IREN · Forward model · Bull case

The Bull case, 20 quarters out

Each segment is projected from its own operating driver, rolled up into consolidated cash flow, and discounted back to a fair value per share. The assumptions below are editable — change them and every number on this page moves with them.

READ THIS FIRST. The basis quarter of this model is superseded by results published on the same day it was built. IREN reported fiscal 2026 Q4 and the full year to 2026-06-30 after the US close on 2026-08-27; nothing from that report is in here. The 2026 Q1 basis is unrepresentative on the company's own telling: it carries $140.4m of transition impairment with roughly $520m more pre-announced to follow, the decision to cease all Childress Bitcoin mining was taken inside it, and Horizon 1 - the first revenue-bearing Microsoft deployment - was accepted on 2026-08-13, six weeks after the quarter closed, so not one dollar of Horizon revenue appears in any actual here. The largest revenue line in the basis quarter, Bitcoin mining at 76.8% of revenue, is one management is deliberately switching off. The model is flagged unreliable for those reasons: its assumptions are worth inspecting and editing, its fair value is not a ranking signal. SEGMENTS. Bitcoin mining and AI Cloud Services are IREN's only two reportable segments; both revenue and cost of revenue are disclosed per segment every quarter, there are no intersegment eliminations, and the two lines sum to consolidated revenue in all twelve quarters carried here - $111.160m + $33.635m = $144.795m at the basis quarter, matching the 10-Q and the release. No split was invented. Four points are derived by subtraction from disclosed annual and nine-month figures and are flagged estimated (2023 Q2 and 2024 Q2 on both lines); the AI Cloud zeros for 2023 Q2 to 2023 Q4 are flagged estimated because the segment did not exist, not because a zero was reported. THE IMPAIRMENT IS NOT A RUN RATE. Mining's 68.21% margin is ($111.160m - $35.336m) / $111.160m, where cost of revenue is the disclosed electricity $30.645m, employee benefits $3.548m and other direct expenses $1.143m. The $140.411m of impairment in the quarter and the roughly $520m estimated to follow sit below that line and are in neither revenue nor margin, exactly as IREN's own adjusted EBITDA excludes them. THE MINING WIND-DOWN. No dated hashrate glide path has ever been published; the disclosure is 'over time'. Installed hashrate has fallen 50 to 46 to 38 EH/s (-12.8% a quarter) and average operating hashrate 45.3 to 43.0 to 36.0 (-10.9% a quarter). Rather than extrapolate a growing business, the model holds installed capacity at 38 EH/s, adds nothing, and glides the energised share from the disclosed 94.7% (36.0 of 38) toward zero at 20% of the gap a quarter - roughly -20% a quarter, retiring 79% of the March run rate by 2027 Q4, the window in which the disclosed retrofit ladder takes AI Cloud capacity from 480MW to 1,210MW gross. That is a chosen path with a stated rationale, not a disclosure. Revenue per exahash is held flat at $3.088m a quarter ($111.160m over 36.0 EH/s): this model does not forecast Bitcoin. SEASONALITY. Recomputed here from the twelve quarters of per-vertical history by ratio to a CENTRED four-quarter moving average, which leaves eight interior ratios and two independent observations per calendar quarter. Bitcoin mining normalises to 1.041 / 0.975 / 0.968 / 1.016 for calendar Q1-Q4, a signal range of 0.073, against a window-to-window spread of 0.134 / 0.114 / 0.688 / 0.111 - the Q3 pair is 0.630 in 2024 against 1.318 in 2025, a spread 9.4x the entire signal. AI Cloud Services normalises to 0.819 / 1.518 / 1.175 / 0.488, a signal range of 1.029, against spreads of 0.293 / 0.332 / 0.634 / 0.750, and its Q4 factor rests on a 2023 Q4 observation of exactly zero because the segment did not exist. In both cases the spread meets or exceeds the signal, so neither vertical carries a seasonality array. What moves these lines is the Bitcoin price, network difficulty, IREN's own hashrate decisions and the date a data centre passes customer acceptance - none of them a calendar. CAPITAL. Capex on the AI line is anchored on disclosure: $5.8bn of Dell hardware for Horizon 1-4 at about 300MW gross is $19.3m per gross MW, netted to about $10.6m by the disclosed ~45% customer prepayment on contracts signed since 2026-06-01, against $2.08m of quarterly revenue per gross MW implied by $8.33m of targeted ARR per gross MW. That is growth capex of about 5.1x each increment in quarterly revenue plus replacement of about 38.7% of revenue on an assumed six-year GPU life; 440% gliding to 50% is the single-parameter fit. It spends $3.054bn across the four quarters to 2027 Q1, so a corporate programme carries the remaining $8.845bn and those four quarters total the disclosed $11,899,054,000 of commitments payable within twelve months of 2026-03-31. Corporate overhead of 10% of revenue is the forward version of a cash SG&A that was $45.4m in the basis quarter ($81.750m of SG&A less $31.5m of stock-based compensation less $4.826m of other operating income), or 31.4% of a $144.8m revenue base; the check is that $75.824m + $29.038m - $45.4m = $59.4m against the reported $59.5m of adjusted EBITDA. Ten percent is what the same corporate function should cost against a revenue base an order of magnitude larger, and it understates the near-term dollar burn. NET CASH IS THE BASIS BALANCE SHEET, DELIBERATELY. -$1,748.814m is cash $2,213.274m less convertible notes payable $3,687.832m less finance leases $274.256m, all at 2026-03-31. It excludes the $3.0bn of 1.00% notes closed 2026-05-14, any draw under the $3.6bn Hardware 3 GPU facility (the drawn balance is not published) and the preliminary, unaudited $7.6bn of June cash of which $1.7bn is restricted. A single honest net-debt number does not exist in public disclosure, so the model uses one date consistently rather than mixing them. Shares are 357,378,674 at 2026-04-30 plus the approximately 12.6m issued for Mirantis on 2026-08-04; about 86m convertible underlying shares, 41m more from the May notes, 30m NVIDIA rights at $70 and 14.8m RSUs are charged nowhere in the fair value. WHAT THE MODEL PRINTS. Base case: 2026 Q2 revenue $173.0m, which with the disclosed nine months of $569.782m makes fiscal 2026 $742.8m against a $722.9m consensus; CY2027 revenue $4.17bn, an exit run rate of $6.08bn annualised against a 1,210MW ladder worth about $10.1bn of ARR, so the model is deliberately below the ladder; $18.67bn across the final four quarters; free cash flow negative in every quarter until 2030 Q3; fair value $24.59 against a $39.58 close. Bear -$17.41, bull $93.64, Roberts $85.00.

Where each case comes from

From cash flow to fair value

Capex outside the verticals

Each vertical's capex is a share of its own revenue, which is the right shape for capacity that scales with sales. These are not that: fixed-size programmes with their own schedule that spend whether or not any segment grows. They are added on top, and they are why free cash flow dips in the years below.

The projected path

Every row is projected. A year-over-year change is shown only where the quarter it compares against exists — an em dash means there is no comparable quarter, not a flat year.

Model revisions

Assumptions are marked to reality as each quarter prints. Every change is appended here, with the fair value the model produced at the time, so the model's own history stays visible.