| Item | Read |
|---|---|
| Score | 6.5 / 10 (v2 7.0 · v1 7.5). Down half a point. Demand, pricing and delivery all improved; the capital intensity re-rated up ~6x (FY27 capex $25–30B vs FY26 $4.3B) and the Red Team surfaced $13.6B of capex commitments payable within 12 months that no lane had priced. |
| Thesis status | INTACT — ON WATCH. No breaker tripped: no Microsoft termination, no discount raise, no founder exit, ARR target raised not cut, Horizon 1 accepted. Watch items: Horizon 2–4 timing inside the grace window, and how the FY27 residual ($3–8B) gets funded. |
| Tier | Conviction-spec, NOT Tier-1. Moat 7.0/10 fails the ≥9 gate (v2 7.5, Qual 6.5, Tracker "7.5→8"; PM lands at 7.0 — see Management & Moat). |
| Macro stamp (Stage 1) | 🔴 for initiating/adding at spot · 🟡 hold-don't-add for an existing position with the exit plan armed. Fed hiked 25bp on 9/16 (first since 2023), 10Y 5.18% (highest since 2007), net liquidity contracting, CCC OAS +81bp/20d, MOVE 104. IREN is GAAP-unprofitable, FCF −$2.2B, needs ~$8B+ of external financing for a $25–30B capex year, with $6.75B of converts out of the money — the exact archetype the late-cycle rule targets. |
| Sizing rule | Framework: ≤10% per position is Tier-1 only (moat ≥9). Conviction-spec = ≤5% of portfolio at full ladder (PM house rule: half the Tier-1 cap; the framework sets only the Tier-1 number). New money only at ladder zones or via cash-secured puts. Macro 🔴 also means portfolio cash at the framework's 25–30% "dicey" level. |
| Next check | Q1 FY27 print 2026-11-05 UNVERIFIED — yfinance/stockanalysis calendar; confirm with company IR: Horizon 2 acceptance, AI Cloud revenue ≥$150M, ATM/convert activity, share count vs 394.06M, first DC-level financing. |
IREN owns ~5.3GW of grid-connected power in places a hyperscaler cannot replicate before ~2030, is converting it from Bitcoin mining into GPU cloud at $11–25M per MW-IT per year, and has $4.0B of contracted ARR against $1.0B operating ARR and $707M of trailing revenue. If Horizons 2–4 land inside Microsoft's grace window (Dec-2026 quarter target, grace to early Q2 CY27) and the Nov-5 print shows the ARR billing through the P&L, the stock re-rates from a "show-me" 6.2x forward sales toward the neocloud band, and an 83M-share short book with two days of cover is fuel. If delivery slips, or FY27's $25–30B capex has to be equity-funded below management's ~$50 issuance floor, the modal outcome is $22–29, not $45. At $44 the market is pricing the base case and paying nothing for the bull: probability-weighted value ≈ $46, i.e., no asymmetry at spot; asymmetry starts at $34–37 and below. Own it through the wheel, at the ladder, sized for a 2027 funding round you may have to sit through.
What it is. A Sydney-founded, Nasdaq-listed owner-operator of power-rich data-center sites (Texas, British Columbia, Oklahoma, Spain, Australia) that is converting from Bitcoin mining to renting NVIDIA GPU clusters ("AI Cloud") to Microsoft, NVIDIA and AI labs on 3–5-year contracts. It owns the land, the grid interconnects, the buildings, the cooling, the GPUs, and (since Mirantis) a software layer. It does not lease shells to others — it is the tenant of its own buildings.
The 8-year-old version. IREN owns big buildings with really cheap, really big electricity hookups in places where nobody else can get them for years. It used to fill them with bitcoin machines. Now it fills them with NVIDIA AI computers and rents those computers by the year to Microsoft, NVIDIA and AI companies, who pay a chunk of the rent up front. The rent is $11–25 million per megawatt per year; the computers cost about $29 million per megawatt, and lenders front ~90% of that. IREN makes money if the rent keeps coming after the loans are paid off and before the computers are obsolete.
Revenue today. FY26: BTC mining $578.2M (82%) / AI Cloud $128.8M (18%). Q4 FY26: mining $66.7M (49%) / AI Cloud $70.5M (51%). Mining 23.2 EH/s (~380MW) at Jun 30; "effectively discontinued by end of December 2026"; zero BTC held.
| Date | Milestone |
|---|---|
| 2018-11 | Founded as Iris Energy by Daniel & Will Roberts (Sydney); Canal Flats BC first site 2019 |
| 2021-11 | Nasdaq IPO; 2022 −90% drawdown; 2023-08 BC court win vs NYDIG SPV lenders (non-recourse) — "IPO curse" survived |
| 2024 | Childress TX build-out; renamed IREN; ~3MW GPU cloud pilot |
| 2025-09-11 | Both co-CEOs sell 1.0M sh each @ $33.13 ($33.1M each) |
| 2025-10-14 | $1.0B 0% converts due 2031 @ $85.63 |
| 2025-11-03 | Microsoft $9.7B/5-yr GB300 contract (200MW IT, Horizons 1–4, 20% prepay); Dell $5.8B GPU order |
| 2025-12-02/08 | 39,699,102-sh registered direct @ $41.12 ($1,631.5M net) funds $1,632M cash repurchase of 2029/2030 converts; $1.15B 2032 + $1.15B 2033-Jun converts @ $51.40 RT-FIX: Qual/Macro dated this May 2026 |
| 2026-01 | Kiowa OK 1.6GW site (power from 2028) |
| 2026-02-27 | MSCI USA inclusion |
| 2026-03-04 | 150k-GPU fleet target; $6.0B ATM program filed |
| 2026-05-01 | Sweetwater 1 (1,400MW) substation energized — on the April promise |
| 2026-05-07 | NVIDIA partnership (up to 5GW DSX); $3.4B/5-yr NVIDIA cloud contract; NVIDIA 30M-sh purchase right @ $70; Mirantis and Nostrum announced |
| 2026-05-14 | $3.0B 1% converts due Dec-2033 @ $73.07 (capped call $110.30) |
| 2026-06 | $3.6B IG GPU financing @ ~6.0% (Microsoft deployment); Nostrum closed 6/15; Russell 1000 6/29 |
| 2026-07-01 | 9,099,328 RSUs each co-CEO; 7/08 Chair letter |
| 2026-07-20 | $2.8B new TCV; ARR target >$4B |
| 2026-08-03 | Mirantis closed (~580 staff) |
| 2026-08-13 | Horizon 1 delivered/accepted by Microsoft; NVIDIA Exemplar Cloud (GB300 NVL72) |
| 2026-08-25 | $2.4B Blue Owl/PIMCO non-IG GPU financing @ 9.0% |
| 2026-08-27 | FY26 results; FY27 capex guide $25–30B |
| 2026-09-08 | 2GW Sweetwater conditionally Base Load in ERCOT Batch Zero |
Childress TX 750MW (AI operating ~40MW IT pre-Horizon; Horizons 1–4 = 200MW IT) · Sweetwater 1 1,400MW (energized 5/1/26; no tenant/financing announced) · Sweetwater 2 600MW (late 2027) · Mackenzie 80 / Canal Flats 30 / Prince George 50 (BC; liquid-cooled retrofits 2027) · Kiowa OK 1,600MW (2028) · Bundey AU 800MW · Badajoz ES 300MW. $4B ARR consumes "<10%" of it.
Delivery status: H1 live; H2 commissioning; H3–4 late-stage construction for Dec-2026 quarter (grace to early Q2 CY27); 2026 ≈ 300MW IT / 480MW gross; 2027 +500MW IT to 800MW IT / 1.2GW gross; Sweetwater 1 first 300MW gross Q4 2027.
Bottom-up from IREN's own unit economics — no third-party GW forecast was sourced this pass, treat as sensitivity.
| Horizon | Company-path revenue (Qual, on-time build) | PM base path (delivery lag + funding friction) | Assumptions |
|---|---|---|---|
| FY26A | $707M | — | 82% mining |
| FY27 (1-yr) | $2.5–3.0B (consensus $2.8B) | $2.2–2.6B | $1B→$4B ARR ramp weighted through the year; H2–4 in Dec-qtr or grace window; mining ~$120M |
| FY29 (3-yr) | $9–12B | $6–9B | 800MW IT end-2027 + NVIDIA 60MW + ~300MW 2028 adds ≈ 1.1–1.2GW IT × $12M; PM haircuts 2027 uncontracted 500MW to 60–70% filled |
| FY31 (5-yr) | ~$22B ARR (2GW IT × $11M) | $12–16B | Requires Sweetwater 1 fully built and financed; renewals at −20–40% $/MW |
| FY36 (10-yr) | ~$27B (3GW IT × $9M) | $18–25B | Pipeline growth (Batch Zero, Kiowa) offsets price decay; GPU refresh capex recurring |
Is it spinning? Yes, nodes 1–5 verifiably turned in FY26 (energization on time; H1 accepted; $6.4B GPU facilities; $1B operating ARR). Node 6 is half-turned (NVIDIA status yes; no DC-level refinancing closed). Node 7 still runs on an equity crank: $3.06B net ATM + $1.63B registered direct + $6.3B converts in FY26. The flywheel is real but not yet self-funding; it becomes self-funding when a shell refinancing closes and ex-prepay OCF/CapEx exceeds ~0.3. Framework note: no flywheel = no long-term hold; this one qualifies, with the crank flagged.
| # | Question | Current answer | What would change it |
|---|---|---|---|
| 1 | Do Horizons 2–4 (150MW IT, ~$1.45B/yr) land inside the Dec-2026 quarter / grace window? | H2 "commissioning," H3–4 "late-stage construction" (8/27). H1 slipped ~6 weeks on NVIDIA equipment. Grace extends to early Q2 CY27. | An 8-K on H2 acceptance before 11/5 (bullish); "commissioning underway" language on 11/5 with no acceptance (bearish; each quarter of slip ≈ $360M revenue pushed right). |
| 2 | How is the FY27 residual funded, and at what price? | Sources ~$12.7B (cash $5.9B + restricted $1.7B + undrawn IG ~$2.7B + Blue Owl $2.4B) vs $13.6B 12-month commitments and $25–30B guided; "$8B targeted" + $3–8B from "DC financing, OCF, corporate debt and equity." ATM $3.51B left; management historically sells ≥$50. | A closed DC-level refi ≥$1B at ≤7% (bullish, kills the equity path); a corporate raise >5% of float below $45 (bearish, confirms N2). |
| 3 | Does contracted ARR convert to GAAP revenue at the claimed rate? | Contracted $4.0B vs operating $1.0B vs Q4 AI revenue run-rate $282M — 14x and 3.5x gaps. Consensus Q1 FY27 $221.8M implies ~$160M AI Cloud. | Q1 FY27 AI Cloud ≥$190M with H2–4 "on track" = converting; <$150M = ramp assumptions ran ahead again. |
| 4 | Is the 2027 uncontracted ~500MW IT priced at scarcity or commodity? | "Late-stage discussions over a significant portion" at $20–25M/MW. B200 expected to commoditize by H1-27; H100 rates already halved. | First 2027 contract announced ≥$18M/MW (validates); a hyperscaler 2027 capex cut or deal <$15M/MW (breaks the $10B ARR hope). |
| 5 | Can IREN run a cloud, not just build one? | Microsoft accepted H1; NVIDIA Exemplar on GB300; but SemiAnalysis ClusterMAX 3.0 (9/23) rates IREN "not recommended" bottom tier vs CRWV/NBIS Platinum; Bernstein "behind on enterprise." Mirantis closed 8/3. | ClusterMAX upgrade or H2–4 acceptance without SLA credits (bullish); disclosed service credits / churn at renewal (bearish → landlord multiple). |
| 6 | Concentration: what happens if Microsoft renegotiates? | Microsoft $1.94B/yr = 48% of $4B ARR (66% with NVIDIA); accounted as ASC 842 operating lease ($11.4B contracted lease value); $1.8B deferred lease revenue must be earned by delivering on spec; termination/credit rights exist (terms undisclosed). | A second hyperscaler-grade anchor at Sweetwater ≥300MW (dilutes concentration); any credit/termination disclosure (thesis breaker). |
| 7 | GPU residual value vs contract term vs debt tenor | Contracts ~4-yr wtd (3-yr for the >$20M deals); GPUs depreciated 5 yrs; Blue Owl amortizes ~30 months; VR200 already on order. | Renewals (Together/Fireworks first) printing ≥70% of original $/MW (accept); −40–50% (equity = scrap value). |
| 8 | Will Sweetwater survive SB6/Batch Zero as base load, and get a tenant? | Conditional base-load status 9/8; energized 5/1/26 = post-cutoff → curtailable; no tenant, no financing announced; first 300MW Q4 2027. | Anchor tenant + project financing (validates 5GW narrative); reclassification or curtailment terms a hyperscaler won't accept. |
| 9 | Does SG&A/SBC build ahead of revenue break the margin story? | Adj. EBITDA 41% → 14% in one quarter; SG&A +$40–50M guided for Q1 FY27; headcount ~3x again; SBC $205M (29% of rev) + ~$208M/yr CEO grants. | Adj. EBITDA margin back >35% by Q3 FY27 on ARR ramp (operating leverage proven); stuck <25% (structural). |
Price: $41.14 → $44.125 (+7.2%). Path: −30% to $28.93 (7/29) then +70% to $49.37 (9/23). BTC $64.1K → $84.0K. All three v2 entry zones printed; no v2 trim band reached.
Structural change v3 makes: IREN is now a US domestic filer (10-K/8-K/Form 4 since Aug 2026; the 6-K search v2 relied on returns nothing for Jul–Sep 2026). v2's "~2 quarters of runway" framing was static and wrong in direction: the company raised ~$19B in 12 months. Runway was never the binding constraint; capital-markets access and price are. v2's per-share targets assumed "~400M fully diluted by 2027" — basic count was already 394M in Aug 2026. Both fixed here.
| Date | Event | Why it matters |
|---|---|---|
| 2026-08-27 | FY26 results + 10-K. Rev $707.0M (+41%); AI Cloud $128.8M (~8x); NI −$702.6M incl. $638.8M impairments; adj. EBITDA $245.7M. Q4: rev $137.2M vs $157.1M consensus (MISS), AI $70.5M > mining $66.7M, adj. EBITDA $19.2M (14%), NI −$684.0M incl. $450.4M impairment | Stock −12.5% next day to $35.45. CFO Lewis on the call, verbatim: "we're guiding CapEx of approximately $25 billion to $30 billion" for FY27; "$14B" secured; "roughly an additional $8B" of GPU financing/prepay targeted. |
| 2026-08-27 (10-K Note 29) | Capital commitments $13,810.0M, of which $13,611.0M payable within 12 months (vs $368.8M a year earlier) | The funding gap in one number. Missed by all five lanes; found by Red Team. RT-FIX |
| 2026-08-13 | Horizon 1 (50MW IT, GB300 NVL72) delivered and accepted by Microsoft; NVIDIA Exemplar Cloud status | First hyperscaler tranche live ~9 months from signing; 10-K discloses the slip was NVIDIA-equipment shortages, inside contractual grace. Operating ARR → $1.0B (8/26). |
| 2026-07-20 | $2.8B TCV new contracts (Prometheus, Perplexity, Together AI, Figure AI, Fal, Hume, Higgsfield, Cohere); ARR target raised to ">$4B", ~85% contracted; ~45% prepay; ~4-yr wtd term | +19.6% that day. Pricing >$20M/MW on 3-yr terms. |
| 2026-08-25 | $2.4B Blue Owl/PIMCO non-IG GPU financing, 9.0% fixed, ~30-month amortizing, Mackenzie air-cooled; funds ~90% of associated GPU capex | Marginal non-IG cost of capital = 9%, amortizing faster than the 4-yr contracts. |
| May–Jun 2026 (missed by v2 at publish) | NVIDIA partnership (up to 5GW DSX), $3.4B/5-yr NVIDIA cloud contract (60MW, ~$0.7B ARR, ramps early 2027), NVIDIA right to buy 30M sh @ $70 (5-yr); $3.0B 1% converts due Dec-2033 @ $73.07 (cap $110.30); $3.6B IG GPU facility @ ~6.0%; Sweetwater 1 (1,400MW) energized 5/1; Nostrum closed 6/15 | ~$10B of capital + demand events that v2's #1 risk ("$2.6B cash vs $3.96B debt") did not reflect. |
| 2026-07-01 | 9,099,328 RSUs to EACH co-CEO (10-K Note 30; ~$832M combined at grant per Qual), service-only, 4-yr vest + 2-yr hold, no further grants to FY31; Chair DEFA14A 7/8 defends rejecting performance hurdles | Zero FY26 expense recognized; ~$208M/yr SBC from FY27. CONFLICT Tracker's Form-4 pull shows 9,651,525 each — 10-K figure used. |
| 2026-08-03 / 06-15 | Mirantis closed ($544M: 12.6M sh + ~$40M cash/RSUs; ~580 staff; k0rdent; NVIDIA-certified hypervisor) · Nostrum closed (837,424 sh + ~€82M; Badajoz ES ~300MW) | Software layer bought, not built. CONFLICT Tracker cites 13.67M Mirantis shares; 10-K 12.6M used. |
| 2026-09-08 | 2GW Sweetwater conditionally included as Base Load in ERCOT "Batch Zero" | +5% day. Conditional; SB6 curtailment rules finalize end-2026. |
| 2026-09-14 → 09-21 | JPM double-upgrade UW→OW $65; Northland init OP $99; BTIG reit $80; Redburn init Neutral $40 (Sell on CRWV/NBIS) | Street went 1 Sell → 0 Sells in 90 days. |
| 2026-09-09/10 | Co-CEO Roberts on X after Communacopia: "The biggest debate on our stock: the gap between $71m of quarterly AI Cloud revenue and $1bn of ARR operating, $4bn contracted for year end"; investors "grown numb" to deal announcements | Management naming the bear case. −3.3% that day. |
| 2026-06-29 / 02-27 | Russell 1000 · MSCI USA inclusion | Passive bid is structural; explains part of the 13F "accumulation." |
| 2026-07-02/15 | CPO Aghili (Oracle OCI), CDO Nudelman (Google/CyrusOne), CISO Hammersley (Nutanix) | First credible hyperscale product/DC-dev hires; CTO seat still legacy colo. |
| v2 claim | Grade | Evidence |
|---|---|---|
| Thesis INTACT — ON WATCH, conviction-spec, Zone 1 active | KEPT (status); score re-cut 7.0 → 6.5 | Demand stronger; funding intensity ~6x larger. |
| $3.1B contracted ARR, target $3.7B CY26 | KEPT / AHEAD | Target raised to >$4B (7/20); "$4B contracted, 2026 largely sold out" (8/27). |
| Horizon 1 to Microsoft "this quarter" (Q4 FY26) | KEPT in substance / MISSED ~6 wks | Accepted 8/13 (Q1 FY27), inside company CY-Q3 target. |
| AI Cloud run-rate crosses $500M | KEPT | $500M at Jun 30 exit; $1.0B operating Aug 26. |
| OCF/CapEx inflects 2 consecutive quarters | AMBIGUOUS — KPI redefined ex-prepay | Q3 0.06 → Q4 0.92 headline / 0.045 ex-prepay. |
| Share growth <+20% YoY | MISSED | +47.8% FY-end to FY-end; +52.7% Jun-25 → Aug-26. |
| NVIDIA exercises $70 right | PENDING | Never above $49.37 since grant. |
| Third anchor signs | PARTIAL | $2.8B batch + unnamed frontier lab; no MSFT-scale anchor. |
| Breaker: MSFT termination / Horizon slip ≥2 qtrs | NOT TRIGGERED | H1 slipped ~0.5 qtr; H2 commissioning; H3–4 late-stage. |
| Breaker: deep-discount raise | NOT TRIGGERED | ATM avg $52.84 under $6B program; Dec-25 direct at $41.12; converts at 32.5% premium. |
| Breaker: BTC <$40K / founder exit / ARR cut | NOT TRIGGERED | BTC $84K; both founders; target raised. |
| Targets Bull $85–100 / Base $48–60 / Bear $12–18 | STALE — rebuilt (Valuation) | Assumed 400M FD; latent converts $6.75B excluded. |
| Entry ladder $38–42 / $33–36 / $26–30 | KEPT — all filled | 38 sessions in Z1, 12 in Z2, one in Z3 (7/29). |
| "P/S TTM >25 = trim regardless" | BROKEN BY DESIGN — retired | Fires at a flat price as mining revenue exits. Replaced by EV/operating-ARR + cap/backlog bands. |
| No-chase "above 50-day (~$54)" | STALE | 50-day now $41.21; re-anchored to $49.50 triple top. |
| Insiders 13.6% | UNSOURCED — retired | yfinance 3.3% / MarketBeat 5.0% / stockanalysis 12.1%; Form-4 math ~7.4% ex-unvested. Pending FY26 proxy. |
| "140K GPUs by end-2026" | NO LONGER RELEVANT | Company reports MW/ARR now; GPU counts not disclosed. |
| Mining = fuel not floor | RIGHT | $450M more impairment; shutdown Dec 2026; zero BTC held. |
| Funding gap = risk #1; dilution IS the model | RIGHT (now dominant) | $13.6B 12-mo commitments; $19B raised TTM. |
| NVIDIA $70 = smart-money strike | RIGHT | Dec-33 converts priced at $73.07 corroborate the band. |
(1) Growth story / execution changed? YES, both directions — demand up (pricing +125%, sold out, $1B operating), capital need up (~$25–30B FY27). The bet moved from "can they deliver 480MW" to "can they fund $25–30B without a punitive raise."
(2) Overlooked, now revealed? YES — (a) prepayment accounting flatters OCF/FCF; (b) $6.75B latent convert dilution; (c) $1.72B restricted cash isn't free; (d) M&A is a new dilution channel; (e) Microsoft is an ASC 842 operating lease ("deferred lease revenue"), which makes IREN a landlord to its largest customer; (f) $13.6B 12-month commitments; (g) revenue-per-MW pricing power is a lever v2 never modeled.
Tally: v2 7 PASS / 5 FAIL / 4 WATCH → v3 7 PASS / 6 FAIL / 3 WATCH. Net margin flipped to FAIL; institutions moved to PASS; three gates got worse in magnitude (FCF, dilution, margins), two improved in direction (insider activity, moat trajectory). The 16 gates are the v2 checklist carried forward so grades are comparable.
| # | Gate | v2 | v3 | Evidence (one line) |
|---|---|---|---|---|
| 1 | Revenue growth 30–50%+ | PASS | PASS | FY26 $707.0M vs $501.0M (+41%); AI Cloud $128.8M vs $16.4M; FY27 consensus $2.8B (+297%, 19 analysts). Q4 YoY −26.7% is the mining teardown. |
| 2 | FCF positive | FAIL | FAIL (worse) | FY26 FCF −$2,232M; ex-prepay ≈ −$4.07B. Q4 −$167M headline only because OCF carried the Microsoft prepayment. FY27 capex $25–30B. |
| 3 | Visible flywheel | PASS | PASS (stronger) | Prepayments now 45–55% of GPU capex + 90% GPU facilities = step 3 literally funds step 4–5. Step 7 (next site) still equity-funded. |
| 4 | Moat ≥9/10 (Tier-1 gate) | FAIL (7.5) | FAIL (7.0) | Real 2–4-yr time-moat on energized power + vertical build; no switching costs beyond contract term, no network effects, GPU layer refreshes every ~4 yrs, ClusterMAX bottom tier. |
| 5 | Founder-led | PASS | PASS | Daniel (41) & William (35) Roberts, co-founders/co-CEOs since 2018; ~44% voting via B-class UNVERIFIED. |
| 6 | Skin-in-the-game ≥5x | PASS (7.9x) | PASS — weak (8.8x) | ~$642M vested equity each ÷ $72.6M FY25 total comp = 8.8x; 218x on cash comp; ~1.5–2.5x on FY26 grant-inclusive comp. |
| 7 | Insider net activity (12 mo) | WATCH | WATCH → improving | Open-market buys $0 / sells $0 in window; grants only. Just outside window: −$66.3M co-CEO sales 9/11/25 at $33.13. No open-market buy ever found. |
| 8 | Institutional adds/trims | WATCH (46.8%) | PASS (caveated) | 13F Q2: 467 buyers / 139 sellers; top adds BofA +12.1M, BlackRock +8.9M, Norges +5.2M. Ownership 56–68% depending on source CONFLICT. Believer cohort trimmed. |
| 9 | Earnings beat history (8 straight) | FAIL | FAIL | 4/8 on street-adjusted EPS; last 3 beats; Q4 FY26 revenue miss −13%. Milestones (ARR/MW) kept or raised with one disclosed slip. |
| 10 | Explainable to an 8-year-old | PASS | PASS | Company 101. |
| 11 | P/S band | WATCH (19.4x) | WATCH — denominator broken | 24.6x TTM ("excited") on revenue being switched off; 6.2x FY27 consensus ("aware"); cap/contracted backlog $16.6B = 1.05x; cap/operating ARR 17.4x. |
| 12 | No shareholder dilution | FAIL (+52%) | FAIL (worse) | 257.2M → 380.2M FY-over-FY (+47.8%); 394.06M at 8/14; $6.75B converts unconverted; $3.5B ATM room; 18.2M CEO RSUs; NVIDIA 30M @ $70. |
| 13 | Current ratio ≥2:1 | PASS (3.72) | PASS (3.55) | CA $7,886M / CL $2,221M; quick 2.66. |
| 14 | Net margin positive | WATCH | FAIL | FY26 NI −$702.6M (−99% of revenue); adj. EBITDA margin fell 41% → 14% in one quarter on SG&A build; FY27 SBC ≈ $400M+. |
| 15 | Cash > debt | FAIL ($2.6B vs $3.96B) | FAIL (~parity incl. restricted) | $5,895.6M unrestricted + $1,723.9M restricted vs $7,592.9M debt (+$243.8M finance leases). Net debt/FY26 adj. EBITDA 6.9x; 22x on Q4 run-rate. |
| 16 | Operational efficiency rising | PASS | PASS | Revenue/MW $10–15M → >$20M/MW between contract vintages; ~2-yr payback claimed; AI Cloud cost of revenue ex-D&A only $16.9M on $128.8M. |
Where the fails cluster: all six are the balance-sheet cost of building three years ahead of revenue — same cluster as v2, numbers 2–3x larger. Whether that is operating leverage or ruin is decided by the Q1–Q3 FY27 prints.
One chart tells the story: revenue shrinks by design while contracted ARR runs 14x ahead of operating ARR; the gap closes only via capex that share count and converts are funding. Quarterly Q1 FY25 → Q4 FY26, plus the Aug-26 operating-ARR point.
| Item | Value | Source / note |
|---|---|---|
| Price | $44.125 | yfinance 9/25 close |
| Shares outstanding | 394,058,648 (+2 B-class) | 10-K cover 8/14/26. Jun 30 2026 380.19M; Mar 31 340.98M; Dec 31 2025 331.76M; Sep 30 2025 283.46M; Jun 30 2025 257.21M (equity statement; 258.10M cover) |
| Market cap | $17.39B | 394.06M × $44.125; reconciles to yfinance |
| Cash | $5,895.6M unrestricted + $1,723.9M restricted | 10-K Jun 30 2026. Restricted = GPU-financing SPV reserves; Microsoft attribution is inference UNVERIFIED |
| Debt | $7,592.9M carrying + $243.8M finance leases = $7,836.7M | Converts $6,745.7M principal / ≈$6,652M carrying after $92.9M discount + ≈$941M GPU "financing facility"; yfinance $7.84B. RT-FIX: quant's "operating leases" label wrong — operating lease liabilities are $2.8M. 10-K Notes 23/26; XBRL |
| EV | $19.09B | cap + debt − unrestricted cash; $17.36B netting restricted cash; yfinance $19.33B |
| Net debt | $1,697M | ≈ −$27M (net cash) if restricted counted |
| Share count YoY | +52.7% vs Jun 30 2025 | +47.8% FY-end to FY-end; +39.0% vs Sep 30 2025. RT-FIX: "+80% (218.9M→394M)" in Macro/Sentiment/Tracker uses a Dec-2024 diluted weighted base — wrong. 10-K/10-Q |
| Short interest | 83.09M sh (9/15) = 21.1% SO / 24.0% float · DTC 2.0 | Float 346.2M; peak 98.2M (7/31). CONFLICT: quant's 27.0%/93.6M is the 8/31 print. Nasdaq via theonlineinvestor; yfinance 21.7% |
| Institutions / insiders | 56–68% inst · 3.3–12.1% insiders | yfinance 59.2%, stockanalysis 56.2%, BusinessQuant 68.2% — definitional; 68.2% reproduces on no denominator. Insiders 3.3% (yfinance) to 12.1% (stockanalysis); Form-4 math: co-CEOs 23.64M each incl. ~9.1M unvested ⇒ ~12% incl. grants, ~7.4% ex UNRESOLVED until FY26 proxy |
FY25 adj. EBITDA are IFRS-era 6-K figures; gross profit = revenue − cost of revenue ex-D&A, company convention. Scroll horizontally.
| Qtr (end) | Rev | BTC rev | AI Cloud rev | GM% | Op inc ex-impair / OpM% | Impairment | Op inc incl. impair | Adj EBITDA (margin) | GAAP NI | GAAP EPS dil | Adj EPS est→act | OCF | CapEx | FCF | Period-end sh (M) |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Q1 FY25 (Sep-24) | 52.8 | 49.6 | 3.2 | 39.6% | −39.3 / −74% | n/a | −39.3 | 2.6 (5%) | −51.7 | −0.27 | −0.07→−0.16 MISS | −3.9 | −383.4 | −387.3 | 197.1 |
| Q2 FY25 (Dec-24) | 116.1 | 113.5 | 2.7 | 72.2% | 16.3 / 14% | n/a | 16.3 | 62.6 (54%) | −21.9 | −0.10 | 0.06→0.09 BEAT | 53.6 | −187.6 | −134.0 | 214.4 |
| Q3 FY25 (Mar-25) | 144.8 | 141.2 | 3.6 | 71.0% | 27.7 / 19% | n/a | 27.7 | 83.3 (58%) | −16.1 | −0.07 | 0.14→0.11 MISS | 93.1 | −443.8 | −350.7 | 223.6 |
| Q4 FY25 (Jun-25) | 187.3 | 180.3 | 7.0 | 71.8% | 15.9 / 8% | n/a | 15.9 | 121.9 (65%) | +176.7 (incl. +147.7 unrealized FV gain) | +0.65 | 0.22→0.21 MISS | 103.1 | −357.9 | −254.7 | 257.2 |
| Q1 FY26 (Sep-25) | 240.3 | 232.9 | 7.3 | 66.4% | −60.1 / −25% | 16.3 | ≈ −76 / −32% | 91.7 (38%) | +384.6 (incl. +665.0 unrealized gain) | +1.08 | 0.14→0.01 MISS | 142.4 | −280.6 | −138.3 | 283.5 |
| Q2 FY26 (Dec-25) | 184.7 | 167.4 | 17.3 | 64.4% | −84.7 / −46% | 31.8 | ≈ −117 / −63% | 75.3 (41%) | −155.4 (111.8 inducement) | −0.52 | −0.25→−0.03 BEAT | 71.7 | −719.1 | −647.4 | 331.8 |
| Q3 FY26 (Mar-26) | 144.8 | 111.2 | 33.6 | 72.4% | −93.3 / −64% | 140.4 | ≈ −234 / −161% | 59.5 (41%) | −247.8 | −0.74 | −0.26→−0.16 BEAT | 75.3 | −1,355.2 | −1,279.9 | 341.0 |
| Q4 FY26 (Jun-26) | 137.2 | 66.7 | 70.5 | 75.7% | −140.6 / −102% | 450.4 | ≈ −591 / −431% | 19.2 (14%) | −684.0 | −1.88 | −0.45→−0.25 BEAT | 1,811.1 (≈89 ex-prepay) | −1,978.2 | −167.1 (≈ −1,889 ex-prepay) | 380.2 |
RT-FIX Quant's table header said op income "includes impairments" but the figures excluded them; the FY26 sum ex-impairment (−$378.7M) + impairments ($638.8M) + disposal loss (~$29M) reconciles to the 10-K FY26 operating loss of −$1,046.7M (−148% of revenue). Both columns shown above.
Revenue $707.0M (+41.1% vs $501.0M; BTC $578.2M, AI $128.8M vs $16.4M); adj. EBITDA $245.7M (FY25 $269.7M); NI −$702.6M (FY25 +$86.9M); OCF $2,100.4M; capex $4,332.7M (PP&E $2,998.0M + hardware $1,335.1M); FCF −$2,232.3M; SBC $205.0M (29.0% of revenue; FY25 $42.6M / 8.5%). Beat record 4/8 on street-adjusted EPS; GAAP EPS is dominated by non-cash FV swings (+$558.5M unrealized gains on prepaid forwards/capped calls in FY26) and impairments ($638.8M) — the P&L is long IREN's own stock.
Deferred revenue $0.9M → $1,842.5M; $1,796.1M non-current = deferred lease revenue under ASC 842, i.e., the Microsoft prepayment RT-FIX label; ASC 606 contract liability is only $219.1M. Including intangibles ($107.6M) and deposits ($203.4M): −$2,543.7M. Ex-prepayment: OCF ≈ $258.7M ⇒ FCF ≈ −$4,074M (−$4,385M all-in). yfinance's −$4.22B resembles the ex-prepay figure by coincidence, not method RT-FIX. SBC is 29% of revenue and rising: the co-CEO grant recognized zero expense in FY26 (10-K Note 30) and adds ~$208M/yr from FY27; options-only unrecognized cost $23.1M RT-FIX: quant's "RSU/option $23.1M" tag is options-only.
| Qtr | OCF/CapEx | FCF margin | Op margin (ex / incl. impair) | Read |
|---|---|---|---|---|
| Q1 FY25 | −0.01 | −734% | −74% | pre-scale |
| Q2 FY25 | 0.29 | −115% | +14% | mining cash |
| Q3 FY25 | 0.21 | −242% | +19% | |
| Q4 FY25 | 0.29 | −136% | +8% | |
| Q1 FY26 | 0.51 | −58% | −25% / −32% | peak mining cash |
| Q2 FY26 | 0.10 | −351% | −46% / −63% | build begins |
| Q3 FY26 | 0.06 | −884% | −64% / −161% | |
| Q4 FY26 | 0.92 headline / 0.045 ex-prepay | −122% / −1,377% ex-prepay | −102% / −431% | prepayment quarter |
Verdict: OCF/CapEx <0.3 in 6 of 8 quarters; the only >0.5 print is the Microsoft prepayment. Gross margin on AI Cloud ex-D&A is 87%, but D&A ($417M FY26) is the real cost of a GPU business and segment EBITDA is undisclosed. Operating leverage exists in the unit (2-yr payback at $20M/MW claimed) but is invisible at the company level because SG&A ($128.3M Q4, +57% QoQ), SBC and depreciation are being built ahead of revenue. FCF-positive requires operating ARR ≥ ~$4B and a capex plateau; with FY27 capex $25–30B (verified verbatim in the Q4 transcript; 10-K MD&A gives no FY27 guide), FCF stays deeply negative through FY27 on every scenario. The framework's "FCF-inflection entry" is a FY28+ event; the 2026–27 re-rate catalyst is delivered ARR, not FCF.
100% organic revenue. Mirantis ($544M) and Nostrum (837,424 sh + ~€82M) are software/ops tuck-ins, immaterial to FY26 revenue. Growth is a swap: BTC $232.9M (Q1) → $66.7M (Q4) as 50 EH/s → 23.2 EH/s; AI Cloud $7.3M → $70.5M (+110% QoQ Q4). BTC dependence: 49% of Q4 revenue; FY26 6,075 BTC at ~$95k realized (arithmetic-consistent with $578.2M) UNVERIFIED realized price; → 0 by CY26-end. Anniversary: Horizon 1 revenue starts Aug 2026 (Q1 FY27); YoY comps turn sharply positive from Q2 FY27 (Dec-2026 quarter).
| Metric | Q1 FY26 | Q2 FY26 | Q3 FY26 | Q4 FY26 | Latest |
|---|---|---|---|---|---|
| Contracted ARR | $1.9B (MSFT, 11/3/25) | $1.9B (target $3.4B) | $3.1B (MSFT 1.9 + NVDA 0.7 + other; target $3.7B) Tracker: "Prince George $0.5B" was never company-disclosed — dropped | >$4B target (7/20) | $4.0B "for 2026 capacity" (8/27) |
| Operating ARR | ~$29M (rev×4) | ~$69M | ~$134M | ~$500M exit (call) | $1.0B (8/26, post-H1) on ~90MW IT ⇒ $11.1M/MW |
| RPO (ASC 606) | $195.0M | $289.4M | $710.3M | $5.1B ($0.9B next 12 mo) | + $11.4B contracted lease value (ASC 842, Microsoft) = $16.6B total backlog RT-FIX |
| Deferred revenue | $23.3M | $46.6M | $120.4M | $1,842.5M | — |
| Cash (unrestricted) | $1,032M | $3,261M | $2,213M | $5,896M + $1,724M restricted | — |
| Capital commitments | — | — | — | $13,810M ($13,611M ≤12 mo) | — |
Microsoft ($9.7B/5yr/200MW IT GB300, 20% prepay; implies ~$29M/MW-IT GPU capex via Dell $5.8B, $9.7M/MW/yr revenue), NVIDIA ($3.4B/5yr, 60MW air-cooled Blackwell, early 2027, ~$0.7B ARR), Cohere, Prometheus, Perplexity, Figure AI, Fal, Higgsfield, Hume, Fluidstack, one unnamed frontier lab; renewals Together AI, Fireworks. Top-2 = 66% of contracted ARR RT-FIX: Qual's 63%. Pricing: 3-yr ">$20M/MW IT" (some slides ">$25M"), 5-yr just under $20M; payback ~2 yrs. Delivery: H1 live 8/13; H2 commissioning; H3–4 Dec-quarter; "significant ARR ramp in March quarter 2027." GPUs: 150k fleet target (3/4/26); installed count not disclosed.
| Series | Principal | Coupon | Conv. price | Maturity | Shares if converted | Note |
|---|---|---|---|---|---|---|
| 2029 | $233.4M (of $550M; $316.6M repurchased Dec-25) | 3.50% | $13.64 | 12/15/2029 | 17.1M | deep ITM — will convert |
| 2030 | $212.3M (of $440M; $227.7M repurchased) | 3.25% | $16.81 | 6/15/2030 | 12.6M | deep ITM |
| 2031 | $1,000M | 0% | $85.63 | 7/1/2031 | 11.7M | issued 10/14/25 @ ref $60.09 |
| 2032 | $1,150M | 0.25% | $51.40 | 6/1/2032 | 22.4M | capped call $82.24 |
| 2033-Jun | $1,150M | 1.00% | $51.40 | 6/1/2033 | 22.4M | capped call $82.24 |
| 2033-Dec | $3,000M | 1.00% | $73.07 | 12/1/2033 | 41.1M | 5/12/26 @ ref $55.15; capped call $110.30 (cost $174.5M) |
| Total | $6,745.7M RT-FIX: Macro $6.45B / Qual $6.3B are proceeds, not face | blended ~0.9% (~$60M/yr) | 127.2M | 10-K maturities: yr1 $177M, yr2 $270M, yr3 $194M, yr4 $573M, yr5 $128M, >5yr $6,363M |
Other debt: ~$941M GPU financing drawn at Jun 30 (of $3.6B IG @ ~6.0%, Microsoft deployment); $2.4B Blue Owl/PIMCO @ 9.0% fixed, ~30-month amortizing from each funding date (10-K Item 9B), availability through 12/31/26; $2.8B non-IG committed in total; finance leases $243.8M. Covenants: converts none financial (verify indenture); GPU facilities secured on GPUs/contracts within SPVs (IE US Hardware 3 LLC, $1,487M pledged net carrying), with parent "guarantees or other credit support in respect of certain performance and shortfall obligations" — terms undisclosed. Dec-2025 repurchase of 2029/2030 paper: $1,623.5M cash, $111.8M inducement expense, $981.0M APIC charge — retired 36.7M convert shares while issuing 39.7M new shares at $41.12 ⇒ net share count up ~3M; an anti-dilution optic, not an act RT-FIX to Qual's capital-allocation credit.
Current ratio 3.55 PASS (quick 2.66). Cash > debt FAIL (net debt $1.70B; ~parity incl. restricted). Net debt / FY26 adj. EBITDA 6.9x; 22x on Q4 run-rate — FAIL any leverage test until AI EBITDA lands. Liquidity vs commitments: sources ≈ $12.7B (cash 5.9 + restricted 1.7 + undrawn IG ~2.7 + Blue Owl 2.4) vs $13.6B due ≤12 months vs $25–30B guided ⇒ $3–8B residual. The 10-K's going-concern sentence relies on "proceeds from financing activities." Convert risk: 2032/2033-Jun ($2.3B) convert at $51.40 (+16.5%); below that they are ~$32M/yr of coupon — cheap; the risk is 2031–33 refinancing if ARR under-delivers, not a near-term wall. Maturities ≤12 months $177M.
Under the current $6B ATM (filed 3/4/26; refiled 6/30/26): 47,165,838 sh for $2,492.1M gross, avg $52.84; $3.51B remaining; zero sales Jun 30–Aug 14 at $29–40. Converts contributed no FY26 shares (repurchased for cash).
| Series | Q1 FY25 | Q2 FY25 | Q3 FY25 | Q4 FY25 | Q1 FY26 | Q2 FY26 | Q3 FY26 | Q4 FY26 | Aug-26 |
|---|---|---|---|---|---|---|---|---|---|
| A · BTC rev ($M) | 49.6 | 113.5 | 141.2 | 180.3 | 232.9 | 167.4 | 111.2 | 66.7 | — |
| A · AI Cloud rev ($M) | 3.2 | 2.7 | 3.6 | 7.0 | 7.3 | 17.3 | 33.6 | 70.5 | — |
| B · Contracted ARR ($B) | — | — | — | — | 1.9 | 1.9 | 3.1 | 4.0 | 4.0 |
| B · Operating ARR ($B, AI rev×4) | 0.013 | 0.011 | 0.014 | 0.028 | 0.029 | 0.069 | 0.134 | 0.282 | 1.0 (8/27) |
| B · RPO ($B) | — | — | — | — | 0.195 | 0.289 | 0.710 | 5.1 (+11.4 lease) | — |
| C · FCF ($M) | −387 | −134 | −351 | −255 | −138 | −647 | −1,280 | −167 (−1,889 ex-prepay) | — |
| C · CapEx ($M) | 383 | 188 | 444 | 358 | 281 | 719 | 1,355 | 1,978 | — |
| D · Shares (M) | 197.1 | 214.4 | 223.6 | 257.2 | 283.5 | 331.8 | 341.0 | 380.2 | 394.1 |
Caption: revenue shrinks by design while contracted ARR runs 14x ahead of operating ARR; the gap closes only via capex that share count and converts are funding. Weekly price file: px.js (105 bars, SMA20/50/200 + BB) drives the Trade Plan chart.
Daniel Roberts (41, ex-Palisade infra/renewables finance) and William Roberts (35, ex-Macquarie/Westpac/Brookfield) — co-founders, co-CEOs, each holds 1 B-class share (15 votes per ordinary share held; ~44% combined voting UNVERIFIED, DEF14A-derived; expires ~Nov 2033). Finance/real-asset operators, not technologists. Survived a −90% 2022 drawdown and a lender fight — the framework's "IPO curse rite of passage" box is ticked.
CTO Denis Skrinnikoff (since 2021; ex-TeraGo colo/network) — a colocation engineer, not an AI-systems/GPU-cluster leader: weak for the seat. CPO Kambiz Aghili (Jul 2026; VP Products, Oracle OCI) — credible. CDO Michael Nudelman (Google/CyrusOne) — credible. CISO Eric Hammersley (Nutanix). Software depth arrives via Mirantis (CEO Alex Freedland retained, ~580 engineers, k0rdent, NVIDIA-certified hypervisor). CFO Anthony Lewis (took over from Belinda Nucifora in FY26; date/reason n/a), COO David Shaw, CCO Kent Draper (runs sales; clearest voice on calls), CLO Cesilia Kim; Chief Innovation Officer John Gross and CMO Chris Parker bios n/a. Read: real technical leadership = Mirantis + Aghili; the CTO chair is legacy; external validation (Microsoft H1 acceptance, NVIDIA Exemplar) compensates, ClusterMAX contradicts.
David Bartholomew (64, Chair since 2021; ex-CEO DUET Group; Comp Chair), Christopher Guzowski (40; renewables developer), Michael Alfred (44; Alpine Fox; 878,067 sh — the only meaningful outside holder), Sunita Parasuraman (52; Meta/VMware/Apple treasury; Audit Chair). Network map: energy-infra + crypto-investor + big-tech finance. Nobody who has run a hyperscale cloud, a GPU fleet or a DC REIT — for a $25–30B capex program that is a gap, not filler. The board unanimously approved no-hurdle mega-grants (FW Cook advised); say-on-pay at the AGM (~Nov 2026) is the next governance read.
Comp: FY25 cash (salary $964k + bonus $1.968M + other) ≈ $2.95M each; FY25 total incl. stock awards $72.6M each (DEF 14A) not re-verified this pass; FY26 total will include the ~$416M/each RSU grant. Gate 6 = PASS, weak tier. Read: aligned by holdings; pay structure is retention-by-dilution with no performance hurdle.
Net insider activity, 12 months (9/25/25→9/25/26): open-market buys $0, sells $0 = 0% of holdings. Grants only (7/1/26). Just outside the window: 9/11/25 each co-CEO sold 1,000,000 sh @ $33.13 ($33.1M each), seven weeks before the Microsoft announcement; 9/16/25 Guzowski sold 11,958 sh ($434k). No open-market insider purchase found in any period. Mirantis recipients' 11.98M resale shares registered 8/4/26 (424B7) UNVERIFIED vs 12.6M issued.
| Dimension | Score | Evidence |
|---|---|---|
| Management (capital allocation, ethics, long-term) | 6.5 | Plus: $19B raised in 12 months at converts struck $51–86 with capped calls; 6% IG GPU debt; 45–55% prepayments; contracted before building; cash-repurchased the $13.64/$16.81 converts. Minus: that repurchase was share-neutral and cost $111.8M of inducement RT-FIX; $66M of founder sales pre-Microsoft; $832M no-hurdle grant with hurdles explicitly rejected; ATM avg $52.84 means management refuses to fund below ~$50 while commitments are due; Q4 revenue miss with the CEO conceding "ramp assumptions ran ahead of guidance." (Qual 7.0; PM −0.5) |
| Operations (customer obsession, innovation, failure tolerance) | 8.0 | Sweetwater 1 energized on the April promise; Horizon 1 delivered ~9 months from signing and accepted; NVIDIA Exemplar on GB300 NVL72; 7 years of continuous self-build; ~3,000 site personnel; closed-loop liquid cooling; ERCOT alignment. Watch: NVIDIA-equipment slip on H1 (disclosed), BC liquid conversions slipped "end 2026" → "during 2027"; ClusterMAX bottom tier. |
| Structure (talent density, meritocracy, decentralization) | 6.0 | FTEs 257 → 685 in FY26 (+580 Mirantis; "similar growth" FY27); credible 2026 hires (Oracle, Google, Nutanix, Mirantis founder); but founder-centralized dual-class control, no cloud operator on the board, integration of a 580-person software company during a 3x headcount year. Talent-gravity signal: the indexed Glassdoor page (109 reviews, 3.5★) is the Italian utility Iren SpA — unusable; Revelio job postings +148% to 214; no low-view technical podcasts from division leaders found (gap). |
| Composite | 6.8 | Framework tie-break for a multi-way unprofitable race ("most relentless founder, highest skin, best talent"): Roberts brothers score high on relentless and skin, mid on talent. |
| Moat source | Grade | Evidence |
|---|---|---|
| Power / land / interconnect (time moat) | Strong, 2–4 yrs | 5.3GW executed grid agreements; Sweetwater 2GW conditional base-load; Roberts: greenfield started today = ~2030 compute. Not permanent — every miner and utility is racing the same clock. |
| Vertical integration / process power | Moderate-strong | Designs, builds, operates own DCs (130–200kW racks, direct-to-chip); now owns the software layer (Mirantis). Captures cloud margin, not landlord margin. |
| Scale economies | Emerging | Procurement (Dell/NVIDIA), financing cost falling from mid-teens private credit to 6% IG. |
| Switching costs | Weak | Contractual only (3–5 yrs, prepay). At renewal the customer can move; GPU refresh every ~4 yrs resets the relationship. |
| Brand / network effects | Weak-moderate | NVIDIA Exemplar + partnership referrals; Microsoft reference. No network effects. ClusterMAX "not recommended" cuts against brand. |
| Contestability | High above power; low at power | GPU cloud is a commodity with hyperscaler and neocloud entrants (10-K names them); power is scarce for ~3 years. |
| Factor | IREN | CoreWeave (CRWV) | Nebius (NBIS) | Applied Digital (APLD) | Cipher (CIFR) | TeraWulf (WULF) | Core Scientific (CORZ) |
|---|---|---|---|---|---|---|---|
| Model | Owns power + shell + GPUs + software (full-stack, self-built) | Full-stack neocloud; leases shells | Full-stack neocloud, software-native | Powered shell landlord + some cloud | Shell landlord (ex-miner) | Shell landlord (ex-miner) | Shell landlord (ex-miner) |
| Mkt cap $B | 17.4 | n/a this pass | 60–65 | 7.5–7.7 | 7.4 | 7.9 | 5.6 |
| Rev TTM $M / latest-Q YoY | 707 / −27% (FY +41%) | Q2 CY26 $2.58B (+112%) | 1,355 / +454% | 611 / +407% | 191 / −43% | 165 / −6% | 440 / +109% |
| Next-FY rev est $B (fwd P/S) | 2.80 (6.2x) | n/a | 12.34 (9.1x) | 0.83 (9.2x) | 0.85 (13.0x) | 0.91 (14.6x) | 1.16 (6.3x) |
| P/S TTM / EV/S TTM | 24.6 / 27.3 | n/a | 48.0 / 49.6 | 12.5 / 18.3 | 38.5 / 63.4 | 47.5 / 63.4 | 12.7 / 18.6 |
| Gross / op margin | 69% / −54% ex-impair (−148% incl.) | n/a | 74% / −37% | 26% / −29% | −10% / −28% | 69% / −207% | 32% / −32% |
| FCF TTM $B | −2.23 | n/a | −5.88 | −2.78 | −1.50 | −2.74 | −0.96 |
| Current ratio (≥2) | 3.55 PASS | n/a | 4.03 PASS | 4.01 PASS | 3.00 PASS | 0.75 FAIL | 0.89 FAIL |
| Net cash $B / D/E | −1.9 / 1.87 | n/a | −2.15 / 0.99 | −3.51 / 1.36 | −4.75 / 9.51 | −2.62 / 35.5 | −2.64 / neg BV |
| Shares chg YoY | +47.8% (FY-end) | n/a | +22.1% | +36.8% | +13.3% | +12.0% | +3.6% |
| Short % float | 24.0% | n/a | 19.8% | 23.3% | 20.9% | 31.8% | 22.5% |
| Founder-led | yes | no | yes (Volozh) | yes (Cummins) | no | yes | no |
| 8-qtr beat record | 4/8 | n/a | 6/8 | 7/8 | 3/8 | 3/8 | 2/8 |
| Contracted backlog (TCV) | $16.6B (RPO $5.1B + $11.4B lease) | $104.2B (+$25B early Q3) | ~$46B (MSFT $17–19B + Meta ≤$27B) | ~$11B CoreWeave UNVERIFIED | ~$8.5B (AWS $5.5B/15yr + Fluidstack ~$3B/10yr) | ~$19B Anthropic/401MW UNVERIFIED | ~$10B CoreWeave UNVERIFIED |
| Cap / backlog | 1.05x RT-FIX from 3.4x | n/a | 1.4x | ~0.7x | 0.9x | n/a | ~0.6x |
| Off 52-wk high | −43% | −9% 3m | −21% | −48% | −41% | −47% | −43% |
| Analyst PT (upside) | $77.97 (+77%) | n/a | $276 (+16%) | $66 (+152%) | $31 (+76%) | $34 (+118%) | $35 (+101%) |
Who is winning. On growth and backlog: CoreWeave, then Nebius (ARR $3B Jun-26 guiding $7–9B YE26, software-native org). On capital-per-ARR: IREN (owned shell + 45–55% prepay + 90% GPU gearing). On duration/risk: the landlords (10–20-yr leases at $2–3M/MW/yr colo vs IREN's ~4-yr $20M+/MW cloud paper). IREN is the only converted miner that went full cloud AND kept the power; it trades at a Nebius-like forward P/S with a landlord-like cap/backlog (1.05x) — the premium to landlords is justified only if cloud EBITDA (undisclosed by segment; blended 14% in Q4) proves out above lease economics. The whole group is −41% to −48% off highs (NBIS −21%): a sector de-rate on financing cost, not IREN-specific.
| Qtr | Revenue | Rev est | Adj EPS est → act | GAAP EPS | Stock reaction | Guidance / milestones |
|---|---|---|---|---|---|---|
| Q1 FY25 (Nov-24) | $52.8M | n/a | −0.07 → −0.16 MISS | −0.27 | n/a | 50 EH/s target; Childress build |
| Q2 FY25 (Feb-25) | $116.1M | n/a | 0.06 → 0.09 BEAT | −0.10 | n/a | Horizon 1 AI DC plan |
| Q3 FY25 (May-25) | $144.8M | n/a | 0.14 → 0.11 MISS | −0.07 | n/a | 50 EH/s reached mid-2025 (KEPT) |
| Q4 FY25 (Aug-25) | $187.3M | n/a | 0.22 → 0.21 MISS | +0.65 | n/a | AI Cloud ARR $500M+ "early CY26" (KEPT late, Jun-26) |
| Q1 FY26 (Nov-25) | $240.3M | n/a | 0.14 → 0.01 MISS | +1.08 | n/a | $3.4B ARR by end-CY26 on 140k GPUs; Sweetwater energize Apr-26 (KEPT); BC conversions by end-2026 (PARTIAL → 2027); H1–4 phased through 2026 (H1 KEPT w/ slip) |
| Q2 FY26 (Feb-26) | $184.7M | n/a | −0.25 → −0.03 BEAT | −0.52 | n/a | $3.4B reiterated; $3.6B IG financing <6% (KEPT); ">4.5GW" secured (KEPT, now 5.3GW) |
| Q3 FY26 (May-26) | $144.8M | n/a | −0.26 → −0.16 BEAT | −0.74 | n/a | ARR target RAISED $3.7B/150k GPUs; H1 handoff Q3 CY26 (KEPT); Mirantis close (KEPT 8/3) |
| Q4 FY26 (Aug-27-26) | $137.2M | $157.1M MISS −13% | −0.45 → −0.25 BEAT | −1.88 | −12.5% next day ($40.53 → $35.45) | ARR RAISED >$4B contracted; H2–4 Dec-qtr; FY27 capex $25–30B; ~$8B more financing; SG&A +$40–50M; mining off by 12/31/26 |
| Q1 FY27 (due 11/5/26 UNVERIFIED date) | est $221.8M | — | est −0.665 (yfinance) | — | — | Watch: H2 acceptance; AI Cloud ≥$150–190M; ATM disclosure |
Street-adjusted EPS 4/8 (last three beats). Revenue vs consensus only verified for Q4 FY26 (miss). Guidance: ARR targets raised three times ($3.4B → $3.7B → >$4B); never cut. Milestones: Sweetwater energization KEPT to the month; H1 KEPT with a disclosed ~6-week equipment slip; $500M ARR KEPT one quarter late; BC liquid conversions SLIPPED to 2027; 140k-GPU framing abandoned for MW. Gate 9 (8 straight beats) FAILS. Because GAAP EPS is noise (FV swings, impairments), judge on MW/ARR milestones — those have been kept or raised with one disclosed slip and a grace clause already in use. Tracker fix: v2's "Q3 −$0.74 vs −$0.22" compared GAAP to adjusted; v3 uses adjusted throughout.
Lenses at price $44.125, cap $17.39B, EV $19.09B.
| Lens | IREN now | Framework band | History / peers |
|---|---|---|---|
| P/S TTM ($707M) | 24.6x | "excited" (>20) | v1 (Jan, $53.74) ~18–20x; v2 (Jul, $41.14, TTM $757M) 19.4x — multiple rose because revenue fell |
| EV/S TTM | 27.0x | peers 18–63x | |
| P/S fwd (FY27 $2.8B; range 2.5–3.3) | 6.2x | "aware" (4–8) | CORZ 6.3x, NBIS 9.1x, APLD 9.2x, CIFR 13.0x, WULF 14.6x |
| EV/S fwd | 6.8x | ||
| Cap / contracted ARR $4.0B | 4.3x (EV 4.8x) | hunting-ground edge if delivered | Jul: EV/ARR 5.2x on $3.1B |
| Cap / operating ARR $1.0B | 17.4x | "optimistic" | |
| Cap / total backlog $16.6B | 1.05x | CIFR 0.9x, NBIS 1.4x, CORZ ~0.6x, APLD ~0.7x | |
| Fwd P/E · PEG · P/FCF | n/a | FY27 EPS −$3.42; FCF −$2.2B | |
| EV / adj. EBITDA FY26 $245.7M | 77.7x | GAAP EBITDA ≈ −$232M ⇒ n/m RT-FIX: "504x" retired | |
| P/B | 4.0x ($11.01/sh) | Jul 5.3x |
Price +7% while contracted ARR +29%, operating ARR ×7, cash ×2.3, debt ×1.9. The market did not re-rate for the Aug-13 delivery; it de-rated on the Q4 revenue miss and the "$25–30B" number. Rich on trailing, fair-to-cheap on forward, cheapest in the group on cap/contracted-ARR, mid-pack on cap/backlog. The multiple is a bet that FY27 revenue ≈ 4x FY26.
Retired rule: "P/S TTM >25 = trim regardless" fires at a flat price as mining revenue exits. Replacement bands (delivered ARR basis): EV/operating-ARR >12x with operating ARR ≥$3B = trim; cap/total backlog >2.0x = trim; cap/total backlog <0.8x with H2–4 accepted = add.
| Bear | Base | Bull | |
|---|---|---|---|
| NTM revenue | $2.0–2.25B | $2.8–3.0B | $4.0B (FY28 run-rate) |
| EV/S band | 4.0–5.0x | 6.0–7.0x | 8.5–10x |
| Net debt | $2.0B | $1.7–2.0B | $3.0B |
| Shares | 470M | 430M | 460M |
| Price | $16 ($13–25) | $40 ($36–48) | $72 ($68–87) |
| Probability | 25% | 50% | 25% |
CONFLICT Quant's cap-based P/S math gave base $42–49 and EV ≈ $50; Red Team's severity-weighted EV ≈ $41 with a modal bear of $22–29. PM lands between: base ≈ spot; the market pays nothing for the bull; Stage 10's 10:1 test passes only from Zone 2 ($34.5–36.5) where the ten scenarios stop costing more than the one pays. The Red Team's reminder stands: the last year's largest seller of IREN stock was IREN, at an average $52.84.
17–22 analysts: 12 Strong Buy / 2 Buy / 3 Hold / 0 Sell (stockanalysis); mean PT $77.97, median $80, range $40–131; FY27 consensus revenue $2.81B (+297%), EPS −$3.42. Last five actions: Redburn init Neutral $40 (9/21: "unit economics and conversion risk fairly priced"; Sell on CRWV/NBIS — IREN the relative favorite); Northland init OP $99 (9/18); JPM double-upgrade UW→OW $46→$65 (9/14) — the Street's last bear flipped; BTIG reit Buy $80 (9/14); Bernstein OP $100 (9/3; but 7/1 "behind on scale and enterprise"). Weiss quant Sell D. UNVERIFIED MarketBeat's "Morgan Stanley downgrade 9/2/2026" — only MS downgrade found is 9/26/2025. Third-party ops rating: SemiAnalysis ClusterMAX 3.0 (9/23) "not recommended," bottom tier with Sharon AI; CRWV/NBIS Platinum — body confirms the tier; the "#1 worst site" phrase is headline-only RT-FIX. No analyst reacted.
Short interest 51.6M (3/13) → 98.2M peak (7/31, ~28% float) → 83.09M (9/15; 24.0% float; DTC 2.0): 15M covered in 6 weeks into a $35 → $48 rally; top-decile short for a $17B cap but liquid — fuel, not a trap. Options: P/C OI 0.92; ATM IV 73–76% front, 81–85% for 11/20 (earnings inside); largest OI Nov-20 calls 70/60/50 (30k/27k/22k) — upside speculation heavy; deep-OTM puts (20/18/35) are tail hedges. 13F (6/30): 673 holders (+173); 467 buyers / 139 sellers; adds BofA +12.1M, BlackRock +8.9M, Goldman +7.0M, Norges +5.2M; trims Defiance ETFs −12.6M, D.E. Shaw −6.1M, Situational Awareness −2.2M (still #5). Top-4 holders are bank broker-dealers — largely convert-arb delta and prime inventory, not conviction. Believer cohort (Situational Awareness, Value Aligned, BIT Capital) flat-to-trimming. Supply overhang: Mirantis resale 11.98M sh (424B7 8/4) UNVERIFIED; NVIDIA 30M @ $70; ATM $3.51B. MSCI (2/27) and Russell 1000 (6/29) passive bids already in.
Who is the marginal buyer/seller. Since August the buyers are short-coverers (15M sh) and the sell-side momentum crowd (JPM day +7%; Northland +2.8%); long-only adds (Norges, BlackRock) were a Q2 phenomenon at $30–45. Marginal sellers: Mirantis recipients, miner ETFs, hedge funds. The incremental "real" buyer is a long-only growth fund pulled in by a Sweetwater anchor tenant — the kind of announcement Roberts says the market has "grown numb" to.
Retail. StockTwits 65/35 bullish, fatigued and $50-anchored ("dead until ARR increases," MM-conspiracy posts, wheelers "just got assigned 1,100 shares"). Reddit flipped from #1 sentiment score (Dec-25) to "hard pass" threads (Sep-26) second-hand, reddit blocked. Substack attention thinned since May. Seeking Alpha split, tilting cautious on dilution.
Contrarian verdict (Stage 7 gate): PASS WITH CAUTION — crowded, not despairing. Sell-side 82% Buy with +77% PT = the "unanimous bulls" caution; 24% short float + 75% IV + 56–68% institutional = maximum positioning, zero apathy. Retail fatigue is a partial despair signal and the roadmap is executing — but "high P/S after a crash = still crowded" applies (24.6x TTM). Edge: fade the Street's ARR arithmetic (M1), own delivery proof at the ladder, treat retail catalyst-chasing disappointment days (e.g., a "Horizon 2 Monday" that doesn't come) as the buy windows. Kill conditions for the sentiment thesis: (a) Microsoft delays acceptance of H2–4; (b) another ATM/convert raise before 11/5; (c) StockTwits >80% bullish on a break of $50 — the walk-away trigger.
IREN is a 90%-geared GPU lessor with ~4-year paper, a 30-month amortizing non-IG loan book, and a $25–30B capex year it has funded less than half of, trading at 24.6x trailing sales into a hiking cycle. The P&L is going backwards while the story goes forward: revenue down four straight quarters, adj. EBITDA $91.7M → $19.2M, GAAP operating loss −$1,046.7M, SBC 29% of revenue with ~$208M/yr more coming from a grant that recognized zero expense in FY26. Operating cash flow is a customer loan: ex-prepayment OCF ≈ $259M vs $4.3B capex, and the $1.8B is a liability earned only by delivering H1–4 on spec. The 12-month math does not close without equity: $13.6B due within 12 months vs ~$12.7B of sources, then $25–30B guided with a $3–8B residual — and management does not sell stock below ~$50 (zero ATM Jun 30–Aug 14 at $29–40). The debt is shorter than the assets and the assets are shorter than the story: Blue Owl amortizes over 30 months at 9%, GPUs depreciate over 5 years, contracts run 3–4 years, and VR200 is already on order — the equity return on a non-IG deployment is the residual value of 3-year-old B300s in 2029. Concentration is structural: Microsoft is 48% of ARR (66% with NVIDIA), accounted for as an operating lease with acceptance, ramp, service-credit, delay-credit and termination rights the 10-K lists and does not quantify; delivery has already slipped once and the grace clause is in use. SemiAnalysis says they cannot run a cloud; Bernstein says they are behind on enterprise. Governance: dual-class, $832M no-hurdle grants, $66M of founder sales before the Microsoft deal, no open-market buys ever, and a P&L carrying $558.5M of unrealized gains on its own stock. Macro is red by the desk's own rule. Every dollar of contracted ARR is matched by ~$1.45 of GPU capex, 90% borrowed on 2.5–5-year amortization against 3–4-year contracts; the equity earns the scrap value until a DC-level refinancing closes and a second hyperscaler signs Sweetwater.
Base cap $17.39B, net debt $1.7B, 394M sh; price = (rev × EV/S − net debt) ÷ shares. Bar = Red Team probability.
On-time H2–4 + 2027 500MW contracted at $20–25M/MW + DC refi ⇒ FY28 revenue $6–8B × 6x − $4B ÷ 460M.
At $44 the ten cost more than the one pays; at Zone 2 ($34.5–36.5) the base case is +15–25% and the bull +100%; at Zone 3 ($30–33) the bear case is roughly priced. That is the whole trade plan in one sentence.
Share growth +52.7% not +80%; FCF −$2.23B not −$4.2B; op income incl. impairments −$1,046.7M; GAAP EBITDA negative; backlog $16.6B / cap-backlog 1.05x; registered direct dated Dec 2025; convert repurchase share-neutral; converts $6,745.7M face; finance vs operating leases; $23.1M is options-only; top-2 66%; 68.2% institutional does not reproduce; short interest 9/15 print; three share-growth bases labeled; restricted-cash attribution flagged; ClusterMAX "worst site" headline-only; Mirantis $544M/12.6M sh; Nostrum consideration filled; $27.05 low unverified; Sentiment M1 corrected; IRR claim unverified; Section 232 unverified; Fed/10Y/VIX verified; FY27 capex verified in transcript; 10-K going-concern language depends on financing; operating AI capacity ~40MW ⇒ $11.1M/MW; Q4 ATM split inferred.
| # | Unknown | Why it matters | Where it resolves |
|---|---|---|---|
| 1 | Microsoft contract mechanics: acceptance criteria, ramp period, delay/service-credit schedule, termination rights; whether the $1.8B deferred lease revenue is refundable on non-delivery | 48% of ARR and the $3.6B IG facility hang on it | Undisclosed; watch 8-Ks and 10-Q risk language |
| 2 | Source and price of the FY27 residual ($3–8B): corporate debt vs ATM vs converts; will the board authorize ATM sales below $50 | Scenario #2 (60%) | 10-Q ATM disclosures (Nov, Feb); any 8-K raise |
| 3 | Whether any data-center-level (shell) refinancing has closed since 8/27 | The only equity-release lever; "entire portfolio unencumbered" | 8-K; Q1 FY27 call |
| 4 | 2027 contracting: MW signed vs "late-stage discussions," counterparties, $/MW, term; any Sweetwater tenant | Scenario #3; the 5GW narrative | 8-Ks; Q1–Q2 FY27 calls |
| 5 | Horizons 2–4 acceptance dates and whether Dec-quarter ">$4B operating" includes non-MSFT deployments at Mackenzie/BC | Scenario #1 (40%) | 11/5 print; 8-K |
| 6 | Contract term structure by customer (3-yr vs 5-yr share of the $4B); 2028–29 renewal $/MW; the GPU residual assumption behind "2-yr payback" | Scenario #5 | Undisclosed; first Together/Fireworks renewal terms |
| 7 | Non-IG GPU debt terms: Blue Owl amortization, DSCR covenants, parent guarantees, cross-default to converts; drawn amount of the $2.8B | Scenario #6 | 10-Q Sep-26 debt note |
| 8 | Installed GPU count and generation mix at Jun 30/today vs 150k target; Hopper impairment exposure | Residual-value risk; ramp capacity | Not in results release; 10-K body not fully parsed |
| 9 | AI Cloud segment profitability after D&A (segment EBITDA, D&A by segment) | Whether cloud margins beat lease economics — the whole premium-to-landlords question | Undisclosed; Q1 FY27 segment note |
| 10 | Customer-level concentration % in FY26/Q4 revenue and operating ARR (10-K says "concentrated," no % found in 820k chars) | Scenario #4/#7 | 10-K Item 1A / 10-Q |
| 11 | Sweetwater under SB6/Batch Zero: curtailment terms, survival of "conditional" status, $50k/MW security proposal, hyperscaler tolerance of curtailable load | Scenario #8 | PUCT rules end-2026 |
| 12 | ClusterMAX specifics: what failed (network/storage/K8s/SLA/security) and whether Microsoft acceptance coexists with it | Scenario #9; M3 | SemiAnalysis primary (not accessed) |
| 13 | Co-CEO FY26 comp table; whether 23,641,221 Form-4 shares include the unvested 9.1M; true insider % (3.3% vs 5.0% vs 12.1%); net share overhang after capped calls and prepaid forwards | Gate 6/7; dilution ceiling | FY26 DEF 14A (~Oct 2026) |
| 14 | Deferred lease revenue recognition pattern (straight-line over 5 yrs vs by MW delivered) | Whether FY27 revenue lags cash | 10-Q revenue note |
| 15 | Mining wind-down cash: ASIC sale proceeds (held-for-sale $72.5M after $110.6M write-down); remaining power obligations on ~380MW | Small cash/impairment item | 10-Q |
| 16 | PFIC determination for FY26 (10-K flags risk) | Changes the US holder base | 10-K/proxy |
| 17 | NVIDIA $2.1B purchase right vesting schedule vs the 600k-GPU milestone; exercisable in a down-tape? | Validator; dilution | Agreement not public |
| 18 | Consensus composition: which brokers' FY27 $2.8B assume Dec-quarter delivery vs the grace window | M1 sizing | Broker notes |
| 19 | Exact 11/5/26 earnings date; MarketBeat's "MS downgrade 9/2/26"; the $27.05 intraday low; cause of the 7/24–29 slide | Housekeeping | Company IR; MS note; exchange data |
| 20 | Power hedge ratio and Childress/Sweetwater $/MWh; Section 232 data-center exemption status after the 7/1 Commerce review | ~2% of AI revenue; equipment cost inflation | 10-K MD&A; Federal Register |
| 21 | Peer backlogs for WULF/APLD/CORZ; third-party GW forecast for TAM; IREN-specific Glassdoor page | Comparison quality; TAM is assumption-based | Next pass |
Contrarian edge stated (framework: "it's not a book report"): the Street models contracted ARR as revenue and ignores ops quality; the bears model GPU capex as equity dilution and ignore that it is customer- and lender-funded. Both are wrong in the same direction on timing: the P&L catches up to the ARR a year later than the bulls think and with less equity than the bears think. That makes the ladder (not spot) the trade and the Nov-5 / Feb-27 prints the referee.
Moat 7.0 < 9 ⇒ conviction-spec, ≤5% of portfolio at full ladder (PM house rule: half the ≤10% Tier-1 cap). 10–15 positions max; portfolio cash at 25–30% while the macro stamp is 🔴/🟡. Size CSPs so full assignment across all three zones ≤ the cap. No leverage. Barbell logic: life-changing if right ($70–95), survivable if wrong (−60–80% on ≤5% = −3–4% of portfolio).
2026-09-25 close $44.12. 52wk high 76.87 (2025-11-05), low 28.93 (2026-07-29). Daily SMA20 43.29 / SMA50 41.21 / SMA200 45.59; SMA50 < SMA200 (death-cross state); only 6 of last 60 closes above SMA200 → 200-day is active resistance. RSI(14) 51.1; MACD 1.308/1.117 positive, flattening. ATR(14) $2.96 = 6.7%/day. BB(20,2): 36.09 / 50.48. RV20 81%, RV60 112%. Week of 9/25: o47.85 h49.37 l43.41 c44.12, vol 179M — bearish reversal candle off the 90-day high. Fib of major swing (5.12→76.87): 0.236 59.94 · 0.382 49.46 · 0.5 41.00 · 0.618 32.53 · 0.786 20.48. 200-week MA 18.16; 50-wk 46.57, 20-wk 46.20 → $45.6–46.6 resistance cluster. Fib of the decline (76.87→28.93): 0.236 40.24 · 0.382 47.24 · 0.5 52.90 · 0.618 58.56 · 0.786 66.61. Volume profile POC 41–43. 8-week base 36–50, higher weekly lows 28.93 → 34.81 → 35.14 → 40.97; triple top 49.19/49.29/49.37; volume NOT confirming accumulation (up/down volume ratio 0.79). Breakout = weekly close > 50.5 on > 250M; breakdown = daily close < 36.
(394.06M shares, cap $17.39B — market cap/shares reconciles; dilution risk from converts means per-share floors are optimistic) On FY26 actual rev $707M: P/S 24.6 (framework "excited/euphoric"). On $1B operating ARR: 17.4x (optimistic). On $4B contracted 2026 ARR: 4.35x (aware band). Contracted ≠ delivered — floors below use $4B and must be haircut if delivery slips.
FCF-inflection note (Stage 9): first FCF-positive print is the rerate trigger; quarterly FCF was −$167M (Jun-26 qtr) after −$1.28B — with FY27 capex guided $25–30B (search result, unverified in 8-K) FCF positive is not near; the rerate catalyst is delivered-ARR, not FCF.
PM overlay: the FY27 capex figure is now verified verbatim in the Q4 transcript. Given macro 🔴, Zone 1 is CSP-only for new money (no stock); stock purchases begin in Zone 2. Red Team's modal bear ($22–29) sits below Zone 3 — that is the price of being wrong on funding sequence, and it is why the ladder halts at $28.93 instead of averaging down.
| Trigger | Action | Why |
|---|---|---|
| $52.9 | Trim 10–15% | 0.5 retrace of the decline |
| $58.6 | Trim 15–20% | 0.618 retrace |
| $66.6 | Trim 20% | 0.786 retrace |
| $76.87 | Remainder decision | ATH (double-top risk) |
| Daily close < $36 | Sell-stop for Zone-1 tranche | Invalidates base — re-buy per Zone 2/3 rules only if the thesis is intact, otherwise it is a stop |
| P/S > 8 on OPERATING (delivered) ARR | Trim | Valuation rule; >10 on delivered ARR → cut to core |
| Time rule | Halt adds and reassess | If operating ARR is not ≥ $3B by the Feb-2027 (FY27 Q2) print, or the $4B contracted number is not reaffirmed on 11/5/26 |
| Thesis breakers | Hard exit | Cancellation/re-negotiation of the Microsoft or frontier-lab contract; failure to close the $2.8B GPU financing; equity raise > 10% of shares below $40; energization slip > 2 quarters at Sweetwater/Childress; BTC-mining revenue still > 30% of total two quarters out |
| Runner (PM overlay) | Keep ~25% uncovered above $66.6 | Framework layer 4 — the bull case is $70–95 and a covered call at $70 sells it away |
| Macro trim trigger (Stage 1) | Trim to core regardless of price | VIX >17 with MOVE >100, 10Y >5.25–5.30%, HY OAS >3.25%, any hyperscaler 2027 capex cut, or an Oct-28 hike |
Liquidity: excellent. Jan-2027: put OI 312k / call OI 392k; Oct-16: 151k / 128k; Nov-20: 96k / 172k; Dec-18: 129k / 141k. Bid/ask 3–7% at $40–45 strikes; 20–60% at $30–33 near-dated (use limit orders at mid). Short interest 21.7% of float (346M float).
IV: ATM 30-DTE ≈ 73% (10/30 exp: 44P 0.734, 45P 0.737). vs RV20 81% / RV60 112% → IV is BELOW realized; premium is fair-to-cheap, not rich. IV rank n/a (no IV history from yfinance). Skew: calls priced above puts (50C IV 0.78–0.86 vs 40P 0.72–0.80) — upside-call demand, sell CCs into it. Nov-20 and later expiries straddle the 11/5 earnings.
| Zone | Strike / exp | DTE | Mid | Cash yield | Annualized | Assigned basis |
|---|---|---|---|---|---|---|
| Zone 1 | 40P 10/30 | 34d | $2.08 | 5.2% | 56% | 37.92 |
| Zone 1 PM PICK | 40P 11/20 | 55d | $3.45 | 8.6% | 57% | 36.55 |
| Zone 1 | 42P 10/30 | 34d | $2.96 | 7.0% | 76% | 39.05 |
| Zone 2 | 36P 11/20 | 55d | $1.99 | 5.5% | 37% | 34.01 |
| Zone 2 | 35P 10/30 | 34d | $0.76 | 2.2% | 23% | 34.24 |
| Zone 2 | 35P 12/18 | 83d | $2.48 | 7.1% | 31% | 32.52 |
| Zone 2 | 37P 1/15/27 | 111d | $3.95 | 10.7% | 35% | 33.05 |
| Zone 3 | 33P 11/20 | 55d | $1.21 | 3.7% | 24% | 31.79 |
| Zone 3 | 32P 1/15/27 | 111d | $2.17 | 6.8% | 22% | 29.84 |
| Zone 3 | 30P 12/18 | 83d | $1.25 | 4.2% | 18% | 28.75 |
| Zone 3 | 30P 1/15/27 | 111d | $1.81 | 6.0% | 20% | 28.20 |
| Band | Strike / exp | Mid | Yield | Annualized |
|---|---|---|---|---|
| 52.9 | 50C 11/20 | $3.78 | 8.6% | 57% |
| 52.9 | 50C 12/18 | $4.95 | 11.2% | 49% |
| 52.9 | 52C 10/30 | $1.72 | 3.9% | 42% |
| 58.6 | 55C 11/20 | $2.58 | 5.8% | 39% |
| 58.6 | 55C 12/18 | $3.73 | 8.4% | 37% |
| 58.6 | 60C 11/20 | $1.77 | 4.0% | 27% |
| 58.6 | 60C 12/18 | $2.81 | 6.4% | 28% |
| 58.6 | 60C 1/15 | $3.80 | 8.6% | 28% |
| 66.6+ | 65C 1/15 | $3.06 | 6.9% | 23% |
| 66.6+ | 70C 12/18 | $1.70 | 3.9% | 17% |
| 66.6+ | 70C 1/15 | $2.51 | 5.7% | 19% |
Wheel sizing note: 6.7% daily ATR on a 21.7%-short name = assignment gaps are the norm; size CSPs so full assignment across all three zones ≤ the position cap (Stage 10: ≤10% per name; moat/conviction-spec sizing unless the moat grade clears 9).
Preferred first tranche = 40P 11/20 ($3.45, basis $36.55, straddles earnings — you are paid for the event and assigned into Zone 2 if it disappoints). Do not sell Zone-3 puts before 11/5; the ladder halts at $28.93 and a 30P assigned into a broken thesis is a stop, not a wheel. Covered calls only above $52.9 on the trimmable 75%; keep the runner naked.
| # | Risk | Mechanism |
|---|---|---|
| 1 | Funding sequence | $13.6B due ≤12 months, $25–30B FY27 capex, $3–8B residual, marginal non-IG cost 9%, management's ~$50 issuance floor vs a $44 stock, converts OTM. Kills via a punitive raise or a capex cut that strands 2027 capacity. |
| 2 | Microsoft delivery/acceptance on H2–4 | Grace already in use; credits/termination rights unquantified; 48% of ARR + IG facility. |
| 3 | 2027 uncontracted 500MW IT pricing | Scarcity pricing ($20–25M/MW) vs B200 commoditization H1-27; hyperscaler 2027 capex is the swing. |
| 4 | Macro / cost of capital | Fed hiking, 10Y 5.18%, CCC widening; each +100bp on ~$10–15B of new GPU debt ≈ $100–150M/yr. |
| 5 | Operational quality at scale | ClusterMAX bottom tier, 3x headcount, Mirantis integration, SLA credits. |
| 6 | GPU residual vs contract term vs debt tenor | 30-month amortization, 4-yr contracts, 5-yr depreciation, VR200 on order. |
| 7 | Concentration | MSFT 48% / top-2 66% / Childress = all hyperscaler capacity through 2026 / NVIDIA is supplier + customer + optional shareholder. |
| 8 | Dilution machinery | $3.5B ATM, 127M convert shares, 18.2M CEO RSUs, 30M NVIDIA, M&A in stock. |
| 9 | ERCOT SB6 / Batch Zero | Curtailable post-cutoff loads; Sweetwater tenant risk. |
| 10 | Governance | No-hurdle grants, founder sales, dual-class; a say-on-pay revolt would be noise, a founder exit a breaker. |
| 11 | Residual BTC beta through Dec-2026 | 49% of Q4 revenue; tape still trades N3. |
| Validator | Check by | Passes if |
|---|---|---|
| Horizon 2 accepted by Microsoft | 2026-11-05 (8-K or print) | "Delivered and accepted" language, no credits |
| Q1 FY27 AI Cloud revenue | 2026-11-05 | ≥$190M (Red Team's ≥ threshold); ≥$150M acceptable with H2–4 "on track" |
| Contracted $4B reaffirmed; operating ARR ≥$3B | 2026-11-05 / Feb-2027 print | Reaffirmed; ≥$3B by Feb-27 |
| First DC-level financing closes | 2027-03-31 | ≥$1B at ≤7% on Horizon shells |
| First 2027 capacity contract | 2027-03-31 | ≥100MW IT at ≥$18M/MW, ≥3-yr term |
| Sweetwater anchor tenant | 2027-06-30 | ≥300MW IT signed, any hyperscaler or IG counterparty |
| ATM discipline | Each 10-Q | Zero sales below $50 (management refusing to fund at these prices is bullish for price) |
| ClusterMAX / ops evidence | Next SemiAnalysis update | Tier upgrade, or H2–4 accepted with no disclosed credits |
| Adj. EBITDA margin recovery | Q3 FY27 (May-2027) | Back >35% on ARR ramp |
| NVIDIA exercises any part of the $70 right | 2027-12-31 | Any exercise = smart-money confirmation |
| Breaker | Check by |
|---|---|
| Microsoft or frontier-lab contract cancelled/renegotiated, or credits disclosed | continuous (8-K) |
| H2–4 slip beyond the grace window (past early Q2 CY27) | 2027-05-08 (Q3 FY27 print) |
| Equity raise >10% of shares below $40 | continuous |
| Failure to draw/close the $2.8B non-IG financing, or a new facility >12% | 2026-12-31 (Blue Owl availability end) |
| Energization/delivery slip >2 quarters at Sweetwater 300MW (Q4 2027 target) | 2028-06-30 |
| Mining still >30% of revenue two quarters out | Feb-2027 print |
| Founder departure | continuous |
| Macro: VIX >17 with MOVE >100 / 10Y >5.30% / HY OAS >3.25% / hyperscaler 2027 capex cut | weekly EOD report |
v3.0 — 2026-09-26 (Liquid Wheel Research · deep-dive team)
Score 7.0 → 6.5; status INTACT — ON WATCH kept; macro stamp 🔴 initiate / 🟡 hold; moat 7.5 → 7.0; gates 7P/5F/4W → 7P/6F/3W; P/S-TTM trim rule retired for EV/operating-ARR and cap/backlog bands; entry ladder re-cut to $39.0–41.5 / $34.5–36.5 / $30–33 with no-chase $49.50; scenarios rebuilt on 430–470M shares and EV/S; Red Team audit applied (30 items); next review 2026-11-05 (Q1 FY27). Six lanes merged (Quant, Qual, Macro, Sentiment, Technicals, Tracker) plus an independent Red Team that pulled the full 10-K text and XBRL and audited the other five. Conflicts between lanes flagged in-line as CONFLICT; unverified figures as UNVERIFIED; Red Team corrections applied marked RT-FIX. Framework: Stages 0–12 of investing-framework.md; the 16 gates are the v2 checklist carried forward so grades are comparable. Design v3: sticky verdict bar, 30-second TL;DR strip, gate-score ring, v2→v3 diff panel, scenario range bar, downside-scenarios grid, known-unknowns ledger, competitor table, KPI scorecards with check-by dates, sources drawer, staleness banner, mobile jump nav. Data as of 2026-09-25.
v2.1 — 2026-07-12 (design pass)
Company 101 section added. Trade Plan tab: live candlestick chart (weekly, 14 months, real OHLC) with Bollinger(20,2), SMA 20/50/200 and the entry/trim zones painted on price. Animated Liquid Wheel replaces the flywheel boxes. Tables now stack into readable cards on phones.
v2.0 — 2026-07-12 (Claude · Liquid Wheel Research)
Full framework rebuild: skin-in-the-game score, 16-gate checklist, earnings tracker, entry ladder + exit plan, dilution & debt gates, flywheel map, dilution-adjusted scenarios. Data refreshed to Q3 FY26 filings + Jul 10 close. Thesis reframed: mining is no longer the floor — it's the fuel being burned for the AI conversion. Status: INTACT — ON WATCH. Design v2: dashboard skin matching the macro-regime board.
v1.0 — 2026-01-31 (Tony 🦞)
Original analysis: dual-engine thesis, Microsoft catalyst, 7.5/10. Price $53.74.
Next scheduled review: 2026-11-05 (Q1 FY27 print date UNVERIFIED).