IREN Ltd
- Market cap
- 15.25B
- P/E (TTM)i
- -17.43
- P/Bi
- 3.64
- EPSi
- -2.22
- Div yieldi
- 0.00%
- 52W posi
- 20%
Anonymous reader poll. Unscientific, not investment advice.
Valuation each multiple against its own 5-year range
Vs. peers Capital Markets
| Company | Market cap | P/E (TTM)i | P/Bi | Div yieldi |
|---|---|---|---|---|
| IREN Ltd (IREN) | 15.25B | -17.43 | 3.64 | 0.00% |
| Morgan Stanley (MS) | 297.95B | 15.32 | 2.80 | 2.11% |
| Goldman Sachs (GS) | 258.33B | 13.70 | 2.35 | 1.92% |
| Charles Schwab (SCHW) | 165.29B | 17.41 | 3.76 | 1.23% |
| Robinhood (HOOD) | 98.46B | 48.46 | 10.39 | 0.00% |
| Interactive Brokers (IBKR) | 39.75B | 34.82 | 6.73 | 0.37% |
Other StockVane-tracked companies in the same industry.
Morningstar
Trading 11.1% below Morningstar's fair value estimate.
Analyst note
IREN is making progress with its artificial intelligence cloud transition. Quarterly AI cloud revenue of $71 million surpassed bitcoin mining revenue of $67 million for the first time. It also landed a multiyear deal with a frontier AI lab and delivered the 50-MW Horizon 1 data center to Microsoft.
Why it matters: IREN is closing its service delivery gap with leading neocloud providers. The latest management update confirmed robust demand trends disclosed by peer neoclouds in their earlier earnings calls. Pricing for 3-year and 5-year client contracts increased by 125% and 70%, respectively, over the past nine months. Customer prepayments are funding 45%-55% of GPU expenditure, but higher customer prepayment means IREN is putting in less equity, which can bring excessive leverage risk. We believe management's two-year GPU payback period estimate might have interpolated short-term elevated GPU pricing to the entire duration of the GPU's useful life. We think a payback period of between four and five years is more realistic based on midcycle rates.
The bottom line: We raise our fair value estimate for no-moat IREN to $43 per share from $41, as the company is catching up with top neocloud firms. Shares look moderately undervalued following the 10% selloff after earnings, which we surmise is linked to the $25 billion-$30 billion capital expenditure projection for fiscal 2027. In our view, the $25 billion-$30 billion capex outlook is reasonable, given that IREN is looking to develop 800 MW of active IT capacity by the end of calendar-year 2027. When we incorporate the latest capex guidance into our model, it hits IREN's free cash flow hard, which explains market concerns. That said, the negative free cash flow impact is offset by higher revenue forecasts from improved revenue per megawatt in the near term and expanded total data center capacity. We expect IREN to completely build out its existing 5 GW secured capacity over the next seven years.
We are glad to see IREN's financing practices maturing, with the company securing its first investment-grade credit facility backed by Microsoft's contract. The 6.0% interest rate is among the industry's best and sufficiently below the 9% return rate benchmark we have for neocloud companies, which means the credit facility should meaningfully boost shareholder returns on the Childress site.
Fair value
Our fair value estimate for IREN is $43 per share, implying a fiscal 2027 enterprise value-to-sales multiple of 4 times. We forecast a 137% five-year compound annual growth rate for the company, which seems very high, but not unreasonable, given IREN’s relatively small revenue base. Revenue growth can accelerate further if more cloud providers purchase capacity from IREN on a wholesale basis, as in the $9.7 billion deal with Microsoft. That said, we do not expect the company’s GAAP earnings per share to turn positive over the next 12 months, and its free cash flow is likely to remain negative for at least another five years.
We expect IREN to fully replace bitcoin miners with GPU clusters at its British Columbia locations by fiscal 2027. For its Childress site, the near-term priority is to deliver Microsoft’s contracted Horizon 1-4 capacity of 200 megawatts by the end of calendar 2026, as well as the 60-megawatt capacity for Nvidia. After that, we expect IREN to land additional bulk capacity sales with major customers for the remaining 400 megawatts. The entire Childress site should operate at its full 750-megawatt capacity before fiscal 2030. IREN has met its goal of energizing the 1.4-gigawatt Sweetwater 1 substation by Spring 2026, and the remaining 600-megawatt Sweetwater 2 substation energization by late 2027 is on track. Overall, we think the company is on track to hit its 800-megawatt IT capacity goal by the end of calendar year 2027 (second quarter of fiscal 2028).
Because IREN mainly competes by providing GPU compute at a lower cost, our long-term adjusted EBITDA forecast of 66% in fiscal 2035 is moderately lower than peers CoreWeave and Nebius, whose unit economics benefit from value-added services such as better customer support and software. Similar to other high-growth firms, our fair value estimate for IREN is largely driven by terminal assumptions. We use a 10% weighted-average cost of capital, an 10% stage II earnings-before-interest growth rate, and a 31% return on new invested capital, consistent with the trend we see in the final years of our Stage I forecast and our terminal forecast for other neoclouds.
Economic moat
We do not think IREN has an economic moat. Having started as a bitcoin miner, IREN will convert all of its bitcoin mining capacity to AI cloud over the next three years and build AI data centers at new sites, too. That said, as a new entrant to the space, IREN is still learning the ropes, and the company’s own cloud service is subpar compared with neocloud leaders like CoreWeave and Nebius. In the near term, IREN’s rich land and power supply in data center hotspots like Western Texas should support a healthy growth trajectory, but it is challenging for us to see a path where IREN develops a lasting competitive advantage that eventually leads to a moat, especially given the very high uncertainty of future demand regarding AI data centers.
There are many benefits to IREN’s decision to pivot from bitcoin mining to AI cloud. For example, AI data centers can bring in up to 10 times the revenue per megawatt compared with bitcoin mining. Also, in the current environment, where demand for AI cloud infrastructure chronically outstrips supply, revenue from AI data centers would not experience the same high volatility as IREN’s bitcoin mining business. Many neocloud companies benefited from the cheap land and power supply they had secured previously for crypto mining, and this is no different for IREN. We believe IREN’s transition from a bitcoin miner to an AI data center operator will enable very fast revenue scaling over the next five years.
IREN provides two types of products to customers: an AI cloud based on Nvidia’s reference architecture and data centers that house customers' hardware to support their cloud infrastructure operations. Given the company’s extremely limited history in cloud infrastructure operations, IREN’s AI cloud lacks value-added services that would allow it to charge premium prices. The company provides Nvidia GPUs in their raw form, and pricing is the only dimension in which IREN can compete with other neoclouds.
We are not necessarily concerned with IREN’s position in a price-based competition because it sits on a massive renewable energy supply across different parts of North America. However, we doubt IREN’s access to low-cost energy can lead to any cost advantage. There are many other neocloud companies with a crypto mining background that can access cheap electricity. More importantly, IREN’s low prices need to compensate for its inferior customer experience, which puts its AI cloud product at a disadvantage. Although IREN’s AI cloud is built on the same Nvidia architecture as other neocloud companies, it has not yet quite nailed nuances in hardware configuration and customer support to let the Nvidia GPUs operate at their best performance, leading to deployment delays and interruptions during training runs. We expect IREN’s AI cloud operational glitches to improve over time, but we see major challenges before the company can deliver an AI cloud on par with leading neocloud companies.
Currently, AI cloud infrastructure is a complete seller’s market. Large language model providers are searching for as much computing capacity as possible in a race to expedite their training and inference workloads. Whatever number of new Nvidia GPUs IREN can secure and bring online will be booked in no time. However, over the long term, we view IREN’s AI cloud as an also-ran that, for the most part, serves as an overflow relief when larger cloud providers experience a demand spike. It is also possible for IREN to exit the cloud business completely and transition into a pure AI data center infrastructure operator.
We think IREN is doing a better job with its colocation and built-to-suit data center business. The skills needed to construct, manage, and deploy bitcoin mines are transferable to AI data centers. This is an area where IREN’s proprietary knowledge for building and running world-class computing facilities can really shine. IREN’s data centers are able to reach a power usage effectiveness of 1.1, which is an industry-leading efficiency level similar to its larger counterparts. In addition, IREN’s existing data center sites in British Columbia and Texas already have the physical infrastructure and power utilities needed for computing, which can save on initial investments.
That said, we think it is unlikely that IREN will earn a cost advantage in data center construction and operation. Lower costs on data center projects usually come from outstanding material sourcing, proper construction management, and efficient use of resources. With only 6 sites under development, IREN’s data center footprint is limited. We need to see more evidence that IREN can replicate a low data center cost structure on a larger scale as the company continues to expand across different geographies.
Bull case
IREN’s access to cheap renewable energy and outstanding data center efficiency allows the company to undercut competitors with low prices to capture more demand.
Landing more wholesale deals like the one with Microsoft can jump-start IREN’s revenue growth and accelerate its data center buildout.
Reduction in the unit price of GPU compute should unlock more use cases around large language models, expanding IREN’s total addressable market.
Bear case
The long-term demand for AI cloud infrastructure is highly uncertain. If IREN is unable to ramp up revenue rapidly, the large amount of convertible notes on its balance sheet will significantly dilute existing shareholder interest.
Customers may hesitate to sign up for IREN’s AI cloud if its service remains subpar compared with other neoclouds, which can lead to lost market opportunities.
IREN’s future expansion plan can fall apart if the company is not able to secure the funds it needs for its cloud buildout.
By Luke Yang, CFA
Quote time 2026-10-08 07:30:15 · For reference only, not investment advice and not tailored to your situation.