I — knowledge base

Overview

IREN is executing a rapid pivot from BTC mining toward AI cloud / neocloud infrastructure, targeting multi-billion ARR by end-2026 via staged GPU cluster deliveries (PG + Horizons H1–H4), greenfield power advantages, and enterprise/frontier-lab contracts priced at parity with peers (NBIS, CRWV). Q2/Q4 FY26 results mixed on near-term AI revenue timing but validated $1B operating ARR mid-Q3 CY2026, ~87% AI gross margins, non-dilutive GPU financing covering the bulk of large CapEx, and a path to 65%+ EBITDA margins as AI mix scales. 2026 build plan: ~480 MW gross / 300 MW IT load supporting $4B contracted ARR (Mackenzie + air-cooled Childress included).

Key facts & figures

AI Cloud revenue Q4 FY26 (quarter ended 30 Jun 2026): $70.5M / ~$71M (missed prior $85–110M expectations due to PG GPU billing delay despite Q1 RPO). [[s:138@00:00:23]]
Operating ARR: exited Q4 ~$0.5B; reached $1B as of 26 Aug 2026 (mid-Q3) after Horizon 1 acceptance; PG supports ~$500M ARR (or ~$125M/quarter) without further GB300s; H1 adds ~$500M. [[s:138@00:00:56]] [[s:138@00:01:48]]
2026 target: $4B contracted ARR operational by year-end from 2026 capacity (largely sold out), including Mackenzie and air-cooled Childress. [[s:138@00:01:13]]
2026 build-out: ~480 MW total/gross load or ~300 MW IT. [[s:138@00:01:30]]
Delivery cadence: Horizon 1 (H1) delivered/accepted mid-Aug 2026 (Microsoft); H2 first half Oct; H3 late Oct/early Nov; H4 early Dec; PG air-cooled fleet fully commissioned. [[s:138@00:00:40]]
Q3 AI revenue guide context: ~$187M with H1 only half-quarter contribution.
Enterprise/mid-term pricing: >$20M revenue per MW IT load on recent 3-year contracts; active discussions ~$25M/MW — matches midterm bare-metal enterprise deals at NBIS/CRWV. [[s:138@00:03:01]]
Customers/logos: Cohere, Prometheus, Perplexity, Figure AI, Fal AI, Higgsfield AI, Hume AI, Flexity, plus multi-year contract with leading (undisclosed) frontier AI lab. [[s:138@00:02:36]]
Financing: Microsoft/HS-backed GPU financing ~6% blended; enterprise/non-IG (e.g. Mackenzie) ~9% ($2.4B of $2.8B at 9% fixed). Closed $2.8B GPU financings funding ~90% of associated GPU capex (with prepayments); prefers debt over converts to limit dilution. [[s:138@00:04:09]] [[s:138@00:04:24]]
FY27 CapEx guide: ~$25–30B; ~$22B expected covered by existing cash + committed GPU financing/prepayments (~$14B) plus targeted additional ~$8B GPU financing/prepayments (non-dilutive). [[s:138@00:05:38]]
Margins: AI gross margin ~87% (Q4 FY26: $70.5M rev vs $9.2M COGS, ex-D&A; full-year AI similarly ~87%). Adj. EBITDA moving from ~40% BTC-mining levels toward 65%+ with rising AI mix (project-level AI e.g. ~85% MSFT); note recent reported overall adj. EBITDA margins were lower (~41% Q3, 14% Q4, 35% FY26) amid transition/impairments. [[s:138@00:06:22]] [[s:138@00:06:29]]
Other: 3x FTE growth viewed as positive for multi-year build; SW1 DSX Nvidia campus collab expected next quarter.

Thesis & bull case

Execution + contracted ramp: Mid-Q3 already at $1B ARR; staged H2–H4 through early Dec plus Mackenzie/Childress support $4B ARR by end-2026 on 300 MW IT; PG can exceed $500M+ ARR with more GB300s. Greenfield power cost advantage underpins expanding margins.
Pricing parity / no tech gap: $20–25M/MW enterprise wins demonstrate IREN competes with NBIS/CRWV on bare-metal midterm deals; strong AI-native customers avoid walled-garden SaaS. [[s:138@00:03:24]]
Capital structure: Debt + GPU financing + customer prepayments blunt equity dilution on very large CapEx; ~$22B of $25–30B FY27 needs covered non-dilutively; MSFT-backed paper at ~6% vs enterprise ~9%. Convertibles/debt/prepayments are interchangeable neocloud tools with tradeoffs — IREN emphasizing lower-dilution path. [[s:138@00:04:49]]
Margin expansion: 87% AI gross margins; mix shift from BTC drives adj. EBITDA from ~40% toward 65%+ (and higher as AI dominates). High project-level margins (e.g. MSFT) support durability.
Demand validation: Named enterprise + frontier-lab multi-year deals; volume delivery and expected Nvidia SW1 campus collab as near-term catalysts. 3x headcount growth aligns with multi-year build scale.

Risks & bear case

Near-term revenue timing: AI revenue miss ($71M vs $85–110M) from delayed PG GPU billing despite prior RPO highlights recognition/lag risk even when hardware is in.
CapEx scale & financing execution: $25–30B FY27 CapEx requires successful additional ~$8B GPU financing/prepayments on top of committed amounts; any disruption increases dilution or slows build.
Margin bridge: Reported consolidated adj. EBITDA margins have been volatile/lower (teens to ~40%) during transition; 65%+ at scale (or at 50/50 mix) remains forward-looking and sensitive to AI ramp, impairments, and cost control. [[s:138@00:06:29]]
Delivery concentration: H2–H4 and remaining 2026 MW must hit Oct–Dec windows to realize $4B ARR exit; customer concentration (MSFT Horizons + enterprise cohort) and power/grid execution remain dependencies.
Competitive/pricing: Parity at $20–25M/MW is positive but leaves limited differentiation if peers match power/cost or if enterprise demand softens.

Timeline of developments

2026-06-30: Q4 FY26 / Q2 CY26 AI Cloud revenue $70.5M; exited quarter ~$0.5B ARR (PG/prior). [[s:138]]
Mid-Aug 2026: Horizon 1 delivered/accepted by Microsoft; PG air-cooled fleet fully commissioned. [[s:138@00:00:40]]
2026-08-26: $1B operating ARR reached after H1. [[s:138@00:00:56]]
2026-08-28 (space): Review of mixed Q2 print; reiteration of 480 MW / 300 MW IT 2026 plan, $4B ARR target, $20–25M/MW pricing, $2.8B GPU financings, ~$22B non-dilutive coverage of FY27 CapEx, margin trajectory, and H2–H4 delivery schedule (H2 1H Oct, H3 late Oct/early Nov, H4 early Dec). [[s:138]]
Expected near-term: H2–H4 deliveries Oct–early Dec 2026; SW1 DSX Nvidia campus collab announcement targeted next quarter; continued enterprise ramp and additional GPU financing.

Open questions

Exact timing and magnitude of remaining FY27 ~$8B GPU financing/prepayments and any residual equity need.
Sustainability of $25M/MW discussions converting to signed contracts and duration mix (3-year vs longer).
Realized consolidated adj. EBITDA margin path as AI mix rises (bridge from recent 14–41% prints to 65%+ target).
Identity and scale of the undisclosed frontier AI lab multi-year deal; further logo additions and utilization rates on delivered MW.
Mackenzie and air-cooled Childress exact contribution timing within the $4B 2026 ARR and any power/interconnection contingencies.
Impact of 3x FTE growth on opex vs. execution speed; details of forthcoming Nvidia SW1 campus collaboration.

Notable predictions to track

$4B ARR operational by end-2026 from contracted 2026 capacity (Mackenzie + Childress included). [[s:138@00:01:13]]
H2 first-half Oct, H3 late Oct/early Nov, H4 early Dec 2026 deliveries. [[s:138@00:00:56]]
Q3 AI revenue around $187M (H1 partial quarter).
Adj. EBITDA margins expanding toward 65%+ as AI mix rises (from ~40% BTC baseline). [[s:138@00:06:29]]
~$22B of $25–30B FY27 CapEx met via GPU financing + prepayments (non-dilutive). [[s:138@00:05:38]]
SW1 DSX Nvidia campus collab expected in the next quarter.
PG able to support $500M+ ARR after additional GB300s; pricing discussions converting at ~$25M/MW.