IREN is an artificial intelligence and cryptocurrency mining company that operates data centers with a focus on GPU-accelerated computing. The company, formerly known as Iris Energy, was founded in 2018 and is headquartered in Sydney, Australia. IREN has transitioned its business model from primarily Bitcoin mining to a diversified approach, allocating significant resources to AI-related workloads such as machine learning and large language model inference.
IREN's core operations involve the development and operation of large-scale data centers designed for high-density computing. The company emphasizes its proprietary immersion cooling technology, which it states enables efficient management of heat generated by high-performance GPUs. This infrastructure underpins both its cryptocurrency mining activities and its AI services, which are offered to external clients and used internally for model training and inference tasks.
History and Pivots
IREN was founded in 2018 as Iris Energy by Daniel Roberts, a former investment banker. The company initially focused on Bitcoin mining using renewable energy sources, developing data center sites in Canada and the United States. In 2021, Iris Energy went public on the Nasdaq under the ticker IREN. The company's strategy involved building modular data centers and securing low-cost power agreements.
In 2022, as the cryptocurrency market experienced a downturn, Iris Energy faced financial challenges, including a US$103 million loan default on mining equipment. The company restructured its debt and continued operations. Starting in 2023, IREN began exploring AI as a complementary revenue stream, acquiring NVIDIA GPUs and repurposing a portion of its data center capacity for AI workloads. In early 2024, the company officially rebranded to IREN to reflect its expanded focus.
AI Infrastructure and Products
IREN operates AI-focused data centers in Childress, Texas, and Sweetwater, Texas, with a combined target capacity of over 1 gigawatt. These facilities utilize immersion cooling, where servers are submerged in a dielectric fluid, to support high-density clusters of GPUs. As of late 2024, IREN reported deploying approximately 10,000 NVIDIA H100 GPUs for AI services, with plans to expand to 50,000 GPUs by 2025.
The company's key AI product is IREN Compute, a cloud-based service offering GPU compute for training and inference of large language models and other AI applications. IREN also develops internal AI models, such as its in-house large language model for code generation, trained on its own infrastructure. In 2024, IREN announced a multi-year partnership with a major cloud provider to supply AI compute, though specific terms were not disclosed.
Financial and Operational Performance
IREN reported revenue of US$190 million for the fiscal year ending June 2024, with a net loss of US$38 million. The company's AI segment contributed approximately US$15 million in revenue during that period, up from zero in the previous year. IREN's total hash rate for Bitcoin mining reached 10 exahashes per second in July 2024.
In August 2024, IREN acquired a 400-megawatt data center campus in West Texas for US$100 million, expanding its total power capacity. The company has also signed long-term power purchase agreements for up to 1.6 gigawatts of renewable energy across its sites. As of November 2024, IREN employed around 200 staff, including engineers and data center operators.
Competitive Position and Strategy
IREN competes with other AI compute providers such as coreweave and crusoe-energy, as well as traditional cloud services like Amazon Web Services and Google Cloud. The company differentiates itself through its immersion-cooled, purpose-built infrastructure, which it claims offers lower operational costs and higher density than air-cooled rivals. IREN also leverages its experience in energy procurement and grid management from its mining background.
The company's strategy involves vertical integration, controlling everything from power generation contracts to GPU deployment and software stack optimization. IREN has not disclosed specific model architectures or training runs publicly, but its technical team has published research on improving inference performance for Transformer (architecture) models. The company aims to capture demand from both enterprise clients and AI startups that require flexible, scalable GPU resources.
Future Outlook
IREN plans to double its AI compute capacity by mid-2026, driven by expected growth in Generative AI workloads. The company is evaluating locations in South America and Europe for future data centers, focusing on regions with surplus renewable energy. IREN has also invested in Large language model fine-tuning tools and is developing proprietary software for scheduling GPU tasks.
Challenges include volatile energy prices, potential supply constraints for NVIDIA GPUs, and competition from established cloud providers. IREN's reliance on a single hardware vendor ([nvidia-gpu]) and its dependence on Texas power grids are noted risk factors. Despite these, management projects that AI services will constitute over half of total revenue by 2026.
Governance and Leadership
IREN is led by co-founder and CEO Daniel Roberts, with co-founder and CTO Daniel Parker overseeing technical strategy. The board includes former executives from Nokia Bell Labs and Xerox PARC. The company maintains a public-facing research blog and has published papers on Model Pruning and Data Augmentation techniques tailored for GPU clusters.