Summary I see IREN Limited as a speculative buy, driven by its pivot from crypto mining to AI data center compute, with a 10x ARR growth in this segment. IREN's AI Cloud business targets $500M ARR by Q1 FY26, with significant GPU investments and strong contracted power capacity. At the end of 2024, ARR was only $32M. Near term, I believe risk is elevated: shares are overbought since early September, valuation is rich at ~154x P/E, and Q1 FY26 could spark profit-taking from short-term traders. I rate this stock as a buy for a 3-5 year horizon. Funding is secured for GPU expansion, with $875M convertible notes recently announced at an attractive conversion price premium. At the moment of writing this article, I am ultra bullish on two themes: anything-AI, and, I thought I would never in my life say it, Bitcoin (yikes!). Therefore, I recently tuned my screeners, and IREN Limited ( IREN ) came up from nowhere. I say "from nowhere" because I am quite reluctant to invest in anything outside of the US. According to the last 20-F , IREN is a Sydney-based company, exposed to both the crypto space and, more recently, the AI-data center space: Our data centers are purpose-built for power dense computing applications and currently support a combination of ASIC’s for Bitcoin mining and GPUs for HPC solutions (including AI Cloud Services). In this article, I review the recent pivot, which will hopefully explain the massive bull run (that I completely missed due to my reluctance to look outside of the US). At the current price level, I believe IREN is a speculative buy. On the positive side, the ARR growth of its AI cloud business is impressive, from just $32M at the end of 2024 to $500M by the end of this year. On the negative side, I believe any news will act as a catalyst for traders to take profits. That's why I would rather wait until Q1FY26 earnings to see the market's reaction following the results. On a longer horizon, let's say 3-5 years, I believe this stock is a buy, and any pullback now will likely be just noise. How The Pivot Unfolded Back in 2024, when IREN was actually called Iris Energy, the company announced its very first cloud-services deal with Poolside for a modest 248 NVIDIA H100s. Just a few days later, management tripled the AI fleet to 816 H100s to meet future demand. The early validation signs of that future demand came in April 2024, with Poolside upsizing the contract to 504 H100s, citing performance and reliability as one of the main drivers. By H2 2024, the pivot was clear. The company rebranded to IREN to reflect the commitment to a multi-year, next-gen data-center focused business. They began GPU buildouts in Texas (Childress) and Canada, purchased 1,080 H200s, and flagged by October 2024 an AI run-rate of $32 million. Looking at 2025, Q2 FY25 (reported in February, as the company is not following a normal calendar year to report financial results) showed total revenues of $119.6M, out of which $113.5M were from Bitcoin mining. AI revenue was a modest $2.7 million. The big announcement in Q2 was Horizon 1, a 75 MW liquid-cooled DC at Childress, with a 50MW IT load, and new Sweetwater 2, a massive 600 MW site to form a 2 GW hub with Sweetwater 1. In Q3 FY25, the company reported $148.1M in revenue, out of which $141.2M came from Bitcoin mining. AI Cloud revenue was a modest $3.6M, still far from the $32M run rate projected in October 2024. The surprise came in on August 28, when the company reported FY25 results. Revenue was $501M, a 168% yoy increase, with adj. EBITDA sitting at $269.7M, while the net income was $86.9M. Notably, the company reported 2,910 MW contracted grid power (+35% YoY) and 810 MW of operating data center capacity (+212% YoY). This was enough for investors to bid up the share price, breaking the historical resistance and leading to a massive bull run, as shown below. Trading View On the ARR side, the company reported "approaching $1.25 bn total annualized revenue with scope for further growth ahead". Out of that, about $200-250M comes from AI Cloud, with management expecting 10.9k live NVIDIA GPUs by Dec 2025. On September 22, the company doubled its AI Cloud capacity to 23,000 GPUs. The new orders include the " crème de la crème " in AI accelerators, including 7,100 B300s, 4,200 B200s, and 1,100 AMD MI350X. Furthermore, the company raised its AI Cloud ARR target to over $500m by the end of Q1’26. So, Is IREN A Buy? I believe starting a long position in IREN at the current price level is risky for investors who have a shorter horizon. The stock is in overbought territory since the start of September, so a profit-taking pullback is likely on the horizon. That said, the company announced today $875M in 0.00% convertible notes due 2031. This is great news for two main reasons. First, it gives the company liquidity to deploy GPUs, so the few skeptics who wondered how the company would be capable of funding 7,100 B300s, 4,200 B200s, and 1,100 AMD MI350X, now have a compelling answer. Second, the initial conversion price was $85.63 (or 42.5% above the close on October 8). That's a very bullish sign. In fact, the stock didn't sell off following the announcement and is now up low single digits on the premarket session of October 9. I already discussed why such a high premium on the conversion price is bullish for the stock in one of my articles on Nebius. With Q1 FY26 earnings yet to be disclosed and the company announcing just a few days back new AI cloud contracts, reiterating the $500m AI cloud ARR, I believe the positives are priced in. From my experience as a market practitioner, I believe any announcement could be a reason to sell, regardless of the flavor (positive or negative). Therefore, considering my shorter investment time frame (3-24 months), I prefer to wait and see. I added this stock to my watchlist, and I will strongly consider adding a small speculative position in any future pullback. That said, if I had a multi-year timeframe of 5-10 years, I do believe the company is a buy, and any pullback in the near term will be just noise on a longer investment timeframe. Why? Simply, due to the high demand for computing, which is significantly larger than the current supply. Furthermore, the company has 2,910 MW contracted with utilities, which is a massive upside considering that the bottleneck of GPU deployments is power. Conclusion Overall, I would treat IREN with caution. Despite my buy rating, which considers a multi-year timeframe of 3-5 years, I believe a near-term pullback is possible during the next earnings release, as all positives may already be priced in. On a valuation basis, the company is trading at a significant premium, as the excitement following the pivot from crypto mining to data center computing has driven the multiples up. On a P/E basis, the stock is trading at 154x earnings, which is expensive to say the least. That said, I would strongly consider adding this stock to my portfolio in any pullback. Q1 FY26 could be a catalyst for some traders to take profits, as the company has pretty much confirmed in a recent press release the ARR goals ($500M) for its AI cloud segment. Any small disappointment could lead to a selloff, which is the main reason I grade this stock as a speculative buy.
