IREN's $5.5B Nvidia Deal Marks Neocloud Shift

IREN Ltd, formerly known as Iris Energy, has pivoted from its origins as a large-scale Bitcoin miner to become a significant player in the artificial intelligence infrastructure sector. On May 7, 2026, the company announced a substantial agreement with Nvidia valued at $5.5 billion in combined economic value [2].

Key takeaways

• IREN announced a $3.4 billion, five-year AI cloud services contract with Nvidia, alongside a $2.1 billion equity warrant [2].
• The deal involves deploying Blackwell GPUs at IREN’s Childress, Texas campus, with a goal of up to 5 gigawatts of Nvidia DSX-aligned AI infrastructure globally [2].
• Originally founded in Australia as a Bitcoin miner, IREN began transitioning its high-density data centers to AI cloud infrastructure in 2024 and 2025 [2].

From Bitcoin Mining to AI Infrastructure

Founded in Australia, IREN initially built its business by acquiring or building data centers in low-cost power markets to mine Bitcoin using application-specific integrated circuits (ASICs) [2]. While this model was capital-intensive and subject to commodity pricing pressures, the company recognized that its physical assets—specifically the power infrastructure, cooling systems, and high-density rack deployments—were suitable for general-purpose parallel processors like GPUs [2]. Consequently, IREN spent 2024 and 2025 converting its mining capacity to AI cloud infrastructure and signing enterprise customers for GPU access [2].

The Nvidia Partnership and Neocloud Growth

The company's strategic shift culminated in a May 7, 2026, announcement detailing a five-year AI cloud services contract with Nvidia worth $3.4 billion [2]. This agreement includes a $2.1 billion equity warrant, bringing the total economic value of the deal to $5.5 billion [2]. The partnership focuses on deploying Blackwell GPUs at IREN’s Childress, Texas campus, with broader ambitions to target up to 5 gigawatts of Nvidia DSX-aligned AI infrastructure globally [2]. This move positions IREN within the "neocloud" category, which aims to provide dedicated, high-density GPU capacity for training frontier models, contrasting with the shared, metered capacity of traditional hyperscalers [2].

Why it matters

The IREN-Nvidia deal highlights a growing trend where specialized infrastructure providers, or "neoclouds," are stepping in to offer dedicated compute resources that major hyperscalers may not provide efficiently [2]. As demand for AI training and inference rises, the ability to repurpose existing high-density data centers for GPU workloads offers a new pathway for infrastructure development, addressing the question of where the necessary compute for future AI models will originate [2].