Market / Stocks

NextDC lost 44 percent in a morning because the power is late

Australia's AI data centre operator wiped about A$3.1 billion off its value in early trade after a delayed grid upgrade in Melbourne forced it to slash its 2027 guidance and shelve every new project.

Melbourne, Australia · October 21, 2026

Forty four percent, gone in a single morning. NextDC, the Australian data centre operator that investors had treated as one of the purest ways to buy the AI buildout, fell off a cliff on Wednesday, wiping about A$3.1 billion from its market value in early trade. It was the biggest loser on the ASX200 and fell to its lowest level since January. Nothing had gone wrong with demand for its data centres. The problem was far more basic: the electricity for its big M3 site in Melbourne is not coming on time.

The details explain the size of the fall. Works by Jemena Electricity Networks to lift the power supply to M3 were suspended in September after community opposition, safety concerns and a missing council consent, and Jemena does not expect to restart before February 2027. The next increase in grid supply is now expected in 2028. That pushes about A$140 million of contracted revenue out of the coming financial year, so NextDC cut its 2027 net revenue guidance to A$475 million to A$500 million, down from A$615 million to A$640 million, and its underlying earnings guidance to A$245 million to A$265 million, down from A$385 million to A$410 million. Every new development project and land acquisition is shelved, and capital spending guidance came down with it.

What makes the story sting is that the customers are already signed. The M3 capacity is contracted, and the revenue was in the old forecast because buyers want the space. But some of those contracts carry service credits for delay, and some carry termination rights, so a late grid connection is not just revenue moved to another year. It is revenue that has to be defended. NextDC still holds substantial liquidity and says it remains within its debt covenants, yet the market did not spend Wednesday debating the balance sheet. It repriced the assumption underneath the whole AI infrastructure trade: that if you build the halls, the power will somehow be there.

Our call is that this is the moment the AI boom's real bottleneck gets priced properly, and it is not chips or capital, it is grid connections. NextDC did everything the market asked, signed the customers and raised the money, and still lost nearly half its value to a local network upgrade stuck in consultation. From here the winners in data centres will be the operators who already hold secured power, not the ones with the longest list of planned sites, and every developer with an unconnected project should expect investors to start asking for the grid paperwork before they ask about demand. The demand was never the question. The wire is.