IREN, AI Infrastructure Champion!
IREN Investment Case!
By now, everyone and their grandmother knows that the world is experiencing a once-in-a-century AI-driven technological revolution. However, if we want AI to be able to do more than just create silly cat videos, it needs to be properly trained on lots of high-quality data.
Furthermore, for this data to be processed, the world needs to build more data centers than it has ever built before!
Researchers at McKinsey forecast that by 2030, data center capex will reach $7T and data center capacity will almost triple to 219 GW!
To meet this insatiable demand for computing power, capacity will come from clever and previously overlooked places. And increasingly, a trend has emerged of bitcoin mining companies turning into AI data center businesses.
For years, aspiring crypto entrepreneurs built ever larger and more sophisticated bitcoin mining operations. These bitcoin farms went from small basement servers to dedicated data centers with thousands of GPUs, custom-designed cooling equipment, and access to cheap energy.
It so happens that an AI data center also needs thousands of GPUs, cooling, and cheap energy!
Thus, many of these clever crypto mining entrepreneurs are taking their crypto mines and turning them into AI data centers.
One of these crypto miners turned AI data center operators is Iren.
Thanks to strong AI-driven demand for their services, Iren’s stock has risen almost 400% YTD.
Furthermore, just a few weeks ago, the company signed a 5-year $9.7B deal with Microsoft. This is a transformational deal that the company aims to use as a stepping stone towards the AI big leagues.
In this report, I will look at their business model, data centers under construction, capex requirements, valuation, and finish off with a 2030 valuation model.
Let’s begin.
1. Business Model
2. Data Center Locations
3. Microsoft Deal
4. Capex Requirements
5. Valuation
6. Valuation Model
7. Conclusion
1. Business Model
The company was founded in 2018 by Australian brothers Daniel and William Roberts with the idea to build efficient, high-performance bitcoin mining data centers that would incur lower operating costs than competitors.
However, after the AI boom began, it became apparent that AI data centers presented a significantly larger opportunity, thus the company changed direction.
Let’s expand on the current business.
1.1. Bitcoin Mining
Iren has an installed Bitcoin mining capacity of 45.3EH/s, with construction ongoing to increase that to 52EH/s. Plans for further expansion were cancelled to focus on the AI business.
Bitcoin miners use EH/s to measure mining capacity, which stands for exahash per second. The Bitcoin mining process works by performing complicated computations that solve digital cryptographic puzzles. These steps are needed to validate Bitcoin transactions and ensure the security of the network.
As a reward for this task, miners are given newly minted Bitcoins!
With this capacity, Iren was able to mine 2,039 Bitcoins in FY Q1 2026, generating $233M in revenues!
In real mineral mining, the type and purity of the end mineral determine the final price. However, in crypto mining, there is no differentiation, as a Bitcoin is a Bitcoin, Ethereum is Ethereum, etc. As there is no differentiation in the final product, all miners sell their bitcoin for the same price.
This means that costs ultimately determine a miner’s profitability!
That is why crypto miners have a lot of experience in not only finding the location with the cheapest electricity but also designing energy-efficient data centers. A key metric used to measure the efficiency of a Bitcoin miner is joules per terahash (J/TH), which essentially shows how much electricity a miner uses to complete Bitcoin mining puzzles.
As we can see in the table above, Iren’s data centers have an efficiency of 15 J/TH, which compares quite favorably to its bitcoin mining peers.
This clearly shows that the company has the skills to build some of the most efficient data centers in the world!
In the above table, we can see what kind of financial results Iren could generate with its existing Bitcoin mining capacity.
With a Bitcoin price of $115K, Iren could generate $638M in Bitcoin ADJ EBITDA!
However, Bitcoin now sits at $88K, which means that Iren could generate $400M in earnings. With a break-even price of $45K, the company is in a good position to generate hundreds of millions of dollars in Bitcoin profits, which are being reinvested in building AI data centers. In 2024, Iren made $87M in net income, which is essentially only from the Bitcoin business, as the AI business was not meaningful.
1.2. AI Data Centers
As I have already mentioned in my analyses of AMD, Nebius, and WhiteFiber, AI requires fundamentally different infrastructure than general-purpose cloud infrastructure.
Simply put, AI data centers use a higher number of powerful GPUs that “eat” a lot of electricity, requiring meaningful grid upgrades. Additionally, these GPUs generate more heat than a conventional data center, demanding improved cooling systems. This means that, unfortunately, using existing general-purpose data centers for AI workloads is just not economically viable.
This is why we are witnessing this extreme data center buildout boom!
However, not all AI data center companies are the same, as there are 3 key ways in which they earn revenue.
Colocation
Bare Metal
AI Cloud
In a colocation model, a company such as Iren builds the data center, secures energy, and equips it with cooling, networking, and other equipment. Then they secure a customer who signs a long-term lease agreement of 5+ years and brings their own GPUs to the facility. This model has the lowest value added and thus has the worst economics, generating the lowest price per MW.
Meanwhile, in a bare metal model, the owner of the data center is responsible for fully equipping the facility with all equipment, including GPUs. It then finds a customer who negotiates access to a pre-agreed number of GPUs for a set per-hour rental price. Bare metal customers use their own software to run AI workloads on these GPUs. As the data center operator provides the GPUs, the bare metal model has a higher value added than colocation, and thus has better economics, generating a higher price per MW.
However, the most profitable AI data center model is the AI cloud provider!
In this model, the data center operator provides advanced AI training and inference software. Purpose-built AI training clusters, model hosting, managed inference services, AI APIs, storage, networking, developer support, and more. An AI cloud doesn’t simply rent out space, like colocation, or rent out GPUs like in bare metal. An AI cloud is essentially an AI development, optimization, inference, and deployment partner for AI start-ups and large enterprises.
As revenues are software-driven, this model has the highest value added and thus generates the highest price per MW of capacity!
Iren operates using the colocation and bare metal model.
Each of the operating models has its own advantages and disadvantages.
In colocation, the data center operator is essentially a landlord, and such a model can be highly profitable if operated well. The key advantage is that it requires less capital to start, is easier to operate, and generates a more stable and predictable rental income from a small number of key tenants.
A bare metal is similar to colocation. But, if we use the analogy of a rental apartment, colocation is comparable to renting an apartment without any furniture, whilst bare metal is a fully furnished apartment. Meanwhile, an AI cloud could be comparable to a fully serviced apartment, with furniture, maids, a personal chef, nanny, tutor, and chauffeur.
An AI cloud might have the highest margins in theory, but it is also the most difficult model to operate. Companies must employ highly skilled software engineers to build the digital environment that AI start-ups need. This requires a lot of upfront capital and years of research and development. Furthermore, AI clouds must have big sales and customer management teams that manage a large number of small customers.
Most importantly, this area of the industry is highly competitive, filled with highly skilled start-ups such as Nebius, and it is led by hyperscaler clouds with hundreds of billions of dollars at their disposal, such as Google, Amazon, and Microsoft.
This is why Iren has no intentions of competing in this area. Their core competency is not AI software, but
1. Finding locations with cheap energy.
2. Building energy-efficient data centers.
In Bitcoin mining, they use J/TH to measure the energy efficiency, but in data centers, PUE is used.
PUE stands for power usage effectiveness and is calculated by dividing the total capacity of a facility by the power actually used by IT equipment. A low PUE indicates that a smaller share of the total energy capacity is dedicated to overhead tasks, such as lighting, cooling, and others.
Essentially, a data center with low PUE uses a larger share of its electricity on revenue-generating GPUs.
Iren’s data centers are some of the most efficient in the business, with a PUE of 1.1. This means that for each KW used by IT equipment, only 0.1KW is used by the overhead. This compares favorably to other operators whose PUE ranges from 1.2 to 1.8.
2. Data Center Locations
Data centers are electricity-guzzling beasts, and AI data centers doubly so. Thus, finding a location with cheap access to energy is an absolute must.
As we can see in the picture above, Iren is currently working to develop 5 data centers with access to 3GW of secured power and more GW after that. 3 data centers are located in British Columbia, Canada, and 2 are in Texas, US.
British Columbia in Canada was chosen as an ideal location for data centers because of its temperate climate and abundant cheap hydroelectric power. We already established that data centers require the best cooling systems because of the heat. Well, locating data centers in a cold climate means that less powerful cooling is sufficient, saving the operator money on the equipment and energy.
Meanwhile, Texas was chosen because of it being the most energy-rich area not only in the US, but possibly the world. In addition to having access to oil and gas, Texas is perfectly suited for solar and wind farms. Thus, the region is experiencing a rapid renewable energy boom, which AI data center companies are eager to capitalize on.
Let’s expand on each data center further.
With a 30MW capacity, the Canal Flats data center is the smallest one. It is fully operational and is used in Bitcoin mining operations.
Next, with a capacity of 80MW, the Mackenzie data center is the largest one, and just like the Canal Flats location is only used for Bitcoin mining.
The 3rd facility in Canada, the Prince George data center, has a capacity of 50MW and is also fully operational. However, unlike Canal Flats and Mackenzie, this data center is not only a bitcoin miner, but also a bare metal AI data center with 2,200 Nvidia GPUs as of September 2025. The company recently purchased an additional 4,200 Nvidia B200 GPUs for $193M, which will be installed late this year or early 2026.
As AI data centers generate more income per MW of capacity than Bitcoin mining, all locations will ultimately be transformed into AI data centers!
Iren estimates that there is capacity for over 60 thousand GPUs. However, it will likely take years, as there is no rush to transform these highly profitable locations.
While the British Columbia locations are important, the Texas ones are really where Iren’s bull case truly lies, as they are orders of magnitude larger.
The Childress, Texas, data center has an operating capacity of 650MW, which is used for bitcoin mining, generating the majority of Iren’s current revenues.
This location is undergoing extensive expansion that will increase the total capacity to 750MW and transform the site from a bitcoin mine into an AI data center. The site will come online in 10 stages, with the first 4 having 300MW of capacity and the remaining 6 having 450MW.
Iren has already signed a long-term 5-year, $9.7B bare metal deal with Microsoft for 200MW from this facility, and I will expand on this agreement a bit later.
When people say that everything is bigger in Texas, they surely mean data centers as well. And Sweetwater is one large data center project, consisting of 2 sites with a planned capacity of 2GW.
This is a brand new project that will exclusively function as an AI data center. Iren expects Sweetwater 1 to be energized in April 2026 and Sweetwater 2 in late 2027. Please note that energization simply means that the site is ready to draw energy from the grid, it doesn’t mean that the data center will be fully operational by then. Construction still needs to take place, and GPUs, cooling, and all other equipment must be installed and tested.
Just with these locations alone, Iren’s hands will be full for a few years!
3. Microsoft Deal
A few weeks ago, Iren signed a large and transformational $9.7B 5-year deal with Microsoft!
The best part of the deal is that Microsoft agreed to pay for 1 year of services in advance, around $1.94B. Moreover, this pre-payment will be credited towards years 3-5, not towards the beginning of the deal. Meaning that the cash flows will probably look like this:
Now: $1.94B
Year 1: $1.94B
Year 2: $1.94B
Year 3: $1.29B
Year 4: $1.29B
Year 5: $1.29B
This is an excellent result for Iren, with the majority of the cash flows coming early in the project lifecycle, allowing the company to use cash flow from this project to build it. And others. Essentially, Microsoft is lending money to Iren at no cost (Although cash flow timing could have lowered the contract price).
The deal amounts to $1.94B per year for access to 200MW of capacity, resulting in $9.7M ARR per MW!
For a bare metal agreement, this is quite a high price, which clearly illustrates the appeal of Iren’s efficient data centers and cheap access to energy. However, Nebius signed an agreement with Microsoft for $11.6M ARR per MW, 20% higher than Iren. As I mentioned before, Nebius is a full-service AI cloud, and it will provide more value-added services in addition to simple bare metal.
Iren estimates that it will cost them $14-16M per MW to build this capacity, resulting in an estimated building cost of $2.8-3.2B!
Meaning that the pre-payment covers around 60-69% of the cost to build the data center!
This doesn’t cover GPUs and other equipment, which will cost $5.8B and will be purchased from Dell, with the company providing $199M in 24-month vendor financing for the first tranche of the equipment.
The exact details of the purchase agreement have not been announced, but most likely the deal bundles Nvidia chips with other networking and server equipment that Dell manufactures. Dell sells this equipment to Iren on credit, with Iren making monthly principal and interest payments to Dell. Dell makes the margin on the equipment and earns interest spread from what it charges Iren and what its lenders charge Dell.
It is quite likely that the $199M vendor financing deal is the first of many, and Dell could issue more incremental vendor financing agreements as Iren meets certain project milestones.
Iren estimates that the Microsoft contract will allow them to achieve an 85% project EBITDA margin!
This demonstrates how profitable a bare metal model can be at this scale. In 5 years, this capacity will allow Iren to generate $8.25B in earnings after paying for direct operating costs. Let’s assume that all GPUs and other equipment purchased for $5.8B from Dell retire after 5 years (they won’t).
Moreover, $2.8-3.2B build-out cost has to be spread over the useful life of a data center building of 20 years, which is around $150M per year and $750M in 5 years.
To conclude, taking into account the 5-year project EBITDA of $8.25B, $5.8B in equipment depreciation, and $750M building depreciation, Iren will generate a project profit before tax and interest of around $1.7B.
This leaves plenty of capital to cover financing, corporate overhead costs, and ongoing capex for expansions!
This is likely the first of many such deals and demonstrates Iren’s core strategy:
1. Find a great location for a data center
2. Design an efficient data center
3. Find a large anchor tenant
4. Negotiate a substantial pre-payment
5. Use that pre-payment to cover 40-70% of the build-out costs
6. Cover the remaining 30-60% with debt, vendor financing, or cash
7. Finance the GPUs, timing repayments with project cash flow
8. Use the project cash flows to cover the debt and build-out of new projects
9. Repeat
4. Capex Requirements
To get all 3GW of capacity online, Iren will need to raise a lot of capital.
As we just discussed, it costs Iren $14-16M per MW to build a data center, but this excludes GPUs and other equipment. If this equipment costs $5.8B for 200MW, that adds an additional $29M per MW, bringing the total per MW cost to around $44M.
This is more than double what I estimated it costs WhiteFiber to build a bare metal data center. This is because greenfields are much more expensive than retrofits, as they use the most advanced AI hardware. However, obviously, this is reflected in Iren’s revenues, which are likely double per MW of WhiteFiber’s smaller-scale retrofit data centers. Both models work, as they serve a different set of AI customer needs.
This means that if Iren wants to scale all 3GW into bare-metal AI data centers, it will need $132B!
There is simply no realistic scenario in which they raise $132B, so not all of this capacity will become bare metal data centers. Most of it will be converted into colocation facilities that have significantly lower capex requirements.
Modeling the $14-16M in capex required to build a data center, excluding GPUs, Iren will need $45B. If we assume an additional $20B for non-GPU equipment, the total capex requirements to build out the capacity into colocation would be around $65B.
Iren will decide how much of the capacity gets turned into colocation and how much into bare metal, depending on customer demand and their ability to raise capital.
Realistically, Iren’s capex requirement could be between $80-100B in the next 5-7 years!
The deal with Microsoft is a clear example of how this will happen. Iren will use customer pre-payments, vendor financing, GPU leasing, real estate mortgages, long-term debt, convertible notes, and equity issuances to raise the required capital.
Parts 5-7
The valuation analysis and a 2030 valuation model are available for Global Equity Briefing Premium subscribers.
A reminder that I have added a full portfolio review to the Global Equity Briefing schedule, where I:
List all my holdings
Track performance
Explain my strategy
Present my stock watch list
Look at potential catalysts
Notify of all buys/sells as they occur
Don’t miss out on my Iren Valuation model, and a full November portfolio review coming next Monday.
Thank you for being a loyal reader of Global Equity Briefing!
Small note, please go to the website to purchase the subscription, as if you do it on the iOS app, you have to pay the 30% Apple Tax. (The plan costs more there)
Here is what my Premium Members can expect:
Portfolio Review - Each month, I will present the portfolio performance and discuss my stock watchlist and my best ideas.
Recent developments.
Unwarranted pullbacks.
Insider activity.
Potential catalysts.
Deep Dives – 8,000+ word detailed analysis of a company, delivered in 3 Parts.
Part 1 – Brief History of the company and its Business Model.
Part 2 – Management, Moats, Competitors, and Risks.
Part 3 – Opportunities, Financial Analysis, and a Valuation Model.
You can expect a comprehensive research report that is educational, interesting, and provides actionable insights!
To see what you can expect, read my Palantir Deep Dive!
Members of the Premium service get access to my library of 11 Deep Dives and to all future Deep Dives, which will be released on semi-monthly basis.
Investment Cases – A short, concise report with actionable insights.
This report is about the size of a single part of a Deep Dive.
Focused Investment Thesis
Main drivers of the Bull Case
Valuation Model
To see what you can expect, read my Oscar Health Investment Case!
Earnings Reviews and Updates – For companies that are of great interest to me and my readers, I will provide regular quarterly or semi-annual updates after earnings reports.
Financial performance
Business Update
New developments
Updated Valuation Model
To see what you can expect, read my Google Q2 2025 Earnings Review!
Equity Research Report List
You can follow me on Social Media below:
X(Twitter): TheRayMyers
Threads: @global_equity_briefing
LinkedIn: TheRayMyers
Disclaimer: Global Equity Briefing by Ray Myers
The information provided in the “Global Equity Briefing” newsletter is for informational purposes only and does not constitute financial advice, investment recommendations, or an offer or solicitation to buy or sell any securities. Ray Myers, as the author, is not a registered financial advisor, and readers should consult with their own financial advisors before making any investment decisions.
The content presented in this newsletter is based on publicly available information and sources believed to be reliable. However, Ray Myers does not guarantee the accuracy, completeness, or timeliness of the information provided. The author assumes no responsibility or liability for any errors or omissions in the content or for any actions taken in reliance on the information presented.
Readers should be aware that investing involves risks, and past performance is not indicative of future results. The author may or may not hold positions in the companies mentioned in the “Global Equity Briefing” report. Any investment decisions made based on the information in this newsletter are at the sole discretion of the reader, and they assume full responsibility for their own investment activities.
The pivot from crypto mining to AI data centers makes a lot of sense when you think abut the infrastruture overlap. The PUE of 1.1 is impressive, especially compared to traditional data centers running at 1.5 or higher. What stood out to me is how IREN structured the Microsoft deal with that upfront payment. How do you see the colocation vs bare metal mix playing out as they scale the Texas sites?
what do you think is more interesting, colocation model (CIFR, WULF..) or bare metal model? For me, CIFR has less risk than IREN, bur IREN model has more potencial...