IREN Missed and Dropped 8%. The Real Bet Is a December Deadline

IREN finished Thursday's regular session up 2.4% at $40.53, then fell about 8.5% in after-hours trading to near $37 once its fiscal fourth-quarter report hit the tape — a long way from the $76.87 record it set last November.

The surface read justified the sell. Revenue of $137.2 million came in around 13% below a Wall Street consensus of roughly $157 million. Loss per share was $0.74 against the $0.55 expected. And IREN reported a $684 million net loss for the quarter.

Look inside that revenue number and part of the "miss" turns out to be a business being killed on purpose.

Bitcoin mining revenue fell from $111.2 million to $66.7 million in the quarter. That was not a market accident. IREN is decommissioning its mining machines — switching them off and selling them — to turn its sites into AI data centers. In the same quarter, AI cloud revenue more than doubled, from $33.6 million to $70.5 million, and for the first time overtook mining as more than half the business. Total revenue still fell because the company switched the old engine off before the new one finished being installed.

The gargantuan loss is the same story in accounting dress. About $552 million of the $684 million — roughly 80% — was non-cash charges for retiring the mining fleet: a $450.4 million impairment plus a $102.1 million markdown on machines held for sale. An impairment is the company admitting machines are worth less on the books because they are being scrapped — ugly, but no cash leaves the building. Adjusted for that, IREN still produced $245.7 million of EBITDA for the full year, a 35% margin, on revenue that still grew 41% to $707 million.

So why did the stock drop 8% anyway? Because the number the market actually prices — the AI cloud — is still small relative to the story. IREN has contracted about $4 billion of annualized revenue against its 2026 capacity, which is largely sold out (ARR: what the signed contracts would earn in a year once running), but only about $1 billion of it was operating at report time, after Microsoft accepted the first 50-megawatt block at its Texas site in August. Management was explicit about the consequence: much of the remaining capacity gets commissioned late in the December quarter, so the revenue it produces won't reach the income statement until the March quarter. "Built but not switched on" is exactly the phase of this sector where expectations get hurt.

What the sell-off leaves standing is demand. Three-year contract pricing is up about 125% since November, past $20 million per megawatt, with new deals under discussion near $25 million — by management's math, a two-year payback. Customers now prepay 45% to 55% of GPU costs, so the buyer, not IREN, fronts most of the hardware bill. IREN also holds what is scarce in this market: secured power and land, about 5 gigawatts of it, largely renewable, from Texas and Oklahoma to British Columbia, Australia and Spain.

The co-CEO compressed the whole trade into one sentence: "The digital world scales almost instantly." The physical world does not. Nvidia can make chips abundant; it cannot make finished, powered data centers appear at the same speed. IREN's answer to that gap arrived last November, when it signed a $9.7 billion agreement to build GPU cloud capacity for Microsoft. The stock's entire remaining argument is that IREN is one of the scarce pieces — an owner of built, operating, contracted data-center supply — sitting in front of an abundant supply of demand.

Now the case against, stated as fairly as I can. This is a $25 billion to $30 billion capital expenditure program for fiscal 2027. IREN ended the year with about $7.6 billion of cash and roughly $14 billion of committed cash, GPU financing and prepayments — leaving on the order of $8 billion more to finance, on top of the $19 billion it has raised in the past twelve months. The share count has climbed with each round. If the next money comes increasingly from equity, each share of a company worth roughly $14 billion at Thursday's close is being asked to fund a construction bill larger than the company's own value. Prepayments and GPU debt keep that affordable; equity offerings quietly consume it. Which way that tilts, not the quarterly loss, is the real risk to per-share value.

The deadline is December because the old business is running out of clock. IREN paused its mining expansion more than a year ago to chase AI, and management now expects the mining business to be effectively gone by the end of this year. That old business was a real one — last September it generated a record $240.3 million in a single quarter. Bitcoin has since fallen from roughly $125,000 to about $80,000, squeezing mining margins across the industry, and a miner's revenue is a commodity. A contracted, prepaid data center is not. IREN is spending tens of billions to trade one for the other, which is exactly why its income statement looks this ugly in the interim.

So the after-hours 8% was the market repricing a timetable and a bill, not a verdict on demand. The demand is already signed at higher prices than anyone was getting in November. What decides this stock over the next few months: whether operating ARR reaches the $4 billion mark it has contracted to have running by December, whether the revenue shows up in the March quarter as promised, and how the rest of that $8 billion-plus gap gets funded. Debt and prepayments are one thing; new shares are another. Watch those three. The quarterly loss — a business being decommissioned on purpose — is the least informative number on the page.