Iren (IREN) Q3 2026 Earnings Call Transcript

IREN reported $3.1 billion of contracted ARR and said it is on track to reach $3.7 billion by year-end, backed by a $3.4 billion five-year AI cloud contract with NVIDIA and a $2.1 billion conditional NVIDIA investment. The company ended April with $2.6 billion in cash and is targeting 480 MW of AI capacity in 2026, scaling to 1.21 GW in 2027 and 5 GW of secured power globally. Near-term results were mixed, with revenue down to $144.8 million and a $247.8 million net loss due to mining asset impairments, but the strategic AI buildout and contracting momentum were materially positive.

Analysis

The market is still underestimating how much of this story is now a financing and scheduling arbitrage, not just a demand story. Once a hyperscaler or strategic anchor customer precommits, the company can convert customer paper into asset-level leverage at progressively lower cost of capital; that creates a reflexive loop where execution lowers funding friction, which in turn accelerates delivery. The important second-order effect is that this should compress the valuation gap between pure-play infrastructure operators and software-like cloud names, because contracted revenue is becoming increasingly visible before the capex is fully spent.

The real competitive moat is not GPU access; it is project sequencing. Anyone can announce power, but very few can repeatedly move from secured megawatts to commissioning without breaking timelines, and that makes the installed base more valuable than the backlog. That matters especially for air-cooled retrofits, which appear to be the bridge product: they are faster to monetize, capital-light relative to greenfield liquid cooling, and likely to absorb the next wave of scarce supply first. The spillover is that adjacent beneficiaries are likely to be equipment vendors and network/thermal infrastructure names rather than just the headline GPU suppliers.

The main risk is that the equity is now pricing a near-perfect conversion of power into contracted ARR over the next 6-18 months, while the legacy mining wind-down still creates accounting drag and potential financing noise. If delivery slips even modestly, the market can punish the stock sharply because the bull case depends on a specific cadence of handoffs, not just long-term power ownership. Conversely, if management proves it can repeat the Microsoft/NVIDIA template across multiple campuses, the multiple could rerate quickly as investors move from EV/EBITDA to contracted-capacity valuation frameworks.

Contrarian takeaway: the headline partnership may matter less for the cash economics than for signaling quality. NVIDIA’s involvement de-risks customer acquisition and financing, but the bigger implication is that it validates the platform as a preferred fulfillment partner for the broader AI ecosystem, which could crowd in enterprise and sovereign demand at better terms than hyperscaler-only exposure. In that sense, the stock is not just an AI infrastructure beta; it is a scarce execution asset with embedded option value on Europe and APAC expansion.

Key Decisions for Investors

• Long IREN on pullbacks over the next 2-6 weeks: buy weakness into any post-earnings volatility, with a 3-6 month target tied to proof of Q3 handoff and fresh contract disclosures; risk is a delivery miss that breaks the financing flywheel.
• Pair trade: long IREN / short a basket of over-extended AI infrastructure names with weaker contracted backlog visibility over 3-6 months; the thesis is that contracted ARR plus financing visibility should command a premium as capital markets reward execution certainty.
• Long NVDA against a small IREN hedge for 1-3 months: NVIDIA’s strategic investment creates asymmetric upside if the partnership catalyzes broader ecosystem demand; hedge with IREN to isolate partner-validation upside while limiting single-name execution risk.
• Buy MSFT calls or maintain long exposure into the next 2 quarters: Microsoft appears to be getting a de-risked, capital-efficient capacity pipeline, which should support long-dated AI infra availability without requiring it to build everything itself.
• Avoid shorting IREN outright until after the next two commissioning milestones: the setup is too execution-dependent, and a successful cadence could force multiple expansion faster than fundamentals normalize.