Date: 2026-02-22 | Period: Q2 FY26 (Dec 31, 2025) | Author: muji (CMF_muji)
/briefs/IREN_earnings-review_2026-02-21/~/.agents/skills/atlas/analyses/IREN/IREN_earnings-review_Q2_FY26.mdcompanies/IREN.mdearnings/IREN/_ROLLING.mdQ2 FY26 looked ugly on the surface. Revenue $184.7M, down 23% QoQ. GAAP operating margin -63%. Net loss 155M.FCF−648M. If you screened those numbers in isolation you'd run.
Here's the thing — every single one of those metrics is explained by an intentional, capital-rational pivot from Bitcoin mining to AI cloud. Management is deliberately destroying short-term revenue to build long-term contracted ARR. And they're doing it with $3.7B in convertible notes at sub-1% coupons and $3.6B in GPU financing at <6%. That's the story. The GAAP income statement is a distraction.
The number that matters: $2.3B ARR under contract. Against TTM revenue of $757M. That's a 3x revenue backlog sitting in signed agreements. I've rarely seen that gap in a growth company outside of hypergrowth SaaS. This is the leading indicator divergence you drill into.
IREN is picks and shovels for AI — and not generic picks and shovels. They have three durable advantages that look increasingly structural:
1. Power as platform. 4.5 GW of secured power is not a KPI — it's a platform asset. Power is the scarcity in AI infrastructure right now. IREN has 4.5 GW secured (810 MW operating, Sweetwater 2 GW in construction, Prince George contracted, Oklahoma 1.6 GW added). They're using ~10% of this to hit $3.4B ARR. The rest is pure optionality. This is "scale in platform" applied to physical infrastructure. Competitors cannot replicate 4.5 GW overnight. It took years of site acquisition and grid agreements. The 2023 Sweetwater ERCOT agreement that protects them from batch delays is exactly the kind of first-mover advantage I look for.
2. GPU financing as moat. $3.6B at <6% is extraordinary. CoreWeave's debt is at 15%+. At $3.4B ARR, IREN's cost of capital for their GPU infrastructure is dramatically lower than any pure-play competitor. This isn't just a balance sheet fact — it structurally enables them to offer competitive pricing to hyperscalers while maintaining 40%+ EBITDA margins.
3. Vertical integration. They own power, they own facilities, they lease GPUs at <6%. They're not renting colocation from Equinix and buying GPUs at retail. This vertical stack is what gives them the "fastest time-to-datacenter" they claim — and the Microsoft $9.7B TCV validates it.
Architecture grade: New build, cloud-native infrastructure. No legacy baggage. Green flag. GTM model: Hyperscaler contract + direct enterprise. Not self-serve, but that's appropriate for this scale. Business model: Building block (infrastructure layer for AI training/inference) vs. process tool. IREN is as "building block" as it gets.
| Q424 | Q125 | Q225 | Q325 | Q425 | Q126 | Q226 | |
|---|---|---|---|---|---|---|---|
| Jun-24 | Sep-24 | Dec-24 | Mar-25 | Jun-25 | Sep-25 | Dec-25 | |
| Revenue ($M) | 53.2 | 53.1 | 78.8 | 148.1 | 240.3 | 240.3 | 184.7 |
| YoY % | — | — | — | +172% | +229% | +342% | +60% |
| QoQ % | — | -0.2% | +48% | +88% | +62% | 0% | -23% |
| Gross Margin | — | 39.6% | 56.6% | 72.2% | 67.6% | 68.2% | 64.4% |
| Adj EBITDA ($M) | — | 2.5 | 28.0 | 74.2 | 121.9 | 86.7 | 75.3 |
| Adj EBITDA % | — | 4.7% | 35.5% | 50.1% | 50.7% | 36.1% | 40.8% |
The YoY deceleration from 342% → 60% looks alarming. NUANCE: The comp is Q2 FY25 ($78.8M) at a much lower base. More importantly, the QoQ story is intentional: BTC mining $167M (down 28% QoQ) + AI Cloud $17.3M (up 137% QoQ). They're cannibalizing their own mining revenue to build the higher-quality AI cloud business. The Adj EBITDA margin recovery (36% → 41%) confirms the mix shift is working economically.
| Segment | Q425 | Q126 | Q226 | QoQ |
|---|---|---|---|---|
| AI Cloud | $7.3M | — | $17.3M | +137% !! |
| Bitcoin Mining | $180.3M | — | $167.4M | -7% |
AI Cloud is still tiny ($17M) but growing triple digits. When Microsoft revenue commences (Q2 CY2026), this segment steps up by orders of magnitude. The 1.94BMicrosoftARRalone= 485M/quarter at full run rate. That's 2.5x IREN's current quarterly revenue from a single customer.
| Metric | Value |
|---|---|
| TTM Revenue (run rate) | $757M |
| Contracted ARR | $2.3B (3.0x TTM) |
| ARR target (end CY2026) | $3.4B (4.5x TTM) |
| Microsoft TCV (5-year) | $9.7B |
| Prince George ARR | $400M (included in $2.3B) |
DING DING DING. This is the leading indicator. When contracted ARR runs 3-4x trailing revenue, you're looking at a future inflection — not a deteriorating company. The market is starting to understand this (stock +5% post-earnings despite the GAAP miss).
The balance sheet is engineered for the build. Cheap debt, massive prepayments, positive OCF. FCF is negative because they're spending $827M on hardware — and 95% of that is financed externally. The actual cash drain from operations is minimal.
Atlas gave this a conviction 3.5/5 with an EV/forward ARR of 5.1x on $3.4B target. I agree with the framework. Where I'd add nuance:
The GPU financing moat is underweighted. Atlas noted <6% vs CoreWeave 15%+. I think this is actually the decisive competitive advantage. In a capital-intensive infrastructure race, cost of capital IS the product. IREN can offer Microsoft competitive pricing AND earn 40%+ EBITDA margins simultaneously. That's structural, not cyclical.
The execution risk is real but manageable. 140k GPUs by end CY2026 requires Sweetwater 1 energization on time. Management has been consistent here — the 2023 ERCOT agreement insulates Sweetwater from batch delays. I'd watch this like a hawk. Slip by one quarter and the ARR target slides.
Customer concentration is the primary risk I'm watching. Microsoft = 84% of contracted ARR. That's not a "diverse customer base" — that's a single dependency. IREN needs to close additional hyperscaler/enterprise contracts in the next 2 quarters to prove this is a platform, not a one-customer project.
The GAAP numbers are noise. Here's what's driving the -$155M net loss this quarter:
Strip those out: Adj EBITDA +75Mat4172M positive. The operations are generating cash. The income statement is distorted by the financial engineering required to fund the buildout cheaply.
Trust but verify: The 10-Q confirms 64.4% gross margins (the transcript had a 10.7% transcription error that got corrected). Always read the filing, not just the call.
Current: Tier 2 (transitioning to Tier 1)
Tier movement trigger: Microsoft revenue beginning + additional customer contracts → Tier 1 upgrade
Going in: Management promised Microsoft contract (✅ delivered), GPU financing at <6% (✅ delivered), $2.3B ARR contracted (✅ delivered), Sweetwater 2 GW power secure (✅ delivered).
Outstanding commitments:
Delivery rate has been excellent. Management credibility is high. But the next two quarters are the real test — actual revenue commencement, not just contract signing.
IREN is the most interesting infrastructure play in the portfolio right now. The thesis is asymmetric: if execution holds (Sweetwater on time, Microsoft revenue starts, additional contracts close), this steps up 3-5x in revenue within 18 months on existing contracted commitments. The <6% GPU financing gives them a structural cost advantage that pure-play cloud competitors cannot match. The 4.5 GW power pipeline is a multi-decade platform asset.
The risks are real — execution risk on construction, 84% customer concentration, 49% potential dilution from converts. But the balance sheet is funded, the operations are cash-generating, and management has delivered every milestone to date.
Hold at 5.7%. Add only when Microsoft revenue begins (Q2 CY2026) — that's the confirmation event. The valuation (5.1x forward ARR on $3.4B) is reasonable if they execute. Don't pay for optionality you can confirm in 2 quarters.
long IREN 5.7%
-muji