NEXTDC ($NXT) has removed affected Melbourne customer contracts from its FY27 outlook because it cannot reliably estimate their contribution. In its 21 October release, the data-centre operator says the next increase in grid supply to M3 is now expected during FY28. A customer contract can establish demand; the operator still needs enough electricity to deliver it.
Revised FY27 net revenue guidance is A$475–500 million, against A$615–640 million previously. Underlying EBITDA guidance falls to A$245–265 million from A$385–410 million. The exclusion covers affected capacity already billing as well as future capacity; service-level credits could reduce EBITDA below the revised guidance. Capacity remains contracted, subject to customer rights, including termination in some circumstances.
NEXTDC attributes the delay to suspended Jemena network works and reports community opposition, access and consent issues. That is the issuer's account, rather than an independent finding about responsibility. Consultation, reliable supply and local consent are part of delivering infrastructure.
The spending cut needs the same accounting basis
Revised FY27 capital expenditure is A$4.5–5.0 billion, including approximately A$230 million of capitalised interest that the previous guidance excluded. NEXTDC puts the comparable reduction before capitalised interest at about A$980 million. September liquidity of A$8.675 billion includes A$1.275 billion cash; most of the remainder is undrawn financing whose availability has conditions. Release, pages 1–4.
For a supplier, postponing an uncontracted development can mean a later order for equipment or construction work. For a shareholder, postponement can conserve cash while also delaying the earnings that were meant to justify the investment. Those outcomes can occur together. A smaller capital budget does not, by itself, establish better returns.
The financing condition deserves particular attention. A developer may have substantial committed facilities and still need to satisfy drawdown and covenant tests as its billing timetable changes. An analyst who adds every facility to cash and stops there misses the question that matters: how much can the operator actually draw, on what terms, while waiting for customers to start paying?
What follows for copper, energy and the economy
The IEA's April 2026 assessment projected global data-centre electricity consumption rising from 485TWh in 2025 to 950TWh in 2030. It also warned that bottlenecks were limiting the more aggressive near-term scenarios. These are dated projections, not October electricity readings. NEXTDC provides a specific example of the delivery risk behind that broader outlook.
Our interpretation is that the demand case for grid investment survives this delay. The spending sequence becomes more important. An investor in copper should ask which grid projects have approvals, procurement schedules and credible energisation dates before assuming a data-centre pipeline produces a matching near-term increase in metal consumption. Equipment manufacturers can receive orders before a campus starts billing, while a postponed expansion can shift later orders into another financial year.
Energy investors face a similar distinction. Planned power demand creates an opportunity for generators, networks and storage providers, but each business earns under different contracts and regulatory arrangements. Extra electricity demand cannot be translated into a common profit uplift across the sector. The IEA itself cautions against expecting a generalised energy-sector valuation benefit from AI.
The economic consequences are conditional. If connection delays become widespread, construction activity and the expected productivity benefits from new computing capacity could arrive later. Developers could carry financing costs for longer and seek sites where power is available sooner. If investment instead accelerates to remove the constraints, equipment and skilled labour could become more expensive. Neither outcome establishes that households must pay the bill: that depends on tariffs, contracts, regulation and who finances the network works.
The connection may yet be brought forward, and the computing demand may remain. One Melbourne development cannot establish a global collapse in AI investment. Equally, a large global demand forecast cannot establish that this particular development will earn an adequate return.
The Take. We remain interested in the physical investment required to supply AI, especially the work that makes a connection usable. NEXTDC's update argues for closer attention to delivery dates and financing, rather than a change from indiscriminate enthusiasm to indiscriminate pessimism. The next useful evidence is an agreed power timetable, quantified customer-contract consequences and the company's capital review. A demand forecast needs all three before it becomes a dependable cash forecast.