IREN Limited (IREN) Business Profile

What it does

IREN provides computing infrastructure for companies developing and using artificial intelligence. It owns and operates data centers, the computers inside them, and software and support for running AI workloads. Its customers include large technology companies, AI developers, frontier labs and enterprises.

The physical layer includes land, power connections, substations, buildings and cooling. The compute layer includes graphics processing units, or GPUs, alongside servers, storage and networking. Customers can rent dedicated computing equipment, known as bare metal, or use managed cloud services that help them deploy and operate their workloads. These services support both AI training and inference, the process of using an AI model to produce results.

IREN also mines Bitcoin using computing equipment at some of its data centers. It is decommissioning mining hardware and reallocating power and data center capacity to AI cloud services. The company aims to substantially complete the transition by the end of the December 2026 quarter.

Its software capabilities include tools for provisioning GPUs, deploying clusters, monitoring infrastructure and supporting customers. IREN acquired cloud software and services provider Mirantis in August 2026. The company has projects across North America, Europe and Asia Pacific, with operating sites in British Columbia and Texas. William Roberts and Daniel Roberts are co-chief executive officers.

How it makes money

IREN earns AI cloud revenue by providing computing capacity and managed services for AI training and inference. It primarily sells capacity under multi-year reserved arrangements. These contracts generally specify the equipment and capacity provided, prices, service levels, deployment schedules, prepayments and customer acceptance conditions. Revenue generally begins after the computing equipment is delivered, commissioned, placed in service and accepted where applicable. IREN also expects to add on-demand offerings, which generally provide shorter-duration or usage-based access.

Bitcoin mining generates revenue when IREN contributes computing power to the Bitcoin network and receives a share of block rewards and transaction fees. It typically sells the Bitcoin it mines daily and converts the proceeds into conventional currency to fund operating and capital spending. Mining performance depends on factors including Bitcoin’s price, network difficulty, equipment efficiency and electricity costs.

For FY2026, Bitcoin mining revenue was $578.2 million, or 81.8% of revenue, while AI cloud revenue was $128.8 million, or 18.2%. IREN is moving mining capacity toward AI services, so the two businesses’ contributions changed in the subsequent quarter.

Total revenue was $707.0 million for FY2026 and $228 million for the September 2026 quarter.

In that quarter, AI cloud revenue was $200 million and Bitcoin mining revenue was $28 million. After quarter-end, IREN signed a five-year agreement with a customer prepayment covering approximately 90% of the associated GPU capital expenditure. Prepayments help fund equipment, but are credited against later service fees under arrangements such as the Microsoft contract.

Customers and how it sells

IREN targets hyperscalers that need dedicated compute, enterprises that may need managed services and support, and AI developers and frontier labs building models and applications. It also targets channel partners that resell IREN capacity as part of their own services. The company develops its sales pipeline through direct engagement, technology relationships, referrals, industry events and marketing.

Microsoft is a customer for dedicated GPU services at the Horizon facilities in Childress, Texas. The contract has an approximately $9.7 billion total value through 2031, and Microsoft accepted Horizon 1 in August 2026. NVIDIA also has a five-year cloud services contract for its internal AI and research workloads. IREN has contracts with AI developers including Prometheus, Perplexity and Together AI, and an unnamed frontier AI lab.

IREN’s customer contracts are concentrated: losing a large customer, a customer default or a reduction in capacity taken could affect revenue and debt service. IREN says it is seeking to broaden its customer base and uses credit assessment and prepayments to help manage the exposure.

Its data center footprint covers North America, Europe and Asia Pacific. As of June 30, 2026, grid connection agreements and equivalents represented approximately 5GW of total power capacity in the United States, Canada, Spain and Australia. These agreements and equivalents represent power capacity; projects are at different stages of development.

Suppliers and costs

IREN develops and operates its data centers and purchases the computing and other equipment installed in them. Its infrastructure suppliers include NVIDIA, AMD, Dell Technologies, Lenovo, Supermicro and Gigabyte. GPUs installed and on order include systems from NVIDIA and AMD. IREN works with suppliers on servers, racks, networking, storage and successive generations of computing equipment.

Electricity is another operating cost. In British Columbia, BC Hydro supplies facilities under regulated tariffs; in Texas, IREN buys electricity under market-based arrangements linked to wholesale prices. The company may use derivatives or other hedges to fix a portion of electricity costs and is evaluating longer-term purchase arrangements. Operating sites have been powered by renewable energy sources or renewable energy certificates since operations began.

Building data centers and buying GPUs require capital. IREN uses funding sources including customer prepayments, equity, convertible notes and equipment financing. It may commit to equipment purchases and begin site development before arranging financing or signing customer contracts. In the September 2026 quarter, IREN spent approximately $5 billion on capital expenditure and planned approximately $15 billion for the remainder of the fiscal year.

Competitive position

IREN competes in AI cloud services with hyperscale providers Amazon Web Services, Google Cloud, Microsoft Azure and Oracle Cloud, and specialized providers including CoreWeave, Nebius, Crusoe, Lambda, Nscale and SpaceX. Some companies can be customers, suppliers or partners in one context and competitors in another.

Competition also takes place at individual layers of the infrastructure stack. IREN competes for land, power, grid connections, equipment, labor and capital with developers and infrastructure owners. It competes with cloud providers for GPUs and customers, and with software providers for tools that manage and monitor AI infrastructure. Relevant factors include access to power and current-generation computing equipment, construction speed, financing, price, reliability, security and customer support. Many competitors have greater resources, longer cloud operating histories or larger customer bases.

Our reading: Ownership of data centers, power connections and compute gives IREN a potential advantage in coordinating deployments and adapting sites, but it also requires the company to build and finance those assets. Our reading: NVIDIA relationships and the addition of Mirantis broaden access to technology and software capabilities, although IREN still competes with larger cloud providers and depends on suppliers for equipment. Our reading: Multi-year contracts can support revenue visibility, but customer acceptance and performance requirements limit how much protection a signed contract provides.

Growth opportunities

IREN’s growth plans center on converting Bitcoin mining capacity to AI cloud use, bringing contracted GPU capacity into service and developing additional data centers. In the September 2026 quarter, operating capacity rose to 0.1 gigawatts. The company aims to substantially complete its mining transition by the end of the December 2026 quarter.

IREN reported more than $20 billion of revenue backlog, with two-thirds of it from investment-grade customers. It has expanded a contract with an unnamed frontier AI lab and re-contracted H100 GPUs at double their original pricing. The company also signed a five-year agreement with a prepayment covering approximately 90% of the related GPU capital expenditure.

IREN is targeting January for the capacity supporting more than $4 billion of operating annual recurring revenue, or ARR. It says the timing depends on suppliers and may extend to the end of February. A customer-requested design change and supplier timing moved the target from year-end.

Beyond near-term deployments, IREN has more than 35 gigawatts in power development. It cautions that many opportunities are early stage and not all will become grid connection agreements. Its NVIDIA partnership is intended to support deployment over time of up to 5GW of AI infrastructure, while Mirantis adds software and customer-support capabilities. IREN plans to pace investment according to funding conditions, customer contracts and other factors.

Structural risks

• Delivery and acceptance: AI cloud revenue generally begins only after delivery and customer acceptance. Construction, equipment, testing or commissioning delays can postpone revenue while costs continue, and may lead to credits, damages or termination rights. IREN says the timing of its near-term ramp depends partly on suppliers.

• Financing and capital needs: Data center construction and GPU purchases require substantial funding. IREN plans additional capital expenditure and may need further financing; the availability and terms of financing depend on market conditions. Debt service and possible share issuance can also affect cash available for other uses and dilute shareholders.

• Customer concentration: Contracts are concentrated among large customers. A customer’s default, loss or reduction in capacity taken could affect revenue and IREN’s ability to service debt used to finance the related infrastructure.

• Power and regulation: Growth depends on securing and energizing grid-connected power. Permitting, utility processes and changing regulations can delay projects or increase costs. Changes to Texas interconnection procedures may delay energization or change energization levels at IREN projects.

• Equipment and technology: GPU, networking and storage availability, pricing and delivery timing affect deployment. New chip architectures can reduce market rates for earlier equipment generations, while supply-chain disruptions and trade restrictions can affect hardware costs and access.

• Mining transition: Until the transition is complete, Bitcoin price, network difficulty, fees and power costs continue to affect mining performance. IREN expects additional noncash impairments as it decommissions mining hardware.

Bottom line

Our reading

IREN is changing from a Bitcoin miner into an owner and operator of AI computing infrastructure, with long-term customer contracts supporting the transition. Our reading: Its progress depends on turning contracted capacity into accepted, revenue-producing deployments while financing substantial data center and GPU spending. Ownership gives IREN control over more of the infrastructure, but also exposes it to construction, power, equipment and funding risks.