IREN (NASDAQ: IREN) dropped 15% last week after quarterly results missed expectations, yet the company heads into the back half of 2026 largely sold out of capacity and carrying $4 billion in annualized run rate revenue. The tension is that the financial print looks bad while the forward book looks good, and which one the tape decides to price is the question worth sitting with.
The June 30 quarter was genuinely messy. AI cloud services revenue more than doubled sequentially to $70.5 million, but total revenue slid 5% to $137.2 million as the company continued unwinding its cryptocurrency mining operations. Adjusted EBITDA dropped 68% to $19.2 million. Those are real numbers, not noise.
What's changed is the pricing environment for the forward book. IREN is in active discussions to provide capacity at roughly $25 million per megawatt, up from recent deals priced at $20 million per MW. Customers have been agreeing to prepay up to 55% of GPU-related costs, an unusual degree of commitment from buyers who typically hold leverage in early contract negotiations. Co-CEO Daniel Roberts framed the structural position plainly in the earnings release: "We have spent years assembling what is difficult to replicate: power, land, data centers, compute, software, and people."
The counterargument is that the earnings are not a distraction. They are the signal. A 68% drop in adjusted EBITDA in a quarter when AI services revenue doubled means the business is spending well ahead of what it is collecting, and that gap can widen for longer than patient investors expect. A $4 billion annualized run rate looks compelling until you notice that total quarterly revenue came in at $137.2 million. The read-through is that the 2026 book is sold, but how fast that capacity comes online and at what cost remains the operative question.
H.C. Wainwright analyst Mike Colonnese has a buy rating on IREN with a $90 price target, a level that would represent gains of over 150% from current prices. On balance, the line to watch is not whether management's thesis about AI infrastructure scarcity is correct. Roberts said as much in the release: "IREN was built for this moment." The risk is timing. A $90 price target means nothing if the path from here runs through multiple quarters of earnings that look like the one the market just punished.