NextDC Crashed 44% in a Day Because the Power Grid Couldn't Keep Up

The AI data center boom just had its worst single day. NextDC, Australia's largest data center operator, saw its shares plunge roughly 44% on Wednesday, October 21, 2026, wiping about A$3.1 billion off its market value and making it the biggest loser on the ASX200. The cause was not demand, which remains strong, but electricity: delays to power grid upgrades at its M3 facility in Melbourne forced a revenue forecast cut and a freeze on all new projects. The NextDC share price crash is the clearest proof yet that the binding constraint on AI infrastructure is not chips or money. It is the power grid.

The NextDC share price crash, explained

NextDC (ASX: NXT) fell to its lowest level since January. The trigger was an announcement that about A$140 million (roughly $98 million) of contracted customer revenue for this financial year would be deferred because the electricity network upgrades feeding its M3 data center in West Footscray, Melbourne, are running late. When nearly a tenth of your expected revenue slips because the power company cannot deliver, the market reprices everything.

What went wrong at M3

The utility, Jemena Electricity Networks, suspended parts of its 66,000-volt line works in September, citing safety concerns and lack of access amid community and stakeholder opposition. Jemena has told NextDC the works will not resume before February 2027. NextDC now expects the next step-up in grid supply to M3 only in fiscal 2028. The brutal detail: even halls already handed over to customers cannot reliably carry their full contracted load, so the affected M3 contracts were pulled from guidance entirely.

This is infrastructure risk in its purest form. NextDC did everything right on the data center side and got stopped by a utility's construction schedule and local opposition to power lines. No amount of GPU procurement solves a 66,000-volt line that is not built.

The guidance cuts

Revised FY2027 net revenue guidance: A$475-500 million, down from A$615-640 million. Underlying EBITDA cut to A$245-265 million from A$385-410 million. Capex guidance reduced to A$4.5-5.0 billion from A$5.25-5.75 billion, with a detailed spending review due at the November 20 investor meeting. And the strategic move: NextDC shelved all new development projects and land acquisitions indefinitely, and is evaluating capital recycling including potential sales of selected development sites. That is a growth company hitting pause on growth because it cannot power what it already promised.

The grid is the business model now

NextDC's collapse validates a thesis this site has been tracking all year: the AI buildout's real bottleneck is power, chillers, and permits, not GPUs. AWS is paying billions for Nvidia GB300 racks it cannot switch on yet for related reasons, and data center demand doubled in six months while permits became the new constraint. The industry keeps discovering that a data center is, functionally, a very expensive way to convert grid electricity into compute, and the grid was not designed for this.

The uncomfortable implication: every AI infrastructure forecast assumes power arrives on schedule. NextDC shows what happens when it does not. Utility timelines, community opposition to transmission lines, and transformer shortages are now first-order risks to data center economics, and they are much harder to solve than buying more chips.

Not just NextDC

The selloff extends a souring mood around AI infrastructure bets globally. A data center IPO in Australia was withdrawn last week, and credit investors are turning selective on AI-linked names. The market spent two years pricing AI infrastructure as a one-way bet; NextDC's 44% day is the repricing arriving all at once.

For operators, the lesson is to treat grid interconnection with the same seriousness as customer contracts: secured power, with dates, or no announcement. For investors, it is simpler. In the AI buildout, the scarcest asset is not the GPU. It is the electrons.