Revenue fell two straight quarters on purpose — ASICs decommissioned ($140M impairment) so the same cheap power feeds GPUs for Microsoft ($9.7B/5yr) and NVIDIA ($3.4B/5yr). At 19.4x trailing sales it looks expensive; at ~4x the contracted CY26 run-rate it's cheap — if they deliver. What it is not: a Tier-1 core hold. Moat ~7.5 (below the 9 gate), FCF deeply negative, debt > cash, share count +52% YoY. High-conviction speculative allocation — earned through research, sized with respect, wheeled for premium (beta 4.28 pays well).
IREN owns cheap renewable electricity and sells it as compute. Founded in Sydney by brothers Daniel & Will Roberts (ex-Macquarie infrastructure investors), the company buys land and locks up hydro/wind power where it's abundant and cheap — British Columbia and West Texas — then builds its own data centers on top. The product has changed twice (Bitcoin hashrate → GPU cloud → hyperscaler colocation) but the business never has: convert ~3.3¢/kWh electrons into the most valuable compute the market will pay for. Fully vertically integrated — they own the land, the power contracts, the buildings, and the machines.
| Year | Milestone |
|---|---|
| 2018 | Founded as Iris Energy — thesis: monetize stranded renewable power with Bitcoin mining |
| 2019–21 | First hydro-powered sites in British Columbia (Canal Flats, Mackenzie, Prince George) |
| Nov 2021 | Nasdaq IPO (~$28) near the crypto top — then a >95% drawdown to ~$1 in the 2022 winter. The company survived; most peers didn't |
| 2023 | Childress, TX campus energized — the 750MW power-bank strategy goes big |
| 2024 | Renamed IREN · first NVIDIA GPU cluster, AI Cloud launched · hashrate scaled toward 50 EH/s (top-tier miner) |
| Nov 2025 | Microsoft signs $9.7B / 5-yr AI contract with ~$1.9B prepay — the pivot gets its anchor |
| 2026 | NVIDIA $3.4B contract + partnership · mining decommissioning begins · 5GW power pipeline secured |
The 2022 near-death experience is load-bearing context: this management has already operated through a 95% drawdown without losing the asset base. That's the résumé line that matters for the current build-out.
The wheel only spins if step 4 outpaces the cost of step 5. Today it doesn't yet — that's the FCF fail. Every Horizon handoff moves it closer. Watch the wheel, not the quarter.
| Item | v1.0 · Jan 31 | Now · Jul 12 | Read |
|---|---|---|---|
| Price | $53.74 | $41.14 (−23%) | Descending channel, below all major SMAs |
| Thesis shape | "Miner with AI kicker; mining = cash-flow floor" | Mining being decommissioned for AI conversion | The "BTC floor" pillar retired — by the company |
| NVIDIA | Hardware supplier | $3.4B/5yr contract + right to buy 30M sh @ $70 | Second anchor; $70 = smart money's strike |
| Contracted ARR | — | $3.1B | The de-risking mechanism |
| Debt | $964M | $3.96B | Leverage tripled to fund build-out |
| GPUs | 23K → "140K by '26" | ~74K deployed | Guidance now in MW: 480MW CY26 fully contracted |
| Power secured | 2,910 MW | 5 GW | The real asset: the power bank |
| Analysts | $69–72 | ~$62–81 · Jefferies $79 | Wide dispersion — execution stock |
No hand-waves. A great setup rarely passes everything; the discipline is knowing exactly which boxes fail and why you're proceeding anyway.
The fails cluster in one place — the balance-sheet cost of building ahead of revenue. That's either the operating-leverage story playing out, or the risk. It cannot be both forever; Aug 27 starts answering which.
| Quarter | Period end | Revenue | Net income | EPS dil. | FCF | Dil. shares |
|---|---|---|---|---|---|---|
| Q4 FY24 | Jun 2024 | $57.4M | −$27.0M | −$0.21 | −$292M | 129M |
| Q1 FY25 | Sep 2024 | $52.8M | −$51.7M | −$0.27 | −$387M | 189M |
| Q2 FY25 | Dec 2024 | $116.1M | −$21.9M | −$0.10 | −$247M | 210M |
| Q3 FY25 | Mar 2025 | $144.8M | −$16.1M | −$0.07 | −$492M | 219M |
| Q4 FY25 | Jun 2025 | $187.3M | +$176.9M | +$0.66 | −$255M | 274M |
| Q1 FY26 | Sep 2025 | $240.3M | +$384.6M | +$1.08 | −$138M | 362M |
| Q2 FY26 | Dec 2025 | $184.7M | −$155.4M | −$0.52 | −$648M | 298M |
| Q3 FY26 | Mar 2026 | $144.8M | −$247.8M | −$0.74 | −$1,280M | 334M |
Q3 FY26 loss drivers: $140.4M non-cash impairments (mining decommissioning) + $23.7M unrealized capped-call losses. Adj. EBITDA was positive $59.5M (41% margin). Cash $2.6B at Apr 30. Sources: company Q3 FY26 release (May 7, 2026); StockAnalysis filings compilation.
Read both ways: it passes the 5–10 bar, but only because FY25 comp was enormous ($72.6M each; 993:1 CEO-to-median ratio — the chairman publicly defended it). Strip equity grants and measure against cash salary, and the score is in the hundreds. The truth: these founders are overwhelmingly paid in their own stock, most of it locked — the June 2026 package (9.1M RSUs each) vests 4 years plus a 2-year hold reaching FY2033. Incentives point long; the dilution bill is real (that grant alone = 5% of shares).
| Trait | Score | Evidence |
|---|---|---|
| Capital allocation | 8/10 | Power bank bought early and cheap; pivot timed to AI demand; funded by heavy dilution |
| Long-term focus | 9/10 | Tore down a profitable mining business mid-bull-market to build something bigger |
| Shareholder friendly | 5/10 | +52% YoY shares, $700M RSU package, 993:1 ratio. Offset: 6-yr lock, no hedging/pledging allowed |
| Execution track record | 8/10 | 2018 startup → $14.7B; 480MW CY26 "on track"; Horizon 1 handoff due this quarter |
| Insider conviction | 7/10 | 13.6% insider ownership; $66.4M sold Sept '25 (≈11% trim) — monitor, don't panic |
Management 7.4/10. Builders with skin in the game who pay themselves like champions before the shareholders have won. Structural traits (talent density, decentralization) unscored — revisit via Glassdoor/interviews next review.
| Promise / target | Made | Status · Jul 2026 | Grade |
|---|---|---|---|
| "AI Cloud ARR $500M+ by early CY26" (v1.0 validator) | 2025 | Contracts signed far beyond ($3.1B) — but deployed AI revenue $33.6M/qtr (~$134M run-rate). Signed ahead, converted behind. | SPLIT |
| 140K GPUs by end 2026 | Q2 FY26 | ~74K as of May; guidance reframed to MW (480MW CY26, fully contracted, on track) | BEHIND |
| Maintain 50 EH/s hashrate | 2025 | Abandoned by design — ASICs decommissioned ($140M impairment) | PIVOT |
| Horizon 1 handoff to Microsoft Q4 FY26 | Q3 FY26 | On track per May 7 update; GB300 installs underway at Childress | ON TRACK |
| Q3 FY26 street EPS | — | −$0.74 actual vs −$0.22 expected (non-cash driven) | MISS |
| 85% EBITDA margin on AI contracts | Q2 FY26 | Unproven at scale; blended adj. margin 41% mid-transition | PENDING |
| $3.7B ARR by end CY2026 | Q3 FY26 | $3.1B contracted (MSFT 1.9 + NVDA 0.7 + Prince George 0.5); needs deployment, not sales | THE BIG ONE |
| Lens | Multiple | Band |
|---|---|---|
| P/S on trailing revenue ($757M) | 19.4x | "Excited" — pricing the future |
| EV (~$16.1B) / contracted ARR ($3.1B) | ~5.2x | "Market aware" |
| Mkt cap / CY26 ARR target ($3.7B) | ~4.0x | "Market aware" — cheap if delivered |
| P/B | 5.3x | Book $7.81/sh — hard-asset floor far below |
All per-share scenarios assume ~400M fully-diluted shares by 2027 (357M + 18M RSUs + partial NVIDIA warrant) — dilution-adjusted, unlike most street math.
| KPI | Why it's the signal | Current → target |
|---|---|---|
| Contracted-ARR conversion rate | The entire thesis is contracts becoming revenue | $134M AI run-rate → $3.7B ARR CY26-end |
| MW energized & handed off | Physical proof; each Horizon = a revenue switch-on | 480MW CY26 → 1,210MW '27 |
| OCF / CapEx ratio | The flywheel-closing metric — crosses 1, dilution ends | Deeply <1 → watch trend |
| Share-count growth rate | Your ownership decays at this speed | +52% YoY → must decelerate |
| AI EBITDA margin at scale | Proves the 85% contract math | 41% blended → 60%+ as AI mix grows |
No-chase rule: no adds above the 50-day (~$54) unless a new contract materially raises contracted ARR. Zones derive from indicator levels + valuation floors — verify on the chart before orders. Wheel-native entry: selling cash-secured puts at Zone 2–3 strikes = getting paid to wait for your price.
| Trigger | Action | Why |
|---|---|---|
| $70–77 | Trim 20–25% | NVIDIA's strike ($70) + prior high ($76.87); recoup basis, ride the rest |
| $95–105 | Trim to core (≤50% of max) | Bull case realized; P/S stretched unless ARR raised again |
| P/S (TTM) > 25 | Trim regardless of price | Euphoric band — protects from narrative highs |
| MSFT termination / Horizon slips 2+ qtrs | Full exit | The contract IS the thesis |
| Punitive financing | Full exit | Means contracted cash isn't arriving; flywheel broken |
| Founders depart | Full exit | Founder-led is a checklist pillar |
| Covered calls (holders) | Sell strength $50+, 21–45 DTE above basis | Beta 4.28 + 24.6% short interest = rich recurring premium |
| # | Risk | Mechanism |
|---|---|---|
| 1 | Funding gap | Q3 burn $1.28B vs $2.6B cash ≈ two quarters of runway at peak CapEx without new capital. Prepays + debt + ATM bridge it — until markets say no. The risk that compounds all others. |
| 2 | Microsoft execution | Terminable on missed milestones; phased handoffs through 2027 = repeated single points of failure. Horizon 1 (this quarter) is the first real test. |
| 3 | Dilution treadmill | +52% YoY; even the bull case can disappoint per-share if the count keeps compounding. |
| 4 | Compute oversupply | Meta builds its own; CoreWeave/Nebius/Lambda racing. By 2027–28 GPU-hours could commoditize — contracts protect 5 years, not forever. |
| 5 | Transitional BTC dependency | Mining ($111M/qtr) still funds the pivot. BTC <$40K (now ~$64K) squeezes the bridge at peak CapEx. |
| 6 | Hardware obsolescence | 2026 GB300s vs whatever ships in 2028 — permanent CapEx refresh treadmill. |
| 7 | Violence of the tape | Beta 4.28, 24.6% short interest: ±15% weeks are normal. Size for them. |
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: Aug 27, 2026 (Q4 FY26 + full-year results).