The transition from Bitcoin mining to a large-scale AI data center operator is exemplified by IREN Ltd, which announced on Tuesday that it has agreed to purchase NVIDIA's air-cooled Blackwell architecture AI computing systems from leading AI server provider Dell Technologies for approximately $16 billion. The objective is to bring more AI computing resources online to meet the globally surging demand for AI computing infrastructure, driven by AI agents. According to an internal statement, this transaction supports a previously announced five-year, $34 billion cloud computing AI services contract with Dell. This substantial procurement of Blackwell systems by IREN serves as another significant indicator of the persistent explosion in global AI computing demand.
Purchases like IREN's Blackwell systems, Microsoft's early reservation of AI GPU cloud computing capacity from IREN, and NVIDIA's investment in data center operator IREN all address the same fundamental issue: AI computing infrastructure demand is now so robust that the market is securing supply for the period around 2027 in advance. Financial reports from the four major tech giants—Microsoft, Amazon, Meta Platforms, and Alphabet—released nearly simultaneously, convey a crucial signal for the AI computing supply chain super-cycle: the arms race in AI computing infrastructure and resources is far from over. Combined capital expenditures for 2026 are projected to reach at least approximately $725 billion, significantly exceeding 2025 spending levels and surpassing the latest Wall Street analyst expectations. Tech giants are willing to endure pressure on cash flow and profit margins rather than fall behind in this computing arms race.
NVIDIA's latest earnings report clearly highlights that the global frenzy for AI computing infrastructure construction continues unabated and is expanding from AI GPUs and ASICs to data center CPUs, high-performance networking infrastructure, enterprise-grade HBM, DRAM, NAND storage, server clusters at the rack level, AI super factories, and large-scale enterprise AI cloud computing systems. On Wall Street, bullish sentiment towards NVIDIA, the global AI leader, is intensifying. The average price target suggests NVIDIA's market capitalization could surpass $7 trillion, with a consensus expectation that spending on computing infrastructure for AI training and inference will reach at least the $3 trillion level by 2030.
The demand for AI computing is generating super-sized orders. IREN's increased investment in Blackwell systems aims to boost its annualized recurring revenue target to $44 billion. Bitcoin mining companies transitioning to AI data centers possess advantages in speed and capital efficiency compared to cloud giants building from scratch, primarily because computing infrastructure, power, and sites are largely already in place. They have grid interconnection points, large substations, and available computing infrastructure parks that, after modification or minor upgrades by cloud giants, can be operational for formal large-model AI data centers. Simultaneously, they can "lock in" cash flow through multi-year AI contracts. This provides them with significant speed and capital efficiency advantages in the current environment of severe supply shortages for AI chips and other computing infrastructure, coupled with long-term constraints on grid interconnection and permitting.
According to IREN's forecasts, once the actual contracts are operational, the company's annualized recurring revenue is expected to increase from $37 billion to $44 billion, reflecting continuous positive progress in bringing GPU capacity online and expanding AI computing infrastructure. These Blackwell AI computing systems will be deployed in IREN's existing AI data centers at its large Childress, Texas campus and are expected to be ready by early 2027. IREN stated that the $16 billion purchase price covers all infrastructure equipment and computing system services, including core AI GPU products, server equipment, storage components, high-performance networking infrastructure, auxiliary equipment, integration services, and warranty services, with payments due upon delivery.
Daniel Roberts, Co-CEO of IREN, stated, "In a market where time-to-compute is everything, ensuring sufficient AI computing infrastructure capacity and accelerating its deployment into operational models is our top priority." Additionally, NVIDIA, the world's highest-valued company and the dominant force in AI chips, announced earlier this month that it would invest up to $21 billion in IREN as part of a broader transaction to deploy up to 5 gigawatts of AI computing infrastructure. IREN has agreed to grant NVIDIA a five-year warrant, allowing it to purchase up to 30 million shares at an exercise price of $70 per share. The two parties will also jointly deploy tens of billions of dollars worth of computing power. As part of the agreement, NVIDIA will invest up to $21 billion in IREN. NVIDIA and IREN plan to jointly deploy large-scale AI data centers, leveraging NVIDIA's exclusive computing equipment and IREN's expertise in securing land, power infrastructure, and other construction elements.
Last November, Microsoft, one of the dominant players in U.S. software and cloud computing, entered into an agreement with IREN worth up to $97 billion, which includes priority access to NVIDIA's most advanced AI computing clusters. The transition from pure "mining" to AI data center campuses represents Bitcoin mining companies riding the AI super-cycle. IREN is essentially a power-intensive data center operator that started with cryptocurrency mining infrastructure, with core capabilities in securing large-scale power, building high-density facilities, and operating Bitcoin mining equipment. Now, it is applying these "power + facility + operations" capabilities to AI cloud computing.
The logic behind IREN's transformation is clear: Bitcoin mining companies already possess power access, land, cooling, grid interconnection, and experience operating large-scale server clusters—precisely the engineering resources most scarce for AI data centers. Therefore, IREN's shift from mining to AI data centers is not merely a change of track but a repurposing of its existing energy and computing infrastructure into AI cloud capacity. The explosive growth of generative AI applications and AI agents, which are set to significantly boost human productivity, is driving a sustained surge in AI computing demand, forcing tech giants to massively build and expand AI data centers. Investors are clearly taking note, pouring funds into Bitcoin mining companies transitioning to AI data centers. IREN's stock has risen approximately 60% year-to-date and a staggering 285% for the full year 2025.
Companies once focused on mining cryptocurrencies like Bitcoin are now securing long-term contracts with tech giants for their massive AI workloads, leveraging their land, power resources, and data center computing infrastructure. For these mining companies, computing infrastructure once dedicated to mining is perfectly suited for the data center computing resources required for AI training and inference. Moreover, they already have grid interconnection points, substations, land resources, and other infrastructure like power distribution, enabling them to adapt to massive AI workloads in an extremely short timeframe.
Microsoft's approximately $97 billion AI computing procurement agreement with IREN last year involved NVIDIA's most advanced AI GPUs. NVIDIA, this month, announced an investment of up to $21 billion in IREN as part of a deal to deploy up to 5 gigawatts of AI infrastructure. This signifies the formation of a closed loop among cloud computing giants, GPU suppliers, and new AI data center operators: Microsoft invests heavily in computing resources, IREN delivers cutting-edge AI computing clusters, Dell provides AI server systems, and NVIDIA supplies the underlying AI chips and capital support.
Considering the latest revenue data from Anthropic, a global leader in AI applications, the demand side of this chain is exceptionally strong. Anthropic anticipates Q2 revenue reaching $10.9 billion, doubling from Q1's $4.8 billion, and may achieve $559 million in quarterly operating profit. It has also agreed to pay SpaceX $1.25 billion monthly for computing resources, indicating that frontier AI companies' demand for AI training and inference computing resources, along with the computing consumption driven by the programming AI agents and enterprise AI tools they launch, is exploding.
Thus, IREN's purchase of Blackwell, Microsoft's early reservation of AI GPU cloud capacity, and NVIDIA's investment in a data center operator essentially answer the same question: AI computing demand is now strong enough for the market to lock in data center computing resource supply for the period around 2027 in advance. Morgan Stanley notes that the AI computing arms race has entered a system-level expansion phase. The institution has significantly revised its 2026 capital expenditure expectations for major U.S. tech giants upward from $433 billion a year ago to $805 billion, with 2027 capital expenditures potentially reaching $1.1 trillion, up from a previous forecast of $950 billion. This latest projection underscores that supply chain bottlenecks at the AI computing infrastructure level have expanded comprehensively from "mass purchasing of GPUs/ASICs" to "striving to simultaneously address the entire AI data center delivery process, including data center power equipment, liquid cooling, data center CPUs, DRAM, NAND, HBM, optical communication/interconnect, high-performance network interconnect, transformers, gas turbines, and more."
Comments