IREN Limited — Blitz Index 56 (C)

The crossover quarter: AI cloud passed mining — $13B in contracts, $137M in revenue.

Score breakdown

Radar axes (0–100): Moat 57 · Speed 68 · Creative 73.

Watch item

Gates: G1 ✓ (AI Cloud $128.8M, Q4 crossover) · G2 ✓ estimated. 9 below threshold — contracted, not yet delivered.

Strengths

• AI Cloud $16.4M → $128.8M (+687%); Q4 crossover $70.5M > mining $66.7M — the pivot is revenue now
• MSFT $9.7B (5yr, 20% prepay) + NVDA $3.4B (5yr) — two hyperscaler stamps of approval
• 5GW+ power pipeline; 2026 capacity largely sold out — power is the scarce resource
• Horizon 1 delivered & accepted (Aug 2026); NVIDIA Exemplar Cloud status

Concerns

• $13.81B capital commitments vs $707M revenue — extreme absolute leverage
• Share count +52.7% in 14 months — dilution as the price of speed
• Q4 missed consensus by 13% ($137.2M vs $157M) — patience wearing thin
• FY2026 GAAP net loss −$703M (incl. $450M impairments on retired miners)

Research notes

1. Current Position

Current position (56/C, mid-blitz pivot act one): AI Cloud ($128.8M, +687% YoY) overtook mining for the first time in Q4 FY2026 — the pivot is now revenue, not narrative. But the thesis still lives in contracts: $13.1B in backlog against $137.2M in quarterly revenue, and a 13% consensus miss showing the market's patience thinning. Speed (68.0) is blitz-grade; moat (56.7) reflects the structural limits of a capacity business.

2. Blitzscaling Lens

Blitzscaling lens: Rule 8 textbook (four differently-priced moneys, 0%–9%, plus "built with the customer's money" via MSFT's 20% prepayment) and a paradoxical Rule 5 (decommissioning a healthy cash cow to light the new fire). The break: data centers cannot be patched like software, and GPU cloud has no true network effects — hence "heavy-equipment blitzscaling."

3. Competitive Landscape

Competitive landscape: Explicit rivals are the neocloud cohort (challenger vs. CoreWeave); the real threats are structural — fellow kin attempting the same disruption and hyperscalers' own GPU expansion, with MSFT and NVDA as both customers and competitors. The Ledger's pie-expansion insight is a tailwind: AI compute is a new pie, not a stolen one. Power constraints themselves are the structural competitor.

4. Inflection Points

Inflection points: Horizon 2–4 delivery within 2026 — a one-quarter delay wakes D1 (−4). Whether the mining→AI crossover widens and total revenue rebounds past the J-curve trough (watch FY2027 Q1 in November). And the health of the $13.8B leverage structure — quarterly cash burn, refinancing of the 9.0% GPU financing, and whether MSFT's prepayment lands in cash flow.

5. One-Line Risk

One-line risk: With $13.81B in commitments (19.5x revenue) riding on 2026 delivery schedules, a one-quarter Horizon delay could shake the entire leverage structure — the fundamental risk of heavy-equipment blitzscaling, where data centers cannot be patched like software.

Current Position

IREN is mid-blitz. The pivot from bitcoin miner to AI cloud became revenue, not narrative — in Q4 FY2026, AI Cloud $70.5M overtook mining $66.7M for the first time (confirmed, the "crossover"). AI Cloud revenue grew from $16.4M in FY2025 to $128.8M in FY2026, +687% YoY (confirmed), among the highest growth rates of any company measured. Total FY2026 revenue of $707M (+41.1% YoY — confirmed) is the basis for G1 (a second curve) passing. 3MW → 480MW (2026 goal) → 1.2GW (2027 goal) — a 160x scale-up plan (company announcement) is the blitz's physical blueprint.

Yet the score is 56/C, 9 points below the cutoff, because the thesis still lives in "contracts." MSFT $9.7B (5 years, 20% prepayment) + NVDA $3.4B (5 years) in contracts (confirmed, company filings/8-K) versus quarterly revenue of $137.2M — the gap between the $13.1B contract backlog and quarterly recognized revenue is this company's everything. That Q4 FY2026 revenue missed consensus ($157M) by 13% (press-reported) shows the market's patience is beginning to be priced in. In the score file's words: "the contracts are signed; the buildings are not yet built" — Horizon 2–4 deliveries all fall within 2026. Speed (68.0) is the highest of the three axes because the velocity of capital raising and contract wins is blitz-grade, while moat (56.7) is low because of the structural limits of a capacity business (B1 network effects at 6 points).

The duality of the capital structure: unpacking the basis of A2 (capital velocity) at 4 points, IREN's capital raise is a combination of "different kinds of money." The June 2026 $3.65B GPU securitization (Fitch A rating — confirmed) is structured finance collateralized by hardware; the October 2025 $875M convertible bond (0%, 42.5% conversion premium — confirmed) is equity-market capital; the MSFT-contract-based $3.6B investment-grade facility (6.0% — confirmed) is project finance collateralized by customer contracts; the $2.4B GPU financing (9.0% — confirmed) is high-rate bridge-style money. Rates spread from 0% to 9% means the capital's quality is not uniform — the most expensive money (9.0%) must be repaid first. The D1 basis of −4 — "$13.81B in capital commitments (19.5x revenue) vs. $5.90B cash + $1.72B restricted cash" (confirmed) — compresses this qualitative difference into a single number. v1.1 candidate #1 (a capital-intensive blitz track) aims at exactly this point — the current flat −10 deduction cannot distinguish project finance (repayment tied to contracts) from indiscriminate leverage. The score file judges IREN's leverage closer to the former than the latter, but the distinction is not yet codified.

The quarterly trajectory of the transition: the FY2026 quarterly revenue flow reveals the anatomy of the transition. Q4 FY2026 revenue of $137.2M is −5.3% vs. Q3 ($144.8M) (confirmed) — the quarter where mining-equipment decommissioning ate into total revenue faster than AI Cloud growth. That total revenue contracted in the quarter the crossover occurred (AI $70.5M > mining $66.7M — confirmed) means the pivot is passing through the J-curve's trough. At the trough, what matters is not its depth but the timing of the rebound — if total revenue rebounds next quarter (FY2027 Q1, expected November), the transition's worst can be judged past. The 13% consensus miss ($137.2M vs. $157M — press-reported) also means the market saw the J-curve trough as deeper than expected. AI Cloud's +687% YoY (confirmed) growth rate gets diluted into total revenue's +41.1% (confirmed) — a pivot company's combined revenue is a veil over the new business's growth rate. v1.1 candidate #3 (G1 criteria for pivot companies) targets lifting this veil.

Two-sided assessment (per the score file): - Positives: AI Cloud +687% YoY and the Q4 crossover; MSFT $9.7B + NVDA $3.4B (the validation stamps of two hyperscalers); 5GW+ power pipeline with 2026 capacity "largely sold out" (company announcement); power cost of 3.8c/kWh and 100% renewable energy (confirmed); NVIDIA Exemplar Cloud certification for Horizon 1. - Concerns: $13.81B in capital commitments vs. $707M revenue (19.5x revenue — confirmed); share count +52.7% in 14 months (258.1M → 394.1M — confirmed); Q4's 13% consensus miss; FY2026 GAAP net loss of −$703M (annual impairments of $638.8M, including $450.4M in Q4 — confirmed); residual mining revenue dependence (if BTC prices fall, transition funding weakens).

The directional assessment is "rising (execution sprint)" — the score's direction hinges on a single variable: the speed at which contracts convert to recognized revenue. The score file's 3 KEY DRIVERS reach the same conclusion: ① the crossover quarter (the pivot's revenue materialization), ② the gap between contracts and recognized revenue is both the thesis and the risk, ③ capital intensity is both moat and threat. D1 at −4 is assessed as "close to the maximum warning the current v1.0 can issue" — v1.1 candidate #1 ("a capital-intensive blitz track: distinguishing project finance from indiscriminate leverage") targets exactly this point.

Gate verdicts: G1 (sustained growth or a second growth curve) is PASS (confirmed) — AI Cloud revenue from $16.4M in FY2025 to $128.8M in FY2026 (+687% YoY), with the Q4 crossover as the basis. G2 (new market's 10-year ceiling ≥ 10x current revenue, bar ≈ $7.1B) is PASS (estimate) — AI compute TAM in the tens of billions of dollars plus IREN's standalone contracts alone (MSFT $9.7B + NVDA $3.4B) as the basis. That both gates were passed on "the size of contracts and markets" summarizes this company's current position: the evidence of growth is in contracts, not revenue. If v1.1 candidate #2 ("reflecting contract backlog in scores") is formalized, G1's passing basis will flow directly into the score (A1) — the current v1.0 only sees recognized revenue, creating a structural window where IREN-like companies are under-scored.

Blitzscaling Lens

Rule 8, "Raise Too Much Money" — textbook embodiment IREN is the textbook for Rule 8. The June 2026 $3.65B GPU securitization (Fitch A rating), the October 2025 $875M convertible bond (0%, 42.5% conversion premium), the MSFT-contract-based $3.6B investment-grade GPU facility (6.0%) — all confirmed (company filings/10-K). Capital commitments of $13.81B (19.5x revenue — confirmed) are extreme relative to revenue scale, but because the repayment source is project finance tied to the MSFT contract, it differs in kind from "indiscriminate leverage" (the basis of the score file's D1 −4 verdict — the absolute leverage level is high, but the repayment source is contract-bound). GPU securitization is Rule 8's new variant — financing against hardware to secure speed without equity dilution. The MSFT contract's 20% prepayment is the blitz's most elegant financing: "built with the customer's money." The price was paid as share count +52.7% in 14 months (258.1M → 394.1M — confirmed). Dilution as the price of speed — the classic blitzscaling tradeoff. That the $2.4B GPU financing carries 9.0% high rates (confirmed) shows the capital's quality is not uniform.

Rule 5, "Let Fires Burn" — paradoxical embodiment Hoffman's fire triage is "extinguish only the fires that can kill the company; leave the small ones." IREN went one step further: it decommissioned a healthy cash cow (mining rigs) early to secure AI land. The $450.4M in mining-equipment impairments (confirmed, part of FY2026's −$703M GAAP net loss) is not "a small fire" but "a big fire set by itself." That the mining era's legacy (power at 3.8c/kWh, 100% renewable — confirmed) became a structural weapon of the AI era is a paradoxical execution of the rule — burning the old business's firewood to light the new business's fire. It is the same "pivot blitz" structure as Airbnb leveraging its existing US network into Europe (the score file's Blitzscaling fit). The verdict that power + land + grid interconnection are a distribution moat (B2 at 12 points) — "in the AI era, the distribution network is the power grid" — runs in the same vein.

Where it breaks — the limits of Rule 4 and a twist in the applicability conditions Blitzscaling says "launch an embarrassing product and patch it" (Rule 4), but a data center cannot be patched like software. This is the IREN research's biggest risk and Rule 4's limit. A single quarter's delivery delay can shake the entire leveraged structure — the fundamental difference of "heavy-equipment blitzscaling" from the typical Silicon Valley blitz (the score file's Blitzscaling fit #4). The Nine Rules say "build fast and fix," but power hookup, GPU delivery, and customer acceptance are schedules, not patches.

IREN is also a variant on the applicability conditions. Of Blitzscaling's preconditions — "a business model where speed pays big, such as network effects, virality, or data advantage" — IREN has neither data advantage nor virality. GPU cloud is a capacity business with no true network effects (B1 at 6 points — "GPU cloud is a capacity business"); rather than winner-take-all, the moat is "the distribution of power, a scarce resource" (5GW+ pipeline, 2026 capacity "largely sold out" — company announcement). The five-growth-lever analysis (score file) — main lever distribution (power), sub-lever platform (Mirantis Kubernetes + GPU cloud stack), weak lever network effects — declares a new type: "heavy-equipment blitzscaling." Supporting B1 are the Mirantis acquisition's 1,500+ enterprise customers plus Kubernetes orchestration, NVIDIA Preferred Partner status, and Horizon 1's NVIDIA Exemplar Cloud certification (confirmed/press-reported), but B1 stays at the bottom of the band (6 points), honestly reflecting a capacity business's limits.

A variant of Rule 6, "Do Things That Don't Scale": Rule 6's archetype is the Airbnb founders photographing hosts' homes themselves — learning manually, then scaling. IREN's variant runs the other way: dismantling an "already scaled" business manually while learning. That 23.2 EH/s of mining capacity is being sequentially decommissioned for AI conversion (the score file's Blitzscaling fit — "mining: in retreat") is reverse-direction learning — deliberately de-scaling a scaled business. The power land, grid interconnection, and operating know-how secured through dismantling become AI Cloud's execution assets. Not "things that don't scale" but "things that de-scale" — a pivot blitzer's own grammar. The $450.4M impairment (confirmed) incurred in this process should be read as conversion cost, not learning cost.

Competitive Landscape

Explicit competitors are internal to the neocloud cohort. IREN is the challenger against CoreWeave (the basis of A4 at 3 points), but marquee contract win velocity — MSFT $9.7B (2025-11) → NVDA $3.4B (2026-05) → an unnamed frontier lab (2026-08), all company announcements/press-reported — shows it expanding share within the cohort. Hyperscalers (AWS, Azure, GCP) are both customers and competitors — the tension being that MSFT, IREN's largest customer, is also the dominant player in the AI compute market. NVDA is likewise both a customer ($3.4B contract) and a GPU supplier — a structure where the supply chain and demand overlap.

Applying the Ledger's lens ③ — "civil war among kin is more fatal than external disruption" — to IREN, the real threat is not the centralized cloud as "the thing being disrupted" but fellow kin attempting the same disruption (neocloud cohort members like CoreWeave and Crusoe) and hyperscalers' own GPU expansion. IREN's C1 (the order being replaced) at 4 points — "industry-terrain replacement from bitcoin miner to AI compute supplier + price disruption versus centralized cloud" (press-reported) — points to the direction of destruction, but with the transition still early in revenue contribution, the depth of destruction is unknown.

The Ledger's insight ① (15 of 16 cases were pie-expansion — competitors need not die for the destroyer to thrive) is a tailwind for IREN. AI compute TAM (tens of billions of dollars — estimate) is creating a new pie, enabling a pie-expansion structure in which the entire neocloud cohort "thrives together" with hyperscalers. Just as NVDA created a new pie rather than "stealing" 6x the combined revenue of Intel and AMD, IREN's $13.1B in contracts are a slice cut from the new AI-compute pie, not stolen from MSFT and NVDA's pies. Against the Ledger's other insight — "destruction happens by structure (platform/business-model conversion), not by declaration" — IREN's means of destruction are the business model (miner→neocloud conversion) + price (price disruption versus centralized cloud), closer to quiet destruction.

The customer-competitor duality: MSFT and NVDA are both structural competitors and the customers of the $9.7B and $3.4B contracts. MSFT is the largest competitor with its own AI cloud (Azure); NVDA is a GPU supplier and the architect of the neocloud ecosystem. This duality resonates with the Ledger's proposition that "destroyer-vs.-destroyer configurations are the real threat" — a structural tension exists as a constant, since IREN's largest customers could at any time return to "doing it themselves." The MSFT contract's 20% prepayment and 5-year term (confirmed) seal this tension by contract — the seal's expiration is 5 years, which is this company's time horizon. If IREN does not achieve customer diversification — "running without MSFT" — within 5 years, the thesis returns to the validation table at contract expiry. The contract with the unnamed frontier lab (2026-08 — press-reported) is the first hint of diversification, though its scale is undisclosed.

Structural competitors are power constraints themselves. AI data centers' bottleneck is not GPUs but grid interconnection and power intake — the reason IREN's 5GW+ pipeline is a moat (B2 at 12 points), and why the execution risk of Horizon 2–4 deliveries is itself "the fight against structural competitors." D4 at −1 (environmental scrutiny of bitcoin-mining power use + the AI data-center power debate — press-reported, no specific lawsuits or fines) is a variable on the same axis. If the bitcoin-mining era's power use was the focus of environmental debate, in the AI era the legitimacy standard becomes "how efficiently power is converted into AI compute" — 100% renewable energy (confirmed) is the shield in this debate.

Inflection Points

① Horizon 2–4 delivery within 2026 — the point that is everything - What to watch: at FY2027 Q1 earnings (November — expected), how much AI Cloud quarterly revenue grows versus Q4 FY2026's $70.5M, and whether the company's delivery milestones (3MW → 480MW 2026 goal — company announcement) remain on track. The speed at which the $13.1B backlog converts to recognized revenue decides the score's direction (stated in the directional assessment — "contract→recognition conversion speed decides direction; Horizon 2–4 deliveries are the variable"). A one-quarter delivery delay is the trigger that wakes D1 (−4). In the score file's words: "the speed at which backlog converts to recognized revenue decides the score's direction." - Connection: IREN is not in the 21-company earnings calendar — as a June fiscal-year company, the November earnings are the observation point. Until v1.1 candidate #2 ("reflecting contract backlog in scores: if the backlog/TTM-revenue ratio exceeds a defined level, add to A1") is formalized, only recognized revenue moves the score. Current A1 at 7 points on recognized revenue — also on record: a backlog-addition rule could structurally lift it.

② The mining→AI crossover's persistence — the transition's total-revenue direction - What to watch: Q4 FY2026 total revenue of $137.2M was −5.3% vs. Q3 ($144.8M) (confirmed) — the transition's face, with mining-rig decommissioning eating total revenue. Next quarter, whether AI Cloud growth outruns mining decline and total revenue rebounds, and whether the crossover (AI $70.5M > mining $66.7M — confirmed) widens. If BTC prices fall, the residual mining-revenue dependence (an item of concern — score file) could weaken transition funding. The consensus miss (13% — press-reported) is the indicator measuring the market's patience. - Connection: FY2027 Q1's quarterly total revenue YoY/QoQ. AI Cloud standalone margin is undisclosed (confirmed — 10-K); whether the transition's blended margin (FY2026 GAAP net loss −$703M — confirmed) also improves will be watched together. Connects to v1.1 candidate #3 ("G1 criteria for pivot companies: combined treatment of shrinking legacy business + rapidly growing new business") — currently only combined revenue's growth rate is viewed.

The veil of margins: the basis of B4 (gross margin) at 6 points — "mining-hardware margin ~70–77% (power cost 3.8c/kWh — confirmed) but transition blended: FY2026 GAAP net loss −$703M (annual mining-equipment impairments $638.8M, including $450.4M in Q4 — confirmed). AI Cloud standalone margin undisclosed" — is this company's thickest veil. The 70%+ mining margin is a hint for estimating AI Cloud's margin floor, but GPU depreciation, power, and network cost structures differ from mining, so direct comparison is impossible. The moment AI Cloud's standalone margin is disclosed, B4's verdict splits — if high-margin, 6 points → the 8-point range; if low-margin, "a capacity business's limits" are confirmed. Whether segment margins are disclosed at FY2027 Q1 earnings is the observation point. For now, "a high-margin business being eaten by conversion costs" (score file) is the best available verdict.

③ The leverage structure's health — the repayment trajectory of the $13.8B commitment - What to watch: quarterly cash burn against $5.90B cash + $1.72B restricted cash (confirmed), whether the $2.4B GPU financing (9.0% high rate — confirmed) gets refinanced, and whether the MSFT contract's 20% prepayment actually lands in cash flow. The $13.81B in capital commitments (19.5x revenue — confirmed) is extreme in absolute terms, but the basis of D1 at −4 is that the repayment source is contract-bound — contract execution is repayment capacity. The two co-founders' September 2025 insider sales (~$66M — press-reported, the basis of D3 at −1) remain a variable of insider confidence. - Connection: each quarter's cash/debt metrics in the 10-K/10-Q. The watch point is whether the GAAP net loss of −$703M (annual impairments $638.8M, including $450.4M in Q4 — confirmed) flows through to cash-flow losses. Also on record as a valuation note (not reflected in the score): ~$16B market cap, P/S ~23x (press-reported) plus a GAAP loss approaches v1.1 research's "expensive and fragile" trap profile.

Observation metrics (quarterly checklist): ① AI Cloud quarterly revenue (Q4 FY2026 $70.5M — confirmed) and the crossover spread ② the QoQ direction of quarterly total revenue (Q4 −5.3% — confirmed) ③ progress on Horizon delivery milestones (480MW 2026 goal — company announcement) ④ quarterly burn of the cash balance ($5.90B + $1.72B restricted — confirmed) ⑤ milestone-fulfillment disclosures on the MSFT/NVDA contracts. Of the five, ①–③ are "execution" metrics; ④–⑤ are "firewall" metrics. If execution metrics rise quarter after quarter while firewall metrics stay intact, 56/C moves upward along its "execution sprint" direction. Conversely, a single quarter of delivery delay turns D1's (−4) warning into real leverage pressure — in a heavy-equipment blitz, one quarter's delay carries a different weight than a software company's one-quarter earnings miss.

One-Line Risk

Under $13.81B in capital commitments (19.5x revenue — confirmed), a one-quarter Horizon delivery delay could shake the entire leverage structure — because a data center cannot be patched like software, and that is the fundamental risk of a heavy-equipment blitz.

BlitzLab v1.0 (frozen). Unconfirmed figures carry estimate/press-reported labels as verified.

Next earnings

Nov 5, 2026 · Expected

Coverage

As of October 2026 · Model v1.0 (frozen) · Confidence: High (listed, SEC filings) · Not investment advice — a scholar's assessment.

BlitzLab scores businesses against a scholar’s standard — not investment advice. No price targets, no buy/sell recommendations. Figures are from SEC filings (XBRL) as labeled in the source scorecard.