Centuria Capital is selling equity at a premium earned by calling itself AI infrastructure. The Australian property investment manager announced a A$300 million raise at A$2.00 per share to fund the expansion of its ResetData data-center platform, taking advantage — Jefferies’ phrasing — of a 31% rise in its share price over the preceding month. The share price rose. Then came the equity raise. The sequence is the whole story.
It is true that the demand for compute capacity driven by the current generation of large language models is real, and that someone has to build the infrastructure to meet it. Centuria’s ResetData platform targets a 23-to-24-megawatt pipeline — one of only three Australian NVIDIA Cloud Partners, with a 1.1-megawatt sovereign AI factory already running and a stated pathway to 200 megawatts across existing Centuria properties. Nobody is writing off the opportunity. But an opportunity that costs A$1.35 billion to build and relies on vendor financing at 70% loan-to-value is an opportunity that demands a cold-eyed look at the equity cash flow, not a warm assumption that AI demand will paper over the gap.
Macquarie did the math that should make every growth-hungry infrastructure investor sit up. Its analysts peg the capital cost of building out ResetData’s pipeline at A$56 million per megawatt. On a 23-megawatt pipeline — the Journal’s report cites 24.1 megawatts — that puts total outlay at nearly A$1.35 billion. Even with vendor financing at a 70% loan-to-value ratio, the equity check required runs to roughly A$400 million. Centuria’s A$300 million raise, in other words, leaves a gap wide enough to drive a GPU cluster through. The firm downgraded the stock to underperform, and while it bumped its price target a token 6% to A$1.88, that still sits a clean 30 cents below the last trade of A$2.18. When the math says the upside is already more than priced in, a capital raise isn’t a catalyst — it’s a confession that the bill is bigger than the balance sheet.
Jefferies, on the other hand, wants you to look past the dilution. Analyst Andrew Dodds retains a buy call, raises his price target to A$2.66 per share, and frames the shortfall as forward investment: the raise “will only hit earnings by less than 1% in FY 2027,” he writes, without accounting for “the potentially higher returning reinvestment of proceeds.” Not zero, but close enough to file under “rounding error” in a presentation deck. It’s the classic growth-capital formula: take a short-term haircut now, collect multiples later when the AI factories come online.
The problem is that “later” has a hardening price tag, and it’s measured in megawatts, not quarters. And here it is worth being precise about what a data center at this scale actually requires, because the public discourse has the habit of treating data centers as unitary things — warehouses of humming servers that materialize when capital is deployed. A 24-megawatt facility is industrial infrastructure. It requires specific electrical-grid connections, specific cooling systems, specific water supplies, and specific regulatory approvals, each operating on its own timeline and its own cost curve. The A$56-million-per-megawatt figure Macquarie cites reflects the current cost of building compute infrastructure in Australia, where grid connections are constrained, construction labor is scarce, and the supply chain for high-performance cooling equipment runs through the same bottlenecks every hyperscaler on the planet is competing for. Centuria is not deploying software. It is building a factory. The factory costs four and a half times what it is raising.
The vendor-financing assumption — 70% loan-to-value on equipment whose residual value depends on someone filling the facility with paying customers through the depreciation cycle — is doing an enormous amount of work in both the bull case and the bear case. If the financing materializes at those terms, Macquarie’s equity figure of A$400 million is the floor. If it does not, Centuria is a property manager with A$300 million of shareholder equity deployed in a market whose capital requirements it cannot meet from the raise.
The deluge of institutional capital pouring into AI infrastructure has conditioned markets to believe that any equity raise aimed at data centers is an act of vision. But Centuria’s raise reveals the other side of that flood: the cost to build has escalated so far and so fast that a A$300 million raise — a number that would have been a headline-grabber two years ago — now looks like a down payment. A property manager that was managing property — industrial, office, retail — six months ago can now raise equity at a premium because it has attached the words “data center” to its investor materials. The 31% share-price rise is not the market pricing in the net present value of future data-center cash flows. The data center, at the scale this raise contemplates, does not yet exist. The market is pricing in the narrative.
Jefferies’ buy case rests, stripped to its working parts, on a chain of conditionals: the proceeds might generate higher returns if the data-center market grows as projected, if the vendor financing materializes at the assumed terms, if the operating costs track estimates, if the AI compute demand holds. Each “if” is a load-bearing member. Remove any one and the structure stands less certainly. Remove two and it does not stand at all.
Macquarie’s downgrade isn’t cynicism. It’s a reaction to a stock that had already run 31% in a month on the very story the raise is supposed to fund, and a recognition that the raise itself — even at A$300 million — falls short of the equity slot the full pipeline requires. When the bear case is simply “the math doesn’t work at this price,” and the bull case is “the math will work later if everything goes right,” the burden of proof shifts. In a market that’s been shoveling capital into AI infrastructure with barely a pause for due diligence, Macquarie’s call feels less like a downgrade and more like a tripwire the rest of the market hasn’t noticed yet.
The capital arrives before the megawatts. The share price rises before the revenue. The analyst upgrades precede the operating history. This is the AI buildout reaching the property market, and the property market doing what it has always done when a new narrative arrives: pricing the story first and checking the math later. That gap, not the raise, is the real number. And the real number is not in the prospectus.