[Investment-Company Analysis] First Study/Article. IREN.

"Hello~ The market isn't good as liquidity is drying up. How is your personal fear index?"

I've actually been investing mostly through listening on YouTube, etc. So, I feel a bit more anxious and less confident. Now, I've written my first study/first post to study myself at a time when IREN, in which I'm currently investing, has fallen about 40% from its high ($76)!

[Stock Price]

Regardless of the stock price, IREN is working hard, and the company's capabilities aren't changing in the short term, but isn't valuation/price also an important factor?

After the April crash, IREN went up the most, and it was also defending the downside well. Although it has recently been slightly reversed by Cipher Mining, I think it has the largest upside and the most stable downside.

I also looked at the recent November decline/adjustment. The recovery of Cipher Mining, the larger crash of Coreweave (perhaps debt?), and the rest seem to be at a similar level of -20%. Should I say it's not bad for growth stocks…

[Key Contents of Past Performance Announcements]

ㅇ Q1 2025

- At that time, there was about 52k in cash profit per BTC, but when comparing ROI, the AI infrastructure side was said to be more profitable, and it seemed they had been preparing for it. They are additionally diverting funds to AI, and they added that they can expand the Bitcoin business again later if necessary. Plan B, if AI demand decreases, they can turn back to the coin side.

- To the question of why they are going all-in on AI infrastructure, they see that the demand for power for AI data centers will increase from the current 44GW to 156GW by 2030, a need of +112GW, and they viewed it as a supply shortage that the current global infrastructure can't keep up with.

- Electricity cost: 3.3 cents/kWh

ㅇ Q2 2025

- Electricity cost: 3.5 cents/kWh

- Cash profit of about 63k per BTC. Cost: $36k per 1 BTC, average selling price: $99k

The other content was similar to that of the first quarter.

ㅇ Q3 2025

- Operating 140,000 GPUs using only 16% of the total power. The remaining 84% is for future expansion. IREN's greatest strengths, power, land, and power grid, are secured, and most of them are still unused → explosive expansion is possible.

- Sweetwater supplies 2GW of power next April. Rack installation and GPU placement will likely be in operation in Q3. Childress (Texas), which will perform the contract with MS, is under construction with 200MW and is scheduled to be in short-term operation during 2025-2026.

- 'Bare Metal GPU' (using GPU exclusively on a server physically allocated without virtualization) companies want it. (OS to driver version can be set as desired, Nvidia driver specific version can be fixed, CUDA, cuDNN, etc., desired version setting, various optimizations are possible, and virtualization is not necessary, so performance can be improved by 5~20% and latency can be reduced. NVDIA's top partner has also been obtained, and they will focus on providing GPUs and infrastructure (power, cooling, network). They think SW will become a commodity someday, and they say they are planning to prepare because they're going to put it on a bit for small and medium-sized companies.

- It was revealed that the third quarter's earnings were good due to the valuation gains from derivatives.

ㅇ Conclusion

They predicted the demand for AI and have been preparing since before, and it seems they have been preparing a lot in accordance with the needs of the company. Also, AI sales are not very high at the moment, and as they stated that AI is more profitable, they are expected to achieve even better results in the future. However, since the sales from coin mining are currently higher, and the situation in the crypto market has not been good, the fourth quarter's earnings may be poor, which may affect the stock price.

[Website]

This is the recently changed website. It has been extensively reorganized for AI cloud & data center sales, and it is showing off its capabilities, such as the location and power of the data center.

- Speed is life since the AI competition is very intense. They want infrastructure that can be run 'right now', not in two or three years, and everything is ready, from small to large-scale, whether it's a data center or a cloud. A structure that allows high margins with low electricity through vertical integration, from ownership to operation.

- AI GPU: Basic model learning, inference, fine-tuning, etc. are all possible, and it's also possible privately. The introduction content supports long-term learning and advanced inference. It supports various models and workloads such as Llama and Stable Diffusion, and also utilizes Nvidia InfiniBand, and also supports its own DBs and models. Bare metal GPU too. So, it's prepared according to the company's wishes and the needs of its customers.

- Data Center: Vertical integration. From ownership to operation. Cooling, various facilities, etc. Infrastructure that MS can also trust.

[Others]

Many of the major figures, including the CEO's brothers, CFOs, and others, are from Macquarie, and the key figures such as the COO all have solid careers in the energy and resource fields. (+ Macquarie is almost a global Tier 1 in the fields of renewable energy, data centers, and power grids)

Recruiting for about 90 job categories. Actively hiring.

[Conclusion]

(Strengths) Securing a huge amount of power of 2.9GW, low electricity costs, completely in-house design & construction, and cost efficiency through almost in-house engineering without outsourcing seem to be great strengths. If you do a lot through outsourcing, there will be many issues, such as difficulties in controlling the cost and other parts, but I think it's being operated very stably, and this ability itself is a great competitive edge and will become a stable downside.

(Risk Factors) It seems that many companies are preparing for the future, from data center construction to AI conversion, depending on the demand for AI. If the demand for AI clouds, etc., decreases due to the completion of the in-house data center, the AI industry focuses on optimization & utilization, and the limitations of performance improvement, the competitive landscape of LLMs decreases, the unit price may decrease, and both profitability and valuation may fall. (There are various AIs such as image/video/physical, but the biggest pie at the moment is probably LLM)

(Final) However, considering various permits/regulations and construction time, it cannot be run at the moment, so it seems it will still be safe until 2026. However, if it receives a high valuation next year, it should show something afterwards, or the demand should be overflowing and the supply should be small enough to maintain a high unit price, but it seems that it should continue to watch. Otherwise, due to the nature of stocks that reflect the future in advance, even if the results of 2026 come out well, it may move to another side. (Sweetwater and Childress seem to contain the future/expectations until 2028-2029, and it seems that this could affect the high valuation until 2026-2027, but it seems that it will become a foregone conclusion afterwards) Anyway, since I think it's safe until 2026, I was thinking of holding it, but I will check the demand/supply of the AI computing part and switch to another company if I find a company with a better upside.

Thank you for reading my first insufficient post! Studying/analyzing is fun and seems to help with peace of mind (?). I will upload it regularly! (I welcome any questions, feedback, etc.!)