From Mining to AI Cloud: A Reassessment of Mining Firm’s Computing Power and Valuation Restructuring Behind Iris Energy (IREN)’s $9.7 Billion Contract

In 2026, the meaning of the term “Bitcoin mining companies” is being redefined.

The traditional narrative has always revolved around a computing-power race: the higher the hash rate, the more mining output, and the higher the company’s valuation. But this logic has loosened fundamentally over the past two years. As the pace of data center construction in North America has fallen far behind the expansion of AI computing demand, Bitcoin mining farms—once viewed as “commodity-like capacity”—have suddenly become the most ideal landing platform for ultra-large-scale AI data centers. They have power interconnection that others can’t obtain; substations that other parties need four years to complete are already in place; land approvals that typically take five or six years from zero site selection have long been completed.

Based on this, the market is repricing a batch of targets. Bitcoin mining companies such as Iris Energy, Core Scientific, and Hut 8 are switching the electricity and data center resources they hold from the Bitcoin network to AI computing clients. The changes this switch brings to financial structure, capital expenditures, and valuation models are no longer mere conceptual hype—they are empirical propositions supported by contracts in the billions of dollars.

Transformation Backdrop: Electricity Becomes the Most Scarce Resource for AI Computing, While Mining Companies Sit on Ready “Powered Facilities”

To understand Iris Energy’s current narrative logic, the prerequisite is to understand the real supply situation in the North American data center market. Bernstein’s research shows that the median time to secure 1 gigawatt of grid power access in the United States is about 50 months. Even in relatively data-center-friendly regions such as Texas, utilities handle applications in batches, causing actual waiting periods to far exceed users’ expectations. Meanwhile, there is a huge gap between the contracted hosting capacity for ultra-large-scale data centers and the actual volume delivered after construction—large volumes of demand are locked on paper but cannot be converted into usable IT capacity in the short term.

In contrast, publicly listed Bitcoin miners collectively control more than 27 gigawatts of planned power capacity. Currently, only about 3.7 gigawatts is associated with disclosed AI contracts, corresponding to contract values exceeding $90 billion. In other words, most miners’ power capacity has not yet been discovered by the AI market—and this portion of capacity happens to be at the very moment when AI data centers most urgently need large-scale expansion.

From the perspective of revenue structure, the economic drivers of the transition are also very clear: in Bitcoin mining scenarios, 1 megawatt of power capacity generates annual revenue of approximately $57 to $129, while the same 1 megawatt used for AI GPU hosting or cloud services can generate $200 to $500. In this transition, power resources are no longer simply computational fuel; they become core infrastructure assets that can be directly monetized.

Bernstein analysts, when recently initiating coverage of multiple miners, referred to them as “AI power landlords.” The industry consensus behind this label is that in the AI computing competition, supply-chain bottlenecks have shifted from chip manufacturing to power interconnection. Mining companies, leveraging their existing brownfield site selection, grid connections, and substation resources, stand at this structural inflection point. Over the past two years, the entire industry has outsourced 6 gigawatts of power production capacity to ultra-large cloud service providers and new cloud operators, involving 17 transactions with a total value exceeding $110 billion—about 10% of the currently under-construction data centers across the United States. Bernstein expects the total AI revenue of the Bitcoin miners it covers to grow from $1.2 billion in 2026 to $10.7 billion in 2030.

Iris Energy: A Real-World Playbook From a Bitcoin Computing Leader to an AI Cloud Service Platform

Among the most aggressively transforming Bitcoin mining companies, Iris Energy is one of the cases with the most complete contract lock-in and the clearest capital structure.

On core contracts, in late 2025 IREN signed a five-year AI cloud service contract with Microsoft with an aggregate value of approximately $9.7 billion. Microsoft will pay IREN 20% of the customer prepayment. The contract is expected to contribute approximately $1.9 billion in annual recurring revenue, with an estimated EBITDA margin of about 85%. As capital backing for this contract, IREN also signed a procurement agreement with Dell Technologies worth approximately $5.8 billion for GPUs and related equipment. The funds are used to purchase NVIDIA GB300 GPUs, and they will be deployed in phases in the form of four liquid-cooled data centers at IREN’s 750 MW campus in Childress, Texas, with the expectation that the main deployments will be completed within 2026.

Capital structure is a key dimension to validate the feasibility of IREN’s transformation. For the above purchases, IREN locked in approximately $3.65 billion of investment-grade debt financing, compressing the blended borrowing cost to 6.00%. Together with Microsoft’s $1.9 billion in customer prepayment, the two are expected to cover about 96% of the GPU capital expenditure related to the contract. In addition, IREN signed a GPU cloud expenditure agreement with NVIDIA valued at approximately $3.4 billion. The arrangement between the two also includes a call option for NVIDIA to purchase up to 30 million shares of IREN at a price of $70 per share within five years. As of the end of Q1 of fiscal 2026, IREN had cumulatively obtained more than $9.2 billion in financing sources, covering multiple channels including customer prepayments, convertible bonds, GPU leasing, and GPU financing. At quarter-end, cash and cash equivalents were approximately $2.6 billion.

On capacity construction, IREN has locked in more than 4.5 gigawatts of grid-interconnected power. Of the newly obtained Oklahoma campus land area of approximately 2,000 acres with capacity of 1.6 gigawatts, the grid studies have been completed, and the company expects phased power connection to start from 2028. The company plans to complete the deployment of 140,000 GPUs by the end of 2026, which is expected to generate approximately $3.4 billion in annual recurring revenue.

In terms of performance, some data for fiscal 2026 Q2 (as of December 31, 2025) has been publicly disclosed. Total revenue for the quarter was $184.7 million, down 23.1% quarter-over-quarter from $240.3 million in the prior quarter, and below analysts’ expectations. Bitcoin mining revenue was $167.4 million, while AI cloud service revenue was $17.3 million, up 137% quarter-over-quarter from $7.3 million. Net loss for the quarter was $155.4 million, mainly driven by non-cash and non-recurring items, including an unrealized loss of approximately $219.2 million on convertible bonds and debt conversion costs, $31.8 million in impairment charges for ASIC miners, and $58.2 million in stock-based compensation, partially offset by $182.5 million in tax benefits. Adjusted EBITDA was $75.3 million, with adjusted EBITDA margin expanding from 38% in the prior quarter to 41%. At quarter-end, the company held approximately $3.26 billion in cash and cash equivalents.

It is worth noting that the increase in the size of the net loss for the quarter was mainly driven by accounting treatment rather than a deterioration in operational fundamentals. However, from the standpoint of free cash flow, the funding pressure during the expansion period is worth attention. Free cash flow for only Q1 2026 was -$1.38 billion, which means IREN’s further expansion is highly dependent on continued external financing and customer prepayments.

On analyst pricing, as of early June 2026, IREN’s stock price was approximately $66.60. Based on the 12-month average target price of 14 analysts of $80.49, with a high target of $126 and a low target of $41, the overall consensus rating is “Buy,” consisting of 10 buy ratings, 3 hold ratings, and 2 sell ratings. On June 4, 2026, Bernstein set IREN’s target price at $96.00, after having previously issued a $100 target price.

It should be pointed out that IREN’s transformation progress is highly dependent on the on-time delivery of its planned 140,000 GPU deployment schedule and the continued absorption of computing capacity within Microsoft’s contracts. The management team needs to exercise prudent control over debt and cash flow during this rapid expansion. Any delays in chip delivery, delays in grid interconnection, or weakening AI computing demand could affect how the market prices this narrative.

Benchmarking the Observation: Differentiated Paths of Core Scientific and Hut 8

Iris Energy is not the only company taking this route. Core Scientific and Hut 8 are transforming into the AI data center market with different asset structures and business models. While the three represent three different configuration logics, the core resources they share—electricity, land, and existing data center facilities—are the same.

Core Scientific has chosen a relatively “light-operations” path. It positions itself as an AI data center hosting provider and leases computing capacity to ultra-large-scale customers rather than operating its own GPU cloud services. In fiscal 2026 Q1, its AI hosting business revenue reached approximately $77.5 million, up nearly 9x year-over-year. For the first time in its history, it surpassed Bitcoin mining revenue. Bernstein raised Core Scientific’s target price from $24 to $32, maintaining a “beat the market” rating, and estimated that about 86% of the company’s enterprise value comes from its AI business, with only about 14% attributed to crypto mining.

Hut 8 went even further. In May 2026, Hut 8 announced that it had signed a 15-year AI factory lease agreement with a high-investment-grade tenant for 352 MW of IT capacity. The base contract value was approximately $9.8 billion, with an annual base rent increase clause of 3.0%. The contract is structured as a triple-net lease with a “pay-as-you-go” mechanism, and the tenant is a confidential high-investment-grade company. Together with the earlier 245 MW contract for the River Bend campus, Hut 8’s total signed AI data center capacity reached 597 MW, with a cumulative base contract total value of approximately $16.8 billion. On the capital side, Hut 8 successfully issued $3.25 billion of senior secured notes with a 16.5-year maturity—an industry first for the data center sector: project financing completed through the investment-grade bond market by a single initiator. The company also disclosed a development pipeline exceeding 8,375 MW. Hut 8’s stock price rose significantly during 2026, and the latest analyst target price is around $124.

From a valuation perspective, Bernstein provides a quantifiable reference: Bitcoin miners with signed AI contracts have an enterprise value of about $600,000 per planned MW, while pure mining companies that have not yet obtained AI contracts are valued at about $300,000 per MW. The roughly twofold valuation premium fundamentally reflects the market’s response to certainty in contracted deals. Power capacity locked under long-term contracts is viewed as scarce infrastructure capable of generating stable cash flows, whereas resources without contract anchoring remain exposed to both Bitcoin price volatility and broader network hash-rate pressure.

How to Directly Allocate AI Computing Track Assets Through Gate Stock Services

For investors focusing on AI infrastructure themes, the aforementioned transformation targets—such as Iris Energy, Core Scientific, and Hut 8—are now tradable and allocatable directly through the Gate platform, without the need to open separate accounts with traditional brokers.

Gate recently announced a strategic partnership with Alpaca, a leading global brokerage infrastructure company, to roll out real stock trading services for eligible users. The service covers more than 10,000 US-listed stocks and ETFs traded on the New York Stock Exchange and NASDAQ. Users can trade stocks and ETFs directly using USDT, with fractional-share trading supported with a minimum initial investment of just $1. As a partner, Alpaca will handle core processes including order execution, clearing, settlement, and asset custody, and it also supports dividend distributions and corporate action handling.

The core advantage of this model is that: users’ USDT can be used directly for US stock trading without repeated fiat currency conversions, and without the need to complete KYC again or open a separate brokerage account. The user interface remains consistent with crypto asset trading. Both market orders and limit orders are available, and the trading experience has almost no learning cost compared with daily spot trading.

In addition, Gate also provides tokenized stock products through its TradFi module, using USDT to participate in asset transactions linked to US stock prices as well. No matter which approach is used, users can complete unified allocation of crypto assets and traditional securities within a single account.

For users holding stocks such as IREN, CORZ, and HUT, Gate supports minimum purchase thresholds as low as 0.01 shares and $1, greatly lowering the cost of allocating each individual asset—making it possible to build positions through small-scale testing without taking on the initial capital pressure of buying whole shares.

Conclusion

The transformation of Bitcoin mining companies into AI computing infrastructure in 2026 is no longer just an industry hypothesis; it is an empirical proposition supported by hundreds of billions of dollars in public-company contracts and quarterly financial reporting data. In this structural change, Iris Energy—backed by its $9.7 billion AI cloud contract with Microsoft and its 140,000 GPU deployment plan—has one of the clearest and most complete AI computing expansion frameworks at present. Core Scientific rapidly shifts its revenue structure through hosting, while Hut 8 lays the groundwork for long-term capacity expansion with massive development pipelines and investment-grade project financing. The common value pivot of the three lies in power and land—elements traditionally regarded as resource-heavy real assets—which, in the AI era, are being transformed into measurable scarce computing supply.

From a valuation standpoint, analysts’ pricing models are also undergoing fundamental adjustments: valuations no longer anchor to Bitcoin price or network-wide computing difficulty; instead, they focus on megawatt scale, contract cycles, and the ability to lock in investment-grade customers. For investors following this track, the quarterly GPU deployment progress, the signing cadence of major AI leasing contracts, and the depth of collaboration with ultra-large-scale customers will reflect changes in these companies’ intrinsic value more than simply tracking Bitcoin price movements. On the practical execution level, through Gate’s stock services and tokenized asset products, the asset allocation process for this wave of AI computing infrastructure development has been greatly simplified.