IREN is transitioning from Bitcoin mining to AI cloud services, with revenue surging but losses and capex still weighing on the story. The report rates the stock a Hold as contracted AI demand offsets major execution and financing risk.
IREN Ltd (IREN) is a high-risk infrastructure transition story moving from Bitcoin mining toward AI cloud services. The thesis rests on three named facts: AI Cloud Services revenue rose from $16.4M in FY25 to $128.8M in FY26, management said $4B of 2026 capacity was contracted, and the company delivered its first 50-megawatt Horizon deployment to Microsoft (MSFT). Those facts support a credible growth platform, but they do not yet support a conventional profitability valuation.
The investment case is therefore balanced. IREN controls power, land, data centers, GPUs, and software across Australia, Canada, the United States, and Europe. Its FY26 revenue reached $707.0M, up 41.1% from $501.0M, while AI Cloud Services grew rapidly. At the same time, FY26 produced a $702.6M net loss, annual free cash flow was negative $2.23B, and the company guided to $25B to $30B of FY27 capital expenditures. The growth engine is being built at exceptional speed, but the financing and execution burden is equally substantial.
For a moderate-risk investor with a medium-term horizon, IREN fits a Hold profile rather than a core portfolio position. The quoted price of $39.58 already reflects much of the transition optionality, while our fair value estimate of $45.00 gives credit for contracted AI demand without assuming that every planned gigawatt becomes a high-margin business.
IREN was incorporated in 2018, listed on Nasdaq in November 2021, and changed its name from Iris Energy Limited to IREN Limited in November 2024. The company operates data centers, owns grid connections and substations, deploys GPUs and networking equipment, and provides managed services through its software layer. It also operated a Bitcoin mining business that generated $578.2M of FY26 revenue.
The business has 685 employees, although management said headcount nearly tripled during FY26 and more than 4,000 people were mobilized across active sites. IREN's reported sector classification is Financial Services, but its operating facts place it in AI cloud, high-performance computing, data centers, and Bitcoin mining. That distinction matters because the relevant investment drivers are power access, GPU financing, contract execution, utilization, and technology refresh cycles rather than custody-bank fee income.
AI Cloud Services is the strategic growth segment. FY26 revenue reached $128.8M versus $16.4M in FY25, an increase of roughly eight times. Q4 FY26 AI Cloud Services revenue was $70.5M, compared with $33.6M in Q3. The Q4 increase came while total quarterly revenue fell from $144.8M to $137.2M, showing that the mix was shifting even before the mining fleet was fully retired.
Bitcoin Mining remained the larger FY26 segment at $578.2M, up from $484.6M in FY25, but Q4 mining revenue declined to $66.7M from $111.2M in Q3. Management expects mining operations to be effectively decommissioned by the end of December 2026. The transition caused a $638.8M FY26 asset impairment and a $450.4M Q4 impairment, primarily tied to decommissioning mining hardware. These charges depress reported earnings, but they also mark the cost of reallocating assets toward AI workloads.
The segment transition changes the quality of the revenue base. Bitcoin mining monetizes power through a volatile commodity-linked activity, while AI Cloud Services uses contracted capacity and customer relationships. The difference is visible in management's disclosure of $4B of contracted 2026 annual recurring revenue, with $1B operating at the time of the FY26 results call.
Get AI research on any stock
Instant reports, daily intelligence, and an AI analyst in your pocket.
Horizon 1 is IREN's clearest proof point. The company delivered the first of four 50-megawatt liquid-cooled deployments at its Childress site to Microsoft and achieved NVIDIA (NVDA) Exemplar Cloud status on the GB300 NVL72 platform. Horizon 2 was commissioning, while Horizons 3 and 4 were in late construction and targeted for delivery in the December quarter.
Horizon matters beyond its initial revenue contribution because IREN is using a repeatable template. Management said each Horizon uses the same design, supply chain, and site team, allowing later phases to inherit construction and commissioning lessons. That repeatability could improve deployment speed, but it also concentrates the operating model around successful delivery of a small number of large projects.
The second product layer is Mirantis, the managed-services and orchestration platform. Mirantis gives IREN a way to serve smaller AI developers, enterprises, and labs that want more than bare-metal compute. It supports reserved managed clusters, potential on-demand compute, monitoring, deployment, and enterprise support. Mirantis was also named an inaugural NVIDIA-Certified Hypervisor, giving the software layer a specific validation from the dominant GPU platform provider.
IREN's advantage is an infrastructure moat rather than a traditional software moat. The company owns or controls the land, power, substations, cooling systems, data centers, compute equipment, and software services. A power connection has greater economic value when it is attached to a working data center, a GPU cluster, and a contracted customer. This vertically integrated model is the central strategic claim management made during the FY26 results call.
The company also said its existing sites have spare power beyond current deployment plans. That headroom could support additional GPUs without requiring new grid capacity. Management identified new liquid-cooled installations at Mackenzie, Canal Flats, and Prince George as examples of using existing power more intensively.
Design flexibility is another advantage. Sweetwater 1 is being used as a reference design for Sweetwater 2, Kiowa, Bundey, and later projects. IREN said those designs include modularization, prefabrication, evolving liquid cooling, and 800-volt direct-current capability developed with NVIDIA. The company also uses closed-loop liquid cooling and said NVIDIA's DSX reference architecture can direct more power toward computing rather than cooling.
IREN's operating footprint spans Texas, British Columbia, Oklahoma, South Australia, and Spain. At Childress, retrofit work and GPU installations were running in parallel. At Mackenzie, GPUs had been racked across the first two buildings. At Prince George, the air-cooled fleet was fully commissioned and liquid-cooled installation was underway. Canal Flats was being converted entirely to liquid cooling for GB300 systems.
The construction schedule is aggressive. IREN targets roughly 300 megawatts of delivered IT load in 2026 and another 0.5 gigawatt in 2027, with gross capacity reaching roughly 1.2 gigawatts in 2027. The broader announced pipeline exceeds 5 gigawatts. Sweetwater 2, Kiowa, Bundey, and Badajoz are identified for 2028 development, with roughly 300 megawatts associated with that group.
Supply chain execution remains central because GPUs, networking equipment, substations, cooling systems, and steel must arrive in sequence. IREN's management said the principal industry bottleneck is bringing GPUs online rather than signing contracts. The company has secured $3.6B of investment-grade GPU financing for the Microsoft deployment at about 6% and $2.8B of non-investment-grade GPU financing, including $2.4B at a 9% fixed rate for Mackenzie.
IREN is targeting the rapidly expanding market for GPU compute used in model training, inference, robotics, enterprise software, and other AI workloads. On the FY26 results call, management cited a nearly 17-fold increase in weekly token usage across eight large language models over eight months. That specific usage trend supports the argument that demand is moving beyond a single consumer chatbot cycle.
The commercial market is shifting toward long-term reserved capacity. IREN said three-year contract pricing had risen about 125% since November and five-year pricing about 70%. Recent three-year contracts were priced above $20M per megawatt of IT load, while active discussions were around $25M per megawatt. These figures support strong revenue potential, but they also raise the cost of building capacity before it begins producing reported revenue.
The market opportunity is large enough to support multiple winners, but capacity must be delivered at the right time and at the right GPU generation. IREN's $4B contracted 2026 ARR target came from less than 10% of its stated 5-gigawatt-plus portfolio of grid connections. That creates expansion potential, while also leaving a substantial portion of the long-term development plan exposed to financing, customer, and technology decisions.
Like what you're reading?
Get full access to AI-powered research reports, market analysis, and portfolio tools.
IREN serves four customer groups: hyperscalers, enterprises, AI developers, and frontier AI laboratories. Named customers and counterparties include Microsoft, Cohere, Prometheus, Perplexity, Figure AI, Fal AI, Higgsfield AI, Together AI, and Fireworks AI. Management said Together AI and Fireworks AI renewed and expanded their relationships, providing evidence of repeat demand alongside new customer wins.
The customer mix is becoming broader than the original large-customer strategy. Prometheus and Figure AI are using compute for robotics and physical-world automation, while Mirantis expands IREN's reach to smaller developers and enterprise users that value orchestration and support. This broadening can reduce reliance on a single buyer, although a small number of large contracts still drive the near-term revenue ramp.
Customer economics depend on more than the headline price per megawatt. IREN evaluates contract term, prepayment, utilization, customer quality, service expansion, and strategic value. Recent customer prepayments funded 45% to 55% of GPU capital expenditures, while the Microsoft arrangement funded about 96% of associated GPU capital expenditures when combined with financing. Those structures reduce initial equity needs but increase the importance of delivery against contractual commitments.
IREN competes with specialized AI infrastructure providers such as CoreWeave (CRWV), Nebius (NBIS), Crusoe, Lambda, Nscale, and SpaceX. It also faces hyperscale cloud competition from Microsoft, Amazon (AMZN), Alphabet (GOOGL), and Oracle (ORCL), along with data-center operators such as Equinix (EQIX) and Digital Realty (DLR). In Bitcoin mining, the named competitive set includes Core Scientific, Cipher Mining (CIFR), TeraWulf (WULF), Hut 8 (HUT), Bitdeer (BTDR), Bitfarms (BITF), CleanSpark (CLSK), MARA Holdings (MARA), and Riot Platforms (RIOT).
IREN's strongest competitive points are power access, owned sites, GPU financing, and the ability to repurpose mining infrastructure. Its weaker points are scale relative to hyperscalers, the newness of its AI cloud platform, and the need to execute a rapid shift away from mining. The company has a meaningful infrastructure position, but its advantage remains dependent on construction quality, GPU availability, customer renewals, and financing access.
The most important macro variable for IREN is the cost and availability of capital. Management guided to $25B to $30B of FY27 capital expenditures and expects to fund the plan with cash, GPU financing, customer prepayments, data-center financing, operating cash flow, and corporate sources. The company has already demonstrated access to GPU financing at about 6% for investment-grade capacity and 9% fixed for a Mackenzie facility.
Power policy and grid interconnection are equally important. IREN is developing in Texas, Oklahoma, British Columbia, South Australia, and Spain, giving it geographic breadth but also exposing it to several permitting and regulatory regimes. Management said a tighter Texas interconnection process could favor well-capitalized projects with real development progress.
The company's stated risk factors include tariffs, import and export restrictions, and broader geopolitical conditions affecting equipment procurement and operations. Those risks matter because IREN's growth plan requires large volumes of GPUs and specialized data-center equipment. A delay in equipment delivery can shift revenue recognition while construction and financing costs continue.
FY26 ended with a $702.6M net loss, negative $2.23B in free cash flow, and a planned $25B to $30B of FY27 capex, underscoring how much financing capacity the transition still requires.
Unlock the full analysis
Premium members get the complete breakdown — pick rationale, financial metrics, and recent earnings detail.
Get Full Access →Revenue rose 41.1% to $707.0M in FY26, but a $638.8M asset impairment and a $450.4M Q4 charge helped drive the year’s earnings deep into the red.
Unlock the full analysis
Premium members get the complete breakdown — pick rationale, financial metrics, and recent earnings detail.
Get Full Access →Management said $4B of 2026 capacity was contracted, with $1B already operating at the results call, giving the AI cloud ramp a visible demand base.
Unlock the full analysis
Premium members get the complete breakdown — pick rationale, financial metrics, and recent earnings detail.
Get Full Access →At $39.58, the shares already price in much of the transition optionality, while the $45 fair value implies only limited upside from contracted AI demand.
Unlock the full analysis
Premium members get the complete breakdown — pick rationale, financial metrics, and recent earnings detail.
Get Full Access →The report’s Hold call reflects a fair value of $45 versus a quoted price of $39.58, balancing AI cloud growth against the financing burden of a $25B to $30B capex plan.
Unlock the full analysis
Premium members get the complete breakdown — pick rationale, financial metrics, and recent earnings detail.
Get Full Access →IREN has moved beyond a simple Bitcoin mining story. FY26 AI Cloud Services revenue of $128.8M, the Microsoft Horizon 1 delivery, the NVIDIA-certified software layer, and $4B of contracted 2026 ARR establish a credible AI infrastructure platform. The company also controls a large power and data-center pipeline that can support additional deployments.
The counterweight is equally concrete: a $702.6M FY26 net loss, negative $2.23B annual free cash flow, a 137.0x forward P/E, and $25B to $30B of planned FY27 capital expenditures. IREN can become a major AI cloud operator if it converts capacity into profitable, repeatable cash flow. Until that conversion is visible, the disciplined position is Hold, with accumulation becoming more compelling near $36.00 and a fair value estimate of $45.00.
Get AI-powered research reports, daily market intelligence, and a personal analyst in your pocket.
Get Full Access →Not ready to subscribe? ·
Cancel anytime
IREN's AI contracts are real, but the market is being asked to pay for capacity that has not yet become durable revenue or cash flow. The $638.8 million mining-related impairment, 886x P/E, and 12.5% selloff keep the transition firmly in prove-it territory.
IREN’s latest jump looks earned because the company now has a much larger contracted AI cloud revenue base, not just an AI narrative. The key shift is simple: management raised its year-end AI Cloud ARR target above $4 billion and says roughly 85% is already under contract.
IREN is evolving from a Bitcoin miner into an AI infrastructure platform, backed by secured power, major customer contracts, and rapid AI revenue growth. The opportunity is compelling, but execution risk, heavy capex, and near-term losses keep the story high beta.