“In this post I’ll cover why I expect this hyper-growth stock to crack $150 over the next 24 months—representing a gain of 1150% from its current price of $12.

I went ‘All-In’ this stock, and for good reason…”

Disclaimer

This piece was originally published in November 2024, when IREN was trading around $12. I initiated at $4.50 and added as the story got clearer. The data, models, and price targets are outdated.

A lot has changed since then. For a current view of the thesis, read our more recent deep dives: A Hyperscaler in the Making and Childress Unlocked.

The fundamentals got stronger. Not weaker.

Introduction

I believe that IREN is THE stock to hold over the coming years, as it is positioned at the center of the two strongest ‘narratives’ of this business cycle & bull market – AI and Bitcoin.

This small cap, with a market cap of just ~$2.3 billion, builds and operates data centers to support high-performance computing applications (HPC), such as BTC mining and AI, leveraging cheap renewable energy sources.

@IREN was founded in the fall of 2018 by the Australian brothers Will & Dan Roberts and went public in November 2021. Based in Sydney, the firm was established with a forward-looking vision to address the growing demand for computing power in an increasingly digitizing world. In its early stages, Iren strategically acquired land in Canada and the U.S., particularly in regions like British Columbia and Texas, where the increasing availability of renewable energy sources such as solar, hydro, and wind is steadily reducing industrial energy costs.

As Iren acquired most of its land before the recent AI boom, it’s likely they purchased it for a fraction of what it would be worth today, potentially paying pennies on the dollar compared to current market values. This foresight has positioned them exceptionally well for future growth as demand for HPC and AI infrastructure increases and land with access to power becomes ever more scarce and valuable.

The company currently holds a very impressive land & power portfolio, consisting of infrastructure sites with grid-connected power capacity of >2.3 GW across the U.S. and Canada, with over 1 GW of additional capacity in the pipeline.

To put this into context, in 2023 the TOTAL global data center capacity was approximately 33 GW, according to a report from Citi Research.

However, a significant portion of Iren’s infrastructure portfolio remains undeveloped, with the company currently operating approximately 450 MW of data center capacity. Iren is actively scaling its operations, rapidly expanding its capacity by developing its land and constructing new data centers at a rapid pace of 50 MW per month.

By the end of the year, Iren aims to reach 510 MW of available infrastructure, with the majority of this power dedicated to Bitcoin mining at a hash rate of 31 EH/s - positioning the company as one of the largest BTC miners in the industry.

In case you are new to the Bitcoin mining sector, hash rate (measured in EH/s) is a measurement of computational power being used to mine Bitcoin. Generally speaking, the higher the hash rate, the more BTC you mine. Since there is a predetermined number of BTC being issued per “block” (every ~10 minutes), BTC miners compete for this reward. The higher the hash rate, the higher the cut of that reward. This makes EH/s growth imperative for every miner.

Achieving the end-of-year target of 31 EH/s would set a new industry record for the fastest hash rate growth in a single year by a public BTC mining company, both in nominal and percentage terms. This would mark a leap from 5.64 EH/s in January 2024 to 31 EH/s by December 2024, representing a ~450% increase within just one year.

$Iren already holds the previous record for the fastest-growing Bitcoin miner in a single year, having increased its hash rate from 1.6 EH/s to 5.6 EH/s last year. This represented an impressive 250% growth rate. However, surpassing this record at such a larger scale would be unprecedented and truly groundbreaking in the industry. This company is quickly establishing itself as the fastest growing $BTC miner of all time with unparalleled growth rates, largely thanks to already owning high quality land on which it can rapidly build its data centers.

Given its fast growth, most of Iren’s mining fleet is equipped with the latest hardware this industry offers. This newer hardware translates into higher energy efficiency and, consequently, lower energy costs. As of today, Iren’s fleet operates at an industry leading efficiency of 16 J/TH, which is significantly better than the industry average, which ranges in the low-to-mid 20s. By year-end, the company aims to achieve a fleet efficiency of just 15 J/TH, solidifying itself as the most profitable $BTC miner in the industry.

The company has also recently dipped its toes into the AI market, offering a small yet promising AI cloud service, which includes 816 NVDA H100 GPUs and a recent acquisition of 1,080 H200 GPUs.

While IREN continues to rapidly gain market share in the BTC mining space, the company is now leveraging this success to strategically position itself for emerging opportunities in the AI sector, aiming to monetize a significant portion of its expansive land and power portfolio in this growing segment.

IREN’s Founders

IREN was founded by the brothers Daniel and Will Roberts, who continue to serve as Co-CEOs to this day. Dan, now in his early 40s, takes on the public-facing role, representing the company and engaging with external stakeholders. In contrast, Will, who’s in his mid-30s, concentrates more on internal matters, such as overseeing operational and strategic initiatives within the company.

Before founding Iren, both brothers gained significant experience in finance, energy, and the realm of digital assets.

In the 6 years prior to founding this company, Dan co-owned and led Palisade Investment Partners, an asset management firm focused on infrastructure investments. Palisade specialized in critical energy sector projects, including wind / solar farms, and gas pipelines. Under Dan’s leadership, the firm grew to manage $6 billion in assets across Australia & parts of Europe.

Earlier in his career, Dan served as a Vice President at Macquarie, a prominent Australian investment bank with a current market cap of ~$60 billion. At Maquire, his focus was on infrastructure, finance, and renewables.

His younger brother, Will, held a similar position at Macquarie as Vice President within the Commodities & Global Markets division. In addition to his work in traditional commodities, Will also co-founded Macquarie’s Digital Assetsteam.

According to a recent Forbes article, Dan and Will Roberts were early participants in Ethereum’s presale. Although the article doesn’t specify the exact date, it’s likely they acquired $ETH during the 2014 presale period, when Ethereum first offered $ETH to early investors.

While it’s probably (currently) not directly relevant to IREN itself, an ‘interesting’ fact from the aforementioned Forbes article is that Dan Roberts almost lost his life due to a life-threatening tumor discovered after a football injury. The full extract is provided below:

Insider ownership

Each co-founder currently holds approximately 5.5 million shares, representing an individual ownership stake of 2.93% in IREN, based on a total of ~187 million outstanding shares.

As of now, they have never reduced their holdings, apart from a single time in December 2023, when each of them sold 1 million shares. While the reasons for this sale remain undisclosed, one could speculate that it might have been related to Dan Roberts’ medical situation at the time, with the brothers potentially raising liquidity for Dan’s family as a precautionary measure. Since then, the Roberts brothers have not made any additional sales of their IREN holdings.

Together, the Co-CEOs have nearly 6% insider ownership and hold 1 Class-B share each. In the case of Iren , Class-B shares act as a 15x multiplier in terms of voting power for the ordinary shares they own. While 1 ordinary share carries 1 vote, each B Class share boosts their voting power to 15 votes per ordinary share, granting them significant control over the company’s decisions. However, their combined voting power is capped at 44.45%, likely due to internal policies designed to balance control and governance.

Although opinions on this structure may vary, I believe it benefits shareholders by ensuring the co-founders retain substantial control to drive their vision forward. A great example is Steve Jobs, who was removed from Apple in the mid-1980s, which led to a period of struggle for the company until his return. Strong leadership can be crucial to a company’s success, and the Class-B shares ensure the founders are empowered to lead.

The co-founders also have a shareholder-friendly compensation structure. Each person stands to receive up to3,689,724 shares through stock options and RSUs, contingent on Iren’s share price achieving specific milestones. This aligns their interests with those of shareholders, as large parts of their compensation is directly tied to the stock’s performance.

In case you don’t know, RSUs (Restricted Stock Units) are “free” shares granted to employees that vest over time. They are typically used to incentivize strong performance and retain top talent.

Additionally, each founder will receive a total of 1,891,753 shares in the form of time-based RSUs over the next two years, as long as they continue in their roles. While this will dilute existing shareholders, it reduces the cash burden on the company by offsetting cash-based salaries with equity compensation.

Salary Breakdown

The total $12.6 million compensation per founder/CEO for FY 2024 may seem excessive to most investors, however, if you dig deeper, only $2,872,343 of it is cash, while the majority ($9,753,105) comes from the amortization of stock options and RSUs that may never vest if the share price doesn’t reach the price levels mentioned earlier ($20 - $1,850).

Amortization spreads the estimated value of these stock options and RSUs over their vesting period, but if IREN doesn’t hit these targets, a large part of this “salary” is essentially just an accounting figure - not actual cash they receive.

Alternatively, if IREN meets all the share price milestones and these awards vest, shareholders should also be very pleased with the outcome, as it would result in significant price increases for the stock.

Infrastructure Portfolio

After gaining extensive experience in the energy infrastructure sector, the Roberts brothers launched $IREN with a clear focus: finding land with access to the cheapest available energy.

Contrary to popular belief, many renewable energy sources are often the most cost-effective forms of energy. This is in part because renewable energy typically generates electricity that must be consumed immediately, as storage technologies are still insufficient (although $TSLA is slowly changing this).

Without sufficient demand, much of this energy goes to waste. In some regions, government subsidies have led to an oversupply of renewable energy, creating a mismatch between supply & demand. Recognizing this market inefficiency, the Roberts brothers strategically sought out data center locations near these renewable energy projects, where providers were keen to supply their power to the local grid at low rates.

Canada

The first site they acquired is located in Canal Flats (30MW / 10 Acres), British Columbia (BC), Canada. This facility was built next to a shutdown pulp mill, taking advantage of the local oversupply of hydroelectric power. The Roberts brothers ‘purchased’ the data center (more on that later) and used the existing infrastructure to support their Bitcoin mining operations.

Shortly after the ‘acquisition’ of the first site in Canal Flats, Iren expanded further by securing 2 additional sites in British Columbia: Mackenzie (80MW / 11 Acres) and Prince George (50MW / 12 Acres). Both locations, like Canal Flats, primarily rely on cheap & renewable hydroelectric power. Prince George is currently being used as Iren’s hub for AI computing. The company is also currently conducting power studies with the intention of potentially expanding this site by an additional 20-40 MW.

These Canadian sites have electricity rates in the mid $0.04 range per kWh, which is decent for commercial purposes but not exceptionally low. All BC sites are grid-connected and receive power from the BC Hydro network (in front of the meter), which is almost entirely hydroelectric. The BC Hydro grid is highly stable, with a reliability rate of 99.931%. Additionally, Iren has installed battery and generator backups for key components such as network, storage, security, and key management systems to ensure redundancy and minimize downtime.

The infrastructure sites also have sub-20ms latency, which is more than sufficient for maintaining smooth operations in most AI workloads, specifically in the context of inference.

Inference refers to the process of using a trained machine learning model to make predictions or decisions based on new data. For example, when you interact with an AI system like ChatGPT, the model performs inference by processing your inputs in real-time to generate responses. Low latency is crucial in this context because it directly impacts the response time, which ensures a smooth & efficient user experience. High latency would result in noticeable delays, hindering the performance of real-time applications.

Texas

Following their success in Canada, $IREN expanded into Texas with the massive Childress site (750 MW / ~550 Acres).This site was initially announced to have a capacity of 600 MW but was upgraded to 750 MW in late July this year.

The majority of Iren’s growth over the coming year will come from this site, especially regarding their Bitcoin mining operations. Out of the 750 MW capacity, 350 MW will be operational by the end of 2024, with the remaining 400MW scheduled to be operational by late 2025 / early 2026. By the end of this year, Childress will make up about 70% of the company’s built-out MW capacity.

Interestingly, Iren states on their website that this site has a size of 420 Acres. However, investment analyst & $IREN investor @FransBakker9812 found proof that Childress is actually 553 Acres large and consists of 3 pieces of land. Given the scale of this site, it’s possible that the company has intentions of increasing its capacity yet again, beyond 750 MW.

As of now, Childress has one 640 MVA transformer and is set to receive a second one by Q2 2025. This addition will enable Childress to achieve its 750MW capacity, but theoretically speaking, these two transformers could support up to ~1.2 GW. Given the recent upgrade from 600MW to 750MW, one could speculate that the site could potentially reach 1-1.2GW in the future.

However, it’s worthwhile noting that any additional expansion beyond 750MW at Childress would likely be contingent on securing new connection agreements & getting approved by ERCOT (Texas grid operator).

Similar to Iren’s Canadian site, the Childress facility is grid-connected, drawing power from the AEP network which has a reliability rate of 99.957%. To ensure redundancy, like all of Iren’s sites, Childress is equipped with backup generators and battery systems, which provide additional resilience and minimize the risk of downtime.

Lastly, the Childress facility is connected to dual physical fiber paths, achieving sub-10ms latency, which is considered excellent for AI inference tasks. Iren states that ALL its sites have this level of ‘best practice’ network redundancy, using *at least* two physically diverse fiber paths with multiple tier 1 Internet Service Providers (ISPs). This means that if one ISP or fiber path encounters an issue, such as a physical cable cut or technical failure, the other path with a different ISP can take over, maintaining connectivity.

This setup addresses a common concern with rural or remote sites, where connectivity issues are more likely due to fewer infrastructure options compared to metropolitan or city-adjacent sites, which often have more robust network infrastructure. Iren’s approach ensures that its sites are well-equipped to handle these risks, effectively negating this concern and providing reliable, low-latency network performance.

1.4GW West Texas site:

IREN’s second site in Texas is the largest asset in their portfolio, featuring a massive 1.4 GW capacity. Currently this site is mostly undeveloped, however, the company has already secured all necessary connection agreements, with the procurement process underway for substantial parts of its infrastructure.

This site is very attractive for potential AI colocation / joint venture deals with hyperscalers like Amazon’ AWS, Microsoft’s Azure, Google Cloud, or even Meta. There are not many sites with >1 GW of freed-up energy capacity and there are even less sites of that scale that would be ready to use as early as H1 2026.

A colocation deal between a data center host like Iren and a hyperscaler typically involves the hyperscaler renting the host’s infrastructure, while bringing their own compute hardware like GPUs. These arrangements can vary significantly; the host might offer a standard, off-the-shelf setup or build a customized facility tailored specifically for the client’s needs. This flexibility allows hyperscalers to quickly scale their operations using the host’s established infrastructure without having to invest time & money in building their own data centers.

As of Iren’s most recent updates, the current target for the energization date of the site’s 1.4 GW substation is in April 2026. It’s likely that Iren would be able to start building data centers already before that date, possibly as early as late 2025 and energize them by April 2026. This H1 2026 energization date is a crucial factor in the context of this site’s value proposition.

The world’s largest tech firms are locked in an AI compute arms race, prioritizing rapid scaling and time to market (TTM) as they compete to expand capacity as quickly as possible. For example, Elon Musk’s xAI recently built its ‘Colossus’ mega cluster of 100,000 H100 GPUs in just 19 days, with plans to add 50,000 H200 GPUs soon. One could criticize Elon for heavily investing in an older generation of GPUs, like the H100s, when the newer H200 GPUs are gradually becoming more available, and the much more powerful Blackwell generation is being launched next year. However, this decision underscores the critical importance of time to market (TTM). Elon is focused on securing as much AI capacity as soon as possible, and currently, H100s are the most readily available option, even if they are not the latest & best. The top firms know that achieving the most AI capacity earliest is key to dominating this new market and securing a potentially unassailable lead.

One of the constraining factors that is becoming increasingly more prominent, as the scale of AI clusters increases, isaccess to power. The new generation of AI purposed GPUs require much more energy than traditional data center hardware. Access to power is rapidly becoming a scarce commodity. Keep in mind that, as mentioned earlier, the global data center capacity in 2023 was ca. 33 GW, with most of it dedicated to well-established needs such as hosting, file storage, and other enterprise cloud services. While some of this capacity will inevitably be adapted for AI compute, it’s unlikely to be a substantial amount, as these traditional services remain critical and continue to drive significant demand. Thus, most of the new AI capacity coming online will require new access to energy.

The challenge with this is that developing greenfield sites for data centers is a complex, multi-year process that often takes upwards of 5 years to become operational. This process consists of:

1. Site Selection and Due Diligence (1-2 years)

2. Power Studies and Connection Agreements (1-2 years)

3. Permitting and Construction (2-3 years)

Iren had the foresight to not only acquire the massive 1.4 GW site in west TX years ago, but also start the process of conducting power studies and getting connection agreements / permits well in advance. Having such a large quantity of new, unpurposed, power available as early as 2026 makes this a true ‘unicorn’ site in this new era of compute.

Iren recently clarified that it has the procurement process underway for 4 x 560 MVA 345 kV/138 kV transformers. This provides the company with plenty of optionality in terms of redundancy and potential for energy expansion beyond 1.4 GW. Four x 560 MVA transformers would provide enough energy for 2240 MW. However, it’s unlikely that the site will increase capacity to beyond 2 GW without adding additional transformers. One must understand the importance of redundancy in the context of HPC cloud hosting.

This site’s current set-up of transformers would be sufficient for a T3 facility of 1120 MW (1.12 GW) - whereby two 560 MVAs would be in use and two are on standby for redundancy purposes.

In the data center industry, uptime & redundancy are classified into four tiers, with Tier 1 being the most basic level and Tier 4 representing the highest level of redundancy and fault tolerance. Tier 3 (T3) is generally considered the standard that most hyperscalers strive for. T3 facilities must provide N+1 redundancy, meaning they have at least one backup component for every critical infrastructure element. This ensures that the data center can remain operational during maintenance or equipment failure.

In the case of Iren’s new west Texas site, the current setup of 4 x 560 MVA transformers allows for two transformers to be actively used while the other two serve as backups. This configuration supports the N+1 redundancy required for a T3 facility, ensuring that if any of the active transformers fail, the backups can immediately take over without disrupting operations. Alternatively, Iren could push for higher MW usage at the expense of redundancy, similar to what they did at their Childress site. For instance, they could operate with 3 out of the 4 transformers, providing upwards of 1.5 GW capacity. This setup would sacrifice the N+1 redundancy standard required for T3 status but would likely still qualify for a T2 facility, as it maintains basic redundancy with one transformer on standby.

I would also like to point out that while redundancy is a critical factor in cloud computing, it’s less of a factor in the realm of Bitcoin mining. Sure, having high uptime is detrimental to mining revenues, but having occasional outages isn’t going to break the camel’s back. You are not going to upset any customers when maintenance work is being done for a couple of hours in your facility, because you have no customers. This is why a facility like Childress, which is primarily being built for the purpose of mining $BTC, can get away with less redundancy.

In any case, their current setup offers plenty of optionality, allowing them to balance between maximizing capacity and maintaining redundancy based on operational needs.

Besides the site’s massive scale and short-dated availability, making it an attractive option for colocation deals, the site also has exceptional latency of under 10ms (just as Childress) with multiple fiber paths & tier-1 ISPs . Located in Texas, the site also benefits from some of the lowest energy rates in the country, making it especially appealing for energy-intensive AI workloads.

In later sections of this post, I’ll delve into the financial implications of colocation arrangements, focusing specifically on how they impact cash flow dynamics for $IREN.

Texas & Declining Energy Costs

As it relates to declining energy costs, Texas is arguably one of the most promising locations on earth, certainly within the U.S.

It already ranks among the 5 cheapest energy states in the US, however, this position is further strengthened by rapid growth in wind and solar power, coupled with significant investments in battery storage. These developments will play a crucial role in driving prices even lower over the coming years.

Solar energy is currently the cheapest source of energy, even outperforming traditionally cheap sources like nuclear, coal, and natural gas from fracking. If you doubt this claim, I strongly encourage you to check out the research published by @tonyseba (especially his presentations on YouTube).

As Elon Musk said: “Once you understand the Kardashev Scale, it becomes utterly obvious that essentially all energy generation will be solar. Just do the math on solar on Earth and you soon figure out that a relatively small corner of Texas or New Mexico can easily serve all US electricity.”

Texas also benefits from the free-market dynamics of its unique electricity market, where competition among providers drives innovation and efficiency. ERCOT, the grid operator for most of Texas, manages real-time electricity supply & demand, allowing prices to be set based on market conditions. This system makes it relatively easy for companies to develop projects like solar farms and sell their generated electricity to the grid, thanks to clear market regulations and streamlined processes.

Moreover, Texas is now moving towards a system that allows private companies to compete for building the transmission network, which could further accelerate energy cost declines in the state. Traditionally, transmission development in Texas has been more restricted. The right-of-first-refusal (ROFR) law gave established utility companies the exclusive right to build new transmission lines, which meant that other companies couldn’t compete for these projects.

However, a recent court ruling deemed this law unconstitutional, opening up transmission projects to competition. This change allows private companies to bid on building transmission lines. By introducing competition in this part of the market, the ruling could lead to more efficient and cost-effective infrastructure development. Just as the free-market model for electricity generation has driven innovation and kept prices competitive, this new approach to transmission development could further lower costs and speed up the expansion of the grid to meet growing demand.

The benefits of Texas’ energy landscape are evident in Iren’’s Childress site, which has a historical average spot cost of just 3.5 cents per kWh, which is significantly lower than the energy costs of most competitors.

For instance, Bitcoin mining competitor $CLSK, primarily operating in Georgia, faces average costs of ~ 4.5 cents per kWh. Similarly, $WULF, with operations in New York and Pennsylvania, incurs rates ranging from high 3s to high 4 cents per kWh. As a reference point, the miner with the industry’s lowest electricity rates is $CIFR, with a cost of just 2.72 cents per kWh. However, this low cost comes from a long-term Power Purchase Agreement (PPA), (set to expire on July 31, 2027), that requires them to curtail up to 5% of their energy use annually. This drawback effectively reduces their operational days and thereby its revenue potential & usability for AI workloads that require constant uptime.

Although Iren’s average spot rate at Childress was just 3.5 cents per kWh from April 2023 to July 2024, the company ended up paying a realized price of 4.3 cents per kWh during this period, due to its use of hedging strategies to mitigate price volatility. However, this changed in August 2024, when the company shifted its approach to primarily using spot pricing, by leveraging the favorable rates in Texas and the flexibility to curtail operations during brief periods of volatile price spikes.

This new strategy allowed the company to secure remarkably low energy costs for its Bitcoin mining operation in Childress, with rates of just 3.1 cents per kWh in August, 3.2 cents per kWh in September, and just 3.06 cents per kWh in October. The fact that Iren’s electricity rates for the past 3 months were significantly lower than the average of 3.5 cents suggests that energy prices in Texas are trending downwards. This trend is likely due to the increasing deployment of solar and battery storage. As a result, over the next nine months leading up to summer, $IREN could potentially achieve average electricity rates in the high 2 cent range to low 3 cent range per kWh at their Childress site. In the realm of Bitcoin mining and AI compute, where margins are primarily influenced by energy costs and mining efficiency, this would act as a significant cost advantage and tailwind for rising margins.

However, I do want to point out that, generally speaking, average electricity rates will likely always be cheaper if you have the optionality to curtail when prices spike. Curtailing involves temporarily shutting down operations for a few hours or even several days when electricity prices surge unexpectedly, often due to grid instability. Usually, you only have this sort of optionality in the realm of Bitcoin mining and not in the (AI/HPC) cloud computing market where constant uptime is required.

This means that while I do expect Iren to achieve electricity rates as low as $2.5-3 cents per kWh at their TX sites, I think this will mostly be in regards to their mining operations. In the AI cloud segment, I expect energy rates to be closer to 3-3.5 cents per kWh (at TX), factoring in the occasional pricing spikes that tend to happen from time to time. That being said, the Texas grid is becoming more stable with the deployment of large-scale battery storage, which will enhance network reliability and price stability, leading to fewer sudden pricing spikes over time.

Speculative / Undisclosed Sites

In addition to the confirmed 2,310 MW of secured power across existing sites, Iren has stated that they have over 1GW of additional land & power capacity in the pipeline.

While the company hasn’t provided much info on this additional ~1 GW pipeline, my good friends @FransBakker9812 & @Brenno2332 discovered that Iren acquired several pieces of land in recent times. They did that by discovering the actual property deeds recorded with the municipality, showing that these sites are indeed owned by the company.

According to their research, Iren recently acquired two additional sites in Texas, each approximately 40 acres in size. They also confirmed the purchase of a 159-acre site in the neighboring state of Oklahoma. Moreover, just a few days ago, Frans discovered that Iren recently signed a Memorandum of Contract (MOC), giving them the option to acquire a new 337-acre land site in West Texas, with a 6-month window to exercise the option.

Big s/o to Frans & Brenno for doing so thorough due diligence & finding these nuggets of information.

In conclusion, Iren is on track to achieve a total capacity of 510 MW by the end of 2024, and with the continued development of the Childress site, this capacity is expected to expand to 910 MW potentially as early as 2025. Not only is their ambitious 1.4 GW project set to break ground shortly before April 2026, with a potential start as early as late 2025, but they also have over 1 GW of additional capacity in development.

CEO @danroberts0101 confirmed that Iren is currently growing its built-out data center capacity by ~50MW / month, far exceeding the industry standard of ~5-20 MW per month. Given this rapid growth rate and the company’s extensive infrastructure portfolio, $IREN is poised for an unprecedented surge in computing power over the coming years.

Iren’s extensive land & power portfolio provides significant flexibility for both Bitcoin mining and its expansion into the AI sector - topics I will explore further in the following sections of this post.

For anyone interested in staying up to date with the company’s infrastructure progress, I highly recommend following @FransBakker9812. He provides frequent updates & insights by publicizing satellite images and sharing his in-depth investigative research.

IREN’s AI Entry

While Bitcoin mining operations currently remain Iren’s primary revenue source, the company is increasingly shifting its focus to high-performance computing (HPC) for AI cloud hosting. It’s worth noting that Iren has publicly expressed its interest in the AI sector since its IPO in late 2021.

The company began actively pursuing this strategy roughly a year ago (late August 2023), starting with the acquisition of 248 H100 GPUs from Nvidia. Iren then began scaling its capacity in February of this year, acquiring an additional 568 H100s. Seven months later, in September 2024, Iren purchased 1080 units of the newest H200 generation, bringing their total capacity to 1,896 H100 & H200 GPUs.

The recently acquired batch of H200 GPUs is currently being installed at Iren’s Prince George facility, alongside its existing H100s, and is expected to come online in the coming weeks.

The company monetizes its GPU fleet through a business model called ‘AI cloud hosting’, which involves renting out the computing power generated by these GPUs to businesses looking to train AI models or run inference tasks. Cloud Service Providers (CSPs) like Iren typically charge for this capacity on a per-hour basis, referred to as ‘GPU hours’. More advanced and powerful GPUs command higher rates per hour, reflecting their increased computational capabilities and efficiency.

CSPs typically differentiate themselves based on pricing and service quality.

Those who own 100% of their infrastructure—from GPUs to data centers and the land they are built on—and have access to low-cost energy gain a clear cost advantage, allowing them to offer more competitive rates. However, service quality is also crucial; CSPs that provide high redundancy (ensuring no outages), rapid response times for customer requests & modifications, or the option for fast scalability can charge a premium compared to their competitors.

Iren, being fully vertically integrated, meets both criteria. By owning all its infrastructure and accessing inexpensive energy, the company significantly lowers its operating costs. This vertical integration also gives Iren complete control over its operations, resulting in excellent response times and high-quality customer service.

Iren is quite unique in this regard, as many CSPs rent infrastructure from data center (DC) hosts through colocationdeals. In most cases, DC hosts operate the facility for the CSP clients. Typically, the DC host is responsible for the physical infrastructure, including power, cooling, security, and maintenance. The CSP, leases this space and infrastructure but focuses on managing and delivering services to their end clients. In this setup, the CSP often acts as an intermediary (middle-man), providing cloud computing services while relying on the DC host to maintain and operate the physical environment. This arrangement allows the CSP to scale quickly without the overhead of managing the physical infrastructure, while the DC host ensures the facility’s operations remain efficient and reliable.

However, this arrangement ultimately results in higher operating costs for the CSP, as the hosts need to make a profit too. Since the host ultimately controls the data centers, this type of arrangement also reduces the CSP’s operational control, which can adversely affect the quality of service towards the end-customers.

Moreover, colocation deals can create misaligned incentives for hosts. The host’s primary customer is the CSP, typically secured through long-term contracts. With revenue guaranteed for multiple years, the host’s main incentive shifts towards maximizing profit margins by minimizing costs. This can result in reduced facility maintenance, leading to more outages and repairs over time. Additionally, it may also cause slower response times for the CSP’s customers, whom the host indirectly serves.

While this type of colocation partnership has its drawbacks, the primary advantage is that it allows the CSP to scale more rapidly to increase market share (fast TTM), as it avoids the time-consuming process of building its own infrastructure.

While Iren won’t operate as a CSP that rents other companies’ infrastructure, they are likely to become a host for other CSPs and hyperscalers, in addition to running their own AI cloud service business. In its recent quarterly earnings call, management stated that they are in active discussions and have signed NDAs with some of the largest companies in the world. Iren also noted in an October business update that there is “indicative interest from hyperscalers”. Thus, a colocation deal materializing appears increasingly likely.

The distinction between Iren’s own cloud services and those provided for other hyperscalers would likely be minimal, essentially creating a ‘white label’ model for CSPs utilizing Iren’s infrastructure. Although this approach might initially seem counterintuitive, it aligns with the priorities of hyperscalers, who primarily focus on time-to-market and rapid scaling. Given Iren’s extensive land holdings and access to energy, it would be impractical for them to utilize all this capacity solely for themselves, as it would require well over $50 billion in GPU investments alone. Colocation agreements would arguably be the most cost-effective way to utilize a large majority of Iren’s energy capacity, allowing them to maximize the use of their resources.

One could also argue that a colocation deal with an infrastructure provider like Iren, which operates its own CSP business, is more advantageous for a hyperscaler, as it effectively addresses the ‘incentive problem’ mentioned earlier. Iren’s dual role as both a CSP and a host for other CSPs means that maintenance and customer support standards are likely to be consistent across both services, functioning as a true white-label business. If discrepancies arise, such as Iren prioritizing its own CSP business over its colocation clients, the hyperscaler may have legal grounds to sue, especially if the agreement stipulates identical levels of service. This arrangement significantly mitigates the risk of misaligned incentives.

Moreover, owning a CSP business that Iren plans to scale rapidly over time gives the company invaluable experience in both building AI/HPC data center infrastructure & in providing great customer service - crucial skills for effectively hosting hyperscalers. Hence, while Iren’s own CSP business segment shows promising potential due to its full vertical integration, this also strategically positions the company for future colocation deals with hyperscalers.

Iren’s Data Centers

Given the similarities between $BTC mining and AI processing—both being highly energy-intensive forms of compute—Iren didn’t have much difficulty in repurposing existing Bitcoin mining infrastructure to accommodate the newly acquired GPUs at its Prince George facility.

Recently, CEO Dan Roberts shared a video online showing Iren staff installing the H200 batch in a data center with BTC mining ASICs visible in the background (right pic), indicating that both GPUs and BTC miners are operational within the same data center unit.

This is primarily possible thanks to Iren’s high quality data center design. Iren’s modular DC design is almost “over-kill” for Bitcoin mining purposes. Iren is known for having some of the highest infrastructure costs per MW in the Bitcoin mining industry, totalling $650k per MW, which is at least 20-30% more expensive than its peers. While much of the $BTC mining industry resorts to cost-cutting methods like repurposing old shipping containers, renting existing facilities, or acquiring failed mining sites to house their ASICs (mining hardware), Iren distinguishes itself as one of the few miners that has consistently built its data centers from scratch, usually beginning with greenfield sites.

As a result, Iren achieves some of the best uptimes in the industry. By housing ASICs in optimal conditions—ensuring perfect cooling and maintaining dust-free environments—Iren’s BTC mining operations experience virtually no outages, a rare accomplishment in an industry where interruptions are very common.

Another trait of Iren’s data centers is their high level of standardization. As mentioned earlier, Iren is currently expanding its capacity at an exceptional speed of 50 MW per month. This rapid growth is largely due to their standardized design and consistent construction methods, which not only accelerate the build process but also maintain consistent quality across all sites.

Iren’s data centers have an average Power Usage Effectiveness (PUE) of 1.05, indicating extremely high efficiency.

PUE measures how efficiently a data center uses energy, with 1.0 being the theoretical best, meaning all energy is used directly for computing power, and none is wasted.

Iren can achieve such a low PUE because it perfected the method of air-cooling. Most modern day HPC data centers use liquid cooling since that is widely regarded as the more effective cooling method, however, that comes at the expense of more energy usage & consequently a worse (i.e., higher) PUE. Nonetheless, Iren engineered a way to maximize the effectiveness of air cooling. The company optimized their cooling design to such an extent that it achieves a rack density of 70-80 kW for their H100 GPU set up.

Rack density refers to the amount of power consumption per rack in a data center. It indicates how much computing power is concentrated in a single rack, measured in kilowatts (kW), and helps determine the cooling and power requirements for the facility.

As a reference, most modern day data centers operate under a rack density of 5-15 kW. However, as AI computing requires more energy, much more kW are required per rack. NVIDIA provides a reference rack density of 40 kW for their H100 GPUs. Iren achieving a rack density of up to 80 kW in their air-cooled data centers is an engineering marvel. Typically, air-cooled data centers can reach a rack density of around 15 kW, with highly advanced setups usually maxing out at 25-30 kW per rack. Achieving 70-80 kW through the use of air cooling alone is extremely rare and highlights Iren’s engineering expertise.

Keep in mind, when Iren repurposed their existing mining facilities to house GPUs, they did so without making any upgrades to the infrastructure. They simply unplugged the ASICs and plugged in the GPUs (along with the other AI-related hardware). This seamless transition is a testament to the quality and flexibility of Iren’s standardized data center design.

Nonetheless, next-generation AI hardware & potentially new-gen Bitcoin mining hardware will likely have increasingly higher thermal outputs which would make liquid cooling a necessity.

Some investors have raised concerns about Iren’s ability to implement liquid cooling at their Texas sites, due to the lack of nearby natural water sources. Traditionally, data centers have relied on open-loop cooling systems, which require a continuous external water supply, often sourced from lakes or rivers. Recently, however, closed-loop liquid coolingsolutions have emerged as the most efficient cooling method for large-scale HPC facilities.These systems often consist of a mixture of water & glycol and are far more water-efficient than open-loop setups. Unlike open-loop systems, where substantial amounts of water are lost to evaporation, closed-loop systems continuously recirculate the same fluid, with only a minimal top-up required of around 1% per year.

This makes closed-loop cooling a far more sustainable & efficient option that significantly reduces water consumption and waste. Iren has publicly announced that they are actively exploring closed-loop liquid cooling systems and have already initiated procurement to accommodate liquid-cooled hardware, including GB200 GPUs, at their existing data centers in Prince George and Childress.

The GB200 is a superchip in NVIDIA’s upcoming Blackwell lineup, combining two B200 GPUs with a Grace CPU for maximum AI and HPC performance. It significantly outperforms a standard B200 GPU, offering up to double the processing power and memory bandwidth, making it ideal for large-scale AI model training and inference​.

Iren’s modular data center designs make upgrading current facilities and retrofitting older sites with closed-loop liquid cooling pretty straightforward. Based on conversations I had with industry experts, this approach is entirely feasible at their current facilities. Given Iren’s expansive land & power portfolio—including a massive 1.4 GW greenfield site—retrofitting existing sites becomes less relevant over time for their AI related HPC data centers. Iren can simply build new facilities with liquid cooling being the go-to cooling method, while maintaining their air-cooled data centers for $BTC mining. If newer Bitcoin mining hardware eventually requires liquid cooling, retrofits can be implemented as needed.

Lastly, Iren’s current data centers are closer to Tier 2 (T2) facilities than the high-end Tier 3 (T3) standard commonly seen in the industry. As previously mentioned, T3 facilities are characterized by very high redundancy, ensuring N+1 redundancy for all critical systems. While Iren’s BTC mining data centers achieve excellent uptime, their redundancy levels are not on par with T3 facilities. The nature of BTC mining allows for cost efficiencies by avoiding the high energy consumption and infrastructure costs associated with T3 redundancy.

However, redundancy becomes more important when providing AI computing services to external clients who expect seamless, reliable operations. Iren’s executives have publicly debated whether T3-level redundancy is necessary for all AI HPC facilities, but they have also affirmed that Iren can enhance power redundancy and resilience at all its sites, achieving T3 level if needed. Adding this extra redundancy certainly seems to be feasible, especially if a large hyperscaler client requires it for a colocation deal.

PodTech Co-Founder & Strategic $IREN Shareholder: @brianfry01

IREN’s CSP Growth Timeline

Soon after $IREN installed its first batch of 248 H100 GPUs, the units were deployed for its first major AI client, Poolside, a startup that develops AI-driven coding tools for software developers. This partnership marked the launch of Iren’s new business segment, establishing Iren as a cloud service provider (CSP), in February 2024.

Poolside recently raised $450 million, valuing the company at around $2 billion.

Two weeks after launching its cloud segment with its first client, Iren tripled its AI compute capacity & purchased an additional 568 H100 GPUs, reaching a total capacity of 816 GPUs. Very content with the service, Poolside quickly upscaled its contract from the initial 248 GPU capacity to 504 GPUs. Poolside even went as far as praising Iren publicly at the ‘NVIDIA GTC 2024 AI’ conference back in March of this year, claiming that they provided the best service out of the ~8 cloud providers they’ve tested. By July, Iren filled the remaining capacity with several smaller clients, reaching full utilization of their 816 GPUs.

However, Poolside and Iren parted ways in September. This move followed Poolside’s $450 million funding round and was likely driven by their need for a significantly larger AI cluster, possibly requiring thousands of GPUs—well beyond what Iren could provide at the time.

Nonetheless, in less than 2 months, Iren managed to close most of the hole Poolside left. The company now boasts an impressive revenue growth of 145 %, growing its AI driven top line from $408k to $1 million in a span of 8 months. While most of this growth occurred between March - July, the recent 1080 large H200 acquisition sets the company up for a promising growth path over the coming months.

As Iren prepares for the upcoming Blackwell generation of GPUs and incorporates closed-loop liquid cooling into their data center designs, the company faces multiple strategic options within the evolving AI-HPC market.

On one hand, Iren could continue expanding its own fleet of GPUs, potentially accelerating growth through GPU financing and turning its CSP business into a true flywheel. On the other hand, the company could secure colocation deals with hyperscalers, acting as infrastructure hosts for much larger CSPs.

Given the scale of Iren’s access to power, it’s likely they will pursue both strategies simultaneously. The financial implications for this relatively small company pursuing either of these two avenues could be significant. In the next section, I will explore the financial impacts of these strategies in great detail.

AI: Financial Metrics & Models

While Iren hasn’t yet disclosed pricing on their upcoming H200 units, they currently charge between $2 and $2.5 per GPU hour for their H100 fleet, positioning them as one of the most competitively priced options in the market. In comparison, CoreWeave, one of the largest CSPs, charges in the low to mid $4 range for H100 GPU hours.

In my opinion, Iren’s low pricing strategy for its AI cloud services is a smart approach, considering its status as a new entrant in the sector. By undercutting competitors, Iren can gain market share more rapidly and establish its presence in this field. The company’s full vertical integration—owning everything from the land and data centers to the energy and GPU infrastructure—supports this strategy by allowing it to maintain lower operating costs compared to competitors who rely on third-party infrastructure. This setup enables Iren to offer lower prices without significantly compromising its margins, effectively leveraging its cost efficiencies to attract clients and build a sustainable business model.

Iren has publicly stated that they anticipate a payback period of approximately 2 years for their H100 fleet. After analyzing their numbers— including hardware costs, GPU-hour pricing, projected electricity expenses, and uptime—I found their estimate to be largely accurate. My calculations indicate a payback period of around 26 months, which closely aligns with their guidance.

Interestingly, Iren acquired its recent H200 batch for roughly the same cost per unit as their current H100 fleet. This is likely because the larger order size for the H200 batch allowed them to benefit from economies of scale. The $NVDAH200 is currently the latest & probably last version of the Hopper architecture (next generation is Blackwell). Given that the H200 is about 50-100% more effective in inference and training, it’s reasonable to expect a higher price per GPU hour compared to the H100.

Although we don’t yet know Iren’s pricing per GPU hour for the H200, it’s reasonable to assume they’ll charge at least 10-30% more, given the improved performance of this new GPU. This would give us an estimated pricing range of $2.5- $2.9 per GPUh.

Model 1: GPU Payback Period

Model 1: Components:

● Pricing H100: Used mid point of ($2-2.5 pricing range) = $2.25

● Pricing H200: 22% premium relative to H100 pricing = $2.75

● Energy Costs per GPUh: Used company’s guidance

● Cost/Unit: Used company’s reported costs per batch (includes their estimated costs for most other hardware expenses needed to run AI clusters (CPUs, InfiniBand, etc).)

● Time Deployed: My own estimate (assuming near 90% usage in year 1 & near 100% uptime in year 2+)

Keep in mind, the biggest unknown variable and risk in the CSP industry is the useful lifespan and depreciation of these GPUs. While the hardware itself can last at least 5-7 years (in optimal conditions), pricing for GPU hours generally declines for older models as new & more capable generations are released.

Nvidia’s track record shows it’s launching a new generation every 2-3 years. The Ampere Generation launched in 2020with the A100, the Hopper generation (H100 & H200) launched in 2022, and now in 2025 we are getting the Blackwellgeneration with the introduction of the B200 GPU model. If you’re wondering why Nvidia is launching its next generation with the B200 instead of the B100, it’s due to design flaws that led to the (likely temporary) cancellation of the B100, prompting Nvidia to prioritize the B200 for release.

The best case study of GPU pricing power over time is the A100, as it’s the oldest AI GPU model. When the A100 launched in 2020, pricing among most CSPs generally ranged between $3 and $4 per GPU hour due to high demand and limited supply. Over the past four years, market prices have dropped by approximately 50%, with current market rates typically between $1 and $2 per GPU hour. However, major hyperscalers like Azure still charge up to $4 per A100 GPU hour, while charging ~$7 per H100 GPUh.

It’s also important to note that the A100 had roughly half the hardware cost of the H100, meaning hardware margins remain nearly on par with the H100, even 4 years after its release and 2 years since the Hopper generation entered the market. Any serious cloud service provider (CSP) that invested in A100s 3-4 years ago has likely seen substantial profits by now.

Given Iren’s great level of service in combination with highly competitive pricing for H100 GPU hours, I believe the company’s H100 & the estimated H200 rates are unlikely to drop significantly over the coming year. While it’s certain that the H100 market rates will eventually decline as the H200 becomes more widely available & the new Blackwell generation rolls out, Iren’s current low pricing offers a solid buffer. Even if a competitor like CoreWeave were to reduce its H100 pricing by 25% (from ~$4 down to $3), they would still be ~33% more expensive than Iren.

However, even with some potential diminishing of pricing power over time, I believe it’s still reasonable to expect Iren’s H100 payback period to remain below 3 years. Meanwhile, their H200 batch, acquired at a significantly more favorable price, is likely to achieve a payback period of under 2.5 years.

Model 2: GPU Cash Flow Model

Below, I present a model featuring two tables. Table 1 illustrates the 7-year cash flow progression for a self-funded $100 million investment of B200 GPUs. Table 2 shows the same investment scenario, but with the use of GPU financing (debt) to fund the purchase.

GPU financing is an asset-based loan where companies use the hardware they are purchasing with the debt as collateral. This type of debt typically carries higher interest rates, around 10%, and has shorter terms (often 3-4 years). The advantage for debtors is that it limits risk to the hardware value itself, without jeopardizing the company’s overall financial health. For example, CoreWeave raised $2.3 billion from Magnetar Capital and Blackstone, using newly purchased Nvidia H100 GPUs as collateral to expand their AI infrastructure. Similarly, Lambda Labs secured $500 million from Macquarie Group to acquire additional Nvidia GPUs, leveraging the same model​.

Cost of Investment

In both tables, the total investment cost is $100 million, most of it being allocated to the GPUs (i.e. AI hardware costs). Nvidia CEO Jensen Huang estimates B200 GPU prices between $30,000 and $40,000, likely influenced by order size. For this model, I used the midpoint of this range ($35k). While slightly optimistic given the batch size, it provides a scalable figure for further projections (later models). As Iren’s CSP business grows, its relationship with Nvidia should help secure prices closer to the lower end of future price ranges.

I estimate an extra cost of $5k per GPU for hardware expenses needed to set up AI clusters (Networking Infrastructure, Storage Systems, CPUs, Memory), bringing the total cost per unit to $40,000. Acquiring 2,163 units will result in a total AI hardware cost of approximately $86.5 million, which leaves us just enough remaining capital to cover the necessary DC infrastructure expenses for this investment.

To calculate the infrastructure cost for building the required HPC/AI data centers, we first need to determine both the total MW demand and the cost per MW of build-out. A full-spec B200 GPU consumes up to 1200 watts, or 0.0012 MWper unit. Due to the high power consumption of this GPU generation, liquid cooling is necessary. Currently, Iren’s air-cooled data centers achieve an average PUE of 1.05, meaning that 5% of the energy used for the computing (AI workload) is used on top for cooling & redundancy. However, liquid cooling is more energy-intensive, and thus I expect the PUE to increase to around 1.2-1.3 to account for both higher cooling demands and increased redundancy. Using the higher end of this PUE estimate, we can calculate the required power for this $100 million setup: 2,163 (Units) x 0.0012 (MW) x 1.3 (PUE) = 3.37 MW.

With the planned redundancy upgrades and the shift to liquid cooling, I anticipate that costs per MW will be substantially higher than the $650k/MW for Iren’s current air-cooled data centers. Based on discussions with relevant people in the field, costs are likely to range from $3 million to $5 million per MW. Using the midpoint of $4 million per MW provides a reasonable estimate for this calculation.

This leads us to a total infrastructure cost of ~$13.5 million (3.37 MW x $4 mill). Adding the AI hardware cost of $86.5 million, we reach the total investment cost of $100 million.

GPU Financing Terms

GPU financing typically comes with high interest rates (8-12%) and relatively short payback periods (3-4 years). For this model, I used an interest rate of 10% and a 4-year payback period. While repayment structures can vary, I applied annuity amortization, where each payment covers both interest and principal, resulting in fixed annual payments. Based on these financing terms ($86.52 million, 10% interest, 4 years), the annual payment calculates to $27.3 million, which leads to a total repayment of $109.2 million.

Iren could also potentially secure alternative repayment structures, such as the interest-only with balloon payment method or straight-line amortization. In the interest-only structure, Iren would pay only the interest annually ($8.652 million per year) and then the full principal of $86.52 million as a balloon payment in the final year. This approach results in a total repayment of $121.128 million. While it allows maximum cash flow for reinvestment over the first three years, the large final payment carries more risk.

Alternatively, the straight-line amortization structure divides the principal evenly, with interest calculated on the declining balance. This method starts with an annual payment of $30.282 million, decreasing each year, with the final year’s payment down to $23.793 million, totaling $108.15 million over four years.

Each approach has its strengths and weaknesses, but all enable a cloud service provider like Iren to scale rapidly with minimal self-funded costs, requiring only the self-funded infrastructure investment of ~$13.5 million to build their own data centers.

Cash Inflow

The cash inflow over the years depends mainly on the rate of decline in GPU-hour pricing and the uptime (operational hours) of the GPUs each year. For this model, I used a 7-year timeframe, reflecting the expected lifespan of these GPUs. Industry consensus suggests a usable lifespan of 5-7 years, but given Iren’s high-quality, AI-specific data centers— which I expect to meet the same quality standards as their BTC mining sites—I anticipate the B200 GPUs will achieve the full 7 years, as they’ll likely operate in optimal conditions.

I started with a conservative B200 GPU-hour price of $3.25 for year 1. Given that major CSPs like CoreWeave still charge around $4-$4.5 per H100 GPU-hour and Azure charges up to $4 for A100 GPU-hours, I believe $3.25 is a very conservative estimate that aligns well with Iren’s strategy of undercutting competitors. From there on, I anticipate a gradual price decline: a 10% drop in year 2, a 20% drop (of year 2) in Year 3, and 30% annually from Year 4 onward, reaching $0.56 per GPUh in Year 7.

For uptime, I modeled a slight decline of 1% per year, starting at 99% in the first year and decreasing to 93% in year 7. This reflects the idea that, in theory, lower GPU-hour prices should sustain demand, helping balance supply & demand. However, achieving this equilibrium price becomes increasingly challenging as demand weakens over time. Companies that respond slowly to these demand drops by keeping prices high are risking reduced uptime, as the misalignment with demand results in lower utilization. Maintaining a dynamic pricing strategy is therefore critical, as it allows Iren to hold uptime relatively steady and thus maximizes the total yearly cash flow & thus ROI.

Interestingly, in this model, even with a GPU-hour price of just $0.56 and an uptime of “just” 93%, Iren still generates positive cash flow in the final year and even maintains margins above 75%. This highlights the advantage of full vertical integration, as Iren avoids infrastructure leasing costs, with primary expenses limited to electricity, customer support, and maintenance.

Electricity Costs

I estimate electricity costs at approximately 3.5 cents per kWh, or $35 per MWh. As detailed in the ‘Infrastructure’ section, I based this figure on the assumption that all AI operations will occur in Texas, where Iren’s largest sites are located and where electricity rates average around 3.5 cents per kWh (without curtailment).

Support & Maintenance Costs

This category includes both technical maintenance and customer support expenses. Customer support here represents the cost of on-the-ground staff configuring and adjusting AI clusters as needed for clients. Maintenance costs are expected to remain stable at approximately $50k per MW. Support costs, however, vary with churn rates and operational scale. From discussions with industry professionals, support and maintenance expenses should range between $250k and $400k per MW, depending on scale & churn. For a conservative estimate, I used an annual figure of $350k per MWin this model.

Model 2: Analysis & Commentary

Comparing the light-blue boxes in each table provides us with some key insights. GPU financing requires significantly less self-funded capital, as most costs are covered by debt, resulting in a higher ROI and shorter payback period. However, it also reduces total net cash flow due to debt’s interest. The payback period in the GPU-financed model is just 5 months, as it’s calculated using self-funded costs rather than the full loan term. This approach also influences ROI, which is also based on the self-funded capital, resulting in a much higher return in the financed model.

In Table 1 (100% self-funded), Year 1 shows a significant negative net cash flow due to the large initial investment. By contrast, Table 2 (GPU-financed) maintains relatively steady positive cash flow each year, although Year 4 stands out as the weakest cash flow year within this 7-year model. This dip is a result from persistently high amortization costs, which continue as cash inflows gradually slow.

If Iren opted for the aforementioned “balloon-style” amortization, cash flow in Years 1-3 would be substantially higher, but Year 4 would experience sharply negative cash flow due to the final large payment. Alternatively, with “straight-line” amortization, payments would gradually decrease, creating more balanced cash flow over the four years; the initial years, especially Years 1 and 2, would see slightly lower net cash flow but avoid the steep drop-off in Year 4.

In my view, GPU financing offers clear financial advantages, though it carries higher risk. If the GPU investment underperforms, debt repayment could become challenging, even with the last-resort option of liquidating assets. This “last-resort” option is far from ideal, as it could disrupt service to existing customers and thus severely damage Iren’s reputation.

That said, given the favorable ROI of the GPU financing model, it appears to be a worthy risk, especially for a smaller company like Iren with limited capital. A reasonable approach might be to start with a relatively small $50-100 million loan. This should be manageable even if the GPU investment underperforms, as Iren’s BTC mining operations could potentially cover the debt payments without the need to liquidate the GPUs.

After observing initial success, Iren could double down on GPU financing, potentially with much larger amounts, and thus expand the CSP business on a much larger scale.

Model 3: CSP Flywheel Model

The Flywheel model builds upon the 1. model (GPU Cash Flow Model) by demonstrating the compounding effect of reinvesting 50% of each year’s operating cash flow into new GPUs, which, in turn, generate additional cash flow. Operating Cash Flow is defined as net cash flow minus the initial investment expense.

This model is, of course, a simplification of reality. In practice, it’s highly unlikely that each GPU purchase will yield identical ROIs and financial performance year after year. Although Nvidia CEO Jensen Huang has stated that Nvidia aims to price its GPUs to allow CSPs to achieve stable ROIs and consistent payback periods, significant fluctuations are inevitable. Predicting supply & demand dynamics with accuracy is challenging, and market conditions, technological advancements, and competitive pressures can all impact the realized returns of each GPU batch.

I used a 50% reinvestment rate instead of 100% to showcase how much additional cash can be generated for other purposes, such as investing in other growth avenues like purchasing additional land or even returning cash to shareholders through share buybacks or dividends. This approach of not reinvesting everything into a single ‘basket’ presents a more balanced and realistic strategy. The ‘saved’ cash figure can be found in the 7/10-year stats tables to the right of each table under the term ‘Cash Accumulated (saved)’.

This model includes three tables: Table 1 represents a fully self-funded version with no debt, building on Table 1 from the previous model. Similarly, Table 2 expands on the prior model’s Table 2, illustrating the flywheel effect that could result from a $100 million GPU-financed investment. Table 3 goes a step further by incorporating a second loan, enabling a $200 million investment in the third year.

The green fields display the annual net cash flow, calculated as the sum of the fields directly above them. The first row in each table replicates the net cash flow figures from the previous model. Pink fields represent the reinvestment of 50% of the prior year’s operating cash flow. Each of these reinvestment rows projects the 7-year cash flow progression of this reinvested amount, following the same cash flow pattern as in Table 1 (self-funded) from the prior GPU Cash Flow Model.

Model 3: Analysis & Commentary

Each scenario in this model demonstrates the potential for sustainable cash flow growth well beyond Year 7, when the first batch of GPUs reaches the end of its usable life. In the fully self-funded scenario, a single $100 million investment can yield nearly $100 million in annual cash flow a decade later, with over $350 million accumulated in saved capital over the ten-year span.

Interestingly, in the ‘1x Financing Table’, Year 4 avoids a sharp cash flow drop thanks to the additional cash flow generated from reinvested capital in previous years. It’s also notable that the self-funded flywheel generates about twice the total net cash flow over 10 years compared to the 1x financed model. That’s because, in the self-funded case, the first 4 years yield significantly more operating cash flow, thanks to not having to pay any amortization expenses, allowing capital to compound earlier.

However, the ‘2x Financing Table’ ultimately takes the lead in returns. Although it carries more risk due to taking on 2 separate loans, this scenario produces the strongest flywheel effect, due to the substantial capital that is deployed in the first three years.

Model 4: Colocation

This model projects Iren’s potential cash flow per 100 MW of colocation with a large-scale CSP (hyperscaler). It includes two sections: the top section models projections for Iren constructing data centers with liquid cooling and redundancy upgrades but maintaining T2 facility standards. The bottom section explores the scenario of Iren building fully customized T3 facilities to meet specific needs of a major hyperscaler.

T2 - Colocation:

For the T2 facility projection, I used the same infrastructure cost of $4 million per MW as in previous models, since the data center type remains identical to those depicted earlier.

I estimated revenue at $1.5 million per MW, based on a recent colocation deal where Core Scientific ( $CORZ ), another $BTC miner, agreed to provide 500MW of colocation to CoreWeave at an annual rate of approximately $1.45 million per MW. In this deal, $CORZ is retrofitting its existing BTC mining data centers at a cost of $1.5 million per MW. Given that this cost is insufficient to develop a full-scale T3 facility, I assume the resulting HPC/AI data centers will likely be T2 standard, with an estimated all-in cost of ~$2 million per MW. Since Iren’s T2 facilities are assumed to be built from scratch (rather than retrofitted) and cost ~$2 million more, their performance is likely superior to CORZ’s DCs.

Consequently, I think the $1.5 million per MW revenue figure used in this model is extremely conservative and I wouldn’t be surprised if the actual revenue per MW ends up being higher than even $2 million per MW. It’s also worth noting that Core Scientific recently emerged from bankruptcy in January of 2024, which may have given CW substantial leverage in the negotiation process (low ball offer). Unfortunately, public examples of recent AI/HPC colocation deals are pretty much limited to this one.

As previously laid out, maintenance & support expenses are estimated to range from $250k to $400k per MW, depending on scale and churn. For this colocation model, I used the lower end ($250k per MW), considering 100MW is substantial, equating to about 60-65k B200 GPUs with a PUE of 1.3, or roughly $2.5 billion in AI hardware. With a hyperscaler partner, churn is likely low since hyperscalers typically cater to large ‘Fortune 500’ type companies rather than AI startups. Thus, an annual cost of $250k per MW for support & maintenance seems reasonable.

For electricity expenses, I applied the same rate as in previous models: $0.035 per kWh.

T3 - Colocation

The main difference between the T2 and T3 colocation models lies in the higher infrastructure cost per MW and, correspondingly, the higher revenue potential per MW. The cost per MW of $12 million for a T3 facility is based on discussions I had with industry experts, while the estimated revenue per MW of $3.5 million is my own estimate.

Some might still consider $3.5 million per MW high, but this would result in a cost of roughly $0.62 per GPU-hour for the hyperscaler (($350 million / (64,200 GPUs x 8,760 hours) = $0.62 GPUh). Given that hyperscalers may charge well over $6 per B200 GPU-hour, an operating cost of less than 10% of its revenue would still yield substantial margins for the hyperscaler.

Model 4: Analysis & Commentary

It’s pretty clear that colocation offers lower ROI and thinner margins compared to the pure-play CSP model. However, it compensates with secure, long-term income streams. For example, CORZ and CoreWeave agreed on a 12-year colocation deal, providing CORZ with steady, predictable cash flow. For Iren, with its large-scale infrastructure portfolio, it would be impractical to use all this capacity solely for its own operations. Colocation deals, therefore, offer the most effective way to monetize a majority of its available MW capacity.

One consideration, however, is the high infrastructure cost involved in colocation. Since Iren would own and operate the infrastructure, it would be responsible for covering these expenses, one way or another. CORZ and CoreWeave addressed this by having CoreWeave, the much larger company, finance the initial infrastructure costs. CORZ then pays off this ‘loan’ through a 50% reduction in revenue from CoreWeave until the debt is settled. Such a financing structure prevents any single year of negative cash flow, instead spreading reduced cash flows across several years until the debt is repaid. I would expect Iren to employ a similar structure for any colocation agreement.

Notably, in the T3 scenario, despite infrastructure costs being 3x times higher than in the T2 scenario, revenue per MW is only about 2.3 times higher. However, the T3 model still has a shorter payback period and higher ROI. This is due to the stable cost of revenue - such as electricity + maintenance & support - which remains consistent across T2 and T3 setups. Consequently, the T3 model achieves higher margins, as fixed expenses stay flat while revenue increases, a benefit referred to as ‘operating leverage’.

One could argue that T3 facilities might incur higher maintenance costs due to increased redundancy requirements. However, of the $250k per MW in ‘maintenance & support’ costs, only about 10-20% typically goes towards maintenance, while most is spent on customer support (e.g., staff managing AI cluster configurations). Even with an adjustment for higher T3 maintenance needs, which could add costs of ~$20-30k per MW, would be too marginal to meaningfully impact the bottom line. The operating leverage of the higher-yielding T3 setup would remain in full effect.

AI Wrap-Up & Closing Remarks:

I want to re-emphasize that the biggest risk to the AI/HPC sector is the potential for GPU-hour prices to decline more rapidly than expected, which would erode ROIs. While gradual price reductions are anticipated and accounted for in my models, the lack of historical data for AI chips makes it difficult to predict the extent of future price drops, posing a risk that the pricing decreases could be steeper than anticipated.

This scenario could occur if the supply of AI computing power significantly outpaces demand, leading to more rapid drops in GPU-hour prices. This oversupply could result from both an increase in total GPU availability and advancements in GPU technology, which make newer models far more powerful while depreciating older generations.

If GPU-hour prices were to drop significantly more than anticipated in my models, the ROIs for the CSP-related projections would be heavily impacted (to the downside). Additionally, such a scenario would increase the likelihood of large-scale CSPs & hyperscalers re-negotiating terms with their colocation partners, i.e. their infrastructure hosts. This would then also reduce ROIs and alter the projections in my colocation model.

Despite the outlined risks, it’s crucial to consider the broader market dynamics. The current power crunch, coupled with the growing demand for AI computing, paints a different picture.

While there could be an oversupply of GPUs, energy constraints naturally limit how much of this additional compute power can be fully utilized. Even if NVIDIA and TSMC were capable of producing an unlimited number of GPUs, the challenge of insufficient power infrastructure would still persist. Simply put, available GPU supply cannot not be fully utilized without the necessary energy infrastructure.

I encourage everyone to watch the brief 1-minute video at the bottom of this chapter, featuring @kevinolearytv explaining the current power shortage affecting the AI/HPC sector.

Meanwhile, demand for AI compute is continuously rising. Google recently reported 35% annual growth in cloud revenues, driven largely by its AI-CSP business segment. Given that offering AI services has become an existential matter for many software companies, I believe that demand for AI computing will continue to grow in the coming years.

Considering these market dynamics, I remain optimistic that GPU-hour price depreciation following the release of the B200 won’t be too severe for at least another five years, or at least until the current power crunch is alleviated—whether through expanded solar generation or, as some suggest, the development of nuclear energy.

Although it’s not my base case, I believe there’s even a reasonable scenario in which compute demand exceeds available supply, causing GPU-hour prices to rise over the coming years instead of decline. In such a case, the likelihood of a temporary market bubble forming in AI-HPC-related stocks would be very high. I’ve observed that markets often overreact—either by correcting downwards or upwards—when reality turns out to be the opposite of consensus expectations. A scenario where GPU-hour pricing increases instead of decreases could certainly lead to such an event occurring, reminiscent of the surge in ‘stay-at-home’ stocks, which benefited significantly from sudden shifts in market dynamics during the pandemic.

Hence, given all this context, I consider the previously presented financial models to be based on reasonable assumptions that illustrate a highly compelling outlook for Iren’s potential in this emerging industry. Through GPU financing, the company could achieve annual pre-tax net cash flows exceeding $100 million within a few years. Alternatively, utilizing its 1.4 GW site for colocation could generate over $1 billion in annual pre-tax net cash flows for T2 facilities and more than $3.5 billion for T3 facilities.

Nonetheless, one thing is certain: the CAPEX requirements in this industry are substantial. While Iren is well-positioned with large-scale infrastructure sites and what appears to be ‘abundant’ energy access, it remains a relatively small company with limited cash reserves. Every cent of operating cash flow from BTC mining is reinvested into infrastructure expansion. At this stage in its business life cycle, Iren still depends on external capital to fuel growth. As an $IREN investor, recognizing this is crucial; while the company has immense potential, fully unlocking it will require additional capital.

While Iren currently has no debt, I believe it would be wise for the company to consider options such as GPU financing, issuing convertible debt, or initiating a new at-the-market (ATM) offering to raise substantial capital. I want to emphasize that I’m generally not a proponent of corporate debt, as it can hinder a company’s agility. Nonetheless, in this CAPEX-heavy industry, taking on debt is almost a necessity, particularly for newer players. With the AI/HPC industry still in its early stages, I believe now is the ideal time for aggressive investment into growth, particularly given Iren’s existing infrastructure setup.

Video Link (important to watch):

BTC Mining: Unprecedented Growth

Iren has rapidly established itself as the fastest-growing Bitcoin mining company in history. In 2023, it set a new industry record by increasing its operating hash rate by an impressive 250% within a single calendar year.

However, this record no longer stands, as Iren has once again broken it. Year-to-date, Iren increased its monthly operating hash rate by 14.26 EH/s, growing from 5.64 EH/s in January to 19.9 EH/s in October—a 252% increase that surpasses the company’s previous record. Iren is also not finished for this year, as it plans to install an additional 10 EH/s by year-end, bringing its total installed capacity to 31 EH/s and setting an industry-leading growth rate of over 400%.

This exponential increase in computing power has translated into a rapid rise in market share. Back in January 2023, Iren’s market share in BTC mining was a mere 0.6%. Fast forward to today, and that figure has grown nearly fivefold to 2.77%, reflecting a remarkable market share expansion in less than two years.

Market share is determined by dividing each miner’s monthly Bitcoin production by the total available BTC rewards distributed globally in that month. These rewards consist of newly issued Bitcoin and transaction fees, both of which are publicly trackable through online tools.

While other public miners have also experienced market share growth this year, Iren’s performance stands out.

Another noteworthy achievement for Iren is its ability to grow monthly BTC production from January to October of this year, despite the significant challenges posed by the Bitcoin halving event in April.

For those unfamiliar, the halving event, which occurs roughly every four years, reduces miners’ block rewards by 50%. While transaction fees are unaffected by the halving, they usually account for only a small portion of total rewards.

Despite having its mining rewards slashed in half this year, Iren still managed to increase production YTD—a feat only one other miner, CLSK, has achieved so far.

Iren could soon become the first major public miner to reach a new all-time high (ATH) in monthly BTC production. The company’s previous ATH was 508 BTC, set in May 2023, a month when transaction fees spiked dramatically and contributed nearly 17% of the network’s block rewards, compared to the usual 3-5%. As shown in the accompanying chart (below), most miners experienced a production surge that month. A similar spike in fees occurred in December 2023, helping miners like $MARA achieve their ATH in monthly production.

Since May 2023, Iren has not reached a new ATH in monthly production. This is largely because its hash rate remained flat from June 2023 to January 2024 while the company prepared for its massive expansion into Childress.

Keep in mind that if a miner’s hash rate remains constant, its production will naturally decline over time, as the global network’s hash rate steadily rises with new and existing mining companies adding more computing power.

Although the halving has delayed the achievement of a new production ATH, I believe we are only a few months away from seeing it. With Iren scaling towards its goal of 31 EH/s by year-end, the company should soon have enough computing power to break its production record.

The above charts illustrate how Iren seemingly emerged from nowhere to rapidly establish itself as the fourth-highest-producing public mining company. It’s also becoming clear that the industry is evolving into a “four-player race” among $MARA, $CLSK, $RIOT, and $IREN. With Iren’s aggressive target of 50 EH/s for H1 2025, I believe the company is poised to become the third, or potentially even the second, highest producing publicly traded BTC miner globally.

Frankly, when considering Iren’s current valuation, it seems the market hasn’t fully caught up to these developments. I’m active on X, and Iren still doesn’t appear to receive due credit for its rapid ascent. However, with an additional 10 EH/s scheduled to come online in the final weeks of this year, I expect that to change pretty fast.

Give it another 1-2 months, and I believe Iren will be widely recognized as one of, if not THE, most capable publicly traded mining companies world.

BTC Mining: Industry-leading Efficiency

Besides being the fastest-growing BTC mining company, $IREN is also currently the most profitable. In October, the company achieved an exceptionally low energy cost of just $20,834 per BTC. Since energy expenses make up the vast majority of a miner’s operational cash costs, this positions IREN as the leanest and most cost-efficient operator in the industry.

To illustrate this, I compiled data from the eight largest BTC miners, comparing their fleet efficiency, average electricity rates, and resulting energy costs per BTC. For each column, I used the most up-to-date information published by each respective mining company.

For context; joules per terahash (J/TH) measures the energy efficiency of a Bitcoin mining fleet by showing how much power (in joules) is used to generate one terahash of computing power. A lower J/TH value indicates higher efficiency, as less energy is required per unit of hashing power. Efficient fleets (with lower J/TH) consume less electricity, reducing operating costs and improving profitability.

Despite already having the lowest power cost per mined Bitcoin, Iren is set to widen its lead with further efficiency gains toward year-end. While Iren’s current mining fleet of 21 EH/s already operates at an industry-best efficiency of 16 J/TH, the company is poised to enhance its edge by adding another 10 EH/s using Bitmain’s latest S21XP ASICs, which operate at an impressive 13.5 J/TH. This addition will reduce Iren’s overall fleet efficiency to just 15 J/TH, translating to an estimated 6.25% reduction in electricity costs, all else being equal.

Furthermore, these additional 10 EH/s will be deployed at Iren’s Childress site in Texas, where energy rates are significantly lower than at its Canadian locations—helping Iren achieve a lower overall average electricity rate. In October, Childress achieved a rate of just 3.06 cents per kWh, while Iren’s Canadian sites averaged a higher rate of 4.51 cents per kWh. Since Canadian sites currently account for around 47% of Iren’s operations, this increased the average rate to approximately 3.74 cents per kWh for October. Adding another 10 EH/s at Childress is expected to bring Iren’s blended electricity rate down to 3.52 cents per kWh, a reduction of 5.88%.

Together, these compounded efficiency gains could reduce energy costs by $2,453, bringing Iren’s energy cost per BTC down to a remarkable $18,381 (all else being equal). This means that, beyond already having the industry’s lowest energy costs by a wide margin, $IREN is on track to reduce power costs per BTC by an additional 10-12% over the next six weeks.

However, as it stands, with Iren’s current direct mining expenses, the company generates $29 million in monthly ‘direct mining cash flow’ (revenue minus energy expenses), which equates to almost $350 million annually.

BTC mining: Diminishing Returns or Market Opportunity?

Despite recent improvements in mining economics (i.e., BTC price making new ATHs), the mining industry has faced severe economic challenges over the past two years. Rising competition and Bitcoin prices that lagged behind network growth have pressured miners’ profitability, making it difficult for market participants to achieve positive free cash flow. One of the largest cash costs, often overlooked, has been the extensive CAPEX needed to capture and defend market share. Although classified as an investment, CAPEX is a substantial cash burden, particularly in a sector where capital requirements for infrastructure, hardware, and energy access are immense.

The recent Bitcoin halving only intensified these pressures by slashing block rewards in half, cutting operating cash flow overnight. The BTC network’s reward system further compounds this challenge: Bitcoin’s fixed issuance schedule means that, as the network’s hash rate rises, the rewards available to each miner shrink unless they expand at the same pace. This dynamic creates an environment of diminishing returns for those unable to match the network’s growth rate.

Today’s landscape is largely shaped by decisions made during the 2020-2021 bull run, a period of exceptional mining profitability. Both public and private miners invested heavily in new ASIC miners, anticipating sustained high rewards. However, with deployment often lagging behind equipment purchases, much of this new hash rate only came online during the bear market when BTC prices declined sharply. This influx of hash rate, combined with declining BTC prices, tightened mining economics through 2022 and 2023, intensifying competition and reducing profitability.

The past two years of tight mining economics, compounded by this year’s halving event, have resulted in a stagnant hash rate over the past 6-8 months. During this period, many inefficient or smaller miners halted operations or exited the industry. Only the most well-capitalized mining companies, primarily publicly traded ones, could afford to continue expanding. By tapping public markets, often through ATMs, these public miners secured funds to push forward with growth plans. This access to capital has allowed the top four public miners to capture a growing share of the BTC network while smaller or private players have struggled to keep pace.

Now, with mining economics improving as Bitcoin prices outpace the network’s hash rate growth, these leading public miners are positioned to capitalize on their scale and modernized fleets. While BTC economics are currently improving rapidly with Bitcoin reaching new highs, a good amount of “idle” hash rate likely remains on standby, ready to activate if mining economics continue to improve. This idle capacity mainly consists of older, less efficient machines shut down after the halving, yet primed to restart if BTC prices rally further. With the network hash rate around 800 EH/s, I estimate that an additional ~100 EH/s could quickly come online as conditions improve.

However, I believe the worst is behind the BTC mining industry. I don’t expect the network hash rate to continue growing at the pace seen over the last five years. Mining Bitcoin has evolved into a highly specialized process requiring substantial technical expertise, capital, and vast energy resources. The era of individuals plugging in ASICs at home has ended; these machines now require optimal cooling, consistent maintenance, and access to substantial energy resources.

Going forward, the network’s hash rate growth will likely come primarily from commercial operations and potentially government initiatives. This expansion isn’t solely reliant on Bitcoin’s price increase; it also demands significant energy resources and capital. The AI boom has made energy access even more challenging, with BTC mining firms now competing against some of the world’s best-funded companies for power. Land and energy costs are climbing, and lead times for essential data center components like transformers are lengthening. Physical limitations are increasingly becoming significant barriers to hash rate expansion.

While Bitcoin’s price can theoretically grow indefinitely, the same cannot be said for BTC’s hash rate, which remains heavily constrained by energy availability. Access to energy is quickly becoming the primary bottleneck for both AI compute and BTC mining. This issue can only be resolved by generating new energy. Although some advocate for nuclear energy, building new reactors takes a decade or more, and reactivating older plants is a 3-5 year process due to safety protocols. In my view, solar will likely play a substantial role in resolving this constraint, though even that will take time. Industry experts, such as Tony Seba, are projecting an era of energy abundance driven by solar generation to commence toward the end of this decade or the beginning of the next—still at least five years away. Thus, the BTC mining sector, much like the AI industry, may face another 5-6 years of constrained energy availability, limiting hash rate growth.

These industry challenges have effectively raised the barriers to entry for new players. Competing in this field now requires hundreds of millions in CAPEX for power studies, land acquisition, power contracts, data center construction, hardware procurement, and, crucially, access to low-cost energy. Today, primarily ‘best-in-class’ public mining companies fulfill the requirements to compete effectively. Private mining companies, and especially individuals, can no longer keep up. Therefore, even if Bitcoin prices rise sharply in the coming years, it’s unlikely that we’ll see a substantial influx of new entrants—and, consequently, a rapid increase in hash rate.

Possibility of Major Nation-States Entering the BTC Mining Space

A primary counterargument to the notion of stalling hash rate growth is the potential entry of nation-states into the BTC mining space, particularly BRICS-aligned countries with cheap and seemingly unlimited access to energy. Countries like Russia, China, and Saudi Arabia could see BTC mining as a strategic opportunity, leveraging their energy advantages to add significant hash rate to the network. Hypothetically, if BTC became a strategic reserve asset, these nations could collectively introduce tens of gigawatts of mining power over the next few years, adding hundreds of EH/s and putting pressure on mining economics by introducing formidable, government-backed competitors.

While this scenario raises valid concerns, the practical execution of such a strategy by nation-states would face several constraints. Governments are typically slower and less agile than the private sector, in part due to bureaucratic complexities. Furthermore, government institutions, regardless of the country, often lack the efficiency and competence of private enterprises because they are generally not profit-driven and do not operate under the pressure of financial viability. Unlike private firms, government agencies rarely face the existential threat of bankruptcy if they fall short of goals, and this lack of pressure dampens the incentive to strive for excellence.

A notable example of this dynamic can be seen in the aerospace industry, where SpaceX has made extraordinary strides. Despite decades of experience, vast resources, and a considerable head start in aerospace, major nation-states have been rapidly outpaced by SpaceX, which now holds over 50% of the industry’s market share in launches and continues to push the boundaries of innovation. This progress has been driven almost entirely by the forces of capitalism, highlighting how the private sector’s agility and profit motive can accelerate advancement and dominate even the most established, government-led industries. The BTC mining sector, like any other industry, is likely to follow a similar trajectory.

Moreover, the economic impact of BTC mining on a nation’s revenue or GDP would likely be minor, especially for countries with substantial economies. The revenue generated from BTC mining, while attractive to private companies, represents only a fraction of what a large country would require to see meaningful economic benefits. If a nation-state were intent on accumulating BTC in large quantities, direct purchases from the open market would be a far more efficient and immediate approach than entering the mining sector.

Where BRICS-aligned nations could derive strategic benefit from BTC mining is in securing their ability to transact freely on the network, particularly if member states face economic sanctions. By collectively controlling a modest portion of the BTC network—likely around 10-20%—these countries could ensure that their transactions remain unimpeded by the influence of other, potentially adversarial, network participants. I believe that BRICS countries would have little incentive to grow their network ownership beyond this threshold, as the ability to transact freely is the only benefit to them. If they aim to hold large quantities of BTC as a strategic reserve asset, this goal can only be achieved by purchasing BTC directly from the markets, given that over 94% of the entire BTC supply has already been mined.

In summary, while nation-state entry into BTC mining could add incremental hash rate growth over the coming years, it is unlikely to drive the same exponential growth seen in the past decade. The barriers to large-scale, rapid deployment and the relatively limited economic upside reduce the likelihood of massive, sustained hash rate expansion purely from state-sponsored mining initiatives. Instead, I expect network hash rate growth to continue relying primarily on the private sector’s agility and expertise. Given these dynamics, I see the potential “risk” of nation-states entering the mining space as relatively low, especially during this current four-year cycle ending in 2028. As a result, I anticipate continued improvements in mining economics driven by a deceleration in hash rate growth.

Implications for IREN

In short, improved mining economics will directly lead to higher margins. Based on my estimates, I expect Iren’s gross margin to increase from its current ~73% to around ~77% over the next 18 months. During this period, I also anticipate Iren’s monthly BTC production to rise substantially, supported by the additional 20 EH/s that Iren plans to install over the coming six months.

In the previous cycle (2020-2024), monthly network hash rate growth averaged 3.9%. Given the current market dynamics, I expect this growth rate to slow to approximately ~2-3% MoM, at least until the next halving. At this rate, the network hash rate would average around ~950-1000 EH/s during the one-year period after Iren reaches its 50 EH/s target (Q2 2025 - Q2 2026), effectively positioning Iren with a >5% market share. Over the same timeframe, I anticipate BTC prices to rise at a much faster pace, leading to an increase in margins and profitability.

To model Iren’s gross profit during this period, I created two tables. Table 1 reflects more conservative assumptions, with an average network hash rate of 950 EH/s and an average BTC price of $120k. In contrast, Table 2 represents my actual expectations, with an average network hash rate of 1,000 EH/s and an average BTC price of $200k for the Q1 2025 - Q2 2026 period.

Given these projections, I view $IREN as an exceptionally “cheap” investment, especially with its market cap at just over $2 billion.

IREN’s Risks

Every investment carries risks, and Iren is no exception. However, I believe its risks are relatively modest in relation to its potential upside. That said, here are the key risks associated with Iren:

BTC Mining Economics Worsening

Although Iren has recently expanded into the AI sector, its mining activities still account for the majority of its revenues and cash flow. A key risk to Iren’s margins is the potential for worsening BTC economics, driven by network hash rate growth outpacing BTC price increases. Given Bitcoin’s inherent price volatility, this scenario could unfold relatively quickly, particularly in the event of a ‘black swan’ event.

However, I view this risk as more short-term, and if it were to materialize, it is likely to subside relatively quickly. As discussed in detail earlier, I believe BTC economics are set to improve over the coming years, driven by factors that will constrain network hash rate growth. Thus, I won’t revisit those points here.

It’s also important to highlight that if mining economics were to deteriorate, such as during a bear market, Iren would likely be the least impacted among publicly traded miners, given its significantly higher margins. In such a scenario, other miners would likely face greater challenges, presenting Iren with an opportunity to capture additional market share.

Share Dilution

The BTC mining industry is inherently plagued by one major issue: share dilution. As a capital-intensive sector, any new entrant must invest heavily to capture significant market share. Given the industry’s relatively young age, all current miners have been new entrants in recent times, requiring substantial capital to get started.

Many of these companies, Iren included, are still in the early stages of their business lifecycle, which means they rely on external capital to fund growth, often enduring losses and negative free cash flow (FCF). I would argue that Iren is currently transitioning out of the “hypergrowth” stage in @BrianFeroldi’s Business Growth Cycle and moving into the “Self-Funding” stage—at least in terms of its BTC mining operations.

Entering the “Self-Funding” stage doesn’t mean Iren won’t continue to experience significant growth, but it does indicate that future growth will primarily be financed through operating cash flow. I expect Iren to have one larger ATM offering to fund its expansion from approximately 35-40 EH/s to 50 EH/s and possibly raise some additional capital to support its entry into the AI space. Nonetheless, as shown in my BTC mining-related cash flow models, Iren should generate well over $500 million in operating cash flow over the coming years, which will likely fund most, if not all, of its future expansions from there on.

Looking in the rearview mirror, there’s no denying that Iren has been highly dilutive in recent years, with its share count increasing by 352% since January 2022. While I’m generally not a proponent of dilution, I must acknowledge that it was a necessary ‘evil’ to reach this point.

It’s also worth noting that, since its IPO, $IREN has been by far the most accretive BTC miner in terms of dilution, at least when measured by growth in BTC production and revenue. In fact, Iren is the only BTC miner whose revenue growth has outpaced its share count growth, meaning revenue per share has actually increased.

Keep in mind that Iren’s expansion from its current 21 EH/s to 31 EH/s by year-end is already fully funded through prior ATM offerings. This means Iren’s accretiveness will rise significantly in the coming weeks, as BTC production increases by approximately 50%. There’s little doubt that Iren will be the first—and potentially the only—miner to achieve production growth outpacing share count growth, resulting in a positive accretiveness in terms of mining output, despite the recent halving.

In summary, while dilution should always be taken seriously, the evidence clearly shows that, so far, this funding strategy has been highly accretive for Iren’s investors. Considering the company’s current position in its business growth cycle, it’s evident that the rate of dilution will likely slow dramatically going forward, potentially evencoming to a complete halt over the next two years.

Not Securing a Colocation Deal

Another potential “risk” is Iren failing to secure an attractive colocation deal with large-scale CSPs or hyperscalers. I place “risk” in brackets because this is more of an opportunity cost than a genuine threat to the business. While not securing a colocation deal would represent a significant missed opportunity, it’s also not the end of the world, as Iren is a company with a lot of optionality.

I initially invested in Iren purely for its growing dominance within the BTC mining sector; the entry into AI/HPC, while promising, was more of the cherry on top. However, this “cherry” is quickly turning into a significant part of the company’s business. The most important factor is that Iren has already secured the land & power, meaning that even if a colocation deal does not materialize, it can still leverage its valuable energy access for other business lines—whether to further expand its BTC mining operations or grow its own CSP business.

This level of optionality also means that Iren is under no rush to close a deal, which may explain why it has not yet secured a major long-term agreement for its 1.4 GW site. The energy crunch is expected to last for at least another 3-5 years, and as access to power becomes increasingly scarce, hyperscalers will likely grow more desperate, giving Iren more leverage in negotiations over time.

Overall, I view the chance of this risk materializing relatively low, likely in the 10-20% range, especially considering that Iren has recently confirmed it is actively engaged in discussions with multiple hyperscalers, with whom it has signed NDAs.

NYDIG & Levi & Korsinsky, LLP lawsuits

Iren is currently involved in 2 ongoing lawsuits, both of which it’s likely to win in my opinion.

Case 1: NYDIG

In early 2022, Iren secured a funding deal with New York Digital Investment Group (NYDIG), a financial firm focused on Bitcoin and digital assets. This loan provided Iren with the capital needed to enter the Bitcoin mining sector, allowing the company to purchase ASIC mining machines. Unlike a traditional loan, this was an asset-secured loan, meaning that the purchased mining equipment itself was the collateral. The deal marked Iren’s entry into the industry and provided the foothold needed to establish its initial operations in this new sector.

However, as Iren IPOed in late 2021 and began its primary operations with the NYDIG-funded ASICs just as the Bitcoin bear market set in during 2022, the cash flow generated from operations quickly fell short of covering the NYDIG loan. This resulted in Iren’s default on the loan. The only reason Iren avoided bankruptcy was due to its strategic choice of an asset-secured loan, limiting liability exclusively to the mining machines purchased with the loan. Ownership of the ASICs was transferred to special-purpose vehicles (SPVs) and subsequently to NYDIG, who ultimately bore the primary financial loss as the mining equipment depreciated rapidly in the bear market.

NYDIG later alleged deceptive practices surrounding the loan terms and subsequently sued. However, so far, Iren has seen very favorable outcomes in Canadian court, which dismissed the claims of fraudulent intent and misrepresentation. While this ruling was in Iren’s favor, some aspects were sent back to the trial court for further review. Additionally, PwC (PricewaterhouseCoopers), a global auditing and consulting firm overseeing the SPVs’ liquidation, is attempting to bring the case to Australia under the Cross-Border Insolvency Act, which is prolonging this case’s process.

In a worst-case scenario, Iren could face liabilities of up to $153 million. However, given the favorable Canadian rulings, I believe that NYDIG is most probably aiming for a settlement. Given that Iren’s executive team is packed with former Macquarie investment bankers with extensive experience in structured finance, I find it highly unlikely that Iren engaged in any wrongdoing. However, as the case continues to drag on, consuming Iren’s time and legal resources, I wouldn’t be surprised if they opted to settle for a small amount just to get it over with.

Even in the extremely unlikely worst-case scenario of paying the full $153 million, this wouldn’t break the camel’s back for Iren. The company is nearing a point where it could generate that amount in operating cash flow in less than one quarter. However, this would likely result in the company having fewer cash reserves to fund future expansions, meaning Iren could require more external capital, for example via ATMs. Nonetheless, while this is a risk, I give it a low chance of materializing due to the factors I just outlined.

Case 2: Levi & Korsinsky, LLP

The class-action lawsuit by Levi & Korsinsky, LLP against Iren was initiated earlier this year, claiming that the company misrepresented its capabilities and potential, especially related to its Texas data center sites. This lawsuit is based largely on claims made by Culper Research, an investment firm known for publishing short reports on companies to drive down stock prices. Culper released its report on Iren on July 11, 2024, alleging that the company’s Texas sites were likely unsuited for AI/HPC applications.

Culper’s report is, however, poorly researched. A closer look at the claims shows that they are easily debunked with deeper analysis. The report suggests that Iren’s Texas sites are unsuitable for AI/HPC because of their remote location, allegedly resulting in inadequate latency and network redundancy for AI inference workloads. Yet, as I detailed in previous sections, Iren’s Texas sites actually have very impressive latency (sub-10ms) and robust network redundancy, supported by multiple ISPs and fiber connections. This makes Iren’s Texas sites arguably better suited for AI/HPC than even their Canadian sites, where ALL of Iren’s current AI operations are based.

Further, Culper claimed that liquid cooling would be challenging at the Texas sites due to the lack of nearby water sources. This assertion overlooks the fact that closed-loop liquid cooling, which doesn’t require external water sources, is fast becoming the industry standard due to its superior operational efficiency. As previously highlighted, Iren has already confirmed that they are actively pursuing this cooling technology for their Texas locations.

In summary, Levi & Korsinsky’s case relies on a poorly substantiated short report by Culper Research, which appears to have been aimed at profiting from a stock drop by scaring uninformed investors. Given the shallow basis of these claims, I don’t believe this lawsuit will gain traction or present any substantial risk to Iren.

Australian HQ

One factor that isn’t necessarily a risk for the company but may act as a headwind for the stock is Iren’s Australian headquarters. While Iren’s operations are currently based in Canada and the US, the company remains headquartered in Australia, making it subject to Australian accounting laws and corporate regulations.

This can create a drag on the stock price, as US-based companies often carry a premium—partly due to the trust investors place in the US judicial and governance systems. Many of the largest investment funds in the world are based in the US, and numerous have internal policies that limit them to investing in US-headquartered companies. I’ve personally spoken to several fund managers, some managing billions in assets, who have expressed frustration at being unable to invest in Iren solely due to its headquarters location.

Moreover, US-based companies are generally subject to stricter reporting requirements, which often benefit investors. For example, public companies in Australia, like Iren, are generally only required to file half-yearly financial reports, as opposed to the quarterly reporting required for US companies. For most of the time since its IPO, Iren adhered to this structure, reporting biannually with six-month gaps between reports. They only recently began issuing quarterly reports, primarily out of goodwill, as they are not legally required to do so. However, some investors and funds could be deterred by the absence of legally mandated quarterly reports. Additionally, as an Australian company, Iren does not report in GAAP terms, which is another drawback for many investors.

Australia’s higher corporate tax rate, at 30%, compared to the US federal rate of 21%, is another factor to consider. While Iren has largely avoided the impact of this due to its unprofitable years, this higher tax rate becomes a more significant issue as Iren moves toward profitability. Given my expectations for Iren to become significantly more profitable in the near future, this higher tax rate could pose a disadvantage.

I understand why Iren hasn’t yet moved its headquarters, as the process of relocation is resource-intensive and time-consuming. The company is currently experiencing significant growth, and management is fully focused on reaching its ambitious expansion objectives. It’s also worth noting that, should Iren move its headquarters, there could be tax implications, such as an “exit tax” imposed by Australian tax authorities. While this is subject to confirmation, it could also be one of the reasons why Iren hasn’t moved to the US yet.

However, while management has not completely ruled out the possibility of relocating to the US, their response so far has been somewhat lackluster. My expectation is that Iren will eventually relocate its headquarters, given the substantial benefits this would bring to the company and its stock valuation. However, I believe this will likely happen after Iren achieves its 50 EH/s target in H1 2025. That said, I would appreciate more transparency and communication from management on this matter.

I strongly believe that Iren’s executive team should publicly acknowledge the potential advantages of relocating its HQ to the US and commit to pursuing this after reaching their current expansion goals. Many investors, as well as the market, would likely appreciate such an announcement.

On the bright side, if Iren were to announce plans to relocate its headquarters, I would expect a markedly positive response from the market. Given the strategic benefits, I anticipate this announcement will occur in the not too distant future, likely leading to a substantial valuation increase and creating significant upside for current investors.

Tornado Alley

Lastly, I want to briefly address the concern about Iren’s Texas sites being located in tornado alley. Tornado alley is traditionally thought to be a region in the United States where tornadoes are frequent, spanning across Texas, Oklahoma, Kansas, Nebraska, and parts of New Mexico and Colorado. This, in theory, could pose a risk to Iren’s data centers, especially if a strong tornado were to strike one of the sites, causing significant damage.

However, in practice, this risk has an extremely low probability of materializing. Even the short seller Culper Research, who scrutinized Iren extensively, did not raise this concern. In reality, only tornadoes rated EF3 or higher pose a threat of significant damage, if they were to pass directly over one of Iren’s data center locations. Over the past five decades, there has been just one notable case of an EF3 tornado passing over Childress, which occurred in 2001. Even if Iren had been operational at that time, the likelihood of the tornado directly impacting the data center would have been relatively low.

Further research by Mark Fishetti, Matthew Twombly, and Daniel P. Huffman, published in Scientific American, shows that tornado alley is actually shifting eastward, making it a less concerning risk for West Texas. This finding is supported by AccuWeather’s Lead Long-Range Meteorologist, Paul Pastelok, who has documented the same trend.

Finally, considering this logically, if tornadoes posed a significant risk to data center operators, why would some of the world’s largest tech companies invest billions in building data centers in Texas? Google is currently constructing a large data center site in Haskell, Texas, which is traditionally considered part of tornado alley. Similarly, data center developers Lancium and Crusoe Energy Systems are building a massive 200 MW site in Abilene, Texas, with plans to expand to 1.2 GW. This site, also located in (the old) tornado alley, is likely to be leased by Oracle in a partnership with Microsoft and OpenAI, based on recent reports.

In short, I classify this risk as extremely unlikely, more akin to a “black swan” event, such as World War III breaking out or a meteor striking Texas.

I’d also like to credit my friend @TheKamaHsutra, who shared much of the research that informed my understanding of this risk. Excellent work, Kama, and thank you again for allowing me to publish these findings!

Valuation & Price Target

I fully acknowledge that valuing a company like Iren is a challenging task, involving significant nuance and ambiguity. BTC miners, in general, are difficult to value, as much of their revenue and subsequent earnings depend on factors outside their control, such as the BTC network hash rate and the price of Bitcoin. Nonetheless, based on the extensive research I have conducted, I will attempt to provide a valuation and present my H1 2026 price target for $IREN in this final section of the report.

Valuation Comparison

First, however, I want to demonstrate how “cheap” $IREN currently trades in comparison to its peers. Below is a chart displaying the forward PS ratio of each of the top mining companies.

Calculation Process: Adjusted Market Cap / Forward 1 Year Revenue. Adjusted Market Cap = Market Cap (as of Nov. 12, closing price) - Value of BTC Holdings (~$87,000 per BTC). Forward 1 Year Revenue = October Production x 12 x $87,000.

As seen in the chart, Iren has the lowest valuation multiple among its peers. Admittedly, $CORZ benefits from a large AI/HPC contract with CoreWeave, which has led to a valuation premium. However, aside from this exception, none of the other miners have any significant additional revenue streams. While the market is currently “hoping” or “expecting” that $WULF and $CIFR will secure similar deals, neither has done so yet, much like $IREN. In my opinion, Iren’s exceptionally large 1.4GW site holds the greatest potential to secure a major AI/HPC colocation deal, given its strong metrics, including scale, network redundancy, and low latency.

Even when excluding the outliers of $CORZ, $CIFR, and $WULF, $IREN still trades at a lower multiple than all other pure-play BTC miners. This discrepancy is perplexing, considering Iren’s industry-leading performance across multiple metrics. Iren is the fastest-growing miner of all time, boasts the highest efficiency (resulting in superior margins), and is the most accretive miner (experiencing the least dilution relative to growth). Furthermore, Iren is set to further extend its lead by increasing production by another 50% and achieving additional efficiencies in the coming weeks, as it expands its hash rate from 21 EH/s to 31 EH/s.

In my opinion, $IREN should trade at the HIGHEST forward multiple based on these factors, at the very least in comparison to the “pure-play” miners. At a minimum, Iren should have a forward PS of 10, which would represent a current share price of approximately $25.

This valuation comparison highlights the inefficiencies in the market. People often think the market is an “all-knowing” entity that values everything at fair value. The same people likely would have dismissed a company like $PLTR when its share price dropped from $25 to $6 (“the market is all-knowing”), yet are now eager to buy the stock at over $60 a year later. I don’t know about you, but I prefer to purchase stocks before the rest of the market catches on—buy low and sell high.

People need to understand that the market is simply made up of individuals. Individual investors, generally speaking, are often irrational or fail to conduct the necessary research that leads to rational conclusions. Fund managers aren’t much better—while they typically have better risk management skills, ~85% of them still underperform the S&P 500, meaning they’re not great at stock picking.

I have analyzed Iren’s risks extensively, and they do not justify a valuation discount relative to its peers. Trust me, Iren’s peers have their own set of risks—every stock carries risks. If you believe the Australian HQ issue deserves to be weighted more heavily against Iren’s stock, fine. Subtract 10-20% from its “fair” forward PS of 10, and you still end up with a PS of 8-10—not 5.

In any case, I’m not complaining. Let the market be irrational. I’m taking advantage of it. $IREN is easy money, in my opinion (NFA).

Eventually, ALL stocks reach “fair value.” The concept of “fair value” is a moving target, and to a great extent, it’s subjective. But at the end of the day, money is money, and some investments are objectively cheaper than others relative to their future earnings. Determining future earnings potential is where the subjectivity comes in, as no one has a crystal ball. The investors who do the best research and arrive at the most rational conclusions are typically the most successful.

For these kinds of discussions, I love to bring up the example of $TSLA. Tesla is the perfect example of a stock that went from severely undervalued to severely overvalued in just a matter of months. Between late 2013 and mid-2019, $TSLA stock remained essentially flat, trading within a six-year range of ~$10 to ~$25. People who believe the market is always rational would assume that Tesla’s revenues must have been flat during this period, right? WRONG. In fact, during that time, Tesla’s trailing twelve-month (TTM) revenues grew tenfold (10x). Then, from May 2019 to November 2021, $TSLA stock surged from just $12 to over $400—an increase of more than 33x in just around two years. This price “correction” (to the upside) occurred while revenues barely doubled.

Sure, TSLA was unprofitable during the 2013-2019 period but became profitable during the massive price surge. You’d think the “rational market” could have anticipated that, since the market is always forward-looking, right? After all, many smart retail investors saw this earnings surge coming, especially in late 2018 and early 2019. However, the broader market didn’t see it coming—because the market is irrational. $TSLA stock rose so aggressively that it became severely overvalued by late 2021, which is ironic given how undervalued the stock was just a few years earlier. Once again, the market was irrational and over-corrected to the upside. Now, $TSLA stock is experiencing the consequences of that overreaction, trading sideways for the past three years.

In conclusion, people shouldn’t be deterred by market inefficiencies. If anything, they should take advantage of them—that’s how you make money in the markets.

IREN Price Target

I am setting my price target for $IREN with a focus on the first half of 2026. This timeframe aligns with my expectation that the current macro bull market, which has been ongoing since 2009, will conclude in H1 2026. While I won’t delve into the specifics of this assessment in this post, I might elaborate on it in future posts. The period between Q2 2025 and Q2 2026 is particularly suitable for forecasting Iren’s financials, as the company aims to reach 50 EH/s by Q2 2025.

The table below outlines my personal assumptions and the resulting price target for H1 2026.

My Assumptions:

I expect a 20% increase in share count through the issuance of one more (likely final) ATM, starting in early 2025, which will raise at least $500 million in cash. This capital will primarily be used to help the company reach its 50 EH/s target. Any remaining cash will be allocated to its expansion into the AI sector. Based on my previous expectations of $BTC averaging $200k and the network hash rate averaging 1000 EH/s, this would result in mining revenues of approximately $1.7 billion during the Q2 2025 to Q2 2026 period.

I also expect Iren to generate at least $50 million in AI revenue from its own CSP business. This assumes a relatively modest investment of less than $50 million into the upcoming B200 GPU line, which will be entirely self-funded by cash from operations. I have tempered expectations for Iren’s CSP business during this period, as I anticipate much of the company’s focus will be on servicing its hyperscaler client through a colocation deal of at least 500 MW at Iren’s West Texas site. According to my colocation model, a 500 MW deal would generate at least $750 million in annual revenues. Since this site is unlikely to be energized before April 2026, much of this revenue will begin in 2027 and beyond. However, once the deal is signed and the cash inflows become predictable, I’m confident the market will start to account for much of the future revenue in a forward-looking manner—much like it did with $CORZ. While the market can be irrational, once revenue becomes predictable (black and white), even Mr. Market will take notice. To fund the capital expenditures for this colocation deal, I expect Iren to either take on debt through the issuance of convertible notes or have the hyperscaler fund it in part or in full, similar to the CORZ/CW deal, with a comparable repayment structure.

If Iren does not secure a large colocation deal, I believe they will refocus on scaling their own CSP business significantly, largely funded through a convertible notes offering or GPU financing. This path could yield similar, if not greater, revenues.

Finally, for this level of growth and exceptionally high-margin business, I believe a TTM PS multiple of 15 is highly reasonable. While it may seem a bit rich, it is justifiable given the strong margins and tremendous growth rate.

This gets us to a H1 2026 price target of $169.

Admittedly, there is considerable nuance in forecasting Iren’s revenue potential. Bitcoin’s average price could fall significantly below my assumed $200k figure, which would severely impact mining revenues. On the other hand, Iren could secure a colocation deal of over 1GW, effectively doubling its forward revenue from that line of business. We could also see a broad-based AI/BTC bubble, driving valuations well beyond my PS target of 15. Alternatively, a broad market correction could lead to strong valuation compressions.

Taking all the nuance into consideration, I am confidently expecting a share price of at least $150 in H1 2026.

Closing Remarks

I want to give a huge S/O to the following individuals who have helped me tremendously over the past few months by providing valuable insights and assisting with my research on this company through private discussions:

@FransBakker9812 @TheKamaHsutra @litigious_dulce @benemodi @Umbisam @Brenno2332 @AlphaTrader00 @alanbialo @nanotitan28

I highly recommend anyone interested in $IREN to follow these fellas (on X).

Over the past couple of weeks, IREN announced a series of developments with major implications for the business and its shareholders.

On one hand, the company procured an additional 50,000 Blackwell GPUs. On the other, that news was quickly overshadowed by the launch of a massive new $6 billion ATM facility, which many investors fear could lead to significant dilution.

Since then, most shareholders have been left wondering what to make of these announcements.

Co-CEO Dan Roberts has also since joined two interviews, offering subtle but highly valuable insights not just on these announcements, but also on IREN’s broader strategy.

In this deep dive, I’ll give you the context you need to properly interpret these developments and what they mean for the investment thesis.

From the new GPU orders and what they imply for IREN’s pipeline, to the $6 billion ATM, key insights from Dan Roberts’ recent interviews, my updated earnings framework, and the risks shareholders should be aware of, this deep dive covers everything.

Given the unusually dynamic circumstances surrounding IREN right now, this is one of the most important write ups we’ve published to date.

Read more

If one solely looked at IREN’s share price performance over the past 2 months without knowing anything about the company, one would think it was going through a major crisis.

Despite landing its first deal with a hyperscaler in early November, the stock was trading at more than a 50% discount just weeks later. The irony is that, fundamentally, IREN has never been in a stronger position and over the last couple of months I have become even more bullish on the company and its stock.

In this deep dive, I’ll break down what makes me so optimistic and why I believe IREN will trade at much higher levels in 2026.

I’ll specifically cover:

• IREN’s near term catalysts

• IREN’s positioning in the broader industry

• Updated valuation framework & price targets

I’m confident that any IREN shareholder, or any investor contemplating taking a position in the stock, will derive significant value from this deep dive.

Enjoy!

Read more

Intro

IREN has been on an unprecedented growth tear within the Bitcoin mining industry. Although the company only IPO’d in late 2021, it now ranks among the largest – and, even more impressively, most profitable – BTC miners in the world.

In this deep dive, I’ll break down how IREN reached this position in record time, becoming the most dominant force in the industry, and how that success sets the stage for its ambitions in the AI/HPC space.

Read more

COMING SOON TO PAID SUBSCRIBERS.

Subscribe now

COMING SOON TO PAID SUBSCRIBERS.

Subscribe now

Over recent months, IREN has drawn significant attention — not only from the retail community but also from Wall Street and the broader financial media landscape. While the spotlight is long overdue, few investors truly understand what IREN actually is or what constitutes its moat.

IREN is not just another AI cloud provider, nor is it “just” a Bitcoin miner. From day one, the company has been architected to become one of the most dominant hyperscalers of the 21st century — one of the first true data center conglomerates.

IREN in a Nutshell

At its core, IREN is a data center company, whose sole prerogative is to monetize grid-connected power in the most lucrative manner — i.e., receiving the highest Dollar value per MW ($/MW) possible.

What truly sets IREN apart is that it owns the entire physical infrastructure stack: the land and interconnects, the high-voltage gear like transformers and substations, the data centers themselves, and the compute hardware inside them.

This level of vertical integration opens up virtually every power-monetization pathway across the sector’s value chain:

With a massive portfolio of 2.91 GW in secured, grid-connected power, IREN holds all the cards in this power-constrained market. Naturally, management is prioritizing one of the highest yielding monetization pathways — AI cloud (CSP) — with the recent $9.7 billion, five-year Microsoft contract marking a pivotal moment in the company’s history.

IREN’s Bitcoin Mining Segment

While most analysts and investors now associate IREN with its rapid emergence in AI cloud, its ascent in Bitcoin mining is just as remarkable.

Over the past couple of years IREN quickly established itself as the most profitable and fastest-growing “BTC miner” of all time.

It now ranks among the top three largest publicly traded BTC miners — alongside MARA Holdings (MARA) and CleanSpark (CLSK) — with 50 exahash (EH) of mining compute.

However, a key differentiator is IREN’s vastly superior profitability and focus on cash flow. For one, IREN is the only “BTC miner” to have achieved multiple consecutive quarters of operating profitability during one of the worst BTC bull markets for the mining sector in history. No other miner has achieved operating profitability during this cycle.1

Note: Chart depicts quarterly data. Q3 data no longer accurately reflects the operating profitability of IREN’s Bitcoin mining segment, due to the company’s strong shift toward AI cloud (CSP) in recent months.In Q3, SG&A rose materially due to stock-based compensation and stock-price-linked performance equity awards — a direct byproduct of IREN’s ~10x share-price increase over recent quarters — which depressed operating income but did not affect operating cash flow (OCF).

IREN’s superior margin profile is a combination of full vertical integration and its “giga-site” strategy — with most of its current operating capacity concentrated at its 750 MW Childress site.

Image by: Frans Bakker

Few companies in the global data center space can claim to operate individual sites at this scale. In fact, IREN is one of the only companies with multiple sites exceeding 500 MW of grid-connected power.

Rather than maintaining a fragmented footprint of smaller sites across various regions — each with its own staffing, security, and operational overhead — IREN has taken a hyper-concentrated approach. By clustering the bulk of its capacity at a small number of giga-sites, IREN reduces duplicated overhead costs like staffing, on-site maintenance, and support infrastructure. Everything from energy routing to fiber paths to network interconnects can be optimized within a single high-density campus. This strategic consolidation drives significantly lower OPEX per MW and enables IREN to achieve meaningful operating leverage as capacity scales — a major driver behind its expanding profit margins.

And importantly, this efficiency carries over across all of IREN’s verticals. Not just in BTC mining, but also increasingly in its AI related segments, where margin structure and operational efficiency are just as critical to long-term competitiveness.

Another way IREN differentiates itself in the BTC mining sector is that it's one of the only companies that does not “HODL”, i.e., holding Bitcoin on its balance sheet. It sells 100% of its mined BTC and reinvests every penny into expanding its vast power portfolio — creating a vicious hyper-growth flywheel.

While IREN initially started as a “BTC Miner”, management never saw it as just that. Since its IPO in 2021, it has always positioned itself as a data center company. BTC mining has always been just step one.

The primary reason IREN chose to start with BTC mining is that it was the most straightforward path to monetizing its large powered land portfolio. Pretty much all of the other monetization options require an end customer and in this industry, credibility and a strong track record of development execution are everything. BTC mining doesn’t have that hurdle; you mine for yourself — i.e., for your shareholders — no customers required.

However, with the full intention to eventually move into client-based data center services, IREN always selected and built its sites to be fully suitable for HPC (high-performance computing) purposes. For example, its massive sites in West Texas are strategically positioned on top of a major fiber backbone, enabling exceptionally fast latency of just 6 ms to Dallas (Infomart carrier hotel) — one of the largest interconnectivity hubs in the U.S. In comparison, Galaxy Digital’s (GLXY) 800 MW Helios site (also located in West-Texas), of which 100% is contracted to CoreWeave, achieves latencies of 10–15 ms to Dallas (according to their own specifications) — still good, but not as elite as IREN’s 6 ms.

In case you don’t know, low latencies — both externally to major network hubs like Dallas, and internally across GPU clusters — are critical for cloud services and AI workloads, where every millisecond can impact performance.

Now that IREN has reached 50 EH of BTC mining capacity, management has officially paused further expansion of this segment for the time being. While future growth in mining hasn’t been ruled out, the company currently sees far greater potential in the AI/HPC market — primarily due to the severe shortage of AI-ready data centers.

Having solidified itself as the most dominant and operationally efficient force in the BTC mining industry, IREN is now ready to fully commit to AI/HPC at an unprecedented scale.

If you want a fully in-depth look at IREN’s ascent in the BTC mining industry — including the hiccups and lessons management faced along the way (as well as peer comparisons) — check out this deep dive → How IREN Conquered the BTC Mining Industry.

IREN’s AI Segment

IREN officially launched its AI segment in late 2023 with the acquisition of a few hundred NVIDIA H100 GPUs — capacity that was contracted to the AI startup Poolside in early 2024.

Since then, IREN has scaled its cloud capacity considerably, with the company now holding a total fleet of 23,300 GPUs:

• 1.9k NVIDIA H100s & H200s

• 19.1k NVIDIA B200s & B300s

• 1.2k NVIDIA GB300s

• 1.1k AMD MI350Xs

Most of IREN’s recent surge in fleet capacity occurred in recent months, and much of it can be attributed to the company’s new CFO, Anthony Lewis, who initially joined IREN in early July 2025 as Chief Capital Officer and was promoted soon after.

Lewis has an extensive financing background, spending 22 years at Macquarie Group — most recently as Co-Treasurer — where he was responsible for global funding, liquidity, and capital management.

The reason he has been so instrumental in recent months is simple: IREN’s AI cloud offering never lacked demand, but servicing that demand is extremely capital-intensive due to hardware costs. Single GPUs can run upwards of $40k, and super-chips such as the GB300 are around $75-80k per unit.

Anthony was brought in to engineer capital-efficient ways to scale IREN’s fleet without over-leveraging the balance sheet — which is exactly what he accomplished with the GPU-leasing model IREN now uses. GPU leasing involves renting hardware from vendors over a fixed period (e.g., 24–36 months) instead of purchasing it outright. At the end of the term, there is a buyout option at fair market value (FMV): either a nominal $1 — meaning IREN effectively paid the full purchase price through lease payments — or a predetermined percentage (e.g., 15–20%) of the initial price. The cost of this capital-light structure is the interest paid to the vendor, which currently sits in the high single digits (~9%).

Fun fact: I’ve heard IREN is gaining a reputation in Australia for “sniping” top Macquarie talent, with much of the executive team having deep ties to the country’s largest investment bank.

Over the coming year, IREN’s cloud capacity will grow considerably, with management guiding to a fleet of 140k GPUs by the end of 2026.

Much of that increase is tied to the company’s $9.7 billion deal with Microsoft at IREN’s Childress campus. This five-year cloud agreement encompasses 76,000 GB300 GPUs, to be installed across the newly built Horizon 1–4 facilities throughout 2026.

For the full breakdown of this monumental partnership between IREN and Microsoft, check out my comprehensive analysis (available very soon) —> Unpacking the $9.7b IREN × MSFT Deal.

IREN also plans to acquire ~40k additional GPUs over the course of 2026 for its Canadian data centers. Management has guided that the vast majority of this capacity will consist of air-cooled B200/B300 models — an expansion few neo-clouds, or even hyperscalers, can pull off in North America for reasons I’ll detail later in this deep dive.

It’s apparent that IREN’s rapid ascent in the AI cloud market is nothing short of impressive — but do they truly have a durable edge, and how competitive is their offering?

In short: extremely competitive.

Running a GPU cloud service has strong parallels to BTC mining — both are highly capital-intensive and plagued by rapid hardware depreciation cycles. In BTC mining, ASICs typically need to be replaced every 3–5 years. In the CSP business, GPUs may remain profitable for a similar timeframe — primarily because newer generations render the old ones redundant. That means you’re constantly racing the clock to earn a return before the hardware becomes obsolete.

In the cloud space, there are two main paths to profitability:

1. Pricing Power — something currently reserved primarily for hyperscalers like AWS, Azure, Google Cloud, and Oracle

2. Cost Efficiency — achieved through low operating and energy costs

Pricing power primarily comes from 2 verticals: reliability (measured in uptime) and added service, usually in the form of software solutions offered to the clients. On the other hand, cost efficiency primarily comes from vertical integration of the physical infrastructure; think owning the data center and thus not having to pay costly colocation fees, as well as running tight operations. Having access to cheap energy also plays a factor, but not nearly as much as being able to avoid colocation rent.

The typical hyperscaler achieves profitability through immense pricing power, derived from a vast ecosystem of software solutions and add-ons. For them, cost efficiency is secondary, since they can often charge more than double the rates of the broader market. Their gross margins are naturally wide due to this pricing advantage, and these megacorps frequently lease white-label cloud capacity from reputable providers (think IREN × MSFT, or NBIS × MSFT) or rent data center capacity from colocation providers (think AWS x CIFR).

The rest of the market competes almost entirely on cost. While a few neo-cloud competitors are trying to differentiate with added software solutions, none has achieved meaningful pricing power so far.

So where does IREN fit in?

IREN is on its way to becoming the hyperscaler for hyperscalers.

IREN is all about scale; its biggest challenge is monetizing its vast power portfolio, as quickly and as capital-efficiently as possible. Idle power earns nothing. There’s no point catering to clients that need software solutions, as that segment usually consists of small to mid-sized enterprises that rarely need tens of megawatts, let alone hundreds. Moreover, the likes of Google, AWS, and Microsoft already dominate that space, so pushing software orchestration would pit IREN head-to-head with the world’s largest tech firms.

IREN is targeting the big fish — large-scale enterprise clients, and even other hyperscalers, that don’t need help on the software side. This target audience needs reliable compute for internal use, or they want to layer their own software on top of IREN’s capacity and resell that repackaged compute to smaller clients who do need ‘extra help’.

In other words, IREN offers bare-metal compute in bulk at very competitive prices. CCO Kent Draper and Co-CEO Dan Roberts articulated this market strategy very well on the recent Q3 earnings call:

Because IREN controls the entire physical infrastructure stack, its gross margins are among the highest in the industry — around ~90%. Of course, gross margins aren’t everything; the biggest expense for any cloud provider is depreciation on short-lived GPUs. But everyone faces that same constraint.

The point is: given heavy depreciation, you survive in this industry via 1) pricing power or 2) cost efficiency. Since IREN avoids major expense items like colocation fees — which can eat up ~20% of gross margin — it holds a distinct advantage in the long-term durability of its model.

To fully assess IREN’s long-term potential, we now need to examine the company’s competitive moat — which is exactly what I’ll do in the following sections.

The Power Advantage

You may ask yourself why IREN’s grid-connected power portfolio is such a big deal. Can’t neo-clouds like CoreWeave or hyperscalers like Oracle just throw their weight & capital around to secure an equal amount of power?

In short: no.

You can’t just build a data center and expect to plug it into the grid. You need permits and approvals from local grid operators. The problem is that there’s a severe shortage of power, with electricity demand (primarily from data center growth) vastly outpacing current supply.

While new energy generation is ramping up, the demand curve is steepening even faster, resulting in a supply crunch that shows no signs of easing anytime soon.

Before the recent AI boom — which dramatically accelerated power demand — data center developers traditionally flocked to established markets like Northern Virginia, Silicon Valley, and Chicago. These regions already had mature infrastructure consisting of dense fiber networks, robust power grids, and proximity to enterprise demand.

But most of these regions are now completely tapped out when it comes to new power allocations for large-scale data center projects. This shortage is forcing developers to go wherever power is still available. That alone marks a major industry shift.

The region that’s emerged as an obvious choice is Texas, particularly West Texas, where there’s still some power “up for grabs”. That’s where Stargate’s (i.e., OpenAI’s) first 1.2 GW mega-site is being developed. At this point, permitting queues across the U.S. — including West Texas — have become extremely congested, and securing new grid-connected power now takes roughly 5-7 years.

Many developers openly claim to have gigawatts of power capacity, but when you read the fine print, that capacity often hinges on pending grid approvals.

In contrast, IREN’s management is more discreet when it comes to disclosing its power pipeline. They don’t publish estimated figures and only confirm that it’s in the “multi-GW” range. Their rationale is simple: they view energy access as a binary — you either have grid approval, or you don’t. As such, they only announce new sites once approvals are locked in. Without grid approval, a site is essentially worthless — you can’t build data centers on it.

They’ve emphasized that, in today’s environment of structural power shortages, any pipeline lacking grid approvals carries high uncertainty. That said, my leads suggest IREN has a pipeline of over 5 GW, on top of the 2.91 GW it already has secured.

It’s also important to distinguish between self-owned power connections and “contracted power”. Many cloud providers claim to have multiple gigawatts of connected power, and while that may be technically true, much of that capacity isn’t actually theirs — it’s leased from third-party powered-land and data center developers.

Many investors still struggle to differentiate between self-owned and leased power capacity — failing to appreciate the advantages of the former.

Take Nebius as an example. The company recently unveiled a major increase in its 2026 guidance for contracted power, aiming to secure more than 2.5 GW by next year.

Some investors may look at this and conclude that IREN’s power-related moat isn’t a meaningful advantage in today’s market.

That couldn’t be further from the truth.

Nebius is currently facing a severe shortage of power it has “on hand” — meaning available today. Don’t take it from me; take it directly from the company’s CEO, Arkady Volozh, who said the following on the most recent Q3 earnings call:

“Capacity today is the main bottleneck to revenue growth. And we are now working to remove this bottleneck.”

This bottleneck is so severe that Nebius’s recently announced $3 billion Meta deal was limited to the power the company currently has available:

“The size of the contract was limited to the amount of capacity that we had available, which means that if we had more, we could have sold more.” (CEO Arkady Volozh)

This case study shows the true value of IREN’s fully owned power pipeline. If it were easy to “just get power”, Nebius would have solved that bottleneck already and would not have been constrained when contracting large-scale deals.

Another key distinction is the margin advantage that comes from owning the power and data center infrastructure rather than leasing it. As noted earlier, IREN’s gross margins are among the highest in the neo-cloud cohort (greater than 90%), while operators such as Nebius and CoreWeave report gross margins in the low-70% range — primarily because they pay significant colocation fees to providers like DataOne or Core Scientific (CORZ).

Getting back to IREN’s pipeline, the company now has a consistent record of annual site additions and capacity expansions. The most recent addition was the 600 MW Sweetwater 2 site near Abilene, announced earlier this year. Prior to that, in 2024, the company secured 1.4 GW at its Sweetwater 1 site (also near Abilene) and expanded Childress from 600  MW to 750 MW through a 150 MW increase in approved grid capacity.

Image by Frans Bakker

To contextualize just how large IREN’s powered land portfolio is, consider this: one of the largest data center markets in North America — Chicago — has an operational capacity of just ~1.2 GW. That’s an entire market made up of hundreds of individual data center sites. The largest data center market in the world is Northern Virginia, specifically Ashburn — also known as “Data Center Alley” — with an operational capacity of around 5-6 GW. IREN’s powered land portfolio is already nearly 60% the size of this entire market, with its Sweetwater 1 & 2 campuses alone equivalent to as much as one-third of that scale.

So how did IREN manage to secure a record-breaking 2.91 GW?

Because the company moved early.

IREN’s management had the foresight to secure massive amounts of power long before the AI boom — years before ChatGPT launched and kicked off the gold rush for compute infrastructure. Co-CEO Dan Roberts has repeatedly spoken about his and his brother’s early vision of the world digitizing at an exponential pace — with the physical power infrastructure failing to keep up.

That vision led to a bold bet: real-world assets like grid-connected land and the data centers on top of it would become the rarest and most valuable resource in tech.

Most people are unaware of just how deep Dan Roberts’ background in power infrastructure runs and how critical that experience has been to IREN’s ability to secure utility-scale capacity. Before co-founding IREN, Dan was a founding executive and Investment Director at Palisade Investment Partners, an infrastructure investment firm managing over $6 billion across 20+ asset businesses, including wind farms, gas pipelines, and other energy infrastructure.

Long before the AI race began, IREN’s founders understood that access to power would become the ultimate bottleneck in a digitizing world. That foresight — paired with Dan’s deep infrastructure expertise — is a big reason why IREN was able to quietly amass 2.91 GW of permitted and grid-approved power.

The company now holds one of, if not the, largest powered land portfolios on the planet.

In short, IREN is much more than just a data center developer — it’s arguably one of the world’s best developers of powered land. That means securing large land parcels, negotiating directly with utilities and grid operators, building high-voltage substations, and navigating multi-year permitting and interconnection processes.

It’s a capital-intensive, highly specialized discipline — one that most modern data center developers no longer touch. IREN does it all in-house, and at massive scale.

To illustrate how unique IREN’s position is as a data center developer with gigawatts of self-owned, fully permitted capacity, let’s consider the following example: Most readers are probably familiar with the Stargate alliance between Oracle, OpenAI, and SoftBank. Their first 1.2 GW site in Abilene is being developed by Crusoe. But Crusoe doesn’t own the powered land — it leases the land from a company called Lancium, which specializes in acquiring and developing grid-connected land.

Lancium’s entire business model is focused on that first step: sourcing and entitling powered land. It monetizes this capability by leasing the land out to operators like Crusoe.

That’s a viable standalone model. But for IREN, this is only step one. Unlike Lancium, IREN prefers to own and build on top of its powered land, moving up the value chain and capturing far more value per megawatt. At the same time, it avoids having to lease or acquire powered land at rising market rates — a major advantage in the face of an intensifying power crunch.

This directly translates into higher margins, faster development timelines, and superior returns on invested capital.

Moreover, IREN has strategically secured power in regions with exceptionally low electricity costs, typically in remote areas with abundant renewable generation like wind or solar. At the company’s flagship 750 MW Childress site, all-in electricity costs consistently range between 2.8–3.5 cents per kWh — remarkably low by industry standards. A key factor here is IREN’s optimized energy curtailment strategy: during brief price spikes (usually lasting just minutes or hours), the company shuts off its BTC mining operations to avoid high-cost intervals.

For AI workloads, where uptime must remain near 100%, curtailment isn’t a viable option — but even without curtailment energy costs are likely to average ~5 cents per kWh at IREN’s West-Texas sites. That’s still extremely low compared to major data center hubs like Northern Virginia (6–7 cents) or Silicon Valley (often over 10 cents).

This power cost advantage is especially meaningful for AI cloud, where the host, i.e., IREN, pays for the electricity. And because the energy is fully sourced from renewables like wind & solar, tenants can also claim that their data center capacity is sustainably powered — another very valuable (and often overlooked) selling point for IREN.

Now that you understand IREN’s unique power advantage, here’s an up-to-date overview of the company’s powered land portfolio & data center capacity:

Development Expertise

While IREN’s power advantage is increasingly well understood by investors, I’d argue that few truly grasp the significance of the second pillar of its moat: Development Expertise.

The term ‘development expertise’ is intentionally broad, as it captures several key capabilities that set IREN apart. For one, there are few, if any, data center companies that are this vertically integrated. From developing the greenfield land and energizing it to building full fletched turnkey data centers & running its own GPU cloud — IREN does it all in-house.

Analysts like have labeled IREN the “Tesla of the data center industry” — and I couldn’t agree more.

Tesla set itself apart by doing nearly everything in-house: from design and battery development to manufacturing, assembly, and even the software layer. This vertical integration allows Tesla to capture more of the value chain than any traditional automaker. Thanks to a world-class engineering culture rooted in first-principles thinking, Tesla has driven efficiency at every level — resulting in lower costs per unit and vastly superior margins per vehicle sold. On top of that, Tesla isn’t afraid to branch out into entirely new verticals, such as Megapacks (utility-scale batteries), FSD (full self-driving), and robotics.

Likewise, IREN does nearly everything in-house and isn’t afraid to branch out into new verticals. They’re not just a BTC miner, or just a cloud provider, or just a colocation operator — they’re steadily evolving into a full-scale data center conglomerate, something that hasn’t really existed in this form before.

IREN is essentially a next-gen infrastructure company that’s breaking the mold, not unlike what Tesla did for autos and energy.

This level of vertical integration naturally compounds cost efficiency. Because IREN builds in-house — rather than relying on third-party developers who need their own margin — its cost structure is essentially materials plus labor. At scale, standardized designs, repeatable workflows, and procurement leverage drive further savings through economies of scale and learning-curve effects, which leads to lower unit costs with each successive build.

Another factor that sets IREN apart from anyone in the industry is its incredibly fast development speed. Being able to build & develop its sites at a very quick cadence is the primary reason why IREN quickly emerged as the fastest-growing BTC miner of all time over the past 24 months. Having the necessary secured power is certainly important, but to truly dominate you need to be able to scale quickly.

IREN has built most of its air-cooled data centers at Childress at a speed of roughly 50 MW per month, a pace that’s 3–10x faster than the typical industry norm — where even experienced developers often average just 5–15 MW/month. Now, as most of IREN’s new builds will be liquid cooled to service next-generation GPU hardware, the construction complexity has increased considerably and development timelines are longer. Nonetheless, IREN still ranks among the very best in this new class of facilities, with Horizons 1–4 (200 MW of critical IT) being developed and delivered to Microsoft over the course of 2026.

It’s worth noting that Horizon 1 (50 MW) began in early 2025, and since the Microsoft signing the facility designs have been materially upgraded. Even assuming Horizon 1 is essentially complete by early 2026, delivering an additional 150 MW of top-tier liquid-cooled capacity within the same year would be a remarkable cadence.

In a compute-constrained market, development speed is crucial.

But IREN’s development edge isn’t just about construction speed or cost efficiency — the real differentiator is the build quality, i.e. the performance of IREN’s data centers. IREN achieved what no other data center provider has in the realm of air cooling — pushing the upper boundaries of what was thought to be theoretically achievable by reaching 80kW of rack density at its Prince George facility in Canada.

Industry norms for air cooling sit well below 30 kW per rack — often in the 15–25 kW range. That’s largely because the traditional data center industry didn’t require this level of extreme cooling capacity prior to the AI boom and thus, as a result, engineering progress in this area plateaued and stagnated over the past decade.

The only notable application that required concentrated power at scale was Bitcoin mining. As a result, a company like IREN, which already pushed the envelope in the mining industry, was uniquely positioned to repurpose its high-performance data centers for GPU clusters.

Industry experts like Brian Fry describe rack density as one of the most critical performance variables in a data center. Higher rack density means more powerful compute packed into tighter spaces — which not only improves efficiency but also directly enhances performance. Lack of adequate rack density is like having sand between gears. If you can’t cool densely packed servers, you’re forced to spread them out. That leads to larger, less compact facilities and creates internal latency problems — the greater the physical distance between compute nodes, the slower the communication between them (= worse performance).

IREN’s Co-CEO Dan Roberts has compared this to how semiconductors improve: as the chips get smaller, transistors are closer together, allowing signals to travel shorter distances and boosting performance. High rack density achieves a similar effect — but at the data center level, where GPU servers need to be tightly packed to maximize throughput.

I’ve heard from highly credible sources that multiple hyperscalers were stunned when IREN told them it was achieving 80 kW per rack in an air-cooled facility.

This edge in air cooling now presents a huge opportunity for IREN’s cloud segment and is one of the core reasons for IREN’s rapid cloud expansion. The previous generation of GPUs — the Hoppers — kicked off a gold rush of new “neo-clouds”, many of whom undercut each other on price. Hoppers required ~40kW per rack, but most of these clouds were renting conventional air-cooled facilities with 20–30 kW rack densities. While this led to sub-par performances across the board, it was still “good enough” to flood the GPU rental market with Hoppers.

However, the newly launched Blackwell generation of GPUs requires far more power. The B300, for example — the most powerful air-cooled Blackwell variant — demands rack densities of 60–70 kW for optimal performance. That effectively locks most of the neo cloud market out of this generation. Their existing infrastructure simply can’t handle the compute intensity of the air-cooled B300 — not even at compromised performance levels of say 50 kW per rack, as there are very few air-cooled data centers on the market today that can support even that level of rack density.

Even high-end providers like CoreWeave, Nebius, or hyperscalers are unlikely to compete in the air-cooled B200/B300 space. Instead, they’ll likely focus on the liquid cooled variants. Liquid cooling enables much higher rack densities, but also comes with significantly higher costs — both in terms of CAPEX and OPEX.

Building liquid-cooled data centers at scale (of which very little capacity exists today) requires enormous capital investment and advanced engineering capabilities. This remains a relatively new concept at the commercial level and only a select group of operators have the technical depth and financial scale to execute it properly. As a result, access to the liquid-cooled GPU variants will be limited to the upper tier of cloud providers.

The new Blackwell generation, in effect, acts as a strong barrier to entry for the entire sector: most providers can’t run the air-cooled B200/B300 models because their existing data centers can’t support the rack densities — and they can’t offer the liquid-cooled GB200/GB300 either, unless they’re willing to spend hundreds of millions building entirely new facilities.

This barrier will most likely constrain supply of the new Blackwell GPUs relative to the previous Hopper generation, leading to higher and more sustained GPU rental prices across the market. In other words, returns will likely be quite lucrative for the few cloud providers that can offer Blackwells.

So why does it make sense for IREN to go after the air-cooled Blackwell segment?

Because, unlike anyone else, IREN already has hundreds of megawatts of air-cooled capacity (once used for BTC mining) that can deliver 80 kW per rack — TODAY. There’s no need for new R&D or retrofitting. The infrastructure is already there. For players like CoreWeave, it makes little sense to build high-end air-cooled data centers from scratch if the industry is likely shifting entirely to liquid cooling over the next 3-5 years anyways. Since IREN already has this air-cooled data center capacity built and available today, it can capitalize on this industry-wide dilemma and be one of the only cloud providers able to offer the far more cost-effective B200/B300 models at scale.

Why are these air-cooled GPU models more cost-effective?

Because air-cooled facilities use less energy, require fewer operational resources, and are generally much cheaper to maintain.

However, IREN is also pushing boundaries in the realm of liquid cooling. Horizons 1–4 are set to house 100 MW superclusters for high-performance AI training with flexible rack densities (130 kW to 200 kW). Few other developers are currently delivering multi-hundred-megawatt campuses capable of up to 200 kW per rack with a flexible architecture that can scale densities up or down to meet client requirements.

In contrast, CoreWeave’s liquid-cooled facilities target rack densities of around 130–140 kW, which is just enough to house the current GB300 Blackwells. However, future GPU generations — like NVIDIA’s upcoming Rubin line — will demand even higher power per rack, meaning CoreWeave’s facilities will need significant upgrades to keep up. Unfortunately for them, just like most neo-clouds, the company is heavily reliant on its colocation partners to carry out those upgrades. They’ll need to knock on the doors of providers like Galaxy Digital or Applied Digital (APLD), to modify infrastructure they don’t control — providers who, in many cases, have little incentive to take on the enormous CAPEX burden without being compensated for it.

And upgrading rack density isn’t a minor tweak. It typically involves retrofitting major elements of the data center, including power delivery systems and cooling infrastructure. In some cases — particularly in older, more rigid facilities — those upgrades may not even be possible without having to tear down most of the existing infrastructure.

CRWV is already feeling the pain of relying on third-party contractors for much of its data center footprint. On the November 10 Q3 call, CEO Michael Intrator flagged delivery delays caused by a developer running behind schedule, which lowered Q4 expectations:

“In our case, we are affected by temporary delays related to a third-party data center developer who is behind schedule. This impacts fourth quarter expectations.”

After the recent Q3 earnings call, CRWV’s CEO went on CNBC and, when pressed by Jim Cramer, admitted that the delays the company was facing were across multiple sites, all tied to a single data center provider: Core Scientific (CORZ).

Notice how CoreWeave’s CEO emphasizes that the hardest part of the cloud business is getting infrastructure online — a point Michael Intrator leans on to justify missed delivery deadlines.

Interestingly, IREN’s CEO Dan Roberts addressed this exact issue on the company’s Q2 (FY Q4) call in late August — comments that, in my view, have aged like fine wine:

While many cloud providers are at the mercy of third-party developers, IREN controls its own destiny: all of its infrastructure is owned and operated in house, with no third-party landlords getting in the way of the development process.

When I visited the Childress site in late May, I asked IREN’s operations executives whether the company is future-proofing its data centers.

Their answer: yes, all of IREN’s infrastructure is highly modular and retrofittable.

For one, they assured me that 100% of the 650 MW (built-out) air-cooled data center capacity at Childress is fully retrofittable for liquid cooling. In fact, retrofitting that built-out capacity — rather than building from scratch — could reduce development time considerably. And even beyond that, all new infrastructure is designed to be both scalable and retrofittable for higher rack density requirements down the line.

IREN’s data center performance edge doesn’t stop at rack density either — it’s also pushing the envelope in internal latency optimization. At the beginning of this post, I explained how IREN strategically locates its data centers on top of major fiber backbones, enabling exceptionally low external latency between its facilities and major interconnectivity hubs like Dallas.

But internal latency — the speed at which data travels between servers and racks inside the data center — is just as important, and in some cases, even more so. In GPU deployments, optimized internal latency is critical because it directly impacts how efficiently data moves between compute nodes. Lower internal latency reduces communication delays, enables faster synchronization, improves throughput, and maximizes hardware utilization — all of which result in better overall system performance for AI workloads. This type of latency optimization depends on careful planning of rack layout, cabling length, and network topology.

And from what I’ve heard, IREN is taking this extremely seriously. Sources tell me they are actively designing their upcoming liquid-cooled Horizon build-outs with internal latency as a core focus.

To illustrate IREN’s competitive edge, I created a visual representation of the 2 layers that make up its moat:

Given the company’s vast power advantage combined with its exceptional development expertise, I expect IREN to emerge as a clear market leader in the data center industry over the coming years.

Wrapping Up

In summary, IREN is far more than just a BTC miner or an AI cloud provider. It is a fully vertically integrated data center conglomerate — something the industry hasn’t seen before.

The company’s massive grid-connected power portfolio of nearly 3 GW — with likely more than 5 GW potentially in the pipeline — acts as a near-impenetrable competitive barrier. This is underpinned by best-in-class development capabilities that enable rapid build outs with industry-leading performance.

Combined, these factors form an incredibly strong moat and pave the way for IREN to establish itself as THE data center company of the coming decade.

While I’m incredibly bullish on IREN, there are lingering concerns and risks across the broader AI ecosystem — a topic I’ll cover in depth in my upcoming deep dive: The AI Bubble (soon available to paid subscribers).

Thank you for reading, Cheers! ✌️

Subscribe now

Disclaimer (NFA): This publication is for informational and educational purposes only — not investment, legal, tax, or accounting advice. Nothing herein constitutes a solicitation, recommendation, or offer to buy or sell any security or strategy. The author may hold — and may buy or sell without notice — securities, derivatives, or other instruments referenced. All opinions are the author’s, expressed in good faith as of publication, and subject to change without notice. Information is believed accurate but provided “as is,” without representations or warranties; errors or omissions may occur. Any forward-looking statements involve risks and uncertainties that may cause actual results to differ materially. Past performance is not indicative of future results. Do your own research (DYOR) and consult a qualified, licensed adviser who understands your circumstances before acting on this content. To the fullest extent permitted by law, the author and publication disclaim liability for any loss arising from reliance on this material.