6 ^[15][23], with NSW and Victoria accounting for more than 85% of it ^[15]. McKinsey's reading is that electricity is the largest single source of operating-cost dispersion in AI colocation ^[16]. So the return on those 565.1 MW is decided less by leasing than by a connection queue and a tariff being drafted after the capex was committed.

The counter-case is decent. If connections get scarcer, capacity already energised or already in the queue reprices upward, and NextDC is sitting on A$8.7 billion of pro forma liquidity ^[12], roughly 2.6 years of building at FY26's pace ^[28], with which to buy that scarcity while slower operators consult. Separately, and cutting the other way, strip the A$128.8 million investment property revaluation ^[9] out of the A$82.1 million statutory profit ^[8] and the year is A$46.7 million in the red ^[20] (crudely, since the gain is not tax-adjusted), which makes the profit line a valuer's view of contracted megawatts rather than cash from them.

This is probably wrong, but I would read the equity as a bet on grid access rather than on AI demand, because the demand is the part already contracted. What would break the thesis: NSW grandfathering existing connections in September, and FY27 billed capacity stepping to something like 275 MW, at which point A$1.42 million per megawatt does the work and the queue is someone else's problem. Missing both leaves A$3.4 billion of capex idle while the interest accrues.