Shares of IREN fell more than 8 percent after the Bitcoin miner turned AI data center operator reported a 684 million US dollar loss in its fiscal 2026 results, a reminder that the industry's pivot from mining to artificial intelligence infrastructure comes with heavy near-term costs.
The stock dropped 8.2 percent in after-hours trading Thursday to 37.19 US dollars, after closing at 40.53 US dollars, as investors weighed the quarterly loss against record growth in the company's AI cloud business.
AI cloud revenue surpasses mining
AI cloud services generated 70.5 million US dollars in the fiscal fourth quarter ended June 30, up from 33.6 million US dollars the previous quarter. The segment surpassed Bitcoin mining revenue of 66.7 million US dollars for the first time and accounted for 51.4 percent of IREN's quarterly revenue, a symbolic threshold in the miner's transformation.
Mining revenue fell 40 percent from the prior quarter as IREN converted mining sites for AI use, replacing mining hardware with graphics processing units for AI workloads. Total revenue declined 5 percent to 137.2 million US dollars.
The milestone comes amid a broader rally in mining stocks, driven by Bitcoin's 23 percent weekly climb. IREN itself had been among the sector's strong performers before the earnings report, making the sell-off more about the pace of conversion costs than the underlying strategy.
The 684 million dollar quarter
The headline loss included 450.4 million US dollars in asset impairments, largely tied to the write-down of mining hardware being retired ahead of schedule as the company accelerates its AI conversion. Excluding impairments, the results reflect a business in transition, with declining mining income not yet fully offset by the scaling AI cloud segment.
Of IREN's 4 billion US dollars in contracted annualized run-rate revenue, 1 billion US dollars was operating as of August 26, according to the company. The backlog provides long-term visibility, but the market's reaction suggests investors want faster conversion of contracts into cash flow.
Bernstein analysts expect IREN to wind down Bitcoin mining by 2030 as the company replaces mining hardware with GPUs for AI workloads, effectively completing its evolution from crypto miner to AI infrastructure provider.
A founding thesis tested by markets
"We started IREN with a simple observation: the digital world can scale almost instantly, but the physical world cannot," Co-Founder and Co-CEO Daniel Roberts said in a statement. "This year, that founding thesis became tangible. Exponential AI consumption growth has fueled demand for compute capacity."
That thesis is now being tested in the market's willingness to fund the transition. The impairment charges underscore a real cost of the pivot: mining hardware purchased at premium prices during bull markets is being retired early, and those write-downs land directly on quarterly results.
The strategy also concentrates exposure to the AI infrastructure cycle at a time when some analysts have begun questioning the capital intensity of the sector. With hyperscalers and dedicated AI providers committing billions to compute capacity, competition for contracts, power, and customers is intensifying.
The miner pivot in context
IREN's report lands at a moment of sharp divergence in the mining sector. Bitcoin's rally above 78,000 US dollars lifted miner stocks broadly this week, with some names gaining more than 40 percent, as stronger Bitcoin economics improved margins for companies still focused on hashing.
But IREN's numbers illustrate the other side of the trade: for miners converting to AI, near-term Bitcoin strength is bittersweet, since every machine redirected to AI workloads forgoes mining revenue that would otherwise be climbing with the Bitcoin price.
The market's 8 percent verdict on IREN suggests investors are split on that trade-off. The 1 billion US dollars of operating annualized revenue and the faster-than-expected growth in AI cloud services argue the pivot is working. The impairments and the pace of spending argue it is expensive.
Either way, the quarter marks a turning point: for the first time, more than half of IREN's revenue came from something other than mining Bitcoin. Whether that becomes the sector's template or IREN's exception is now the central question for mining investors watching the AI gold rush from their racks of GPUs.
Disclaimer: This article is for informational purposes only and does not constitute financial advice.
cloud revenue passing mining at IREN, the stock still dumps 5 percent. market wanted the transformation free of charge
684m loss but ai cloud revenue doubled qoq to 70.5m and passed mining for the first time. market sold the headline, story is actually decent imo
70.5m doubling qoq is real growth but a 684m annual loss means the GPUs need to keep doubling for years before the math closes. market isnt wrong to ask questions
Mining fell 40% because they’re ripping out rigs for GPUs. You don’t get the AI multiple without the transition pain showing up somewhere.
story is decent but a big chunk of that 684m is impairment and transition cost booked up front. steady state margins next quarter are the real number
Mining down 40 percent quarter over quarter while they rip out hardware for GPUs. That transition cost is real and the market priced it today.
same story as hut 8 and core. the miners that moved early on ai hosting are the ones still standing
137.2m total rev and everyone acts like the company is dying lol. its literally pivoting to ai while profitable
profitable? the fiscal year was a 684m loss. the quarterly revenue mix changed, the bottom line didnt
GPUs amortize over 5 plus years, the loss is mostly upfront rig writeoffs. steady state margins are the number to watch
its a 684m loss on paper but the writeoffs are booked up front for the rig conversions. watch steady state margins next quarter, not the transition noise
which quarter is steady state though. theyre converting sites all year, every print comes with its own writeoffs. the goalpost moves each time
fair on the moving goalposts, but cloud at 70.5m passing mining is a one way door. once sites convert they dont convert back, the crossover compounds
137.2m revenue against a 684m annual loss is the opposite of profitable. the mix shift is real, the cash burn is realer
cash burn is realer except the writeoffs are non cash. look at the actual opex line before calling the patient dead
cash burn is realer except the writeoffs are non cash and the opex line barely moved. two more clean quarters and the 684m becomes a footnote
37.19 after hours on a revenue mix crossing 51% ai feels overdone. could be wrong but that’s a rough number for the sector leader
the 51 percent ai crossover is symbolic but the multiple already priced in a flawless pivot. an 8 percent dump on a record quarter is the market saying show me profits
agreed the multiple was ahead of the pivot, but mining down 40 percent while cloud doubles to 70.5m is the mix you want if you have to pick one
mining at 66.7m and cloud at 70.5m, the crossover people penciled in for 2027 happened a year early. the dump is transition noise, not a thesis break
record quarter by every operating metric and the stock still gets hit 8.2 percent. markets dont pay for transitions, they pay for finished ones
finished transitions exactly. core scientific ate two years of dumping before the ai multiple showed up, iren is early in the same purgatory
37.19 after hours is the market pricing years of capex before the revenue catches up. classic transformation discount, every miner that survived 2022 went through it
every miner that made it through 2022 survived a transition discount first. cloud doubling to 70.5m and passing mining is worth an 8 percent haircut imo
Core Scientific also returned 5x once the market believed the mix. someone has to hold the bag during purgatory, that is where the discount comes from
cloud at 70.5m doubling while mining drops 40 percent is the pivot working. the 684m loss is tuition, question is how many more quarters of it
37.19 after hours on the exact quarter cloud passed mining is the market wanting the pivot pre loaded. core scientific ate two years of the same discount before it paid