Phase Space AI

Valuation

IREN [IREN]

IREN Ltd [IREN] — Valuation

As of: 2026-07-29 · Spot: $29.325 · EV (ex digital assets): $12,107m · Framework: v1.5.1 / criteria.md 2026-07-29 Cluster context: CLUSTER_Analysis.md

Two outputs are required on every name: a 12-month target and the implied-path test. Reporting only one is a defect. Both are addressed below, and where an input does not exist it is declared UNIDENTIFIED rather than substituted with a peer median.

---

Exit multiple: UNIDENTIFIED

IREN's latest-quarter total revenue growth is −0.02%. No comparator in AI compute is flat — CoreWeave grew +111.6% YoY in Q1 2026; the mature contracted-datacenter set (EQIX, DLR) grows 5–15%. There is no comparator set whose growth brackets IREN's at the exit year in either direction, because IREN is mid-transition between two businesses moving in opposite directions.

Per valuation.md, the multiple is declared UNIDENTIFIED rather than defaulted to a peer median. The mining leg is separately scored UNIDENTIFIED with a floor at zero — and for IREN that floor is close to literal, because management has announced it will retire the mining fleet entirely and take ~$520m of impairment doing so. A stream the operator has decided to shut down has no positive exit multiple.

What today's price requires

Holding the pure-play anchor of 6.6x EV / annualised AI revenue (CoreWeave, 2026-07-29) and today's EV of $12,107m fixed:

• Required: AI Cloud revenue of $1,834m annualised.
• Demonstrated: $134.5m annualised (Q3 FY26 rate of $33.635m).
• Margin (demonstrated − required): −$1,700m, i.e. −92.7%. IREN must grow AI Cloud revenue 13.6x.
• Implied compression: IREN trades at 90.0x annualised AI revenue. Convergence to 6.6x with no revenue growth is −92.7% multiple compression.

Context on achievability, in IREN's favour: AI Cloud went from $9.4m (9M FY25) to $58.3m (9M FY26), a 6.2x in one year. To reach $1,834m it must do that again roughly twice more. That is not absurd — it is the reason IREN sits mid-table on required uplift rather than at the HUT/BTDR end. But it must happen while mining revenue, currently 77% of the total, is deliberately retired.

The contract ratio — where IREN looks worst, and the ambiguity that matters

Because it is a contract-value ratio rather than an earnings multiple, EV/RPO requires no growth match and is directly comparable across the cluster:

IREN is priced at 31x CoreWeave per dollar of signed contract. There are only two possible explanations and they have opposite investment consequences:

1. RPO materially understates the Microsoft Agreement's economics — entirely plausible, because RPO under ASC 606 excludes tranches that are not yet unconditionally committed, and the Microsoft Agreement is explicitly structured "in tranches." If the full contract were unconditional, RPO would be far larger and the ratio far lower.
2. The price is wrong.

This memo cannot adjudicate between them and does not guess. The 10-Q sentence giving the Microsoft Agreement's average term is truncated at a page break in the filing text, and the total contract value is not in the extract reviewed. Obtaining the Microsoft Agreement's total contract value and term is the single highest-value follow-up in this cluster — it is the difference between IREN being the cheapest converting miner and the most expensive.

Corroborating evidence for interpretation (1): deferred revenue rose from $884,000 to $120,388,000 in nine months. Customers are prepaying $120m against $710m of RPO — an unusually high ratio that suggests the contracted book is real and being funded in advance.

Sensitivity over the exit multiple — the highest-variance parameter

The 4th row is the interesting one. At a 40x exit multiple — high, but not unreasonable for a business with a Microsoft anchor contract, 4,510MW of grid-connected power and hedged power costs — IREN needs to roughly double AI Cloud revenue from a rate that just grew 839%. The valuation is defensible at a high multiple and indefensible at the pure-play multiple, and nothing in the filings settles which applies because the multiple is UNIDENTIFIED.

12-month target: UNIDENTIFIED

Both required inputs fail:

1. No consensus. Alpha Vantage quota shared and not attempted. Consensus Criteria INDETERMINATE, blocks nothing.
2. The name's own multiple history spans multiple regime changes, which valuation.md explicitly says requires UNIDENTIFIED rather than a peer-median substitute: an IFRS→US GAAP accounting-basis conversion (20-F to 10-K), zero-to-$3.69bn of convertible debt in 21 months, and an announced exit from the business that is currently 77% of revenue.

And the near-term direction is knowable and adverse. Management has announced it will retire the mining operations; ~$520m of further impairment is flagged; mining is 76.8% of the latest quarter. Revenue falls before it rises. Any 12-month target that ignored the air pocket would be wrong in a predictable direction.

Declared UNIDENTIFIED. No target is manufactured.

---

Why the mining leg and the AI leg cannot share an exit multiple

This is the methodological core of the cluster and it applies directly to IREN:

A single blended exit multiple applied across these two streams is the exact defect that inflated ADMA by 51pp and INOD by 32pp. This analysis therefore ranks on EV / annualised AI-HPC revenue and EV / RPO — a price-for-what-exists and a price-for-what-is-signed — and scores the mining leg at UNIDENTIFIED with a floor at zero, because a stream with a negative gross margin at one operator and a halving margin at another has no defensible positive revenue multiple.

---

Required disclosures per valuation.md