NextDC's 44% ASX Crash: Grid Fiasco Wiped A$3.1B in a Day

By Daniel Rao, TechPulse USA

Two days after freezing its own shares, NextDC walked back onto the ASX with the worst kind of news, and the market wasted no time showing it the door. Shares in Australia's biggest pure-play data-centre operator shed 44% in early trade on Wednesday, erasing about A$3.1 billion of market value in a single morning. The culprit was not chips, customers, or interest rates. It was the grid — or, more precisely, the grid that may not arrive until 2028. If you want to know what finally punctures an AI boom, watch this: the wreckage wasn't caused by demand disappearing. It was caused by power poles that never got built.

The fall completes the arc Monday's trading halt warned about, when NextDC halted its own stock pending an announcement on its M3 Melbourne site. Wednesday's announcement delivered the bill, itemised in excruciating detail.

Photo: a data centre server hall. NextDC's M3 Melbourne facility was designed to reach 225MW of IT capacity by late 2027 — if the power lines to feed it ever get built. Credit: Victor Grigas / Wikimedia Foundation, CC BY-SA 3.0.

A$3.1 billion gone in a morning

Reuters reporters Scott Murdoch and Rajasik Mukherjee, reporting from Sydney on Wednesday, said NextDC "shed nearly half of its market capitalisation" after sharply cutting its revenue forecasts and putting future developments on hold. The wire pegged the fall at 44% in early trade, with about A$3.1 billion of market value gone before lunch — figures that require the time qualifier, because the ASX was still open when the news broke and no closing number had been set.

The intraday snapshots tell their own story about the panic. CNBC's tracker had the shares down 39.5% at one point in the session; Motley Fool Australia recorded a fall of as much as 45%, to A$5.10, at the low. The stock was the biggest loser on the broader ASX200 benchmark, and Reuters noted it had sunk to its lowest level since January 18. NextDC is the market's 28th most shorted stock, with short interest of 8.52% as at October 14 — and MarketIndex estimated the day's only winners, the shorts, banked a one-day profit of roughly $250 million. On any other Wednesday that would be the story. Today it was a footnote.

What actually broke: Jemena walked away from the works

In the company's own telling, the problem is brutally specific. Craig Scroggie, NextDC's chief executive and managing director, said in the company's ASX statement: "Jemena has suspended parts of the network works required for the next increase in grid supply to M3 Melbourne. Jemena has attributed the suspension to community and stakeholder opposition to the works and to the absence of local council consent pending further consultation. On Jemena's current indicative estimate, we expect that increase during FY28."

Read that again. The next increase in grid supply to one of Australia's most important AI data centres now lands sometime in the financial year ending June 2028 — at the earliest, on Jemena's "indicative estimate." The delay strands about A$140 million of contracted customer revenue this financial year, which NextDC says will be deferred (Reuters put the figure at roughly US$98 million). The Australian Financial Review, relayed by MarketIndex, reported Jemena paused the connections because of "safety concerns and lack of access to work sites due to community and stakeholder opposition to the work."

Scroggie was at pains to say the customers themselves haven't walked. "The capacity remains under contract, subject to the customer's rights, and construction of the data halls is continuing. We are working with our customer on a revised power ramp and with Jemena to bring power forward." Note the qualifier doing heavy lifting there: subject to the customer's rights.

Illustrative: a standard wooden utility pole on a suburban Sydney street — not M3 itself. Jemena wants to replace poles like these with 15-metre concrete structures carrying 66,000 volts through Melbourne's west. Public domain.

The guidance got cut to the bone

NextDC yanked the affected M3 contracts out of its numbers entirely. Scroggie said: "We have taken the affected M3 contracts out of FY27 guidance until we can reliably estimate their contribution. The August guidance included approximately A$140 million of each of net revenue and Underlying EBITDA from those contracts. The timing and scale of their contribution remain uncertain."

The new numbers, excluding the affected M3 contracts: FY27 net revenue of A$475 million to A$500 million, down from A$615 million to A$640 million; FY27 underlying EBITDA of A$245 million to A$265 million, down from A$385 million to A$410 million. We pulled the full ASX announcement text and the figures match Reuters' wire copy to the dollar — A$475–500m revenue, A$245–265m EBITDA, both excluding the affected M3 contracts.

Strip the M3 contracts out and the valuation looks colossally worse. MarketIndex ran the arithmetic: Goldman Sachs had pegged the average pure-play data centre stock at 22 times FY27 EV/EBITDA; NextDC was already one of the priciest names at 26 times. After the downgrade, its forward multiple has ballooned to roughly 37 times. A growth stock that stops growing at 37 times EBITDA is not a stock — it's a wish.

The growth engine is switched off

The most alarming line in the announcement wasn't the guidance cut. It was the halt: NextDC said it would shelve all new development projects and land acquisitions for the foreseeable future, and would evaluate capital recycling — including potential sales of selected development sites. A hyperscale developer that has stopped developing is a company that has stopped believing its own expansion story, at least until the grid cooperates.

Scroggie framed it as discipline: "With M3 revenue deferred, we have paused projects not backed by customer contracts and put that capital behind the capacity our customers have already contracted. We have approximately A$8.7 billion of liquidity at 30 September 2026 and we are in compliance with our covenants." FY27 capex was cut to A$4.5–5.0 billion from A$5.25–5.75 billion; the company promised the outcome of a detailed capex review by its AGM on November 20. The Victorian government's planning decision on M3 Stages 6 and 7 will now come only after the November 28 state election — a reminder that this story now runs on a political timetable.

There is a debt subplot American investors should not miss. MarketIndex, citing UBS analysis, flagged that NextDC is loss-making until at least FY31, with long-term debt peaking at $11.17 billion in FY30 — and there is roughly $6.7 billion of undrawn senior debt whose covenants could come under pressure if earnings keep slipping. That is the same AI-infrastructure leverage machine we examined in our piece on AI's $113 billion debt machine hitting the wall: the AI buildout is being financed with borrowed money against contracted revenue that can be deferred by a planning decision.

The residents were never the official reason. They still won.

Here is where the story gets delicate, and where the honesty matters. No source in this week's reporting explicitly names last week's Yarraville pole blockade as the cause of the suspension. What the company and the AFR say is "community and stakeholder opposition" plus absent council consent. That is the same community opposition of the kind seen at Yarraville last week, when residents physically blocked Jemena's 15-metre concrete pole rollout — but the company has not confirmed the link, and neither has Jemena. The framing from Monday's trading halt, where analysts merely "tipped it was likely" the blockade was behind the halt, still holds: opposition is the context, not the confirmed cause.

What is confirmed is grim enough for shareholders. Citi analysts wrote in a note carried by CNBC that NextDC's power delays could affect earnings beyond fiscal 2027, citing delays to approvals for the M3 expansion and a longer-than-expected timeline for transformer upgrades. The AFR's reporting on safety concerns and inaccessible worksites suggests this is not a dispute that gets fixed with a phone call. Community opposition plus council consent plus transformer lead times is a three-body problem no AI forecast had priced in.

Australia joins a global AI-sentiment rout

NextDC did not fall in a vacuum. Reuters, in its Wednesday piece, noted that "sentiment has weakened in recent weeks as investors question whether massive AI spending can justify soaring computing, land and power costs," with credit investors growing "increasingly selective toward AI-linked companies."

The American datapoints are uglier. Reuters reported on October 16 that CoreWeave's shares dived more than 15% in a single week while yields on its 2030 bonds jumped 500 basis points to near 16% — a bond market screaming that AI-linked credit is being repriced. Mike Intrator, CoreWeave's chief executive, wrote in an internal memo seen by Reuters that markets can sometimes "misunderstand and act irrationally." Maybe. But James Athey, a fixed-income portfolio manager at Marlborough, put the deeper fear plainly to Reuters: "The problem with this stuff is that we're talking about huge interconnectedness, not just circular financing but the fact that these assets are all held in the same places, in the same portfolios, by the same types of investors. And leverage is always that thing which turns a storm into a teacup into something more nefarious."

The Philadelphia Semiconductor Index fell 4.2% on October 14, its sixth straight losing session — its longest losing streak in nearly two years. And in Australia, the mood music was already sour: Firmus Technologies withdrew its planned ASX listing on October 9, killing what would have been a ~A$7 billion raise at a ~A$44 billion valuation — the biggest Australian float since Telstra in 1997. "Firmus will now pursue capital from the private markets and consider alternative public and private market options," a Firmus spokesperson said. As we wrote when Firmus killed its IPO, the private markets are where overvalued AI assets go to be quietly marked down. NextDC's crash is the same storm, arriving by grid cable instead of bond desk.

Our verdict: the binding contract is hostage to grid politics

Here is the judgment the market has spent Wednesday arriving at, and it belongs in the AI-bubble debate we have been tracking: the AI trade's most sacred assumption — that signed customer contracts equal future revenue — just got broken in half by a power-pole consent process. NextDC's capacity is still "under contract." The contracts just can't be fulfilled, because the electricity to fulfil them has been deferred to 2028 by a community revolt and a council that hasn't consented. A binding contract without electrons is a piece of paper.

M3 was supposed to be safe. Analysts noted at the time that the site was strategically located close to major electricity substations; proximity to the grid was part of the pitch. The substations were close. The poles to reach them never got built. That is the kind of failure no spreadsheet catches: not a demand problem, not a technology problem, a botched sequencing problem in the most basic infrastructure of all — moving electricity from point A to point B. NextDC's community engagement fell short so badly that Jemena publicly admitted it.

Investors are learning, one A$3.1 billion morning at a time, that the AI boom is not a technology story with an electricity problem. It is an electricity story with a technology layer on top. And electricity, unlike chips, requires the consent of the people who live next to the poles.

Analysts are already pricing in the precedent. RBC Capital Markets’ Tom Beadle said ‘whilst out of the company’s control, delays to power supply and development applications relating to NXT’s M3 data centre is evidence that community concerns around data centre developments present a real risk to the timing of delivery of contracted capacity for data centre owners.’ The damage spread beyond one stock: Australia’s tech sector fell roughly 5.6% on the day (Finimize), and Reuters noted the selloff ‘extends the burst of negative sentiment towards global AI stocks’ — the grid bottleneck is becoming the AI trade’s common denominator.

The crash, answered: the questions investors are asking

Could customers cancel the delayed M3 contracts?

That is the next risk after the deferred revenue. The affected contracts include service-level credits for delays and, in some cases, termination rights. NextDC said the capacity at M3 remains contracted, but every month the power does not arrive increases the chance a customer reaches for one of those clauses. The ~A$140 million of deferred revenue is the best case; the worst case is customers walking.

When will Jemena restart the grid works?

Not before February 2027, per CEO Craig Scroggie’s reading of Jemena’s position — and Jemena has said it does not expect to resume the suspended works until community consultation is further along. The next increase in grid supply to M3 is now expected in fiscal 2028, a year later than planned.

Why does Victoria’s election matter to a data centre?

Because the planning decision is now on a political timetable. Victoria enters caretaker conventions on 3 November ahead of the 28 November state election, and the government decision on later stages of the M3 development is expected in 2027. For at least the next five weeks, the approvals that matter are frozen by electoral convention.

Sources: Reuters (Sydney, by Scott Murdoch and Rajasik Mukherjee, Oct 21); NextDC ASX statement via Motley Fool Australia (Oct 21); MarketIndex (Oct 21, relaying the AFR and Goldman Sachs research); CNBC (Oct 21, relaying Citi); Reuters on CoreWeave (Oct 16); Sydney Dispatch/Guardian Australia on the Firmus IPO withdrawal (Oct 9).

Daniel Rao covers AI infrastructure, semiconductors and energy for TechPulse USA. He writes about the physical side of the AI boom — the grids, the chips, and the debt paying for both.