By Bert Hochfeld | February 25, 2026
IREN delivered a quarter where every headline metric looked wrong and the underlying business made its most important strategic advance yet. Revenue of $184.7M fell 23% sequentially, GAAP net loss was $155.4M, and the stock missed the SA consensus revenue estimate by nearly $40M. None of that is the story. The story is that IREN closed $3.6B in GPU financing from Goldman Sachs and JPMorgan at under 6%, which combined with Microsoft's $1.9B prepayment covers 95% of the $5.8B GPU CapEx required to deliver the $9.7B, 5-year contract that starts generating revenue in April 2026. The execution risk that most concerned the market — where does the capital come from? — has been largely extinguished. What remains is operational: Sweetwater 1 energizes on schedule in April, Prince George completes its GPU fit-out, and Mackenzie/Canal Flats transitions from ASICs to GPUs across Q3 and Q4 FY26.
I want to be direct about the analytical framework here. I haven't covered IREN in the newsletter before, and it requires adaptation of my standard GARP/SaaS lens. This is not a software company — it is a real asset infrastructure business with contracted revenues, debt-financed CapEx, and utility-like economics at scale. The correct valuation frame is EV/contracted ARR and forward EV/EBITDA, not trailing EV/S. On those metrics, the stock is inexpensive for an infrastructure platform with 4.5 GW of secured power and a $9.7B contracted revenue anchor. The risk is binary and near-term: April 2026. If Sweetwater energizes on schedule and Microsoft revenue commences as guided, the thesis is validated and the market will need to re-rate.
Capital milestones:
The sequential decline from $240.3M to $184.7M requires context to interpret correctly. IREN is deliberately migrating power and capacity from Bitcoin mining (which has declining economics and is being phased out of BC sites) to AI cloud infrastructure. Bitcoin mining revenue fell $65.5M QoQ — hash rate was reduced as ASICs were removed from BC sites to make way for GPU fit-outs. AI Cloud revenue grew $10.0M QoQ (+137%), which partially offset but did not replace the mining decline.
The YoY comparison of +59.1% is suppressed by this transition mechanics. The correct forward comparison point is what happens in Q3 FY26 (April–June 2026) when Microsoft contracted revenues commence and the Prince George deployment is complete. The transcript is explicit: "Q2 CY2026 is when initial revenues will commence." CEO Roberts used the phrase "crystal clear" three times in describing this timeline — unusual emphasis that reads either as frustration at repeated analyst skepticism or a deliberate attempt to anchor expectations.
What the revenue number does not reflect is the ARR value already contracted. The gap between $184.7M quarterly revenue and $575M quarterly equivalent on $2.3B ARR contracted is the transition valley. Q3 FY26 is where the valley inflects.
One qualitative flag: The CFO inadvertently mentioned "Verizon data centers" during the call before correcting course. This is not confirmed, but pattern recognition says this was a live engagement at the time of the call or a recently concluded one. Combined with the reference to an "unnamed multibillion-dollar software contract" in active negotiation, customer concentration risk is being actively managed even if the disclosure doesn't show it yet.
The blended gross margin of 64.4% masks the most important structural dynamic in this business: AI Cloud margins are dramatically superior to Bitcoin mining margins.
| Segment | Q2 FY26 Revenue | COGS | Gross Margin |
|---|---|---|---|
| Bitcoin Mining | $167.4M | $63.4M | 62.1% |
| AI Cloud | $17.3M | $2.4M | 86.1% |
| Blended | $184.7M | $65.8M | 64.4% |
As AI Cloud scales from 9.4% of revenue toward the majority, blended margins should move toward the 85%+ level that IREN has guided for the Microsoft contract specifically. The gross margin trend (39.6% → 72.2% → 71.8% → 66.4% → 64.4%) has been slightly pressured during the transition quarter, but the directional destination is higher margins, not lower. This is the opposite of typical mining companies: transitioning away from mining is margin-accretive.
The 64.4% gross margin is robust for an infrastructure company. The relevant comparison is not cloud software but data center REITs and neocloud peers. At scale, IREN's Adj EBITDA margins (guided 85% for Microsoft contract) will far exceed anything available in traditional colocation.
This quarter was fundamentally about capital formation, not quarterly revenue. IREN secured three things simultaneously:
1. GPU Financing: $3.6B at <6% Goldman Sachs and JPMorgan provided a delayed-draw term loan secured by the GPUs themselves plus Microsoft's contracted cash flows. At an expected blended rate of approximately 3% (combining the <6% debt rate with the zero-cost Microsoft prepayment), IREN has locked in exceptional leverage economics. The GPU asset is self-funding: $9.7B revenue over 5 years on $5.8B of cost (net of $3.5B financing) produces returns that don't require equity capital to execute.
2. Microsoft Prepayment: $1.9B received This is extraordinary. A customer paying nearly $2B upfront for future compute services eliminates the working capital risk that typically constrains infrastructure build-outs. Microsoft's counterparty credit quality effectively backstops the debt.
3. Convertible Notes: $2.3B issued in December The balance sheet now holds $3.26B cash (Dec-25) vs 3.69Binconvertiblenotes.Theinterestincome(15.8M in Q2) partially offsets debt service. The capital structure is levered but purposefully so — this is infrastructure capital, deployed against contracted multi-billion-dollar revenue streams.
Coverage math:
The remaining 5% comes from cash on hand. Execution risk on the Microsoft contract is now primarily operational, not financial.
IREN's power portfolio deserves independent valuation attention. The company now holds:
| Site | Capacity | Status | Revenue Potential |
|---|---|---|---|
| Childress, TX (ERCOT) | 810 MW | Operating | Microsoft Horizons 1-4 |
| Sweetwater 1 & 2, TX (ERCOT) | 2,000 MW | Under construction; Sweetwater 1 energizing Apr 2026 | Near-term deployment |
| Prince George, BC | Part of BC portfolio | GPU fit-outs complete | $0.4B+ ARR contracted |
| Mackenzie & Canal Flats, BC | Part of BC portfolio | Converting ASICs → GPUs | ~$1B ARR target (40k GPUs) |
| Oklahoma (Southwest Power Pool) | 1,600 MW | Grid-studies complete; 2028 ramp | Future |
| Total secured | 4,500 MW |
The $3.4B ARR target by end of CY26 represents only the first 10% monetization of this 4.5 GW portfolio. If AI infrastructure demand persists — and every signal in the market suggests it will — IREN has a multi-decade compounding asset base that is essentially irreplaceable. Securing 4.5 GW of grid-connected, renewable-rich power in today's market requires years and billions in development capital. It cannot be replicated quickly by new entrants.
Management's comment that "data center capacity is now scarcer than power" reflects a market dynamic that works in IREN's favor. They have both.
I track whether management delivers on what it says. This is the most important qualitative input to my investment conviction:
| Promise | When Made | Due | Outcome |
|---|---|---|---|
| GPU financing ≥$2.5B | Q1 FY26 | Q2 FY26 | Delivered: $3.6B (44% beat) |
| Sweetwater 1 energization Q2 CY26 | Q4 FY25 | April 2026 | Pending — on track |
| Microsoft revenue Q2 CY26 | Q2 FY26 | April 2026 | Pending |
| 140k GPU deployment end CY26 | Q1 FY26 | December 2026 | Pending |
| $3.4B ARR by end CY26 | Q4 FY25 | December 2026 | $2.3B contracted, pending |
The GPU financing delivery is the single most important completed promise: management committed to ≥$2.5B and secured $3.6B from two of the world's most sophisticated lenders. Goldman and JPMorgan did their own due diligence on the Microsoft contract and the GPU asset quality. Their willingness to lend $3.6B at <6% is itself a third-party validation of the thesis.
The remaining promises are all execution-dependent and timing-specific. The key watch point is April 2026 — if Sweetwater energizes and Microsoft revenue commences on schedule, management credibility reaches its highest point in the company's public history.
One credibility headwind I note: IREN has missed SA consensus revenue estimates in 5 consecutive quarters. The market has trained itself to fade the stock on earnings because the headline number always disappoints. This is a structural issue with how consensus models are built for a transitioning business. The fixes are (a) April 2026 revenue commencement changes the consensus model permanently, and (b) management eventually begins providing more specific quarterly guidance once the contracted revenue base is established. For now, the misses are a perception problem, not a thesis problem.
IREN does not fit neatly into the EV/S cohort framework I use for software companies. The correct comparisons are AI infrastructure peers (CoreWeave, Lambda Labs, Crusoe Energy) and, increasingly, cloud hyperscalers for EV/EBITDA benchmarking. Let me present the valuation multiple ways:
Current market statistics (as of February 25, 2026):
Trailing EV/S (TTM revenue ~$757M): 17.2x — Appears expensive. Meaningless for valuation purposes given the transition.
EV/Contracted ARR ($2.3B): 5.6x — A reasonable multiple for AI infrastructure with hyperscaler counterparty credit. The contracted ARR is not at risk unless Microsoft defaults.
EV/Target ARR end CY26 ($3.4B): 3.8x — If management delivers on the $3.4B ARR target, this is cheap for a platform with 4.5 GW of power assets and a demonstrated capital-raising capability.
Forward EV/EBITDA (projected 2026 run rate): At 3.4BARRwiththeMicrosoftcontractearning 851.65B) and BC cloud at ~65% ($975M), total EBITDA run rate approaches $2.6B. At $13.0B EV that is approximately 5.0x forward EV/EBITDA — inexpensive for infrastructure with this growth profile and asset quality.
For context, CoreWeave went public at EV/S multiples well above 10x on a smaller contract base. IREN is trading at a fraction of that on a contracted ARR basis.
The bull case is not priced in. The market is valuing IREN at trough transition metrics. The contracted fundamentals say something very different.
Dilution reality check: Shares outstanding grew from approximately 180M to 332M over 15 months — 84% dilution. This is the cost of financing the transformation. On a per-share basis, the 9.7BMicrosoftTCVrepresents 29/share in contracted revenue across 5 years. Management's capital allocation discipline post-April 2026 (when Microsoft cash flows begin) will determine whether dilution slows. This is a risk I'll continue to monitor.
The Q2 FY26 quarter delivered the most important capital milestone in IREN's history — $3.6B in GPU financing from Goldman and JPMorgan that de-risks the Microsoft contract execution — while reporting financial results that look poor on the surface due to the mining-to-AI transition valley. The underlying Adj EBITDA of $75.3M at 41% margin shows a business with real economics even during the trough.
The thesis is intact. The binary event is April 2026: Sweetwater energization and Microsoft revenue commencement. If those land on schedule, Q3 FY26 will be a dramatically different financial picture, and the market will need to re-rate a business with $2.3B+ in contracted ARR, $2.6B+ in projected annual EBITDA run rate, and 4.5 GW of power assets representing 90% untapped upside.
On a 12-month view, this is one of the most asymmetric setups in the AI infrastructure space. The risk is that April delays (ERCOT batch permitting, GPU delivery, construction timeline) create another quarter of poor P&L optics. The reward is a re-rating from 5-6x contracted ARR to something reflecting the full asset quality and contracted revenue profile.
I would initiate a position here in the high-growth portfolio at 3-4% weight with a plan to add on any April-related execution confirmation. This is not a small, safe bet — it's an infrastructure company in a capital-intensive, execution-dependent business — but the contracted revenue base and capital structure de-risking make the risk/reward favorable at current prices.
I haven't covered IREN in the newsletter before. This is a first look applying my framework to an infrastructure business rather than SaaS/software. The valuation methodology required adaptation from EV/S cohort analysis to EV/contracted ARR and forward EV/EBITDA, which is appropriate for a company with this revenue structure.
Filed: February 25, 2026 | Q2 FY26 (December 31, 2025)