Summary IREN is rapidly transforming from bitcoin mining to AI and HPC data center infrastructure, leveraging its energy assets for higher-margin growth. AI services revenue at IREN surged 137% quarter-over-quarter to $17.3 million, with gross margins of 86%, while bitcoin revenue and margins declined. IREN projects $3.4 billion in ARR for 2026, with $2.3 billion already under contract, anchored by a major Microsoft deal used as financing collateral. Expansion plans target 4.5 GW installed capacity, positioning IREN to capture significant AI infrastructure demand beyond the current 810 MW. As a bitcoin ( BTC-USD ) investor since 2018, when I first bought it driven by the belief in the future of digital transactions and its acceptance as a medium of exchange, I have a strong interest in the industry. This extends not only to cryptocurrency mining companies, but also to other cryptocurrencies operating in the market. However, that personal bias will no longer cloud my investment decisions (or at least that's what I hope). Today I choose to reward the transformation of cryptocurrency miners that seek to move into the AI and HPC data center business, leveraging their existing energy infrastructure. As I explained in previous articles, my view is that the most lucrative segment of the AI business is not in the data centers themselves but in the companies that provide both the infrastructure and the power required to operate them. To put it in numbers, for every $1 billion invested in data centers, $125 million of investment in energy is needed. Based on projected 2025 investment in data center equipment of $475 billion, that would imply an additional $56.25 billion in energy related investment. And in the same sector, we find bottlenecks that represent a significant opportunity: projects worth more than $750 billion could be delayed through 2030 due to these constraints, according to Coface . On that path, full of challenges and large capital investments for the reconversion, I believe IREN Limited (IREN) is one of the most stable and profitable companies among its peers. I have several reasons to think this way. One of the reasons, in my view, is the speed of adaptation. IREN appears to be well positioned in this race (perhaps not by much, but clearly in the areas that matter most). That adaptability is evident in something as simple as reporting a dedicated revenue line for AI cloud services. It may seem like something minor, but it isn't for me. For example, crypto miners transitioning towards this model, such as Riot Platforms ( RIOT ), Bitfarms ( BITF ), Cipher Mining (CIFR), or CleanSpark ( CLSK ), don't yet have effective recurring revenue. Companies with more advanced infrastructure, including Core Scientific ( CORZ ), MARA Holdings (MARA), and Hut 8 Corp. ( HUT ), have already reported revenue from AI hosting. However, the growth trend at IREN can be seen in Q2 FY26 , the most recent quarter presented at the beginning of this month: AI services revenue increased 137% (from $7.3 million to $17.3 million) compared to the previous quarter, while bitcoin revenue fell 28% (from $233 million to $167.4 million). The gross margin for AI services was 86% versus 62% for mining Bitcoin. Next I'll show you the trend in AI services compared to bitcoin mining over recent quarters. Author's Tabulations What I liked the most about IREN and its transformation is its 2026 projection to reach $3.4 billion in ARR, driven by a planned expansion of 140,000 GPUs. Of that $3.4 billion, $2.3 billion is already under contract. IREN's largest customer is Microsoft ( MSFT ), which announced a $100 billion capital investment plan through 2026 in its latest earnings presentation. Microsoft signed a deal totaling $9.7 billion, with $3.6 billion in financing secured under that contract at an interest rate below 6%. IREN What I liked most of Q2 FY26 The details of the agreement with Microsoft were explained in more detail during the Q2 FY26 earnings conference call. This is something that excites me, especially because it is a contract that has already been used as collateral for credit financing. I mean, there is a real long-term revenue projection behind it. But beyond Microsoft, I wanted to analyze the Q2 FY26 in detail, excluding non-recurring items, which led to a net loss of $155.4 million compared to the $384.6 million in the previous quarter. This was mainly due to a non-cash loss on financial instruments of $107.4 million, compared to an accounting gain of $665 million in the previous quarter. This significantly impacted the quarter's final result, along with other non-cash expenses related to debt conversion of $111.8 million. I've highlighted these items in the image below. IREN I also noted the $31.8 million impairment charge related to mining hardware due to the ongoing transition from ASICs to GPUs, specifically in British Columbia. When these temporary and often volatile accounting items are excluded, I believe adjusted EBITDA becomes more relevant. Adjusted EBITDA was $75.3 million, down from $91.7 million in the previous quarter, but up from $62.4 million in the same period last year, as shown on slide 25 of the earnings presentation . Beyond the accounting noise, the drop in bitcoin revenue also impacted the results, falling 23% quarter-over-quarter to $167.4 million. Bitcoin, in addition to the price decline so far in 2026, didn't show encouraging signs in the last quarter of 2025 either (not only on the revenue side, but also due to increased mining difficulty, which reached 155.9T last November). Price dynamics also played a role, falling from a high near 124k in October to the 88k-90k range in December. Now the bearish scenario is deeper, with prices below 70k. The decline in the last quarter of 2025 affected IREN, as it did all mining companies. In addition, the company reported a higher net electricity cost per bitcoin, rising to 34.4 compared to 21.4k in the same period of the previous year. Although bitcoin production increased from 1,347 to 1,664 over the same period, margins shrank due to the lower price. IREN My disillusionment with Bitcoin and the IREN's shift Bitcoin price estimates for 2026 remain highly dispersed between minimum and maximum projections. So far this year, it has fallen 21%, in line with broader 2026 market trends, where large technology companies are also experiencing corrections and there appears to be a rotation of capital toward more defensive sectors. Data by YCharts Last January, some analysts projected a minimum price of $75,000 for 2026, but now others are venturing to set a floor of $40,000 . I don't like to estimate commodity prices or try to predict the future (and Bitcoin isn't a commodity either). I don't like playing that game in finance. What I do prefer is understanding trends. In this case, my analysis is focused on identifying the long term trend. As I explained at the beginning of this article, I first invested in Bitcoin in 2018, but recently I concluded that it has failed to establish itself as a currency, at least not in the short term or within the timeframe I had imagined. Nor is it a metal that can be demanded by industries or held as a central bank reserve asset. I believe it functions more as a risk asset, driven by expectations of greater or lesser liquidity, and as a kind of global financial thermometer. There are other factors, such as uncertainty about the stance of the new Fed Chair or the heavy selling in the technology market so far this year, that could also negatively affect Bitcoin. The downward trend in Bitcoin, then, could be a problem for IREN's profitability in the short term. However, it's not entirely negative. On the contrary. In my opinion, IREN is progressing correctly in its conversion process. Of course, it would still benefit from strong income from Bitcoin mining for some time. But an ARR of $3.4 billion for 2026, I believe, validates the investment in fundamental infrastructure within the AI value chain. In my opinion, the most important segment is data centers. That's why I'm offsetting my disillusionment by investing in crypto miners that are making a strong bet on hosting the next generation of data centers. In the case of IREN, this is supported by its strong agreement with Microsoft, expectations of new contracts with hyperscalers, an upward trend in AI services revenue, and a solid portfolio of installed capacity. IREN's portfolio on its expansion plan towards 4.5 GW The $3.4 billion ARR projection would represent the use of 460 MW, which would imply only 10% of IREN's estimated future installed capacity of around 4.5 GW. Currently, installed capacity stands at 810 MW, but with the construction of two large facilities, the company expects to significantly increase its capacity, in line with other competing companies that have long term expansion plans. IREN These two large facilities are located in Texas and Oklahoma. In Texas, the facility is called Sweetwater and will include a 1.4 GW site, expected to be completed this year, as well as another phase projected for 2027, with an additional 600 MW. Texas will, then, represent 2 GW of capacity, with a direct connection to the ERCOT grid. Texas has been a key state for long term data center projects. It is also the state with the highest number of data centers in the country , followed only by Virginia. RIOT, while it doesn't have the same aspirations for a full transition to AI data centers, also bases its business transition on two large facilities in Texas, located in Corsicana and Rockdale. The other key state for IREN is Oklahoma, which would represent the remaining 1.6 GW following the acquisition of a new 200-acre site where a new data center will be built, also reflecting a healthy energy diversification. The power potential of nearly 4.5 GW positions IREN on a key hosting platform for AI data centers. It wouldn't be just any hosting provider; it would be the most powerful. I analyzed the business of the eight largest cryptocurrency miners undergoing this industry transformation. And IREN's capacity is the largest. Company GW IREN Limited 4.5 HUT 3.8 Cipher Mining 2.4 Bitfarms 2.1 CleanSpark 1.8 MARA 1.8 RIOT 1.8 Core Scientific 1.3 Each installed capacity figure includes its corresponding source link, based on the company's future projections. In HUT's case I was actually very optimistic, as I included 1.25 GW from exclusive site agreements that hadn't yet been officially announced. If I had excluded that capacity, HUT's projection would be 2.6 GW. So, IREN holds the strongest projected position in infrastructure terms. I believe the deal with Microsoft already reflects that leadership position. This is not a minor detail, because if Microsoft is only using 10% of the projected capacity, that implies significant room for additional contracts. Why would IREN build a 2 GW hyperscale data center in Texas, such as Sweetwater, without secured customers? Microsoft alone, in my view, is sufficient to support baseline profitability. Of course, I don't expect the company to rely on a single client. But I believe this is not a bad starting point. This concern was raised during the earnings call , and CEO Daniel Roberts responded that they are waiting for the right moment and the right customer to maximize long-term value. I consider this to make sense because there are long-term contracts that should guarantee stable revenue to offset the decline in bitcoin mining and ongoing capital investment. They should also be sufficient to generate solid long-term profitability. The market has yet to price in IREN's potential The deal with Microsoft and the projections to reach a solid medium term position in installed capacity have, I believe, already rewarded the company. In fact, IREN's stock price broke its historical high after the Microsoft contract was made public in early November, reaching a level above $75. In the last quarter of 2025 the stock lost some momentum due to the decline in Bitcoin prices. So far this year, IREN is down 1%, after a brief rally ahead of the earnings release at the beginning of the month. The post-earnings selling selloff, which I believe was driven by missed expectations and the non-cash costs I previously discussed, created an attractive correction and a more compelling entry point. Furthermore, the stock now trades at 30x earnings , almost in line with the industry average of 29.9x. When compared with some of the companies I've mentioned before, IREN trades at a relatively similar P/E ratio. HUT's P/E stands at 28.7x, while RIOT's is 37.8x. There are other companies in the sector, but since they are not profitable, the ratio can't be calculated. In this regard, IREN holds a relative advantage, as it remains profitable despite reporting net losses in the most recent quarter. What I'm trying to say about IREN is that, despite having the largest market capitalization and the greatest installed capacity among mining companies transitioning toward AI hosting, the stock doesn't appear to trade at a meaningful premium. One more piece of data: IREN's forward revenue growth of 147.3% is the highest among its peers. The closest competitor I found is CIFR, at 49.9%. Ergo, I believe IREN could revisit its November peak in the $70-$75 range, once it begins to realize its 2026 ARR projections and further monetizes its AI related revenue streams. Execution risk in the AI transition Since IREN aims to lead the market in installed power capacity for AI data centers, at least within the cryptocurrency mining industry, the risks associated with this strategy relate to execution timelines, capital costs, and profitability during the transition period. Fortunately, IREN maintains a strong financial position, with $3.26 billion in cash at the close of Q2 FY26, and relatively stable adjusted EBITDA. However, if the decline in Bitcoin deepens, margins could narrow further in the short term, putting pressure on cash flows while capital expenditures for new facilities continue. Ergo, bitcoin remains a relevant risk factor for the company. The operational transition and the buildout of digital infrastructure to establish an AI hosting platform may continue to generate volatility in non-monetary cash items, including potential hardware asset impairments, as observed in the last quarter. Such accounting adjustments could lead to earnings volatility and, at times, result in EPS coming in below consensus expectations. During this transition phase, as I've said before, I think it will be important to closely monitor adjusted EBITDA, particularly this year, as the Sweetwater facility is expected to be energized by the second quarter. The longer IREN takes to monetize its AI-related revenue streams, the greater the risk of eroding market confidence, while simultaneously exposing itself to competitive advances. I'm not concerned about whether IREN will be able to monetize its infrastructure or not. I simply want to add the risk of delaying that monetization. Conclusions IREN leads projected future installed capacity with 4.5 GW, ahead of its peers. The company has secured a long-term contract with Microsoft for $9.7 billion, with $2.33 billion in ARR already contracted, and an estimated $3.4 billion in ARR projected for 2026. His gradual exit from the bitcoin-focused model should be approached with caution, which is why I have incorporated this into the risk assessment. But I believe this transition represents the appropriate strategic shift, moving from volatile revenues tied to an intangible asset, toward more stable, long term revenue streams with hyperscalers (currently led by Microsoft). The planned 2 GW projected for Sweetwater in Texas, signals IREN's capacity to secure additional large scale contracts. Moreover, the 137% quarter-on-quarter increase in AI services revenue reinforces the view that this is only the beginning of the company's expansion. Based on all the factors discussed, I rate IREN as a Strong Buy, thinking of it as a long-term investment.
