Tech & AI
IREN's AI cloud revenue nearly triples to $200m, but $4bn run-rate target slips to January
The Microsoft-linked data-centre operator blamed a customer design change and supplier timing as it laid out about $15bn of remaining capital spending.
HedgeWeekly Tech & AI Desk ·
IREN, the Australian-founded operator that has pivoted from bitcoin mining to renting out AI computing power, reported fiscal first-quarter results on 5 November. AI cloud revenue for the quarter to 30 September was $200m, up from $71m in the previous quarter. Total revenue rose to $228m from $137m, and the net loss narrowed to $263m from $684m.
The company said it now has more than $4bn of annualised recurring revenue (the yearly value of signed customer contracts) contracted and funded, while operating annualised revenue, which counts only capacity already in service, doubled to $1bn. It moved its target for reaching $4bn of operating run-rate from the end of 2026 to January 2027, with a possible slip to February, citing a customer-requested design change and supplier timing.
Horizon 1, the first phase of its Texas build-out, is in service for Microsoft and earned Nvidia's Exemplar status for GB300 NVL72 systems, a rack-scale design for training large AI models. Horizons 2 to 4 are in late-stage construction. IREN estimates about $15bn of capital spending remains in its 2027 fiscal year; roughly $9bn is covered by cash, committed GPU financing and customer prepayments, and about $6bn depends on further financing and market conditions.
Why it matters: so-called neocloud operators are valued on how fast contracted capacity turns into billable revenue, so a slipped delivery date matters even where no penalty is flagged. The roughly $6bn still to be financed lands as credit markets scrutinise AI infrastructure debt, and the timing of Horizons 2 to 4 is the next test for both IREN and the hyperscaler customers waiting on that capacity.