NextDC cut its earnings forecast by more than a third and the shares fell as much as forty four percent in a day, the worst session the company has had since it listed in 2010, before clawing back part of the loss. This is an Australian data centre operator in the middle of the largest infrastructure build the industry has ever attempted, in a year when demand for computing capacity has never been less in doubt. The problem was not demand. The problem was permission, power and the people who live nearby.
That distinction is the whole story. For three years the data centre trade has been sold to investors as a demand story, which made it the easiest investment case in the market: artificial intelligence needs compute, compute needs buildings, buildings need land, therefore buy the landlord. Every part of that chain is true. What the chain omitted is that the conversion of signed demand into billable revenue runs through electricity connections, grid upgrades, local planning authorities and residents who did not sign anything and have no commercial reason to cooperate. Revenue recognition in this business begins when a facility energises, not when a customer commits. NextDC has just demonstrated that the gap between those two moments is where the equity value lives.
Anyone who has fitted out a cafe recognises the shape of it immediately. You can have the lease, the roaster, the trained staff, the suppliers and a queue of people genuinely desperate for a decent espresso, and none of it produces a single dollar until three phase power arrives at the back wall. The demand was never the hard part. The hard part is a utility connection, an inspector and a landlord’s consent, in a sequence you do not control, on a timetable nobody will put in writing. The machine is beautiful. It is also, until that day, furniture.
What makes the forty four percent drop instructive rather than merely brutal is what it reveals about how the sector was being valued. A share price that halves on a delay rather than a cancellation was priced for perfect execution, which means the market had quietly assumed that the only variable left was how fast the company could build. Timing risk was being treated as noise. It is not noise in infrastructure. It is the single largest determinant of return, because every month of delay carries financing cost on capital already deployed against revenue that has not started. The recovery during the session tells you the long term thesis survived intact. The fall tells you that nobody had modelled the thesis arriving two years later than planned.
There is also a genuinely new factor at work, which is that data centres have become locally unpopular in a way they were not five years ago. A building full of servers consumes serious power and water, employs very few people once it is running, and is difficult to explain to a community being asked to host it. That is a poor political trade for a local authority, and it is now showing up in approval timelines across several countries. The industry has spent its energy convincing capital markets and almost none convincing suburbs, and the suburbs turn out to hold a veto.
NextDC will build the sites, energise the halls and fill them, because the computing demand behind all of this is real and growing. The company has simply been handed the industry’s new price list, and the item at the top is consent.
The cloud has to sit somewhere, and somewhere has an address, a council and a view it rather liked as it was.
Source: Financial Times, “NextDC shares plunge after Australian data centre group cuts earnings forecast”