Iris Energy Limited -- How the Business Works
Iris Energy is an $11.5B Australian-listed digital infrastructure company operating a
dual business model: Bitcoin mining and AI/HPC cloud services. Founded in 2018, IREN
controls 4.5 GW of secured power capacity -- one of the largest portfolios among
pure-play AI infrastructure companies. The company operates 50 EH/s of Bitcoin mining
hashrate (~6% of global network) while rapidly scaling a GPU cloud platform anchored
by a $9.7B Microsoft contract. Revenue has grown from ~$57M (Q1 FY25) to $240M
(Q1 FY26, +355% YoY), though Q2 FY26 dipped to $185M as mining revenue fell during
the AI transition. IREN has raised $9.2B in FY26 YTD via convertibles and GPU financing
to fund 140,000 GPU deployments by end of CY2026. TTM EPS $0.87 (noisy with derivative
mark-to-market). FYE June 30. Composite score 5.6/10 (Watchlist -- borderline HOLD/AVOID).
Price / Composite Score
$34.77 / 5.6
Watchlist -- borderline HOLD/AVOID
Q1 FY26 Revenue (Peak)
$240M
+355% YoY | Q2 FY26 dipped to $185M
Secured Power
4.5 GW
810 MW operating | 4-7 years to replicate
Short Interest / Beta
17.3% / 4.31
Extreme volatility | -55% from 52-wk high
Dual business model -- Bitcoin mining funding the AI/HPC pivot
Revenue Streams -- BTC Mining + AI Cloud + AI Colocation
Pillar 1: Bitcoin Mining
50 EH/s Hashrate
~6% of global network hashrate (~800+ EH/s)
All-in cash cost $36K-$41K vs $93K-$99K realized BTC price
Hardware-level margins 75%+ | Majority of current cash flow
Commodity business -- no pricing power, halving risk
Pillar 2: AI/GPU Cloud (Bare Metal)
140K GPUs Target
NVIDIA H100/H200/GB300 fleet | NVIDIA Preferred Partner
AI Cloud rev: $3.2M (Q1 FY25) to $7M (Q4 FY25) -- still ramping
Hardware-level margins >95% (rev less electricity)
Targeting largest GPU fleet in North America by end 2026
Pillar 3: AI Colocation (Data Center)
$9.7B Microsoft
Anchor contract | 85% projected EBITDA margin
Horizons 1-4 at Childress, TX for Microsoft AI workloads
ARR target $3.4B by end CY2026 (~$2.3B contracted today)
Revenue recognition starting Q2 CY2026 (now)
Revenue Evolution -- Quarterly Trajectory (FYE June 30)
Q1 FY25
~$57M
Q2 FY25
~$117M
Q3 FY25
$148M
Q4 FY25
$187M
Q1 FY26
$240M
+355% YoY
Q2 FY26
$185M
-23% QoQ
Vertical Integration -- Transmission Line to Compute
4.5 GW Secured Power
Grid connections, substations, permits
4-7 year lead time to replicate
→
Data Center Build
810 MW operating capacity
Horizons 1-4, Sweetwater, Prince George
→
GPU Fleet Deployment
NVIDIA H100/H200/GB300
140K GPUs target by end CY2026
→
Contracted Revenue
$9.7B Microsoft + additional contracts
$3.4B ARR target by end CY2026
IREN operates a dual-engine model: Bitcoin mining generates near-term cash flow
while the AI/HPC business scales. Unlike peers winding down mining (CORZ),
IREN maintains 50 EH/s of hashrate as a cash-generating bridge. The AI pivot centers
on two models: (1) bare-metal GPU cloud for sophisticated customers (AI labs, enterprises)
and (2) colocation hosting for hyperscalers (Microsoft anchor). The $9.7B Microsoft
contract validates the platform, but the AI cloud business is still nascent at ~$7M/quarter.
The core differentiator is the 4.5 GW power portfolio -- genuinely scarce in a market
where new grid connections take 4-7 years. Revenue growth has been exceptional (+355% YoY
at peak), though Q2 FY26 showed the transition gap as mining revenue fell while AI
colocation had not yet ramped. The next 12 months are critical: Microsoft revenue
recognition begins Q2 CY2026, 140K GPUs must deploy on schedule, and the $3.4B ARR
target must prove achievable. Total capital raised in FY26 YTD: $9.2B.
Revenue and operational data from IREN quarterly reports (Q1 FY25 through Q2 FY26), earnings call transcripts, and investor presentations. Microsoft contract from public filings.
Infrastructure portfolio -- geographic diversification across North America
Site Portfolio -- Power Capacity and Status
| Site | Location | Capacity | Use Case | Status |
|---|---|---|---|---|
| Prince George | British Columbia, Canada | ~50 MW | BTC Mining + AI Cloud | Operating |
| Childress (Horizons 1-4) | Texas | ~760 MW | Microsoft AI Colocation | Building -- Rev starting Q2 CY2026 |
| Sweetwater | Texas | ~600 MW | AI Cloud / Colocation | Energization Q2 CY2026 |
| Oklahoma | Oklahoma | ~1,600 MW | Future expansion | Energization ~2028 |
| Other / Pipeline | Various | ~1,500 MW | Prospective capacity | Various stages |
4.5 GW
Total Secured Power
Among largest pure-play portfolios
810 MW
Operating Capacity
Across existing data centers
2,000+
Team Size
Scaled from startup in 3 years
$9.2B
Capital Raised FY26 YTD
Convertibles + GPU financing + prepayments
Infrastructure data from IREN investor presentations, 10-K/10-Q filings, and Q2 FY26 earnings call commentary. Power capacity figures from management guidance.
Competitive landscape -- no oligopoly in either segment
Competitive Position Assessment (Oligopoly Gate: FAIL -- Capped at 5/10)
| Competitor | Type | AI/HPC Exposure | Scale / Moat |
|---|---|---|---|
| BTC Miner Pivots (Direct Peers) | |||
| Core Scientific (CORZ) | BTC miner / AI colo | $10B+ CoreWeave deal, 1.2GW+ power | Post-bankruptcy pivot, similar strategy |
| RIOT Platforms | BTC miner / exploring AI | Building AI-ready facilities | Large power capacity, earlier in AI transition |
| MARA Holdings | BTC miner | Primarily BTC-focused | Largest hashrate, less AI pivot progress |
| Neo-Cloud / GPU Platforms | |||
| CoreWeave | GPU cloud provider | Pure-play GPU cloud at scale | NVIDIA-backed, hyperscaler contracts |
| Lambda / Crusoe | GPU cloud providers | AI-focused cloud services | Niche competitors, smaller scale |
| Hyperscalers (Building Own Capacity) | |||
| MSFT / GOOG / AMZN / META | Hyperscale cloud | $50B+ each in annual AI capex | Unlimited capital, custom silicon (TPUs, Trainium, Maia) |
~6%
BTC Mining Share
50 EH/s of ~800+ EH/s global hashrate
Fragmented
AI Data Center Market
Hundreds of developers competing for power
$5T+
AI Infra TAM (5-Year)
44 GW to 156 GW demand by 2030
No Moat
Software Differentiation
Bare metal is commodity-like, no lock-in
Oligopoly gate: FAIL. Neither Bitcoin mining nor AI cloud offers defensible
market position. Bitcoin mining is a commodity business -- IREN holds ~6% of
global hashrate with no pricing power. The AI data center market is highly fragmented
with hundreds of developers competing for power, land, and customers. IREN differentiates
on vertical integration (substation to GPU) and the 4.5 GW power portfolio, but these
advantages are replicable given sufficient capital and time. The Microsoft contract
validates credibility, but the bare-metal cloud offering has no software lock-in --
management themselves acknowledge software will be commoditized. Hyperscalers building
custom silicon (TPUs, Trainium, Maia) could reduce demand for NVIDIA GPU cloud over time.
Thematic score capped at 5/10 per rubric.
Competitive data from company filings, earnings calls, and industry research. Hashrate data from blockchain analytics. TAM estimates from industry reports.
Key risks to the business model
| Risk | Timeframe | Severity | Detail |
|---|---|---|---|
| Execution on Massive Build-Out | Near-term | Critical | Must deliver 140K GPUs + Horizons 1-4 + Sweetwater simultaneously by end CY2026 -- a 10x scale-up from current operations |
| Bitcoin Price Dependency | Ongoing | High | Mining still generates majority of operating cash flow. At $60K BTC, adj. EBITDA drops from ~$588M to ~$270M. Beta of 4.31 reflects crypto correlation |
| Dilution / Capital Structure | Structural | High | $3.7B+ convertible notes, $3.6B GPU financing, multiple ATM raises. Shares ~200M to ~330M+ in 18 months (~50%+ dilution) |
| Microsoft Concentration | Structural | Moderate-High | $1.9B of $3.4B ARR target (56%) from Microsoft. Remaining $1.5B ARR depends on contracts still in negotiation |
| Technology Obsolescence | Medium-term | Moderate | $5.8B in GB300 GPUs could lose value rapidly. NVIDIA 2-year generational cycles. Custom silicon (TPUs, Trainium) could reduce demand |
| Regulatory / Power Market | Medium-term | Moderate | ERCOT batching for Sweetwater not confirmed. Oklahoma 1.6 GW not energized until 2028. Tariff risk on imported equipment |
| Competition / No Software Moat | Structural | Moderate | Bare metal has no lock-in. CoreWeave, hyperscalers, and other miners (RIOT, MARA) all building competing AI capacity |
Risk assessment from IREN 10-K/10-Q, earnings calls, and analyst commentary. Capital structure data from company filings and convertible note prospectuses.