IREN (IREN) Q4 2026 Earnings Call Transcript

IREN reported Q4 revenue of $137.2M (down $7.6M QoQ) as it decommissioned Bitcoin mining hardware ahead of GPU installs, but AI Cloud Services revenue rose to $70.5M (up from $33.6M). Net loss was $684M, including $450.4M of non-cash impairments, while AI cloud contracting remains strong with $4B of 2026 contracted ARR (of which $1B is already operating). Management guided FY27 CapEx to $25B–$30B and secured $6.5B of GPU financing in the last three months (including $3.6B investment-grade at ~6% and $2.8B at 9% fixed), alongside contracted pricing up to ~$25M per MW for new discussions.

Analysis

This is less a demand print than a capital-allocation signal: IREN is converting scarcity in power/grid access into a financed infrastructure annuity. The market will likely reward the backlog/financing combination immediately, but the more important mechanism is that customer prepayments are turning end-demand into quasi-project finance, which lowers incremental equity needs while also making the equity more sensitive to any funding-market wobble.

Second-order winners are upstream GPU suppliers and anchor tenants. NVDA benefits because every megawatt contracted implies another wave of accelerator demand, while MSFT gains optionality from having a scarce compute channel without building the full stack itself. The competitive threat is that larger balance sheets at ORCL, GOOGL and AMZN can choose to outbid on power adjacency and managed-service bundles, which could cap IREN’s long-run pricing power even if near-term megawatt economics stay firm.

The real risk is timing, not thesis: commissioning slippage, GPU delivery bottlenecks, or 100-200 bps of spread widening in GPU/data-center financing would pressure the equity even if bookings remain strong. Consensus may be overvaluing ARR as if it were already free cash flow; this is still an execution-heavy, capex-intensive story with a 6-18 month financing catalyst path. Falsifier: any reset in 2027 capacity pricing, weaker prepayment ratios, or a material delay in Horizon-style deployments.

Key Decisions for Investors

• Buy IREN only on pullbacks after the print, sized as a high-beta 1-3 month event-driven long; thesis is backlog monetization plus financing flywheel, but keep risk tight because execution slippage can re-rate it fast.
• Prefer NVDA over IREN as the cleaner expression of the same demand cycle: long NVDA for 6-18 months, since it captures incremental AI spend without project-finance or construction risk.
• Use MSFT as the lower-volatility beneficiary of scarce compute access; any weakness tied to broader AI capex skepticism is a buy-the-dip opportunity over 3-6 months.
• Avoid chasing ORCL/GOOGL/AMZN here unless they show a matching capacity-funding advantage; they are the main competitive overhang if they decide to bundle power, cloud, and software more aggressively.
• Set a watch item on financing spreads and prepayment terms: if GPU financing moves materially above current levels or prepayments fall below roughly half of capex, reduce exposure to IREN immediately.