IREN announced on August 27 that revenue for the fourth quarter of FY2026 was $137.2 million, down 5.21 TP3T from the previous quarter, and that it reported a net loss of $684 million. On the same day, the company projected a contracted ARR of $4 billion for its 2026 capacity. We are lowering the 12-month baseline price target from $53 to $45 (approx. 62,190 KRW) while maintaining a Buy rating (high risk).
The exchange rate was standardized to 1,382 won per dollar (as of August 28, 2026). The stock price is based on the closing price of the regular session on August 27, and earnings were announced after the market closed on the same day.
- FY2026 Earnings Summary: From the Conclusion
- What kind of company is Airen currently?
- FY2026 Q4 & Annual Earnings Details
- Cash Flow and Financial Statement
- ARR of $4 billion and capacity roadmap
- Valuation: Which Multiple to Trust
- Target Price Scenario
- Risks and Bearish Triggers
- Investment Opinion Conclusion
- Frequently Asked Questions
- Sources and References
IREN FY2026 Earnings Summary: From the Conclusion
Airen's annual revenue for FY2026 increased by 41.11 TP3T to $707 million (approx. 977.1 billion KRW), but the net loss was $702.6 million (approx. 971 billion KRW). This marks a shift from a profit of $86.9 million a year ago to a loss, with the majority of the cause being a non-cash impairment loss of $638.8 million (approx. 882.8 billion KRW) resulting from the disposal of Bitcoin mining equipment.
The figures for the fourth quarter show the middle of a transitional period. AI cloud revenue was $70.5 million (approximately 97.4 billion KRW), an increase of 1,101 TP3T from the previous quarter that accounted for 51.41 TP3T of total revenue for the first time; however, as Bitcoin mining revenue decreased by 401 TP3T, total revenue actually fell by 5.21 TP3T.
The key criterion for evaluating the stock price is not earnings, but rather when and to what extent the contractual ARR of $4 billion (approximately 5.53 trillion won) converts into GAAP revenue. The current enterprise value is 4.11 times the contractual ARR, but based on the actual operating ARR of $1 billion, it is 16.4 times. The gap between these two multiples represents the volatility of this stock.
| characteristic | Q4 FY26 | Q3 FY26 | change |
|---|---|---|---|
| Total sales | $137.2 million (approx. 189.6 billion won) | $144.8 million (approx. 200.1 billion won) | -5.2% |
| AI Cloud Revenue | $70.5 million (approximately 97.4 billion won) | $33.6 million | +109.8% |
| Bitcoin mining revenue | $66.7 million (approx. 92.2 billion won) | $111.2 million | -40.0% |
| AI Cloud Proportion | 51.4% | 23.2% | +28.2%p |
| Administrative expenses | $128.3 million (approx. 177.3 billion won) | $81.8 million | +56.8% |
| impairment loss | $450.4 million (approx. 622.5 billion won) | $140.4 million | +220.8% |
| Net loss | $684 million (approx. 945.3 billion won) | $247.8 million | Widening deficit |
| Adjusted EBITDA | $19.2 million (approx. 26.5 billion won) | $59.5 million (approximately 82.2 billion won) | -67.7% |
| Adjusted EBITDA margin | 14% | 41% | -27%p |
What kind of company is Airen currently?
Irene is a data center operator headquartered in Sydney, Australia. Originally named Iris Energy, it started with Bitcoin mining. The business involved securing land and substations in areas with cheap renewable energy, purchasing power, and running mining rigs; however, it is now in the process of converting those sites and power sources into a GPU cloud for AI training and inference.
The structure the company describes itself as three-tiered. The first tier is data centers (land, power, substations, buildings, and cooling), the second tier is compute (GPUs, storage, servers, and networking), and the third tier is software (managed services and enterprise support). Mirantis, acquired in 2026, is responsible for the third tier, and the company secured European sites through the acquisition of Spain's Ingenostrum Group.
Brothers Daniel Roberts and William Roberts, the co-founders, serve as co-CEOs. In this call, Daniel Roberts summarized the founding premise as follows: “The digital world can scale almost instantly, but the physical world cannot.” This means that while token throughput can double in a few weeks, it takes years to build a data center, and he argues that this gap represents a business opportunity for the company.
Terms appearing in this text
- ARR (Annual Recurring Revenue) — It is calculated by multiplying the contracted GPU hourly rate by 8,760 hours per year and adding ancillary revenue such as storage. It is an operational metric, not an accounting revenue, and the company states that recognized revenue may be significantly lower than this.
- Adjusted EBITDA — It is a figure excluding taxes, interest, depreciation, stock compensation, impairment losses, foreign exchange gains and losses, etc., from net profit. Airen has a very large number of excluded items, so it diverges significantly from GAAP profit.
- Deferred revenue (advance payments) — It is money paid by the customer in advance before receiving the service. It is a liability where cash has already been received but has not yet been recorded as revenue.
- Neocloud — This term refers to emerging providers specializing in renting GPUs for AI computing, rather than comprehensive cloud services like AWS or Azure. Coreweave, Nevius, and Airen are representative examples.
- MW(IT) and MW(Gross) — Gross is the total power entering the site, and IT is the power actually consumed by servers. After deducting cooling and losses, IT capacity is 60 to 701 TP3T of Gross.
- Commissioning — This is the stage where the customer takes delivery after the equipment is installed and undergoes trial operation, and billing begins. Airen's revenue is generated only at the time of commissioning.
- impairment loss — It is an accounting treatment that writes off the difference as an expense when the recoverable amount of an asset falls below its book value. No cash is outflowed.
IREN FY2026 Q4 & Annual Earnings Details
Fourth-quarter revenue of $137.2 million fell short of market expectations. Consensus estimates vary by agency, ranging from $139.5 million to $157.1 million, but the figure falls short regardless of which estimate is used. On the other hand, the adjusted net loss per share of $0.41 was $0.12 better than the consensus loss of $0.53. It was a combination where revenue fell short but profit exceeded expectations, leading to divided interpretations.
As the company explained, the reason for the revenue decline was voluntary. To install GPUs, the mining rigs occupying the space must first be removed; the moment they are removed, mining revenue ceases, and GPU billing has not yet begun. The fourth quarter was precisely that gap period. While mining revenue decreased by $44.5 million, AI Cloud increased by $36.9 million to make up for most of the difference, but it fell short by $7.6 million.
The breakdown of the $684 million net loss is as follows. The operating loss was $620.4 million, of which $450.4 million was an impairment loss due to the scrapping of mining equipment and $25.1 million was a loss on the disposal of tangible assets. Additionally, the fair value of assets held for sale decreased by $102.1 million. These three items alone amount to $577.6 million, accounting for 841 TP3T of the total loss, and are all non-cash items. Essentially, the accounting bills for shutting down the mining business have all come together in this quarter.
A more noteworthy point is the adjusted EBITDA. Even after removing impairment losses, the adjusted EBITDA for the fourth quarter was $19.2 million, down 67.71 TP3T from the previous quarter's $59.5 million, while margins fell from 411 TP3T to 141 TP3T. The company cited rising labor costs and expanded platform investments prior to the AI cloud revenue ramp as reasons. Indeed, selling, general, and administrative expenses jumped 56.81 TP3T in just one quarter, rising from $81.8 million to $128.3 million. The workforce is projected to triple in FY2026 and triple again in FY2027.
| Adjusted EBITDA Trend | June 25 | September 25 | December 25 | March 26 | June 26 |
|---|---|---|---|---|---|
| Adjusted EBITDA (million dollars) | 121.9 | 91.7 | 75.3 | 59.5 | 19.2 |
| Adjusted EBITDA margin | 65% | 38% | 41% | 41% | 14% |
| Stock compensation costs (million dollars) | 18.7 | 72.4 | 58.2 | 31.5 | 42.9 |
| Impairment loss (million dollars) | -2.4 | 16.3 | 31.8 | 140.4 | 450.4 |
This marks five consecutive quarters of decline. This means that the profits generated by mining are disappearing faster than the profits generated by AI cloud, and when this curve rebounds is the key point to watch in FY2027.
Looking at the annual figures, the picture is slightly different. Revenue of $707 million represents 41.11 TP3T growth, and if we isolate AI Cloud alone, it grew 7.9-fold from $16.4 million to $128.8 million (approximately 178 billion KRW). The problem is that costs increased even faster. Selling, general, and administrative expenses rose 3.3-fold from $136.5 million to $449.1 million, while depreciation increased 2.3-fold from $181.1 million to $417.7 million. As a result, operating profit, which was $17.3 million in FY2025, reversed into an operating loss of $1.0467 billion in FY2026.
There is one more point to note regarding the quality of earnings. FY2026 stock compensation expenses amount to $205 million (approximately 283.3 billion KRW), representing 29.01 TP3T of revenue. This figure is nearly equivalent to the adjusted EBITDA of $245.7 million (approximately 339.6 billion KRW), and the adjusted EBITDA is calculated by excluding this expense. In other words, the adjusted EBITDA emphasized by the company does not reflect the actual cost of shareholder equity dilution.
Cash Flow and Financials: Stagnation of 2.9 Trillion Won Operating Cash Flow
Cash flow from operating activities in FY2026 was $2.104 billion (approx. 2.90 trillion won), 8.5 times the previous year's $245.9 million. It seems strange at first glance that a company that posted a net loss of $702.6 million generated $2.1 billion in operating cash. The answer lies in a single line: Deferred revenue increased by $1.8417 billion (approx. 2.55 trillion won).
This is money paid in advance by customers for services they have not yet received. Notably, $1.7222 billion (approximately 2.38 trillion won) came in during the fourth quarter alone. The company revealed that customer advances in recent contracts amount to 45 to 551 TP3T of GPU CAPEX, and this figure is the result. Excluding the increase in deferred revenue, operating cash flow for FY2026 decreases to approximately $258.7 million.
This is where opinions diverge on how to view the advance payment. From a negative perspective, it is still a liability rather than revenue. From a positive perspective, it means a client of Microsoft's caliber is paying half of the capital expenditure upfront, and there are few ways to prove the authenticity of the contract more definitively. I leaned toward the latter view, but chose not to fully reflect the money in the valuation until it is converted into revenue.
$4.723 billion (approximately 6.53 trillion won) was spent on investing activities. This includes $2.998 billion on tangible assets and $1.3351 billion on computer hardware (GPU). Consequently, free cash flow is minus $2.6226 billion, and minus $4.4643 billion when the effect of advances is removed. Cases where net income and operating cash flow diverged in NVIDIA's (NVDA) second-quarter earningsIt is a question that is the exact opposite in direction to when we dealt with [that], but is essentially the same. Where do the ledger numbers diverge from actual cash?.
The gap was filled by financing activities. Cash inflows from financing activities in FY2026 amounted to $9.6801 billion (approx. 13.38 trillion won). The company raised $4.7428 billion (approx. 6.55 trillion won) through equity issuance and $6.2996 billion (approx. 8.71 trillion won) through convertible bonds, while spending $1.6235 billion on inducing conversions. The company stated that it has raised a total of $19 billion (approx. 26.26 trillion won) over the past 12 months through a combination of advances, GPU financing, convertible bonds, and equity.
| Financial Statement (June 30, 2026) | FY2026 | FY2025 |
|---|---|---|
| Cash and cash equivalents | $5.8956 billion (approx. 8.15 trillion won) | $564.5 million |
| Total cash including restricted cash | $7.6195 billion (approximately 10.53 trillion won) | $564.5 million |
| Tangible assets | $6.7532 billion (approx. 9.33 trillion won) | $1.9306 billion |
| Total assets | $15.79 billion (approximately 21.82 trillion won) | $2.9403 billion |
| Total borrowings | $7.593 billion (approximately 10.49 trillion won) | $962.8 million |
| Deferred revenue balance | $1.8426 billion (approximately 2.55 trillion won) | $900,000 |
| Equity | $4.1856 billion (approximately 5.78 trillion won) | $1.8175 billion |
In just one year, total assets have increased 5.4-fold and borrowings 7.9-fold. Dividing the combined total of borrowings and financial leases, $7.8368 billion, by equity of $4.1856 billion yields 1,871 TP3T. However, if unrestricted cash of $5.8956 billion is excluded, net borrowings decrease to $1.9412 billion, and if unrestricted cash is included, it drops to $217.3 million. It is important to fairly point out that while the scale of debt is large, cash has also accumulated to a corresponding extent.
IREN $4 billion ARR and 2026–2028 Capacity Roadmap
The key figure in this announcement is the ARR, not the financial statements. The company stated that it has secured a contracted ARR of $4 billion (approximately 5.53 trillion won) for its 2026 capacity, and that the actual ARR in operation was $1 billion (approximately 1.38 trillion won) as of August 26. The quarterly targets are to exceed $500 million by the end of June, $1 billion by the end of September, and $4 billion by the end of December.
It is not an exaggeration to say that the single quarter spanning from $1 billion to $4 billion constitutes the entirety of this investment. Let’s look at what is concentrated in that single quarter. Microsoft-bound Horizon 2–4 (50 MW IT each), Childress Air Cooling 50 MW, McKinsey 80 MW, and Prince George 50 MW are all targeted for commissioning in the fourth quarter of 2026. The fact that Horizon 1 has already been delivered and has even received NVIDIA Exemplar Cloud certification based on the NVIDIA GB300 NVL72 is proof of execution capability. However, completing the first one is a different matter from completing the remaining three simultaneously in a single quarter.
The customer base has expanded significantly. This quarter, a new multi-year contract was signed with an unnamed leading Frontier AI Lab, and Cohere, Prometheus, Perplexity, Figure AI, Pal AI, and Hicksfield AI were recently added to the contract list. This signals a shift away from a structure that relies on a single hyperscaler.
The contract unit prices are even more impressive. According to the company, the unit price for 3-year contracts has risen by 1,251 TP3T and for 5-year contracts by 701 TP3T in recent months. The recently signed 3-year contract exceeds $20 million per MW based on IT standards and offers a return on investment in about two years, while negotiations are reportedly currently underway at the $25 million per MW level. This data directly contradicts market concerns that GPU rental prices will fall.
The capacity roadmap is 0.3 GW (IT) in 2026 and 0.5 GW (IT) in 2027, and for 2028 and beyond, remaining capacity is being developed at Sweetwater 1 & 2 (1,700 MW), Kiowa (1,600 MW), Bundy (800 MW), and Badajoz, Spain (300 MW). The announced pipeline alone exceeds 5 GW. The company emphasized that its year-end ARR target of $4 billion corresponds to less than 101 TP3T of this pipeline, which, conversely, means that neither contracts nor funding have been determined for the remaining 901 TP3T.
Funding amounted to $6.5 billion (approximately 8.98 trillion won) in GPU financing over the past three months alone. The $3.6 billion (approximately 4.98 trillion won) for Microsoft deployment is at around 6.01 TP3T per annum under investment-grade conditions, while of the $2.8 billion (approximately 3.87 trillion won) for non-investment-grade customer deployments, $2.4 billion (approximately 3.32 trillion won) is being invested in McKinsey air-cooling expansions led by Blue Owl and PIMCO at a fixed rate of 9.01 TP3T per annum. Including down payments, this amounts to raising over 1,001 TP3T of the related GPU CAPEX. The fact that borrowing rates vary by 31 TP3T points depending on customer creditworthiness clearly illustrates the nature of this business model. Convertible stocks originating from the same Bitcoin mining sector Core Scientific (CORZ)Ina Applied Digital (APLD)Unlike the rental model, Airen chose a much more capital-intensive path of directly purchasing GPUs and selling the cloud.
And here is the largest number in this announcement. FY2027 CAPEX forecast: $25 billion to $30 billion (approx. 34.55 trillion to 41.46 trillion KRW)The plan is to execute an amount nearly double the current market capitalization of approximately 20 trillion won within one year. The financing plan consists of cash and committed GPU financing and advances totaling $14 billion (approximately 19.35 trillion won), an additional targeted GPU financing and advances totaling $8 billion (approximately 11.06 trillion won), with the remainder coming from data center financing, operating cash flow, corporate bonds, and stocks. Mathematically, the remaining portion of $3 billion to $8 billion has not yet had its funding source determined.
IREN Valuation: Which Multiple to Trust
Based on the closing price of $40.53 (approx. 56,012 won) on August 27, the market capitalization is $14.485 billion (approx. 20.02 trillion won). Adding $7.8368 billion in borrowings and financial leases and subtracting $5.8956 billion in unrestricted cash results in an enterprise value of $16.426 billion (approx. 22.70 trillion won). Following the earnings announcement, the stock price fell to $37.11 (approx. 51,286 won) in after-hours trading, down 8.451 TP3T.
The conclusion changes completely depending on how this corporate value is divided.
| denominator | amount | drainage | analysis |
|---|---|---|---|
| FY2026 Performance Revenue | $707 million | 23.2 times | Very expensive |
| Currently operating ARR | $1 billion | 16.4 times | expensive |
| FY2027 Projected Revenue | $2.9 billion | 5.7 times | commonly |
| Year-end contract ARR | $4 billion | 4.1 times | Cheap |
The projected revenue of $2.9 billion for FY2027 is an estimate provided by BTIG following this earnings call, and I also obtained a similar value by ramping up my ARR quarterly. Bulls talk about 4.1x, while bears talk about 23.2x. Both figures are from the same company on the same day.
Comparison of the 3 NeoCloud Companies
The figures below are standardized for all three stocks based on the closing price of the regular session on August 27, 2026, and the financial statements for the quarter ended June 30, 2026.
| enterprise | Market capitalization | Net borrowings | Enterprise Value (EV) | Sales over the past 12 months | EV/Sales | Sales growth rate |
|---|---|---|---|---|---|---|
| Coreweave (CRWV) | $47.9 billion (approximately 66.2 trillion won) | $46.1 billion | $93.9 billion (approximately 129.8 trillion won) | $7.59 billion | 12.4 times | +112% |
| Nevius (NBIS) | $59.4 billion (approximately 82.1 trillion won) | $2.2 billion | $62 billion (approximately 85.7 trillion won) | $1.355 billion | 45.8 times | +454% |
| IREN | $14.5 billion (approximately 20.0 trillion won) | $1.9 billion | $16.4 billion (approximately 22.7 trillion won) | $707 million | 23.2 times | +41% |
There is one point to note when reading the table. While Nevius has the largest market capitalization at $59.4 billion, Coreweave has the largest enterprise value at $93.9 billion. The reason for this reversal is debt. Coreweave's net debt stands at $46.1 billion, nearly equaling its market capitalization, whereas Nevius's is only $2.2 billion. In terms of debt-to-equity ratio, Coreweave stands at 1,027.1 TP3T, Irene at 1,871 TP3T, and Nevius at 991 TP3T. Since enterprise value is the sum of shareholder equity and creditor equity, companies that have acquired assets with debt tend to have an enterprise value greater than their market capitalization. The reason for comparing the three companies using EV/Revenue rather than market capitalization is precisely to correct for this difference.
All three companies are operating at a loss and have high short selling ratios. Short selling accounts for 18.91 TP3T of Coreweave's outstanding shares, 23.41 TP3T of Nevius's, and 27.21 TP3T of Airen's. Airen has the highest short selling ratio. While Coreweave is overwhelming in terms of revenue size, it carries $46.1 billion in net debt; Nevius, on the other hand, has the highest multiple but also the steepest growth rate and the cleanest financial structure among the three. Airen falls in the middle between the two in terms of leverage. However, it must be taken into account that this comparison makes Airen appear more favorable than reality, as its revenue over the past 12 months includes $578.2 million in disappearing mining revenue. When covering Nevius (NBIS) Q2 earningsThe conclusion was the same. In this industry, the key factor in judgment is not contracts, but financing.
I have the appropriate drain 6.5 times the contract ARRIt was set at [this value]. There are three reasons for setting it lower than Coreweave's 12.4x. First, ARR is an operational metric defined by the company, and the company itself warns that actual recognized revenue may be significantly lower than this. Second, execution risk remains as $3 billion out of the $4 billion is not yet operational. Third, the unconfirmed remainder of FY2027 CAPEX is between $3 billion and $8 billion.
IREN Target Price Scenario: Why $67 Becomes $45
Let's perform a calculation here that best reveals the nature of this stock. An appropriate multiple of 6.5 times Today's Financial StatusIf applied directly, subtract the enterprise value of $26 billion and net debt of $1.9412 billion, then divide by the current number of shares of 357.38 million. $67.32 per shareIt appears. It is 66% higher than the current price.
However, my 12-month baseline target is $45. The difference of over $22 comes entirely from one thing. The cost of raising $25 billion to $30 billion in CAPEXIn the meantime, net borrowings increase and the number of shares rises. Achieving a $4 billion ARR and the extent to which shareholder equity is diluted in the process are separate issues, and the target price must reflect the latter.
| scenario | Realization ARR | Applied Drainage | Expected net borrowings | Expected number of shares | Target price | Compared to the current price | weight |
|---|---|---|---|---|---|---|---|
| stress | $4.4 billion | 9.0 times | $4.9 billion | 380 million shares | $91 (approx. 125,762 won) | +125.0% | 23% |
| basic | $3.8 billion | 6.5 times | $6.9 billion | 395 million shares | $45 (approx. 62,190 won) | +10.9% | 45% |
| Weakness | $3 billion | 5.0 times | $7.9 billion | 420 million shares | $17 (approx. 23,494 won) | -58.5% | 32% |
The bullish scenario assumes that fourth-quarter commissioning concludes as planned, a portion of 2027 capacity is contracted early, and advance payments and GPU financing sufficiently cover capital expenditures, keeping stock issuance to a minimum. The bearish scenario assumes that commissioning is delayed by one or two quarters and GPU rental rates decline, leading to the raising of the remaining $3 billion to $8 billion through stock issuance at a low share price. Due to high leverage, the stock price moves to extremes both upward and downward. The very fact that the swing range reaches 5.4 times is the stock's honest profile.
The weighted average target price is $46.59, and the expected return based on a closing price of $40.53 is +15.01 TP3T. Based on an after-hours closing price of $37.11, it becomes +25.51 TP3T. The bearish weight was set at 321 TP3T, higher than the standard 251 TP3T, due to three factors: a short selling weight of 27.21 TP3T, a beta of 4.30, and unconfirmed funds.
I would also like to point out the differences with Wall Street. The consensus among 16 analysts is a Buy (1.59), with an average target price of $80.19 (approx. 110,823 KRW), a high of $131, and a low of $41. The average of $80.19 corresponds to applying approximately 10.2 times to a contracted ARR of $3.8 billion. Given that Coreweave is currently trading at 12.4 times based on actual revenue, I view applying 10.2 times to an ARR that is not yet operational as a calculation that barely reflects execution risk. On the other side is Richard Chow of JP Morgan. Citing falling GPU lease prices, uncontracted capacity, and high debt, he issued a Sell recommendation with a target price of $46 (approx. 63,572 KRW) and included this stock in his list of high-confidence short ideas for the third quarter. Coincidentally, my target of $45 is at almost the same level as his $46, but the path taken and the assessment of direction are opposite.
IREN Risk and Bear Trigger
| Risk | Influence mechanism | Monitoring Indicators |
|---|---|---|
| Q4 Commissioning Focus | Horizon 2–4, Mackenzie, and Prince George are all clustered in a single quarter. If even one of them falters, the year-end ARR target of $4 billion will be shaken, and the entire basis for valuation will shift. | Discrepancy between actual operating ARR and target at the end of the quarter |
| Unconfirmed funds $3 billion to $8 billion | Raising through stocks for a company with a market capitalization of 20 trillion won results in double-digit dilution. This dilution is even greater, especially when raising funds at a low stock price. | Quarterly number of shares issued, disclosure of new convertible bond issuance |
| Drop in GPU rental unit prices | If major operators like Meta increase their own capacity, NeoCloud's overall pricing power weakens. While corporate data is still moving in the opposite direction, this applies only to new contracts. | Contract unit price per MW(IT), composition of new contract period |
| The gap between ARR and GAAP revenue | ARR is a value based on 8,760 hours of full operation. Actual revenue is recognized lower than this due to capacity utilization, acquisition delays, and contract structure. The company also specifies this. | The ratio of quarterly AI cloud revenue multiplied by 4 to the ARR at the end of the previous quarter |
| Mining Revenue Gap | A business that generated $578.2 million (approximately 799.1 billion KRW) in FY2026 is disappearing. It fell by 401 TP3T in the fourth quarter alone. Revenue and EBITDA will be suppressed simultaneously until AI Cloud fills this gap. | Point of Rebound in Quarterly Mining Revenue and Adjusted EBITDA Margin |
| additional impairment loss | The 2027 plan includes the removal of 250MW of ASICs from Childress. Additional losses are recognized each time remaining mining equipment is liquidated. | Quarterly impairment loss, balance of assets held for sale |
| Extreme volatility | With a beta of 4.30, the short selling price is 27.21 TP3T relative to outstanding shares. The stock price moves by double digits on a single piece of news regarding earnings or a contract. | Short selling balance trend, 52-week fluctuation |
The bearish logic can be summarized in a single sentence as follows: This scenario assumes that the operational ARR at the end of December is limited to $2.5 billion instead of $4 billion, and that the company raises a large amount of capital through equity in January or February 2027, once this fact is confirmed. If that combination becomes a reality, the $17 scenario opens up. Conversely, if the ARR at the end of December meets the target and the financing is finalized through advance payments and GPU financing, the market will immediately shift its focus to 0.5GW in 2027.
Investment Opinion Conclusion
Investment opinion Buy (High Risk), confidence is low. The 12-month baseline target price is lowered to $45 (approx. 62,190 won) from the previous $53.
The path to this conclusion is as follows. Based on the contract ARR, the current multiple of 4.11x is 371 TP3T lower than the 6.5x I consider appropriate, and the expected return of +15.01 TP3T exceeds the buy threshold of +121 TP3T. Both the valuation and expected return signals point to a buy. However, the reason for maintaining a low confidence level is also clear. The bullish and bearish target price range is 5.4x, and most of the variables determining that range depend on execution events occurring within the next four months. I believe it would be dishonest to claim high confidence in a stock like this.
What has changed from the July post?
On July 1st, I We rated Irene at a target price of $53 and a Buy (High Risk) rating.. At that time, the basis was the remaining ARR of $3.1 billion and the contracts with Nvidia and Microsoft. Over the course of two months, the contract ARR increased to $4 billion, and the contract unit price rose by 1,251 TP3T over a three-year period. Looking solely at the business, it has improved compared to then.
However, I am lowering the target price. The reason is simple: the $25 billion to $30 billion FY2027 capital expenditure figure, which was disclosed for the first time. In July, I did not anticipate the scale of financing to be this large. Given the plan to spend nearly double the market capitalization within a year, there is no way to avoid raising the assumptions regarding dilution and net debt. I have also factored in the adjusted EBITDA margin falling from 411 TP3T to 141 TP3T. The business has improved, but the equity remaining for shareholders is thinning—this is the entirety of my judgment that I have changed over the past two months.
Personally, what impressed me most in this data was the 31 TP 3T point difference in funding rates. The Microsoft batch was 6.01 TP 3T, while the non-investment-grade customer batch was 9.01 TP 3T. This means the market is lending money while already pricing in which client Airen has contracted with. I thought that if one wanted to understand the true nature of this company, it would be faster to look at the rates of the next funding round rather than the client names in the press release.
Conditions for changing opinions
- Conditions for increasing confidence — When ARR of operation exceeds $3.5 billion by the end of December, and the unconfirmed remainder of FY2027 CAPEX is closed with GPU financing and advances without issuing shares. When adjusted EBITDA margin rebounds for two consecutive quarters.
- Conditions for lowering by holding — When Q4 commissioning is delayed by more than one quarter, or when the unit price of a new contract per MW(IT) falls below $20 million.
- Conditions to lower by selling — When raising more than $3 billion in equity at or below current prices, or when the year-end ARR target is officially lowered.
- Weight management — This is a stock with a beta of 4.30 and a short selling ratio of 27.21 TP3T. If you do not reflect the fact that the gap between the upper and lower target prices is 5.4 times in the position size, you may incur losses even if the analysis is correct.
Frequently Asked Questions
What is IREN's 12-month target price?
The base scenario is $45 (approximately 62,190 KRW), which is 10.91 TP3T higher than the closing price of $40.53 on August 27. The bullish scenario is $91, the bearish scenario is $17, and the weighted average expected return is +15.01 TP3T. The target price was lowered from the previous $53 due to an upward revision of dilution and leverage assumptions following the disclosure of $25 billion to $30 billion in FY2027 CAPEX. For reference, the average target price of 16 Wall Street analysts is $80.19, while JPMorgan has a sell recommendation of $46.
Why did sales decrease when performance fell short of expectations?
This is because Bitcoin mining rigs were dismantled first to make space for installing GPUs. While mining revenue ceases immediately, GPU billing only begins after commissioning is complete, creating a gap in between. In the fourth quarter, mining revenue decreased by $44.5 million, and although AI cloud revenue increased by $36.9 million to offset most of the shortfall, a shortfall of $7.6 million resulted in a 5.21 TP3T decrease in total revenue. This is akin to the cost of an intended transition process.
Does the contract ARR of $4 billion become revenue as is?
That is not the case. ARR is calculated by multiplying the contracted hourly unit price of GPUs by 8,760 hours per year, and it is an operational metric based on full operation. The company also stated in its press release that ARR is not a metric under U.S. accounting standards and that recognized revenue could be significantly lower than this. Furthermore, the $4 billion target is set for December 31, while the actual ARR in operation as of August 26 is $1 billion. The remaining $3 billion depends on fourth-quarter commissioning.
How will the $25 billion to $30 billion FY2027 CAPEX be raised?
The company's plan consists of three phases. It includes $14 billion in cash already secured and committed GPU financing and customer advances, an additional $8 billion in targeted GPU financing and advances, and the remainder from data center financing, operating cash flow, corporate bonds, and stocks. Calculations show that the funding source for $3 billion to $8 billion remains undecided. Since filling this shortfall with stocks would result in double-digit dilution for a company with a market capitalization of approximately 20 trillion won, the trend in the number of outstanding shares over the next few quarters will become an indicator as important as earnings.
Sources and References
- IREN Limited Investor Event and Presentation — FY2026 Earnings Presentation (August 27, 2026)
- IREN Limited Official Press Release — IREN Reports FY26 Results
- StockTitan — IREN FY26 Results: Contract ARR $4 billion, Net Loss $703 million
- Benzinga — IREN Q4 2026 Earnings Conference Call Full Text
- Investing.com — IREN Q4 FY26 Presentation Slide Analysis
Financial figures are based on the company’s FY2026 Form 10-K filed with the SEC on August 27, 2026, and earnings presentation released on the same day, while the stock price and analyst consensus are based on the close of regular trading on August 27.
※ This article was written based on company disclosures and public data with the assistance of AI research tools, and the author personally reviewed the figures and logic prior to publication. It is intended for informational purposes only and does not constitute a recommendation to trade any specific stock. Stock prices, exchange rates, and estimates are based on the time of writing and are subject to change; investors bear full responsibility for their investment decisions and the consequences thereof.