Loading research website
IREN · AI infrastructure and digital assets
This model estimates the cash IREN could generate from operating and planned sites, subtracts construction and hardware-replacement costs, reduces the value of uncertain projects, subtracts net debt, and divides what remains among diluted shares.
Base-case value estimate
US$38.96
Based on the conditions listed in “Near-term 1.21 GW stage.”
Blended scenario result
US$40.92
This combines the displayed scenarios using the model weights. The weights are judgment calls, not statistical guarantees.
This model estimates the cash IREN could generate from operating and planned sites, subtracts construction and hardware-replacement costs, reduces the value of uncertain projects, subtracts net debt, and divides what remains among diluted shares.
What could one IREN share be worth at different stages of its AI and Bitcoin-mining buildout?
A standard DCF often treats a company as one set of yearly totals. This model separates sites and GPU buildout groups because each can start at a different time and carry different contracts, costs, financing needs, and completion risk.
The value per share is conditional on the selected buildout stage and assumptions. It is not a prediction that IREN's market price will reach that value by a specific date.
A DCF estimates present value from future cash flows. This page may also use project probabilities, cap rates, net asset value, or sum-of-the-parts methods where they fit the asset better. The model type is shown above and explained here before the detailed tables.
This is a current preliminary scenario model, not a precise forecast. The public workbook contains more detail than the headline page, but several economically important inputs remain estimates. The sections below make those dependencies visible before a reader uses the valuation output.
Facts or company statements tied to filings and formal disclosures.
Editable inputs used to translate disclosed plans into cash flow.
Areas where the current framework remains simplified or incomplete.
Valuation-date reference
$38.26
Workbook market reference on July 30, 2026; not a current quote.
Base diluted shares
385M
About 7.7% above the April 30 filed share count before further financing paths.
1.21 GW base estimate
$38.96
Probability-weighted stage value using the published base assumptions.
Beyond explicit forecast
82.5%
Share of 1.21 GW enterprise value attributable to cash flow after 2030.
The Microsoft-linked cohort produces an 18.7% headline project-EBITDA yield on the modeled $8.8 billion project cost. After reserving enough annual cash to recover the GPU investment over five years, net of a 12.5% residual, the modeled yield falls to approximately 7.2%.
Public WULF and CIFR projects show approximately 17.5% and 17.9% infrastructure yields on cost, respectively. Those are not direct valuation equivalents: their leases are longer and the tenant or compute operator bears most GPU-obsolescence risk, while IREN owns the compute layer.
A useful downside test raises the required return, lowers revenue density and utilization, adds a delay, reduces debt funding, increases diluted shares, and cuts completion probabilities. These links load transparent examples that can be changed further.
Selecting a buildout stage does not automatically switch every input to a separate Bear, Base or Bull economics column. The stage anchor and the adjustable controls are intentionally independent.
Ephesus Research authored the DCF. Company filings and formal disclosures anchor reported facts. Public IREN Community code supplies reproducible site and GPU input references where company detail is unavailable; those inputs remain clearly marked as community-derived or analyst assumptions. The live spreadsheet, source archive, assumption evidence, model checks and dated change log are all public.
Use these sections in order or jump directly to the part you need. The plain-language explanation comes first; the detailed assumptions, formulas, sources, and change log remain available underneath it.
Use the spreadsheet to inspect formulas or change assumptions. JSON and CSV versions are available for structured review and data reuse.
Model overview
Main question
What could one IREN share be worth at different stages of its AI and Bitcoin-mining buildout?
Technical approach
Discounted cash flow and buildout valuation
Company or asset
IREN
Estimate date
Aug 12, 2026
Years modeled
2026E–2030E plus replacement-adjusted terminal value
Currency
USD
Diluted shares used
385 million
Net debt / (cash) used
US$4,700 million
Research status
Current preliminary source-backed model
Model author
Ephesus Research
Last updated
Aug 12, 2026
Linked sources
17 records
Version
1.2.0-preliminary
3 possible outcomes
The base case is the central set of assumptions, not a guaranteed result. Compare every named scenario to see how delays, weaker economics, stronger execution, financing, or other major changes affect the estimate. Any edge case is deliberately kept separate from the central forecast. The weights are model judgments and should change when the evidence changes.
| Scenario | Business valueEnterprise value | Value for shareholdersEquity value | Estimated value per share | Weight in blended result | What has to happen | Main risks |
|---|---|---|---|---|---|---|
| Current 480 MW platform | US$15,126m | US$10,426m | US$27.08 | 30% |
|
|
| Near-term 1.21 GW stage | US$19,700m | US$15,000m | US$38.96 | 50% |
|
|
| Official-capped 5.0 GW stage | US$30,329m | US$25,629m | US$66.57 | 20% |
|
|
Project timing
A large announced power figure is not the same as usable capacity or current revenue. This table separates total power, computing load, timing, build cost, ownership, and the chance that each phase is completed.
| Site and phase | Total power | Usable computing power | Power ready | Revenue could begin | Expected use | Annual revenue per MW | Operating margin | Build cost | How it may be funded | Company ownership | Chance of completion | Value after risk adjustment | Estimated value per share |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Operating and contracted portfolioCurrent 480 MW stage | 480 MW | 360 MW | 2026 | 2026 | 85% | US$7.72m | 84% | US$0.00m | Treated as the opening/current platform for stage analysis; subsequent growth capital is modeled in later cohorts and the capital-structure bridge. | 100% | 100% | US$10,426m | US$27.08 |
| Childress and near-term expansion cohortsIncrement to 1.21 GW | 730 MW | 503.6 MW | 2027 | 2027–2028 | 85% | US$7.68m | 84% | US$19,060m | Includes explicit 2026–2027 cohort build capital; Microsoft prepayment and GPU financing are modeled separately from residual corporate funding and dilution. | 100% | 82% | US$4,574m | US$11.88 |
| Sweetwater and medium-term cohortsIncrement from 1.21 GW to 2.31 GW | 1,100 MW | 733.4 MW | 2028 | 2028–2029 | 75% | US$7.11m | 84% | US$25,168m | Community site timing is probability-weighted; customer contracting, project financing, GPU procurement, and construction remain required. | 100% | 55% | US$5,144m | US$13.36 |
| Sweetwater and later strategic pipelineIncrement from 2.31 GW to official-capped 5.0 GW | 2,690 MW | 1,793.3 MW | 2029–2030+ | 2029–2031+ | 70% | US$6.78m | 85% | US$61,547m | Substantial external funding, contracts, equipment, permitting, and construction are assumed; Kiowa is capped so the official strategic stage does not exceed 5.0 GW. | 100% | 30% | US$5,486m | US$14.25 |
Inputs behind the result
Start with inputs marked low confidence. Those are the assumptions most likely to need more evidence and can be more important than the final headline number. Each row shows whether the value is a fact, company statement, analyst estimate, model assumption, or inference.
| Input | Value used | Case | Evidence type | Where it came from | Information date | Confidence | Why it matters or what remains uncertain |
|---|---|---|---|---|---|---|---|
| Official near-term AI Cloud capacity milestones | 480 MW in 2026; 1.21 GW in 2027 stage-end AI Cloud MW | Base | Company Guidance | IREN Q3 FY26 results and buildout update | Jul 30, 2026 | High | High confidence for aggregate targets; exact site allocation, commissioning, financing, and customer acceptance remain cohort-specific execution risks. |
| Official strategic AI Cloud framework | 5000 MW maximum modeled official-capped stage | Bull | Company Guidance | IREN and NVIDIA strategic partnership disclosure | Jul 30, 2026 | Medium | The up-to-5 GW framework is a strategic ceiling, not a fully contracted backlog or proof of the community model's exact site schedule. |
| Microsoft contracted AI Cloud economics | $9.7B contract, $1.94B annual revenue, 85% project EBITDA, $5.8B GPU capex disclosed contract anchors | Base | Fact | SEC-filed Microsoft contract materials | Jul 30, 2026 | High | Anchors the contracted Horizon cohort. Detailed operating costs, remedies, delivery acceptance, renewal, and end-of-term residual value remain material. |
| Microsoft GPU financing bridge | $1.94B prepayment plus $3.65B debt at 6.00%; 3.31% disclosed blended funding cost; approximately 96% nominal GPU-capex coverage cohort financing reference | Base | Fact | IREN GPU financing announcement and filing | Aug 12, 2026 | High | The customer prepayment and senior debt are modeled separately. The 3.31% average financing cost is not treated as the debt coupon, and future cohorts are not assumed to obtain identical financing. |
| Equity discount rate | 11 % | Base | Model Assumption | Live workbook base assumption | Jul 30, 2026 | Low | The workbook ranges from 9.5% to 14.0% across scenarios. This is a required-return assumption, not company guidance. |
| Terminal replacement-adjusted FCF multiple | 12.5 x | Base | Model Assumption | Live workbook terminal-value assumption | Jul 30, 2026 | Low | Applied after deducting a normalized five-year GPU replacement reserve net of residual value; terminal dependence remains material. |
| Merchant and contracted utilization | 75% merchant; 100% contracted % | Base | Model Assumption | Community GPU defaults and live workbook adjustments | Jul 30, 2026 | Low | Realized utilization, net pricing, discounts, downtime, and customer mix are not fully disclosed for the modeled merchant fleet. |
| AI project EBITDA margin | 82% merchant; 85% contracted % | Base | Model Assumption | Microsoft contracted margin anchor plus merchant haircut | Jul 30, 2026 | Medium | The contracted anchor is source-backed; the merchant margin remains an analyst assumption before corporate overhead and replacement capital. |
| GPU primary life and Year-5 residual value | Five-year primary life; 12.5% unsupported residual; Years 6-8 tail excluded from core base case | Base | Model Assumption | Live workbook replacement-capital and GPU Capital Duration tabs | Aug 12, 2026 | Low | The five-year period is the core accounting and economic normalization, not a hard physical shutdown date. Any explicit post-Year-5 cash-flow tail must replace part of the residual assumption rather than being added to it in full. |
| Later-site execution probabilities | 85% Horizon 5-6; 80% Childress 7-10; 55% SW1; 35% SW2/Oklahoma; 25% Nostrum; 20% Kiowa % by cohort | Base | Inference | Official evidence mapped to community site cohorts | Jul 30, 2026 | Low | Operating and contracted cohorts receive 100%. Later probabilities should change only with contracts, power, permits, financing, construction, or commissioning evidence. |
| Base diluted share count | 385 millions | Base | Model Assumption | Filed share count plus modeled dilution | Jul 30, 2026 | Medium | The filed starting point is approximately 357.4 million shares. The model adds equity awards, converts, financing dilution, and other potential claims to reach the base denominator. |
| Adjusted net debt | 4700 USD millions | Base | Model Assumption | Filed capital structure plus financing bridge | Jul 30, 2026 | Medium | Must be refreshed for post-quarter cash, restricted cash, convertible debt, GPU financing, lease liabilities, prepayments, and subsequent issuance. |
| Mining and legacy EBITDA runoff | $250M / $120M / $60M / $20M / $0M 2026E–2030E | Base | Model Assumption | Operating history plus live workbook runoff assumption | Jul 30, 2026 | Low | Bitcoin price, difficulty, power cost, hardware efficiency, conversion timing, and future mining strategy can materially change this contribution. |
| Data-center and GPU capital cost | $15.0M per critical IT MW for hyperscaler infrastructure; $3.2M per IT MW for merchant infrastructure; GPU cost varies by cohort USD millions per IT MW plus GPU systems | Base | Model Assumption | Live workbook site inputs and public community model | Jul 30, 2026 | Low | The Microsoft $5.8B GPU-system anchor is disclosed. Later-site construction scope, escalation, contingency, equipment mix, customer funding, and overruns remain conditional assumptions. |
| Corporate and platform operating costs | $220M fixed SG&A plus 3.0% of revenue base case | Base | Model Assumption | Scenario Inputs and Consolidated DCF | Jul 30, 2026 | Low | The model applies explicit fixed and variable corporate costs outside project EBITDA. Future organizational scale, support costs, stock compensation, and overhead allocation remain uncertain. |
| Data-center maintenance capital | 0.2 USD millions per IT MW-year | Base | Model Assumption | Scenario Inputs and Consolidated DCF | Jul 30, 2026 | Low | This sustaining reserve is separate from modeled GPU replacement capital. Actual maintenance spending may differ by site, design, age, and customer responsibility. |
| Base cash-tax schedule | 3% / 3% / 5% / 10% / 15% 2026E-2030E | Base | Model Assumption | Scenario Inputs and Consolidated DCF | Jul 30, 2026 | Low | A normalized cash-tax schedule is used rather than a jurisdiction-by-jurisdiction tax model. NOL use, tax incentives, financing deductions, and geographic profit mix can materially change cash taxes. |
| Power and direct operating-cost treatment | Embedded in project EBITDA margins; not forecast as separate line items model treatment | All | Model Assumption | Site Inputs and Consolidated DCF | Jul 30, 2026 | Low | The model does not expose a separate power-price, networking, cooling, support, or downtime forecast on the website. Those costs are captured indirectly through the 82% merchant and 85% contracted project EBITDA-margin assumptions. |
| Working-capital treatment | Not separately modeled model limitation | All | Model Assumption | Workbook scope review | Jul 30, 2026 | Low | Customer prepayments and major financing flows are modeled, but ordinary receivables, payables, inventory, and other working-capital changes are not projected as a separate cash-flow schedule. |
| Value beyond the explicit forecast period | 82.5 % of 1.21 GW probability-weighted enterprise value | Base | Inference | Consolidated DCF reconciliation | Jul 30, 2026 | High | Approximately 82.5% of the 1.21 GW probability-weighted cohort enterprise value comes from value after the explicit 2026-2030 cash-flow period. The comparable figure for the 5.0 GW probability-weighted stage is approximately 82.1%. |
| Microsoft cohort yield after GPU capital recovery | 18.7% headline EBITDA yield; 7.2% after five-year GPU capital recovery yield on $8.8B modeled project cost | Base | Analyst Estimate | GPU Residual & NOI and Peer Cap Rates | Jul 30, 2026 | Medium | The adjustment reserves enough annual cash to recover the $5.8B GPU investment net of a 12.5% year-five residual. It is more comparable with long-duration infrastructure yields than headline project EBITDA, but the asset layers still differ. |
| Post-contract GPU economic-tail sensitivity | Base: one Year-6 tail at 50% price retention, 60% utilization and 10% broad revenue-share stress non-core sensitivity | Base | Model Assumption | GPU Capital Duration tab | Aug 12, 2026 | Low | The Base tail produces approximately $524M of net revenue, $367M of EBITDA and $196M of present value in the workbook, but it is excluded from the core DCF. It may be used only instead of an equivalent portion of Year-5 residual value. |
| NVIDIA-supported residual-value sensitivity | 0% IREN-specific support in core; 25% support ceiling; 18.75% expected residual at 50% eligibility % of eligible GPU capex | All | Model Assumption | NVIDIA financing-platform announcement and GPU Capital Duration tab | Aug 12, 2026 | Low | No public evidence establishes IREN eligibility, support recipient, covered value, legal enforceability, term, revenue-share basis, ownership or end-of-term rights. The 25% figure is a sensitivity ceiling, not a Base residual assumption. |
| NVIDIA revenue-share sensitivity | 0% for signed contracts absent disclosure; 0% / 5% / 10% / 15% for future supported cohorts % of applicable revenue | All | Model Assumption | GPU Capital Duration revenue-share offset sensitivity | Aug 12, 2026 | Low | A broad 10% share of the $1.94B annual Horizon revenue would surrender approximately $194M per year, more than the $145M annual capital-recovery benefit of moving residual value from 12.5% to 25%. Actual cost may be lower if sharing applies only to backstopped capacity. |
| Legacy GPU redeployment versus replacement gate | Replace when new-fleet risk-adjusted cash contribution per critical kW exceeds the legacy fleet after upgrade capex and downtime decision rule | All | Model Assumption | GPU Capital Duration power opportunity-cost gate | Aug 12, 2026 | Low | An older GPU can remain profitable yet be economically obsolete when a newer system produces more contribution from scarce powered capacity. NVIDIA performance claims are used only as a stress, not realized IREN economics. |
Change the assumptions
Each grid changes two assumptions at the same time. Find the row and column matching your view, then read the value where they meet. A wide range of outcomes means the estimate is highly sensitive to those inputs.
Read across or down to choose two assumptions. The value where the row and column meet is the model result under that combination.
The 5.6 GW row is the raw community configuration and exceeds the official up-to-5 GW framework; it is retained only as a reference case.
#terminal-value-sensitivity| Change Stage-end AI Cloud MW ↓ and Valuation treatment → | Base full | Base probability weighted | Normalized reserve floor |
|---|---|---|---|
| 480 | 27.08 | 27.08 | 10.75 |
| 1210 | 44.05 | 38.96 | 10.92 |
| 2310 | 69.91 | 52.32 | 18.79 |
| 3910 | 100.22 | 62.93 | 26.24 |
| 5000 | 116.58 | 66.57 | 29.44 |
| 5600 | 125.59 | 68.37 | 31.02 |
Read across or down to choose two assumptions. The value where the row and column meet is the model result under that combination.
Community outputs are stabilized forward earnings-power references, not discounted present-value conclusions. The live workbook reproduces the public code separately from the DCF.
#revenue-utilization-sensitivity| Change Stage-end AI Cloud MW ↓ and Model output → | DCF probability weighted | Community earnings-power reference |
|---|---|---|
| 480 | 27.08 | 94.63 |
| 1210 | 38.96 | 236.44 |
| 2310 | 52.32 | 340.26 |
| 3910 | 62.93 | 435.72 |
| 5000 | 66.57 | 442.54 |
| 5600 | 68.37 | 494.27 |
Read across or down to choose two assumptions. The value where the row and column meet is the model result under that combination.
The live workbook displays dashes where the bear cohort-cash result is non-positive; zero is used here only as the machine-readable placeholder for that displayed dash.
#delay-and-dilution-sensitivity| Change Official stage ↓ and Scenario → | Bear probability weighted | Base probability weighted | Bull probability weighted |
|---|---|---|---|
| 2026 / 480 MW | 0 | 27.08 | 63.39 |
| 2027 / 1.21 GW | 0 | 38.96 | 113.83 |
Read across or down to choose two assumptions. The value where the row and column meet is the model result under that combination.
Uses the same underlying 1.21 GW probability-weighted enterprise value. It isolates how the capital stack and denominator change common-share value.
#capital-structure-sensitivity| Change Diluted shares (millions) ↓ and Adjusted net debt (USD billions) → | 0 | 2 | 4 | 4.7 | 6 | 8 |
|---|---|---|---|---|---|---|
| 350 | 56.29 | 50.57 | 44.86 | 42.86 | 39.14 | 33.43 |
| 375 | 52.53 | 47.2 | 41.87 | 40 | 36.53 | 31.2 |
| 385 | 51.17 | 45.97 | 40.78 | 38.96 | 35.58 | 30.39 |
| 410 | 48.05 | 43.17 | 38.29 | 36.59 | 33.41 | 28.54 |
| 435 | 45.29 | 40.69 | 36.09 | 34.48 | 31.49 | 26.9 |
Read across or down to choose two assumptions. The value where the row and column meet is the model result under that combination.
Based on the disclosed $1.94B annual contract-revenue and $5.8B GPU-capex anchors plus $3.0B of modeled data-center capex. Residual value includes possible resale or continued use; it is not guaranteed cash proceeds.
#gpu-residual-sensitivity| Change Year-five GPU residual ↓ and Capital-adjusted metric → | Cash margin | Yield on modeled cost |
|---|---|---|
| 0 | 25.2 | 5.6 |
| 5 | 28.2 | 6.2 |
| 10 | 31.2 | 6.9 |
| 12.5 | 32.7 | 7.2 |
| 15 | 34.2 | 7.5 |
| 20 | 37.2 | 8.2 |
| 25 | 40.2 | 8.9 |
| 30 | 43.1 | 9.5 |
Read across or down to choose two assumptions. The value where the row and column meet is the model result under that combination.
Longer amortization lowers annual debt service but can increase total interest and leaves debt outstanding during post-contract tail years. These are financing sensitivities, not disclosed IREN terms outside the Microsoft cohort.
| Change Financing case ↓ and Capital-service metric → | Annual debt service | Total interest |
|---|---|---|
| 5 years / 7% | 237.6 | 188.1 |
| 6 years / 6.5% | 201.7 | 210.3 |
| 7 years / 6% | 175.3 | 227.1 |
Read across or down to choose two assumptions. The value where the row and column meet is the model result under that combination.
The six- and seven-year rows hold project EBITDA constant and are intentionally aggressive architecture sensitivities. The five-year / 12.5% row remains the core Base treatment.
| Change Primary-life / residual case ↓ and Capital-adjusted metric → | Annual cash flow | Cash per critical kW-month |
|---|---|---|
| 5 years / 12.5% | 634 | 264 |
| 5 years / 25% | 779 | 325 |
| 6 years / 20% | 876 | 365 |
| 7 years / 25% | 1,028 | 428 |
Interactive audit tool
Change the major valuation inputs below. Every state is written into the page URL, so a reader or AI system can link directly to the exact scenario rather than citing a generic model page.
Published buildout-stage anchor
Choose the published stage being valued. The WACC, terminal multiple, revenue, utilization, delay, funding and dilution controls remain independent; selecting the 5.0 GW stage does not automatically apply every Bull input from the full workbook.
Custom value per share
US$38.96
-US$0.00 versus the selected published scenario
Implied enterprise value
US$19,700m
Published anchor: US$19,700m
Implied equity value
US$15,000m
Adjusted net debt: US$4,700m
Funding mix
70% debt
30% equity; diluted shares are controlled separately
Calculation bridge
These relative factors show how the custom controls move the selected published scenario. A factor above 1.00 increases the modeled value component; below 1.00 reduces it.
Execution
1.00x
Unit economics
1.00x
Terminal value
1.00x
Required return
1.00x
Delay
1.00x
Changes the present-value factor applied across the configured valuation duration.
Adjusts the residual or stabilized-value portion of the selected scenario.
USD millions/MW-year
Applied with unit economics to the execution-weighted project portion of value.
Discounts the selected scenario for the additional time before value realization.
30% equity funding at the current setting. Share issuance is modeled separately.
Only phases included in “Near-term 1.21 GW stage” are shown and included in the calculation. Later phases do not affect this anchor.
Aggregate factor: 1x
Current 480 MW stage
Operating and contracted portfolio · 360 usable MW
Published probability: 100%
Increment to 1.21 GW
Childress and near-term expansion cohorts · 503.6 usable MW
Published probability: 82%
Calibration: 80% execution-linked value, 20% residual value, 5-year required-return duration, and US$15,629m of risk-weighted, ownership-adjusted construction capex included in the selected stage.
The interactive output is a conditional audit aid, not a prediction, recommendation, or claim of current market value. Use the source map, assumptions, limitations, and downloadable spreadsheet before relying on any result.
These are not footnotes to ignore. They identify missing evidence, simplified calculations, or events that could make the displayed value incomplete or too high.
Evidence
A source may confirm a reported fact or management plan without proving that the forecast will occur. Open a source card to see what it supports and what remains uncertain.
IREN · May 7, 2026 · United States
Relevant finding
Reports IREN's quarter-end financial position and operating disclosures, including its AI Cloud, data-center power portfolio, GPU deployment, Bitcoin-mining operations, and capital commitments.
Review notes
Primary financial source for the March 2026 quarter. Use the filed statements and notes to reconcile cash, debt, commitments, segment results, share count, and risk disclosures before changing the illustrative model.
Relevant pages: Overview; unaudited financial statements and notes; commitments and contingencies; management discussion; risk factors.
IREN · May 7, 2026 · Global
Relevant finding
Describes a five-year AI Cloud contract and strategic partnership with NVIDIA, the staged AI-capacity buildout, secured-power pipeline, current financial results, and management's financing assumptions.
Review notes
Company-furnished material. Reconcile contracted ARR, customer terms, NVIDIA investment rights, delivery schedules, and funding requirements to executed agreements and financial statements. Forward-looking targets are guidance, not realized results.
Relevant pages: Pages 1-5 for the NVIDIA relationship and buildout plan; pages 8-12 for financial statements and reconciliations.
IREN · Jul 25, 2026 · Global
Relevant finding
Frames IREN as part of an NVIDIA-aligned independent cloud layer and discusses the strategic significance of NVIDIA's investment rights, GPU relationship, and use of IREN capacity.
Review notes
Secondary research reviewed from the project's private IREN research archive. Use it for strategic interpretation and question formation; verify contractual, financing, capacity, and valuation claims against primary filings and agreements.
Relevant pages: Saved-PDF sections discussing NVIDIA's independent-cloud strategy, IREN's NVIDIA relationship, and the neocloud competitive landscape; pagination varies by export.
IREN · Jul 20, 2026 · Online
Relevant finding
Provides a site-level buildout and valuation cross-check that helped identify capacity definitions, phase timing, and assumption differences requiring reconciliation.
Review notes
Secondary community model reviewed from the project's private IREN research archive. It is useful for site-level reconciliation and identifying missing questions, but every material capacity, timing, cost, financing, and valuation input must be checked against primary sources.
Relevant pages: Interactive site-level model and supporting methodology pages; accessed July 2026.
IREN · Jul 20, 2026 · Global
Relevant finding
Supports capping the official strategic model stage at 5.0 GW while keeping the community model's raw 5.6 GW configuration as a separate reference case.
Review notes
Primary company disclosure used for the up-to-5 GW strategic AI Cloud framework, NVIDIA relationship, and official capacity ceiling. The framework is not treated as a fully contracted five-gigawatt backlog or a fixed site-level delivery schedule.
Relevant pages: Announcement sections covering contracted AI Cloud revenue, strategic partners, the 480 MW and approximately 1.2 GW scale-up, and the longer-term power-backed expansion framework.
IREN · Nov 3, 2025 · United States
Relevant finding
Provides the strongest disclosed unit-economics and financing anchor for the contracted hyperscaler cohort in the site-level DCF.
Review notes
Primary contract-economics anchor for approximately $9.7 billion of five-year contract value, the customer prepayment, GPU-system capex, Childress deployment, and contracted project-margin framing. Detailed remedies, final delivery, operating costs, renewal, and residual value remain material.
Relevant pages: Contract announcement and presentation sections covering the five-year Microsoft agreement, approximately $9.7 billion of contract value, 20% prepayment, approximately $5.8 billion of GPU and ancillary-equipment capex, and Horizon 1-4 delivery.
IREN · Jun 1, 2026 · United States
Relevant finding
Supports separately modeling GPU financing and customer prepayments rather than treating gross contracted revenue as unlevered equity value.
Review notes
Primary financing source for the $3.65 billion GPU facility and the disclosed Microsoft GPU-capex funding bridge. Financing reduces immediate equity needs but creates senior claims, interest, amortization, and refinancing risk.
Relevant pages: Financing highlights, facility components, blended cost, customer prepayment, collateral, and the disclosed percentage of Microsoft GPU capex funded.
IREN · May 7, 2026 · United States
Relevant finding
Supports a distinct contracted NVIDIA cohort and a five-year primary monetization period while leaving post-contract and residual economics unverified.
Review notes
Primary company disclosure for the approximately $3.4 billion, five-year managed GPU-services agreement with NVIDIA across approximately 60 MW at Childress. The public materials do not disclose a separate NVIDIA revenue share, residual-value guarantee, final ownership allocation, or end-of-term hardware rights.
Relevant pages: Business-update sections covering the NVIDIA contract value, five-year term, deployment capacity, staged acceptance, and related Blackwell systems deployment.
IREN · May 26, 2026 · United States
Relevant finding
Provides a separate GPU-capex anchor for the NVIDIA cohort without establishing a different accounting life or guaranteed residual value.
Review notes
Primary purchase-cost anchor for approximately $1.6 billion of air-cooled Blackwell systems associated with the NVIDIA managed-services deployment. Final financing, system mix, installation cost, ownership and residual arrangements remain incomplete.
Relevant pages: Announcement sections covering the Dell systems purchase, payment timing, deployment capacity and relationship to the NVIDIA contract.
NVIDIA · Aug 10, 2026 · Global
Relevant finding
Supports a non-core financing and residual-support sensitivity ceiling, not an automatic increase to IREN's Base residual value or useful life.
Review notes
NVIDIA described financing platforms with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR intended to mobilize more than $500 billion over time, with possible case-by-case support of up to 25%. Final legal terms, IREN eligibility, support recipients, revenue-share economics, asset ownership and timing remain undisclosed.
Relevant pages: Announcement and reporting sections covering platform scale, independent underwriting, possible NVIDIA support, residual-value framing and the absence of final transaction allocations.
NVIDIA · Jan 5, 2026 · Global
Relevant finding
Supports testing whether older GPUs should be redeployed or replaced based on risk-adjusted cash contribution per constrained critical kilowatt.
Review notes
Official NVIDIA platform claims used as a technology-cycle and contribution-per-kW stress. They do not establish realized IREN rental rates, utilization, deployment cost, power density or customer demand.
Relevant pages: Platform-performance sections comparing Rubin with Blackwell for inference token cost, GPU requirements and large-model training.
IREN · Jul 30, 2026 · Global
Relevant finding
Allows exact reproduction of the community earnings-power outputs while preserving a clear distinction between public code and current intrinsic-value analysis.
Review notes
Public user-generated model used as a reproducible site-level operating engine and cross-check. It is not IREN guidance, audited information, or a substitute for the Ephesus cohort-cash DCF.
Relevant pages: Repository source code, site constants, GPU defaults, calculation engine, and scenario outputs reviewed through July 30, 2026.
IREN · Jul 30, 2026 · Global
Relevant finding
Provides the site-level configuration underlying the 2.31 GW, 3.91 GW, and raw 5.6 GW community stages; official evidence and execution probabilities are applied separately.
Review notes
User-generated site and cohort schedule used to operationalize official aggregate capacity milestones and to expose later-stage optionality. Exact annual timing, GPU configurations, capex, and later sites are not all company commitments.
Relevant pages: Site objects for Canadian campuses, Horizon, Childress, Sweetwater, Oklahoma, Nostrum, Kiowa, and the annual community buildout configuration.
IREN · Jul 30, 2026 · Global
Relevant finding
Provides transparent, editable inputs for the merchant and contracted GPU operating engine while keeping confidence low until company-specific realized economics are disclosed.
Review notes
User-generated GPU hourly rates, hardware prices, utilization, financing, and residual assumptions used for reproducibility and sensitivity analysis. These are not company guidance or reported realized economics.
Relevant pages: Default profiles for Vera Rubin, GB300, B300, B200, MI350X, pricing, utilization, purchase cost, financing, and residual assumptions.
IREN · Aug 12, 2026 · Global
Relevant finding
Makes the calculation chain reproducible, separates GPU accounting life from financing and post-contract economic life, and identifies which inputs are facts, guidance, analyst assumptions, sensitivities or unresolved limitations.
Review notes
Canonical public workbook for formulas, site and GPU inputs, explicit cohort cash flow, financing and dilution, peer project-yield checks, assumption evidence, model checks, and revision history. It is a research model, not independent evidence of future performance.
Relevant pages: Scenario Inputs, Site Inputs, Consolidated DCF, GPU Residual & NOI, GPU Capital Duration, Financing & Dilution, Peer Cap Rates, Sources & Notes, Model Checks, Revisions, and Assumption Evidence tabs.
WULF · Aug 14, 2025 · United States
Relevant finding
Provides a long-duration infrastructure yield-on-cost reference while illustrating that the tenant or compute operator, rather than the landlord, bears most GPU obsolescence risk.
Review notes
Public project-economics reference for more than 200 MW of critical IT load, approximately $3.7 billion of initial contracted revenue, an approximately 85% expected site NOI margin, and an $8-$10 million project-cost range per critical MW. It is an infrastructure-hosting contract, not a direct GPU-compute comparison.
Relevant pages: Transaction highlights covering contracted revenue, initial lease term, expected NOI margin, project cost, Google support, and delivery schedule.
CIFR · Feb 4, 2026 · United States
Relevant finding
Provides a long-duration landlord cash-flow reference that is not directly comparable with IREN's shorter-lived GPU-compute layer.
Review notes
Public infrastructure-landlord reference for 216 MW of critical IT load, a 15-year AWS lease, approximately $5.5 billion of contracted lease payments, a $9.5 million maximum development cost per critical MW, and tenant reimbursement of qualifying cost overruns.
Relevant pages: Transaction overview, financing and credit support, lease economics, construction-cost cap, rent commencement schedule, and operating-expense pass-through sections.
Revision history
Material updates are recorded so readers can see which assumption or conclusion changed, why it changed, and how the result was affected.
Aug 3, 2026 · IREN
Replaced the IREN placeholder source map with reviewed primary filings, SEC-filed company materials, and secondary research from the project's private IREN research archive.
Previous
Three illustrative placeholder records
Revised
Four reviewed records with exact public originals and source-quality notes
Reason
The public source library should distinguish filed facts and company guidance from independent interpretation and community modeling.
Source
Private IREN research-archive review and public-original verification
Estimated effect
No direct change to the illustrative valuation outputs; materially improves auditability and identifies evidence that should be incorporated in a future model rebuild.
Aug 2, 2026 · IREN
Moved illustrative AI Phase 2 energization six months later.
Previous
2028-H1
Revised
2028-H2
Reason
Demonstrates how schedule uncertainty should flow through discounting, revenue ramp, and financing needs.
Source
Sample methodology change; no company disclosure implied
Estimated effect
Reduced illustrative base-case value per share by approximately 6%.
Aug 2, 2026 · IREN
Raised illustrative AI construction and equipment capex.
Previous
$6.4 billion across sample AI phases
Revised
$7.4 billion across sample AI phases
Reason
Shows the effect of higher equipment, cooling, electrical, and construction costs.
Source
Illustrative stress test
Estimated effect
Reduced equity value and increased modeled financing dilution.
Aug 2, 2026 · IREN
Reduced the base terminal EBITDA multiple.
Previous
11.0x
Revised
10.0x
Reason
Demonstrates multiple-compression risk when required returns rise or durability is less certain.
Source
Illustrative terminal-value sensitivity
Estimated effect
Reduced base-case value and the share of value attributable to terminal cash flow.
Jul 24, 2026
Added separate source records for each peer rather than relying on presentation summaries.
Previous
One aggregated peer table
Revised
Company-level filing and presentation mapping
Reason
Peer comparisons require consistent definitions of gross MW, usable IT MW, ownership, capex, and contracted capacity.
Source
Sample source-library expansion
Estimated effect
Reduced false precision in the comparative ranking.
Aug 4, 2026 · IREN
Replaced the website's IREN demonstration scaffold with the current Google Drive site-level cohort-cash DCF, official capacity milestones, probability-weighted buildout stages, source-backed contract and financing anchors, community-model reconciliation, and assumption-level evidence mapping.
Previous
0.9.0-sample illustrative placeholder model
Revised
1.0.0-preliminary current source-backed model mapping
Reason
The live workbook contains a current stage-level DCF, site and GPU inputs, official-pipeline classifications, capital-structure assumptions, source notes, 27 passing model checks, a revision log, and an assumption-evidence tab.
Source
IREN Q3 FY26 buildout update and live IREN Google Sheet
Estimated effect
Changes the public model status, stage estimates, assumptions, buildout schedule, sensitivities, source coverage, limitations, interactive controls, and article interpretation. Later-stage values remain conditional and probability-weighted.
Aug 7, 2026 · IREN
Made the workbook's hidden model dependencies visible on the public page, including unit capex, corporate costs, maintenance capital, taxes, power-cost treatment, working-capital omission, terminal-value dependence, mining runoff, dilution sensitivity, GPU capital recovery and peer project-yield checks.
Previous
The website exposed selected assumptions and linked the full workbook, but readers had to search separate spreadsheet tabs to understand several material operating and capital assumptions.
Revised
A plain-language IREN audit brief, expanded assumption table, capital-structure and GPU-residual sensitivities, clearer provenance, and stage-scoped interactive calculations.
Reason
External reader feedback correctly identified that the public page did not make the full calculation chain and model limitations easy to find, even though most of the detail existed in the live workbook.
Source
IREN live public workbook and assumption-evidence map
Estimated effect
Does not change the published $27.08, $38.96 or $66.57 stage estimates. It changes interpretation and custom-scenario mechanics by preventing phases outside the selected buildout stage from influencing that stage's interactive result.
Aug 12, 2026 · IREN
Added a public GPU-capital-duration and post-contract economics framework reflecting the revised live workbook, including cohort financing, non-core tail cash flows, NVIDIA support and revenue-share sensitivities, and a contribution-per-kW replacement gate.
Previous
The public model showed five-year GPU capital recovery and residual sensitivity but did not separately expose financing duration, explicit Years 6-8 monetization, NVIDIA support, revenue share or power opportunity cost.
Revised
Five-year core life and published stage values remain unchanged; the new sensitivities show how financing relief, residual support, revenue sharing and post-contract monetization interact without double counting.
Reason
The live model and private research ledger were revised after NVIDIA's financing-platform announcement and review of IREN's Microsoft and NVIDIA cohort structures.
Source
IREN live workbook v1.1 GPU Capital Duration update and approved evidence map
Estimated effect
No mechanical change to the $27.08, $38.96 or $66.57 probability-weighted Base stage estimates. The effect is improved interpretation, sensitivity analysis and evidence discipline.
Further reading
IREN
A current preliminary cohort-cash DCF mapped to the live Google Drive workbook. The analysis distinguishes IREN's official capacity stages, cohort financing, five-year primary GPU life, non-core post-contract tail economics, residual support, revenue sharing, and public community-model reference outputs.
IREN · NUAI · WULF · CIFR · APLD
A reader's guide to phase-level DCFs, execution probabilities, timing, financing, dilution, and terminal value.
IREN · NUAI
A precise explanation of why a full-build output can be valid without implying that the company is worth that amount today.
Report a formula error, unsupported assumption, missing source, or unclear explanation. Material corrections are added to the public revision history rather than silently overwritten.