Milestone Hit-Rate
78%
Hit on Restated Targets
Key Snapshot
52 EH/s
Operating Hashrate (self-mining)
750 MW
Childress TX Max Capacity
$9.7B
Microsoft AI Cloud Contract (5-yr)
$1.94B
Annualized Revenue Run-Rate (AI)
140K
GPUs Targeted by End CY2026
$501M
FY25 Revenue (+168% YoY)
100%
Renewable-Powered (net-zero Scope 1–2)
<$0.03
Blended Power Cost ($/kWh)
The Pivot: Bitcoin Miner → AI Cloud Operator (2024–2026)
IREN started life as Iris Energy — a pure Bitcoin miner listed on NASDAQ in Nov 2021. First H100 pilot Q1 2024. Rebranded to IREN Nov 2024. Struck $9.7B Microsoft AI hosting deal Nov 2025 — the single largest contract ever announced by a Bitcoin-era miner. Today, more than half of forward revenue is AI cloud, and the story has flipped from BTC price leverage to GPU yield + hyperscaler concentration.
Hashrate & Capacity Milestones
Announced Dec 2022 · Target FY23 end
10 EH/s by Jun 2023 ❌
Only 5.6 EH/s achieved. First big credibility hit — caused by miner delivery delays and BC power curtailment. Target slipped 6+ months. Management acknowledged in Sept 2023 FY23 results.
Restated target · Mar 2024
10 EH/s by Apr 2024 ✓
Hit on restated schedule after Childress Phase 1 (75 MW) went live Oct 2023. Rebuilt some credibility.
Announced Q4 FY24 · Target Dec 2024
30 EH/s by year-end 2024 ✓
Reached 30.4 EH/s end Dec 2024. Childress Phase 2 (225 MW) + Canal Flats optimization delivered on-time.
Announced Jan 2025 · Target Jul 2025
50 EH/s by mid-2025 ✓
Crossed 52 EH/s by mid-July 2025. Childress Phase 3 + efficiency upgrade on S21 Pro fleet. Management then paused at 52 to redirect capex to AI.
Announced Sep 2024
First H200 fleet ($43.9M, 1,080 units) ✓
Deployed in Prince George 15 MW AI data center. Marked the shift from "can they do AI?" to "they already are."
Announced Nov 2025
Microsoft $9.7B / 5-yr anchor ✓
Single largest contract in IREN history. $1.94B annualized ARR. Childress 750 MW buildout gates rollout Q1 FY26–FY30.
Q2 FY26 update · Feb 2026
23,000 GPUs operating/on-order ✓
Mix of H100 + H200 + B200, with first GB200 NVL72 racks delivered Q2 FY26. Ramp to 140k targeted end CY2026.
Target · End CY2026
140,000 GPUs deployed 🕒
Requires Childress Phase 4 (250 MW) + Sweetwater TX 500 MW first energization. Biggest execution risk in the plan.
Target · FY27 exit
$3.4–3.7B AI ARR 🕒
Implies Microsoft fully ramped + ~$1.5B additional hyperscaler contracts signed. Bear case says this is where customer-concentration risk shows up.
Target · 2027
Sweetwater TX 500 MW live 🕒
Second mega-site for AI hosting. Permits and ERCOT interconnect largely secured; turbines and substation in construction Q1 2026.
Financial Guide vs. Actual — FY23 to FY26
| FY |
Metric |
Guide / Target |
Actual |
vs Target |
Verdict |
| FY23 | Hashrate exit | 10 EH/s (Dec 2022 plan) | 5.6 EH/s | −44% | Missed |
| Revenue | ~$90M (Street) | $75.5M | −16% | Missed |
| Adj EBITDA | ~$10M (Street) | $1.4M | Large miss | Missed |
| FY24 | Hashrate exit | 15–20 EH/s | 15 EH/s | Low end | Met |
| Revenue | ~$160M (Street) | $187.2M | +17% | Exceeded |
| Adj EBITDA | ~$35M (Street) | $54.7M | +56% | Exceeded |
| FY25 | Hashrate exit | 50 EH/s by Jul 2025 | 52 EH/s | +4% | Exceeded |
| Revenue | ~$430M (Street) | $501M | +16% | Exceeded |
| AI Cloud ARR exit | ~$300M (Q3 FY25 plan) | >$500M secured | +67% | Exceeded |
| BTC mined | ~4,800 (plan) | ~5,200 | +8% | Exceeded |
| FY26 | AI ARR exit | $1.9B+ (Microsoft anchor) | — | Tracking | Pending |
| GPUs deployed | 60K+ by Jun 2026 | 23K operating/on-order | On path | Pending |
| Hashrate | Held at 52 EH/s | 52 EH/s | Flat (by design) | Met |
Note: IREN historically guided on operational metrics (hashrate, MW live, BTC mined) more than financial ranges. Revenue/EBITDA comparisons use consensus estimates where explicit company guides were absent.
Financial Trajectory
$1.94B
Nov 2025 ARR
AI Cloud Annualized Run-Rate
~$200M
FY25 (est)
Adj EBITDA
~$1.9B
Q2 FY26
Customer Prepayments (MSFT)
The Roberts Brothers on Strategy
"We built 750 MW of next-generation data centers at Childress in under 24 months from a greenfield site. That execution speed is the reason Microsoft chose us — they don't buy PowerPoints, they buy live megawatts."
— Daniel Roberts, Co-CEO · Q2 FY26 Earnings Call · Feb 2026
"Our model is simple: we lock in the cheapest renewable power in North America on long-dated PPAs, we build Tier-III-ready AI-capable data centers with our own balance sheet, and we lease compute to customers who will pay a premium for speed-to-market. We are not a Bitcoin miner that happens to do AI — we are a vertically integrated power-to-compute company."
— Will Roberts, Co-CEO · Investor Day · March 2026
"Bitcoin mining is our foundation and our cash machine — it funded the first 250 MW of infrastructure that AI is now being deployed into. We do not see Bitcoin and AI as competing; we see them as co-located workloads sharing the same power, land, and cooling advantage."
— Daniel Roberts · Q1 FY26 Earnings Call · Nov 2025
"Our first AI data center cost $700,000 per MW — industry comp is $10–15 million per MW. That's a 15x capex advantage that compounds. Anyone who tells you this is a commodity business hasn't built one."
— Will Roberts · Bloomberg interview · Dec 2025
The Bull vs Bear Debate
🐂 Bull Case
1. Power Arbitrage That Nobody Can Match
750 MW Childress site sits on stranded ERCOT wind at <$0.02/kWh, vs CoreWeave / Applied Digital industry average of $0.04–0.06/kWh. At hyperscale, 2c/kWh advantage = ~$175M EBITDA per GW per year. This is structural, not a trade.
2. Microsoft Anchor = 5-Year Revenue Visibility
$9.7B committed over 5 years — the longest and largest Tier-1 hyperscaler deal ever signed by a miner. ~$1.9B in customer prepayments already on balance sheet reduces equity dilution needed to fund capex.
3. Execution Speed (750 MW in 24 Months)
Greenfield to Tier-III AI-capable in under 2 years is an outlier in data center industry. Every new MW is ~100x more valuable in AI lease mode vs Bitcoin mining mode, so converting fleet from mining to AI hosting is pure margin expansion.
4. 100% Renewable-Powered Narrative
Net-zero Scope 1-2 positioning matters for ESG-constrained hyperscalers. Microsoft's Science Based Targets commitments specifically require renewable-powered hosting — IREN is one of few suppliers that qualifies at scale.
🐻 Bear Case
1. Massive Equity Dilution (~271%)
Share count grew 271% from FY23 to FY25. A $6B ATM offering is authorized — if fully drawn, another 50%+ dilution. Even if the operating story works, per-share value compounds slowly.
2. Microsoft Concentration = ~80% of AI ARR
Single-customer risk is extreme. Renewal risk at end of year-5, pricing pressure at contract refresh, and the optionality MSFT has to build in-house (Fairwater, Arizona) caps IREN's long-term pricing power.
3. AI Hosting Is Commoditizing Fast
CoreWeave IPO'd. Applied Digital, Cipher, TeraWulf, Galaxy, Hut 8 are all pivoting. Bloom Energy, Crusoe, Lambda are chasing the same PPA arb. Gross margins in GPU-as-a-service are already compressing 300–500 bps year over year.
4. Grid & ERCOT Constraints Cap Upside
Childress is capped at 750 MW by substation capacity. Sweetwater is on the critical path but ERCOT interconnect queue is crowded. Expansion beyond ~1.5 GW of hosting is gated by grid physics, not capital.
5. Bitcoin Leverage Still Material
Mining still ~40% of revenue through FY26. If BTC falls to $30k, mining gross profit goes negative and pressure on the AI ramp (compute capex vs FCF) intensifies.
The Microsoft Anchor — Why It Matters
Deal structure (Nov 2025): 5-year AI cloud hosting agreement. $9.7B total contract value. $1.94B annualized revenue run-rate at full ramp. Childress Phase 3 + 4 gated to Microsoft workloads. ~$1.9B customer prepayment on balance sheet as of Q2 FY26 — reduces equity dilution needed for GPU capex.
The deal transformed IREN from a small-cap miner into a credible Tier-2 AI infrastructure operator overnight. Three structural implications: (1) the customer prepayment de-risks capex — Microsoft is effectively funding a portion of the GB200 buildout; (2) the 5-year term sets a floor under revenue visibility in a category where most peers have 12-24 month contracts; (3) it validates IREN's Tier-III build spec for other hyperscalers (Google, Meta, Oracle) who may want redundant supply.
The unresolved question: does Microsoft's success with IREN cement them as a preferred supplier (bull) — or does it give MSFT a blueprint for building comparable infrastructure themselves at Fairwater/Arizona and letting IREN's second-customer pipeline stall (bear)?