At 4,510 MW total pipeline, IREN is valued at ~$3.0M/MW today. Purpose-built AI data centers trade at $10-20M/MW. If IREN successfully converts even half its pipeline (2,250 MW) to AI/HPC at $15M/MW, that alone implies ~$34B enterprise value. The $9.7B Microsoft contract over 5 years provides revenue visibility. Add the mining cash flow bridge plus the Oklahoma expansion, and the 10x case requires: (1) Sweetwater comes online on schedule, (2) Microsoft deal executes smoothly, (3) additional hyperscaler contracts for remaining capacity. In an AGI world where every MW of data center capacity is priceless, a company with 4.5 GW of power pipeline is a scarce asset.
| Dimension | IREN | CORZ | RIOT | CLSK |
|---|---|---|---|---|
| AI Pivot Model | Own GPUs + Cloud Services | Colocation (CoreWeave's GPUs) | Minimal AI pivot | Pure mining |
| Power Pipeline | 4,510 MW | ~1,400 MW | ~1,000 MW | ~900 MW |
| Hyperscaler Contract | Microsoft $9.7B | CoreWeave 590 MW | None | None |
| Balance Sheet | Positive equity ($2.5B) | Negative equity | Positive | Positive |
| Market Cap | $13.7B | $5.1B | ~$4B | ~$3B |
| $/MW (total pipeline) | $3.0M | $3.6M | $4.0M | $3.3M |
| Value Chain Position | Higher (compute provider) | Lower (space/power provider) | Mining only | Mining only |
| Land Ownership | Freehold (owns land) | Mixed | Mixed | Mixed |
| Renewable Energy | 100% (RECs) | Partial | Partial | Partial |
IREN occupies a higher position in the AI infrastructure value chain than CORZ. Where CORZ provides empty data center shells (colocation) for CoreWeave to fill with GPUs, IREN owns the full stack: power contracts, land, data centers, AND the GPUs themselves. This means IREN captures GPU cloud margins (60-80%) rather than colocation margins (30-40%). The risk is also higher -- IREN bears GPU obsolescence risk and must manage the cloud operations -- but the reward is proportionally larger. The Microsoft contract validates this model: Microsoft is paying IREN for dedicated GPU compute, not for rack space.
1,800+ acres · ~40 miles from Abilene, TX · ERCOT grid connection
Sweetwater 1 (1,400 MW): Grid connection targeting Q2 CY2026
Sweetwater 2 (600 MW): Grid connection targeting Q4 CY2027
Capable of supporting 700,000+ liquid-cooled GPUs
Direct fiber loop designed between Sweetwater 1 and 2
2,000 acres · Grid connection via Public Service Company of Oklahoma
Contractual rights to land and power · $112M acquisition cost
Connection rights amortized over 12 years, beginning Q4 FY2028
Newest site, adds geographic diversification beyond Texas/BC
| Category | MW | % of Pipeline | Status |
|---|---|---|---|
| Operating (TX + BC) | 910 MW | 20% | Live and generating revenue |
| Under Construction (Sweetwater) | 2,000 MW | 44% | Grid connection 2026-2027 |
| In Development (Oklahoma) | 1,600 MW | 36% | Land + power secured, early stage |
| Total Pipeline | 4,510 MW | 100% |
| Metric | Value | Note |
|---|---|---|
| Total contract value | $9.7B | Over 5-year average term |
| Implied annual revenue | ~$1.94B/yr | $9.7B / 5 years (when fully deployed) |
| Prepayment (20%) | ~$1.94B | Received before delivery of each tranche |
| Financing facility | $3.6B | Secured by GPUs + contracted cash flows (in negotiation) |
| Current RPO (as of Dec 31, 2025) | $289.4M | Only delivered/accepted tranches included |
| Deferred revenue (Dec 31, 2025) | $46.6M | Up from $0.9M at June 30, 2025 |
The Microsoft deal represents a massive concentration bet. If Microsoft decides to reduce GPU spending, renegotiate terms, or if IREN fails to deliver the GPU tranches on time, the revenue impact would be severe. This mirrors the CORZ/CoreWeave dynamic -- early-stage AI infrastructure providers often start with a single large customer. The critical question: can IREN diversify into multiple hyperscaler relationships before becoming dangerously dependent on one?
However, the 20% prepayment structure reduces execution risk -- IREN gets capital upfront to procure GPUs, and the $3.6B financing facility further de-risks the capex required.
| Period | BTC Mining | AI Cloud | Total Revenue | AI % | YoY Growth |
|---|---|---|---|---|---|
| FY2023 (Jun 2023) | $75.5M | $0 | $75.5M | 0% | -- |
| FY2024 (Jun 2024) | $184.1M | $3.1M | $187.2M | 1.7% | +148% |
| FY2025 (Jun 2025) | $484.6M | $16.4M | $501.0M | 3.3% | +168% |
| H1 FY2026 (Dec 2025) | $400.3M | $24.6M | $425.0M | 5.8% | +152% |
| Quarter | BTC Mining | AI Cloud | Total | AI Cloud Margin | BTC Mining Margin |
|---|---|---|---|---|---|
| Q1 FY26 (Sep 2025) | $233.0M | $7.3M | $240.3M | 90% | 66% |
| Q2 FY26 (Dec 2025) | $167.4M | $17.3M | $184.7M | 86% | 62% |
| Q4 FY25 (Jun 2025) | $180.3M | $7.0M | $187.3M | 93% | 71% |
| Q3 FY25 (Mar 2025) | $141.2M | $3.6M | $144.8M | 91% | 71% |
| Q2 FY25 (Dec 2024) | $113.5M | $2.7M | $116.1M | 90% | 72% |
| Q1 FY25 (Sep 2024) | $49.6M | $3.2M | $52.8M | 93% | 36% |
AI Cloud Services cost of revenue in Q2 FY26 was just $2.4M on $17.3M revenue -- an 86% gross margin. This is the GPU cloud model at work: once GPUs are installed and running, the marginal cost of serving compute is primarily electricity. Bitcoin mining gross margins are 62%, which is healthy but not exceptional. As AI Cloud becomes a larger share of revenue, blended margins should improve significantly -- especially once the Microsoft contract ramps.
| Metric | FY2023 | FY2024 | FY2025 | H1 FY2026 |
|---|---|---|---|---|
| Revenue | $75.5M | $187.2M | $501.0M | $425.0M |
| Operating Income | $(157.2M) | $(27.2M) | $17.3M | $(192.8M) |
| Net Income | $(171.8M) | $(28.9M) | $86.9M | $229.2M |
| EBITDA | $(123.5M) | $19.3M | $278.2M | $418.8M |
| Adjusted EBITDA | $1.1M | $54.4M | $269.7M | -- |
| Adj. EBITDA Margin | 1% | 29% | 54% | -- |
| Operating Cash Flow | $5.7M | $52.2M | $245.9M | $356.4M |
| CapEx | $(116.1M) | $(141.9M) | $(573.5M) | $(1,117.5M) |
| Free Cash Flow | $(110.3M) | $(89.6M) | $(327.5M) | $(761.1M) |
IREN is spending aggressively on GPU procurement, data center construction, and the Sweetwater/Oklahoma buildout. H1 FY26 capex was $1.12B vs. $356M operating cash flow. This capital intensity is the cost of the pivot. The question is whether the Microsoft contract revenue ($1.94B/yr at full deployment) will be enough to make IREN FCF positive. If GPU capex moderates after initial deployment and recurring cloud revenue comes in, the math works. But the current cash burn rate means IREN depends on continued access to capital markets.
| Assets | Dec 31, 2025 | Jun 30, 2025 | Change |
|---|---|---|---|
| Cash & Cash Equivalents | $3,260.6M | $564.5M | +$2,696.1M |
| Other Current Assets | $122.8M | $76.6M | +$46.2M |
| Total Current Assets | $3,383.4M | $641.2M | +$2,742.2M |
| PP&E, net | $3,170.5M | $1,930.6M | +$1,239.9M |
| Intangible Assets (Oklahoma) | $107.6M | $0 | +$107.6M |
| Derivative Assets | $215.7M | $333.7M | $(118.0M) |
| Other Non-Current | $150.4M | $34.8M | +$115.6M |
| Total Assets | $7,027.6M | $2,940.3M | +$4,087.3M |
| Liabilities & Equity | Dec 31, 2025 | Jun 30, 2025 | Change |
|---|---|---|---|
| Accounts Payable & Accrued | $576.3M | $144.1M | +$432.2M |
| Finance Lease Liabilities | $155.9M | $0 | +$155.9M |
| Convertible Notes Payable | $3,685.3M | $962.8M | +$2,722.5M |
| Other Liabilities | $98.9M | $15.9M | +$83.0M |
| Total Liabilities | $4,516.4M | $1,122.8M | +$3,393.5M |
| Total Stockholders' Equity | $2,511.2M | $1,817.5M | +$693.7M |
| Book Value / Share | $7.57 | $7.04 | +$0.53 |
| Price / Book | 5.5x | -- | -- |
IREN has funded its expansion primarily through convertible notes and equity issuances -- notably, no traditional bank debt. This is important because convertibles provide cheap capital (0-3.5% interest) and convert to equity at premium prices, avoiding the covenant restrictions of bank debt.
| Note | Principal | Rate | Maturity | Conv. Price | Status |
|---|---|---|---|---|---|
| 2030 Convertible Notes | $212.3M | 3.25% | Jun 2030 | $16.81 | Deep in-the-money |
| 2029 Convertible Notes | $233.4M | 3.50% | Dec 2029 | $13.64 | Deep in-the-money |
| 2031 Convertible Notes | $1,000.0M | 0.00% | Jul 2031 | $85.63 | Out-of-the-money |
| 2032 Convertible Notes | $1,150.0M | 0.25% | Jun 2032 | $51.40 | Near the money |
| 2033 Convertible Notes | $1,150.0M | 1.00% | Jun 2033 | $51.40 | Near the money |
| Total | $3,745.7M | -- | -- | -- | -- |
Current shares: 331.8M. If all convertible notes convert at their respective conversion prices:
Total potential dilution: ~86M shares (+26%). Fully diluted share count would be ~418M shares. At $41.37, fully diluted market cap is ~$17.3B. The in-the-money 2029/2030 notes (~30M shares) will almost certainly convert, adding ~7% dilution. The $51.40 conversion notes are near current price levels.
Additionally, IREN issued $1B under an ATM facility and did a $1.63B direct equity offering in Dec 2025. Shares outstanding grew from 186M (Jun 2024) to 332M (Dec 2025) -- a 78% increase in 18 months. This dilution is the cost of funding the pivot.
| Metric | FY2023 | FY2024 | FY2025 | H1 FY2026 |
|---|---|---|---|---|
| Bitcoin Mining Revenue | $75.5M | $184.1M | $484.6M | $400.3M |
| Mining Cost of Revenue | $(39.4M) | $(86.7M) | $(157.7M) | $(143.3M) |
| Mining Gross Profit | $36.1M | $97.4M | $326.9M | $257.0M |
| Mining Gross Margin | 48% | 53% | 67% | 64% |
| Installed Hashrate | -- | -- | 50 EH/s | 50 EH/s |
Bitcoin mining serves as the cash flow bridge during the AI pivot. With 50 EH/s of hashrate and 67% gross margins, mining generates ~$500M+ in annual gross profit at current Bitcoin prices. This is substantially better than CORZ's 5% mining margins -- IREN benefits from lower power costs (BC hydro, TX wholesale) and newer-generation mining hardware (Bitmain S21 XP).
Key difference from CORZ: IREN sells all Bitcoin daily (no HODL strategy), which means zero Bitcoin balance sheet exposure. Revenue is immediately converted to USD/CAD. This is conservative and reduces volatility -- but means IREN doesn't benefit from Bitcoin price appreciation on holdings.
The next Bitcoin halving is expected in April 2028 (block reward drops from 3.125 to 1.5625 BTC). If Bitcoin price doesn't double by then, mining margins will compress. But by April 2028, the Microsoft deal should be generating ~$1.94B/yr in AI revenue, making the mining bridge largely irrelevant. The key risk window is H2 2026 through H1 2027 -- after mining hashrate may be curtailed for Horizon GPU deployments but before Microsoft revenue fully ramps.
| Quarter | Shares | Value | Change |
|---|---|---|---|
| Q1 2025 | 3,366,130 | $20.5M | New position |
| Q2 2025 | 6,400,384 | $93.3M | +90% shares |
| Q3 2025 | 7,220,421 | $338.9M | +13% shares |
| Q4 2025 | 8,700,621 | $328.6M | +20% shares |
Leopold entered IREN in Q1 2025 and has increased his position every quarter. At 8.7M shares, IREN represents ~6.3% of his ~$5.2B portfolio (his 3rd or 4th largest position). Unlike CORZ where he filed a 13D (activist), IREN appears to be a passive holding -- he's below the 5% ownership threshold.
Leopold holds both CORZ (activist, 9.4%) and IREN (passive, ~2.6%). His thesis across both: physical AI infrastructure is the bottleneck, and Bitcoin miners who already own power contracts + land are the cheapest way to get exposure to AI compute demand. IREN is his more expensive bet (IREN at $13.7B vs CORZ at $5.1B) but arguably the better-positioned company: IREN owns GPUs (higher value chain), has a Microsoft contract (not CoreWeave), positive equity, and a larger power pipeline (4.5GW vs 1.4GW). The steady building over 4 quarters suggests high conviction -- he kept adding even as the stock ran from ~$6 to ~$47.
| Strategy | Details | Margin Profile | Capital Intensity |
|---|---|---|---|
| AI Cloud Services (Current) |
Own GPUs, sell compute. Currently 1,900 H100/H200 at Prince George, scaling to 10,900. Microsoft contract for Childress Horizon. | 86% gross margin | Very high (GPUs cost $25-40K each) |
| Colocation Services (Future potential) |
Powered shell, build-to-suit, turnkey data centers. Sweetwater's 700K+ GPU capacity could also serve colocation customers. | 30-50% gross margin | Moderate (buildings + power, no GPUs) |
| Date | GPU Count | GPU Types | Location |
|---|---|---|---|
| Jun 2025 | ~1,900 | H100, H200 | Prince George, BC |
| End CY2025 (target) | ~10,900 | H100, H200, B200, B300, GB300 | Prince George, BC |
| CY2026 (Microsoft tranches) | TBD (massive) | Latest NVIDIA (likely GB200/GB300) | Childress Horizon 1-4, TX |
The progression from 1,900 GPUs to 10,900 at Prince George is a 5.7x increase in AI compute capacity. But the real scaling comes from the Microsoft deployment at Childress -- 200MW of liquid-cooled GPU capacity at 100-200kW per rack implies thousands of racks, potentially 50,000+ GPUs. Sweetwater, capable of hosting 700,000+ GPUs, represents an order-of-magnitude further expansion.
| Scenario | Capacity Valued | $/MW | Implied EV | Implied Share Price |
|---|---|---|---|---|
| Current trading (operating only) | 810 MW | $17.0M | $13.7B | $41 |
| Bitcoin miner valuation (full pipeline) | 4,510 MW | $3-5M | $13.5-22.6B | $41-68 |
| Transitioning (blended) | 4,510 MW | $5-8M | $22.6-36.1B | $68-109 |
| AI/HPC infrastructure | 4,510 MW | $10-15M | $45.1-67.7B | $136-204 |
| Premium AI infrastructure | 4,510 MW | $15-20M | $67.7-90.2B | $204-272 |
Note: Implied share price uses fully diluted count of ~418M shares (including all convertible note conversions), subtracting $490M net debt.
| Scenario | Annual Rev | Multiple | Implied EV | Assumption |
|---|---|---|---|---|
| Current run-rate | $850M | 16x | $13.6B | H1 FY26 annualized |
| Microsoft at full run-rate + mining | $2.7B | 10x | $27B | $1.94B MSFT + $800M mining |
| Full deployment (Sweetwater online) | $5-8B | 8x | $40-64B | Additional hyperscaler contracts |
| Bull case (4.5GW fully monetized) | $8-15B | 8-10x | $64-150B | Full AI data center operation |
At ~$13.7B market cap, IREN is near the low end of the 10x entry zone under the bull case. At 4,510 MW total pipeline, the bull case implies ~$14-33M/MW -- within the "premium AI infrastructure" range. This is achievable only if:
Probability: In an AGI-by-2030 world, this is plausible (maybe 20-30% probability). The constraint is execution -- building out 3,600 MW of new capacity while simultaneously deploying hundreds of thousands of GPUs is an enormous operational challenge. But if AI compute demand continues its current trajectory, every MW matters, and IREN has 4.5 GW of it.
| Risk | Severity | Probability | Detail |
|---|---|---|---|
| GPU Obsolescence | HIGH | Moderate | IREN owns GPUs -- if NVIDIA releases faster chips, existing fleet depreciates rapidly. The 10-Q shows $48M in impairment charges in H1 FY26 already. This is the cost of owning vs. hosting. |
| Microsoft Concentration | HIGH | Moderate | $9.7B contract = majority of future revenue. If Microsoft reduces AI spend or renegotiates, IREN has limited alternatives for that capacity in the short term. |
| Execution Risk (Sweetwater/Oklahoma) | HIGH | Moderate | Building 3,600 MW of new capacity is an enormous construction project. Delays, cost overruns, grid connection issues are all possible. 257 employees is a very small team for this scale. |
| Dilution | MEDIUM | High (certain) | Shares grew 78% in 18 months. Convertible notes add 26% more. If stock drops below conversion prices, notes become true debt instead of equity. |
| Bitcoin Price Crash | MEDIUM | Moderate | Mining is 96% of current revenue. A sharp Bitcoin decline would cut the cash flow bridge before Microsoft revenue ramps. The daily liquidation policy limits balance sheet exposure but not revenue exposure. |
| Capital Markets Closure | HIGH | Low-Moderate | IREN is FCF negative and depends on continued access to debt/equity markets. If credit markets tighten, the expansion stalls. The $3.6B Microsoft financing facility is not yet finalized. |
| Power Cost Increase | MEDIUM | Low-Moderate | Texas (ERCOT) wholesale power prices are volatile. BC Hydro contracts can be terminated with 6 months notice. Rising power costs would compress both mining and AI margins. |
| SBC Dilution | MEDIUM | High (ongoing) | $130.6M in stock-based compensation in H1 FY2026 alone (vs. $16.2M in H1 FY2025). SBC is 31% of revenue -- very high. |
| Australia Tax / Regulatory | LOW | Low | IREN is incorporated in Australia with operations in US/Canada. Cross-border tax complexity and Australian regulatory changes are tail risks. |
| Name | Title | Age | Compensation (FY2025) |
|---|---|---|---|
| Daniel Roberts | Co-Founder, Co-CEO | 40 | $2.95M + $11.3M unexercised options |
| William Roberts | Co-Founder, Co-CEO | 35 | $2.95M + $11.3M unexercised options |
| Anthony Lewis | CFO | 49 | Not disclosed |
| David Shaw | COO | 58 | $711K |
| Denis Skrinnikoff | CTO | 40 | Not disclosed |
Key observation: The company is run by two brothers (Daniel and William Roberts), both co-founders. The team is very small -- 257 full-time employees managing 810 MW of operations and 3,600 MW of development. Insider ownership is 3.0%, which is relatively low for a founder-led company, though the $11.3M in unexercised options per co-CEO suggests significant skin in the game via options rather than direct shares.
| Metric | Value | Context |
|---|---|---|
| Employees | 257 | Very lean for $7B in assets |
| Revenue per Employee | ~$2.9M | TTM basis, excellent efficiency |
| Assets per Employee | $27.3M | Capital-intensive, infrastructure model |
| Operating Hashrate | 50 EH/s | ~6% of global Bitcoin hashrate |
| Mining Hardware | Bitmain S21 XP/S21 Pro | Latest generation, energy efficient |
| Energy Source | 100% Renewable | BC hydro + TX RECs |
| Short Interest | 13.4% of float | 44.5M shares short -- significant bearish bet |
| Beta | 4.32 | Extremely volatile |
In the AGI scenario, every watt of data center power becomes priceless. IREN's 4.5 GW pipeline represents ~5% of the estimated incremental US data center power needed by 2031. Grid interconnection queues average 55 months -- meaning IREN's existing grid connections are a 4-5 year head start over new entrants. The Sweetwater sites (2,000 MW targeting 700K+ GPUs) could be among the largest AI compute clusters in the world.
The counter-risk: AGI could design chips so efficient that power demand per computation drops dramatically. But this is a 5-10 year risk, and the demand for total compute likely outpaces efficiency gains in the near term.
| Asset | Book Value | Est. Liquidation | Notes |
|---|---|---|---|
| Cash | $3,260.6M | $3,260.6M | 100% of face value |
| PP&E (land, buildings, hardware) | $3,170.5M | $1,500-2,000M | 50-63% of book -- land + electrical infra retain value, GPUs/ASICs depreciate |
| Derivative Assets (warrants) | $215.7M | $0-100M | Highly dependent on stock price; volatile |
| Other Assets | $380.8M | $150-200M | Deposits, prepayments, receivables |
| Total Liquidation Value (est.) | -- | $4,910-5,560M | |
| Less: Total Liabilities | -- | $(4,516.4M) | Convertible notes may convert to equity rather than be repaid |
| Net Liquidation Value | -- | $394-1,044M | |
| Per Share (332M shares) | -- | $1.19-3.15 | If convertibles don't convert |
The floor is very low relative to the current $41 price. This is a growth/option-value stock, not a value stock. The tangible liquidation value barely covers liabilities. If the AI pivot fails and Bitcoin crashes, the equity could trade in single digits. The $3.75B in convertible debt is a significant obligation -- if the stock drops below conversion prices, these become real debt that must be repaid, potentially forcing dilutive equity raises or even restructuring.
This is not a stock where "you can't lose money." IREN is a high-conviction bet on AI infrastructure demand. The downside is real and significant. The upside is also enormous.
IREN is the most ambitious play in the Bitcoin-miner-to-AI-infrastructure space. Unlike CORZ (which rents space to CoreWeave), IREN is building a vertically integrated GPU cloud with a $9.7B Microsoft contract as its anchor tenant. The power pipeline (4.5 GW) is the largest among miners and the freehold land ownership provides real asset backing. But this is a high-risk, high-reward bet: $3.75B in convertible debt, massive dilution, FCF deeply negative, and execution of a 3,600 MW buildout with 257 employees is an extraordinary challenge.
At $41 ($13.7B market cap), the market is pricing IREN as an AI infrastructure company on operating capacity ($17M/MW on 810 MW) or as a transitioning miner on the full pipeline ($3.0M/MW on 4,510 MW). The 10x case requires all 4.5 GW to come online and be monetized at AI data center multiples ($10-15M/MW), which would imply $45-68B enterprise value. In an AGI-by-2030 world, every megawatt is priceless -- but IREN must survive the capital-intensive transition period to get there.
Data Sources: SEC EDGAR (CIK 1878848), 10-K FY2025 (filed Aug 28, 2025), 10-Q Q2 FY2026 (filed Feb 5, 2026), XBRL financial data, yfinance market data, Leopold 13F filings (Situational Awareness LP), IREN investor relations website.
Generated: March 12, 2026 | Not investment advice. Analysis for research purposes only.