Stock Analysis · Bulls vs Bears

IREN Ltd IREN

🐂 Bulls vs Bears 🐻Report #212 Oct 2026
★ Willem owns this stock · 1.8% of the portfolio
69
high risk▲ +2 vs last
🐂 9vs12 🐻
Bears lead
40%
▲ Chance of the biggest tailwind the forecasts miss
35%
▼ Chance of the biggest headwind the forecasts miss
The verdict

The demand is real and filed with the SEC. Contracted value is $16.6B, ~79% of it Microsoft and Nvidia. Microsoft cannot walk away for convenience, and late delivery only costs credits or one tranche. But the next 12 months' commitments ($13.6B) are larger than the funding we can identify (~$11B). The two anchor contracts are priced at about half the "$20M+/MW" management now quotes, and the BC sites just got…

The Bulls vs Bears podcast is being made
Two AI hosts debate this report, NotebookLM-style. It appears here automatically within a few hours.

Written by Claude (Anthropic’s AI) from public sources · Willem owns IREN · AI can make mistakes — numbers in brackets point to the sources · Not financial advice

🐂 Bull arguments 9

  1. 1$16.6B of filed contracts, led by Microsoft and NvidiaStrong
  2. 2Energised Texas power while the state freezes newcomersStrong
  3. 3The Microsoft contract protects IREN better than fearedMedium
  4. 4New contracts quoted at roughly twice the anchor-deal priceMedium
  5. 5$19B raised in 12 months, including investment-grade debtMedium
  6. 6AI cloud cash costs are only ~13% of revenueMedium
  7. 7Milestones met so far: Horizon 1 accepted, Sweetwater energised on scheduleMedium
  8. 8Nvidia is customer, supplier and a possible $2.1B investorMedium
  9. 9At ~$35, every published target is above the priceWeak

🐻 Bear arguments 12

  1. 1Next-12-month commitments exceed the funding in sightHigh
  2. 2Shares up 45% in a year, with 50% more possibleHigh
  3. 3The anchor contracts pay half today's quoted rateMedium
  4. 4The December quarter must roughly triple running capacityMedium
  5. 5GPUs are depreciated over 5 years but contracted for about 4Medium
  6. 6Lowest ClusterMAX tier and a "worst site" verdict for BCMedium
  7. 7Leverage is building, at a higher real cost than quotedMedium
  8. 8Four straight EBITDA misses while costs run ahead of revenueMedium
  9. 9Two customers hold ~79% of contracted valueMedium
  10. 10Texas froze data-centre approvals with no end dateMedium
  11. 11Anthropic's Australian anchor went to a rival developerLow
  12. 12Founders granted ~$320M each, with no insider buyingLow

Bull arguments show how well they survive the devil’s advocate (strong / medium / weak); bear arguments show the damage they could do after the counter-argument (high / medium / low).

Bull case

🐂 Bull #1$16.6B of filed contracts, led by Microsoft and Nvidia

  • The point: IREN's 2026 build is pre-sold under multi-year take-or-pay contracts. The backlog is now visible in audited numbers, not just in press releases.
  • Evidence:
    • At 2026-06-30, remaining performance obligations were $5.1B and lease arrangements ~$11.4B, together ~$16.6B [1]. $0.9B of the $5.1B falls in the next 12 months and $1.3B in months 13–24 [1].
    • Microsoft ($9.7B over 5 years) and Nvidia ($3.4B over 5 years) are "a substantial majority" of contracted revenue [1]. That is ~$13.1B, or ~79% of $16.6B (my arithmetic).
    • "$4bn contracted ARR for 2026 capacity; $1bn operating" at 2026-08-26 [8]. In July it was "~85% under contract" [12].
    • The weighted contract term is ~4 years [12]. Customers prepay 45–55% of GPU capex on recent deals [8,12].
    • Deferred revenue (prepayments received) was $1.84B at 2026-06-30 [1].
  • Why it matters: contracted revenue is what lets IREN borrow against GPUs instead of selling shares. The 10-K backlog makes the >$4B ARR target checkable.
  • 😈 Devil's advocate:
    • Revenue starts only on acceptance, and Microsoft recognised no revenue in FY26 [1].
    • About 65% of the $4B ARR comes from the two contracts priced lowest per MW (Bear #3).
    • The non-hyperscaler remainder sits partly in the BC halls that SemiAnalysis just panned (Bear #6).
  • Verdict: Strong the backlog is audited and the counterparties are top-tier. The risk is delivery and price, not demand.

🐂 Bull #2Energised Texas power while the state freezes newcomers

  • The point: Texas has moved from "slow" to "frozen" for new large loads. IREN already holds energised capacity there.
  • Evidence:
    • On 2026-09-08 IREN said its 2 GW Sweetwater Hub (Sweetwater 1 at 1,400 MW, Sweetwater 2 at 600 MW) was conditionally included in ERCOT Batch Zero as Base Load. 300 MW gross is under construction for delivery in Q4 2027 [41].
    • The Sweetwater 1 substation was energised "on schedule" by May 2026 [9]. Childress (750 MW) runs today [1].
    • Gov. Abbott ordered an audit of all ERCOT data-centre queue projects on 2026-08-03, with a pause on Batch Zero [44]. On 2026-09-21 he ordered TCEQ to halt data-centre permits, and no agency may advance data-centre approvals until the audit is reviewed [42,43]. Queue: ~474 GW, ~90% data centres [44].
    • On the Q4 call, Co-founder Draper said IREN's practices "align" with the directive [21].
  • Why it matters: if new Texas capacity is delayed into 2028+, energised megawatts become scarcer and pricing power for 2027–28 contracts rises.
  • 😈 Devil's advocate:
    • The same freeze can hit IREN's own expansions. The 10-K warns that the PUCT's 2026-08-20 good-cause exception for ERCOT "could delay energization" of IREN projects [1].
    • The inclusion is "conditional", and Batch Zero results (due 2027-04-09) are now uncertain because the study was paused [44].
    • IREN has not said whether the permit halt touches Sweetwater or the Childress conversions (Bear #10).
  • Verdict: Strong energised power is the scarcest input and IREN has it. The upside from the freeze is the part that is uncertain.

🐂 Bull #3The Microsoft contract protects IREN better than feared

  • The point: our last report treated the Microsoft termination right as a major risk. Reading the full contract shows that Microsoft's remedies are narrow.
  • Evidence (Partner SOW, 2025-11-02 [6]):
    • Late delivery earns daily credits (rate redacted). Missing the redacted "Delivery Delay Window" lets Microsoft terminate that tranche only, with the prepayment refunded.
    • Credits and tranche termination are Microsoft's "sole and exclusive remedies" for lateness. Ending one tranche does not end the agreement.
    • Late delivery and service-level misses are expressly not a material breach. Material breach carries a 60-day cure.
    • Microsoft's general termination for convenience is switched off (Exhibit F disapplies PO Terms §7–17).
    • If the contract is terminated for reasons other than IREN's breach, Microsoft owes total contract value less fees paid (§6.c).
    • Excluded delays (force majeure, Microsoft-caused, documented Nvidia supply failures) extend the date with no credits.
  • Why it matters: the worst realistic case for a Horizon slip is credits plus, at most, one ~$2.4B tranche re-marketed. A $9.7B contract cannot unravel on a schedule miss.
  • 😈 Devil's advocate:
    • The credit rate, window length and liability cap are all redacted [6].
    • If a funded tranche is terminated and not re-let, IREN's parent must cover the allocated GPU debt (net of GPU sale proceeds) within 5 business days [1].
    • Microsoft may substitute VR200 GPUs with 4 months' notice, which reopens price and dates [6].
  • Verdict: Medium the contract is sturdier than the headlines suggested. The parent guarantee keeps a tail risk.

🐂 Bull #4New contracts quoted at roughly twice the anchor-deal price

  • The point: pricing for AI compute has risen sharply since IREN signed Microsoft. Every new contract adds more value per MW.
  • Evidence:
    • Management: recent 3-year contracts are above "$20m revenue per MW (IT)" with a ~2-year payback, and talks are near $25M/MW [8].
    • 3-year pricing is up ~125% since November 2025, and 5-year pricing ~70% (call summary) [22].
    • The Microsoft and Nvidia anchors work out at ~$9.7M and ~$11.3M per MW per year ($9.7B ÷ 5 ÷ 200 MW; $3.4B ÷ 5 ÷ 60 MW; my arithmetic from [1,9]).
    • JPMorgan's upgrade cites neocloud pricing moving from $10–15 to $15–20+ per watt [34].
    • Outside data confirms the direction: the SemiAnalysis 1-year H100 index rose from $1.70 (Oct 2025) to $2.35/hour (Mar 2026) [52]. CoreWeave raised some prices ~25% [37]. Nvidia told customers to expect >15% higher prices on 2027 Blackwell/Rubin servers (Bloomberg via secondary) [54].
  • Why it matters: FY28 consensus revenue of ~$7.3B [31] on ~0.8 GW of IT capacity implies only ~$9.1M per MW (my arithmetic). Each block of 2027 capacity signed at $20M/MW would sit well above what analysts appear to model (Tailwind 2).
  • 😈 Devil's advocate:
    • The $20M/MW is management's unaudited number, quoted per MW rather than per GPU-hour [8].
    • SemiAnalysis says IREN sells "rock bottom prices on HGX B200 and B300" at the BC sites [24], which fits poorly with premium pricing there.
    • H100 spot fell from $1.66 to $1.46 in May 2026 (vendor blog) [51].
    • Redburn's sector call is falling GPU rents [36,37].
  • Verdict: Medium direction confirmed, magnitude unverified. The old book dilutes it (Bear #3).

🐂 Bull #5$19B raised in 12 months, including investment-grade debt

  • The point: IREN has shown it can raise very large sums from several kinds of investors.
  • Evidence:
    • The CFO says ~$19B was raised in 12 months: ~$16B from prepayments, GPU financing and converts, ~$3B from equity [8,21].
    • The $3.645B Microsoft GPU financing was rated investment grade, with $2.1B of 5.96% notes and $1.545B DDTL at SOFR+2.25% [1].
    • Mackenzie: $2.4B at 9.0% fixed from Blue Owl and PIMCO for non-investment-grade customers, signed 2026-08-25 [1,8].
    • $6.75B of converts at 0–3.5% coupons, with no maturity before December 2029 [1].
    • The data centres, including Horizons 1–4, are unencumbered [22].
    • Co-CEO Roberts: "The $25–30bn of forecasted FY27 capex isn't an equity number" [23] (secondary).
  • Why it matters: capital is the raw material. The cheaper and more plentiful it is, the less each MW dilutes shareholders.
  • 😈 Devil's advocate:
    • ~$4.7B of FY26 funding came from share sales [1].
    • No data-centre financing has been signed in the six weeks since the call; EDGAR shows no 8-K after 2026-08-27 other than a 13G [20].
    • The stock is now ~$35, against the ATM's average sale price of ~$52.8 [1].
  • Verdict: Medium an impressive record, but the next tranche has to be raised in a worse tape.

🐂 Bull #6AI cloud cash costs are only ~13% of revenue

  • The point: the first audited segment data show very low direct operating costs for GPU rental.
  • Evidence:
    • FY26 AI cloud cost of revenue, excluding D&A, was $16.9M on $128.8M of revenue (electricity $4.5M, staff $8.5M, other $3.9M) [1]. That is a ~87% direct cash margin (my arithmetic).
    • Quarterly AI cloud cost of revenue: $0.7M, $2.4M, $4.6M, $9.2M [3,4,5,1].
  • Why it matters: if direct costs stay near 13%, almost every dollar of contracted revenue is available for debt service, depreciation and overhead. That is why the 1.05x DSCR covenant can work.
  • 😈 Devil's advocate:
    • It excludes D&A, the main cost of a GPU business, and SG&A, which jumped to $128.3M in Q4 (from $81.8M in Q3) and is guided up another $40–50M in Q1 FY27 [1,22].
    • Group Q4 adjusted EBITDA was only $19.2M (14% margin) [8].
    • Segment EBITDA and segment D&A are not disclosed [1].
  • Verdict: Medium a genuine positive data point. It says nothing yet about margins after depreciation.

🐂 Bull #7Milestones met so far: Horizon 1 accepted, Sweetwater energised on schedule

  • The point: IREN has mostly hit its physical milestones, and it has raised its ARR target every quarter.
  • Evidence:
    • The Sweetwater 1 substation targeted April 2026 and was energised "on schedule" [11,9].
    • Horizon 1 was delivered and accepted by Microsoft around 2026-08-13, with Nvidia Exemplar Cloud status on GB300 NVL72 [13].
    • The 2026 ARR target rose from $3.4B (Nov 2025) to $3.4B reaffirmed (Feb) to $3.7B (May) to >$4B (July) to $4B contracted (Aug) [11,10,9,12,8].
    • KPMG gave a clean internal-controls opinion for FY26 [1].
  • Why it matters: in a market short of delivered compute, credible delivery earns the next contract.
  • 😈 Devil's advocate:
    • The earlier ">$500M ARR by Q1 2026" target was reached only around June 2026, about one quarter late if "Q1" meant calendar Q1 [22] (the target's meaning is unverified).
    • Nvidia component shortages delayed Horizon 1 [1].
    • The FY24 accounts were restated, and the previous auditor's report had a going-concern paragraph [19].
  • Verdict: Medium a decent record on smaller steps. The December quarter is the real exam.

🐂 Bull #8Nvidia is customer, supplier and a possible $2.1B investor

  • The point: Nvidia controls GPU allocation and steers demand, and it has tied itself to IREN three ways.
  • Evidence:
    • Nvidia has a ~$3.4B, 5-year cloud contract within 60 MW at Childress, ramping from early 2027 [9].
    • It holds a right to buy 30M shares at $70 (up to $2.1B), vesting with GPUs supplied and expiring 2031-05-07 [1]. Its grant-date fair value of $793.4M was capitalised into GPU cost [1].
    • Nvidia and IREN have a non-binding "5 GW" partnership centred on Sweetwater [1].
    • SemiAnalysis expects "the $2.1B investment from NVIDIA will help them cut less corners" at the new builds [24].
  • Why it matters: a privileged supplier relationship matters most when chips are scarce.
  • 😈 Devil's advocate:
    • JPMorgan called the structure "circular" before its upgrade [39].
    • The $70 strike is about twice today's price, and nothing beyond the $3.4B contract is binding [1].
    • Nvidia backs many neoclouds.
  • Verdict: Medium Weak — valuable but optional.

🐂 Bull #9At ~$35, every published target is above the price

  • The point: the share price already discounts a lot of bad news.
  • Evidence:
    • Close $34.80 (2026-10-12), against a 52-week high of $76.87; market cap ~$13.7B [32,33].
    • Average target $78.14 from 20 analysts, with a low of $40 (Redburn, Neutral) [32,36].
    • Ratings: 15 Buy or Strong Buy, 5 Hold, 0 Sell [32].
    • Enterprise value ~$14–15.5B (depending on whether restricted cash is counted) is ~5x FY27 consensus revenue of ~$2.8B and ~2x FY28 ~$7.3B (my arithmetic from [1,31,32]).
  • Why it matters: if delivery comes good, the gap between price and consensus is large.
  • 😈 Devil's advocate:
    • Consensus has a poor record here (four EBITDA misses, Bear #8), and FY27 revenue consensus fell ~8% since June [31].
    • Short interest is ~21.8% of float [33].
    • Price targets lag the stock.
  • Verdict: Weak a sentiment point, not a fundamental one.

Bear case

🐻 Bear #1Next-12-month commitments exceed the funding in sight

  • The point: IREN has signed for more spending in the next year than it has identified money for, before counting the rest of the FY27 plan.
  • Evidence:
    • Contractual commitments at 2026-06-30 were $13.81B, of which $13.61B is due within 12 months [1, Note 29].
    • Identified sources total ~$11.0B (my arithmetic):
      • $5.9B unrestricted cash.
      • $2.707B undrawn Microsoft GPU financing, conditional on tranche acceptance.
      • $2.4B Mackenzie loan, funded as GPUs are accepted [1].
    • That leaves a gap of ~$2.6B before new prepayments or ATM sales.
    • FY27 capex guidance is $25–30B. Funding: ~$14B "in hand" (which includes the $7.6B June cash, partly spent since) and ~$8B "targeted", leaving $3–8B with no named source [22,23].
    • FY26 free cash flow was −$2.23B, or −$4.08B excluding prepayments (my arithmetic from [1,3,4,5]).
    • No data-centre or corporate financing has been filed since 2026-08-27 [20].
    • The 10-K adds a new risk factor: GPUs and sites may be committed "before financing is arranged and, in some cases, before customer contracts are signed" [1].
  • Why it matters: the gap will be filled by prepayments, debt or shares. At $35 a share, every $1B of equity is 29M shares (7%).
  • 🛡️ Counter-argument:
    • Commitments include GPUs for contracts that carry 45–55% prepayments, so new deals partly fund themselves [8].
    • The data centres are unencumbered collateral [22].
    • The 10-K states that funds are sufficient for 12 months, with no going-concern doubt [1].
  • Severity: High the counter-argument covers part of the gap, but the plan relies on financing that is not yet signed.

🐻 Bear #2Shares up 45% in a year, with 50% more possible

  • The point: dilution has been the price of growth, and the overhang is large.
  • Evidence:
    • Shares went from 271.98M (2025-08-15) to 394.06M (2026-08-14), up 44.9% [2,1].
    • FY26 issuance [1]:
      • ATM: 66.7M shares for $1.0B, plus 47.2M for $2.49B.
      • 39.7M shares in a registered direct offering at $41.12.
      • 12.6M shares for Mirantis.
    • Overhang as % of 394M (my arithmetic from [1]):
      • Converts: 127.2M shares (32.3%) at initial conversion prices. The 2029/2030 notes (29.7M shares at $13.64–16.81) are deep in the money and their conversion trigger is met.
      • Nvidia right: 30M (7.6%).
      • RSUs, options and the co-CEO awards: 44.4M (11.3%).
      • Total ~201.6M, or ~51%, before the capped-call offset (cap prices not found).
    • ~$3.5B of ATM remains, ≈100M shares at $34.80 (25%) [1].
    • SBC was $205.0M in FY26 versus $42.6M in FY25 [1].
  • Why it matters: the business can hit its ARR targets and per-share value can still lag.
  • 🛡️ Counter-argument:
    • No ATM shares were sold after 2026-08-14 as far as filings show [1,20].
    • The big 2031–2033 converts ($6.3B) convert only at $51.40–85.63.
    • The co-CEOs' awards carry 2-year post-vest holds [14].
  • Severity: High the history and the overhang both point the same way.

🐻 Bear #3The anchor contracts pay half today's quoted rate

  • The point: the two biggest contracts were signed before prices rose. They lock ~65% of 2026 ARR at roughly half the rate management now quotes.
  • Evidence:
    • Microsoft: ~$9.7M per MW per year. Nvidia: ~$11.3M per MW per year (my arithmetic from [1,9]).
    • Microsoft's ~$1.94B a year plus Nvidia's ~$0.68B is ~$2.62B of the $4B ARR, or ~65%.
    • An unnamed bearish note on 2026-10-08 flagged the "older contract book priced below current rates" [27].
    • Bernstein estimated revenue of about $10.4M per MW for FY24–25 (secondary) [28].
  • Why it matters: the "~2-year payback" claim applies to new 3-year deals, not to the bulk of contracted revenue. Returns on the Microsoft GPUs depend on the investment-grade debt and prepayments that fund ~95% of their cost.
  • 🛡️ Counter-argument:
    • Five-year terms with the best credits deserve lower prices.
    • Microsoft's 20% prepayment plus the ~7% financing means little IREN equity sits in those GPUs.
  • Severity: Medium High — it caps the blended return on capital through 2031.

🐻 Bear #4The December quarter must roughly triple running capacity

  • The point: IREN must go from ~$1B operating ARR to $4B in about four months.
  • Evidence:
    • Status [8]: Horizon 2 commissioning; Horizons 3–4 in late-stage construction, all targeted for the December quarter. Mackenzie's first two buildings racked; Prince George fully commissioned.
    • The CFO says most December-quarter capacity comes online late, so revenue lands mainly in the March quarter [21].
    • Draws on the Microsoft financing ($2.7B undrawn) and the Mackenzie loan depend on customer acceptance [1].
    • Microsoft's remedies: credits, plus tranche termination after the delay window with prepayment refunded [6].
    • No acceptance 8-K for Horizons 2–4 yet [20].
  • Why it matters: acceptance unlocks revenue and the debt that pays for the next phase. A slip delays both.
  • 🛡️ Counter-argument:
    • Horizon 1 was the template and was accepted [13].
    • Delay remedies are narrow (Bull #3).
    • The grace period runs into early 2027.
  • Severity: Medium High.

🐻 Bear #5GPUs are depreciated over 5 years but contracted for about 4

  • The point: IREN carries residual-value risk on fast-ageing hardware.
  • Evidence:
    • Hardware is depreciated over 5 years, against a ~4-year weighted contract term [1,12].
    • The 10-K adds a risk factor on GPU obsolescence and shorter useful lives [1].
    • Rubin production shipments start Q3 2026, ramping in Q4 and widely deployed in 2027 [54].
    • Microsoft can swap GB300 for VR200 with notice [6].
    • Silicon Data: H100 hyperscaler median fell from ~$9.34/hour (H2 2024) to ~$6.26 (H2 2025). Marketplace median ~$1.95 (H2 2025). Neocloud median held ~$3 [51].
  • Why it matters: if second-life rents are low, the 5-year depreciation flatters earnings and impairments follow. IREN has already written off $638.8M of mining hardware in FY26 [1].
  • 🛡️ Counter-argument:
    • Prepayments of 45–55% plus a ~2-year payback on new deals recover most of the cost inside the contract [8].
    • Neocloud H100 pricing held ~$3/hour through 2025 [51].
  • Severity: Medium the risk is real but mostly lands from 2029.

🐻 Bear #6Lowest ClusterMAX tier and a "worst site" verdict for BC

  • The point: the industry's most-read GPU cloud rater says IREN's BC halls are unreliable and its managed-cloud claims overstated.
  • Evidence:
    • SemiAnalysis's ClusterMAX 3.0 (2026-09-23) put IREN in "Not Recommended – Underperforming", the lowest tier [24].
      • Quotes: "Multi-day power outages, network upgrades, storage failures, air quality controls leading to link flaps... The #1 worst site in the industry according to users."
      • It called Prince George "N+0" with a single ISP, said IREN sells "rock bottom prices" on B200/B300 there, and recommended IREN "stop pretending to offer managed clusters and inference endpoints" [24].
    • Its X thread on 2026-10-07 coincided with −6.3% and −7.7% days. The stock is down >20% in ten sessions [26,27,28].
    • No IREN rebuttal was found as of 2026-10-12. The only public response was a restated Mirantis/Nvidia hypervisor validation, already mentioned on the August call [30].
    • Exposure:
      • Prince George carried ~$0.5B of contracted ARR in May [9].
      • Mackenzie backs the $2.4B, 9% loan, which funds as GPUs are accepted [1,8].
      • Together AI resold a B300 cluster traced to Prince George [24].
  • Why it matters: non-hyperscaler customers buy on reliability and price. A bad reputation can cost renewals or force discounts on ~35% of ARR outside Microsoft and Nvidia.
  • 🛡️ Counter-argument:
    • The evidence is customer anecdote and reseller testing, with no audited uptime [29].
    • SemiAnalysis says Childress and Sweetwater "look much better" [24].
    • No customer termination has been disclosed [29].
    • The "Underperforming" tier is defined as able to "quickly rise to Bronze by fixing one or more critical issues" [24].
  • Severity: Medium the BC capacity is a minority of ARR, but silence from management is a signal in itself.

🐻 Bear #7Leverage is building, at a higher real cost than quoted

  • The point: IREN is now a leveraged GPU financier, and the effective rates are higher than the headline.
  • Evidence:
    • Debt principal is $7.71B ($6.77B recourse to the parent), plus $270M of guaranteed GPU finance leases [1].
    • Net debt is ~$1.8B, ~7.4x FY26 adjusted EBITDA of $245.7M (my arithmetic). Q4 EBITDA annualised is only ~$77M [8].
    • The "6.0%" Microsoft financing has effective rates of ~7.1% after fees and swaps [1].
    • Mackenzie is 9.0% fixed and amortises, with each draw due 30 months after funding [1].
    • Covenants [7]:
      • Minimum DSCR (cash flow ÷ debt service) of 1.05x, with a cash trap below 1.10x.
      • Parent hedge, remarketing and managed-services guarantees.
    • The 10-K flags possible cross-default across financings [1].
  • Why it matters: with a 1.05x DSCR, a few points of utilisation or credits can trap cash in a financing silo. Short 30-month Mackenzie draws must be repaid or refinanced quickly.
  • 🛡️ Counter-argument:
    • Scheduled principal is only $177M in FY27 and $270M in FY28 [1].
    • No convertible matures before December 2029 [1].
    • The forward EBITDA consensus (~$683M in the March 2027 quarter alone) would cut leverage quickly if delivered [31].
  • Severity: Medium

🐻 Bear #8Four straight EBITDA misses while costs run ahead of revenue

  • The point: reported profit keeps undershooting forecasts while overhead rises ahead of revenue.
  • Evidence:
    • Q4 adjusted EBITDA was $19.2M against $41.8M consensus, a miss in each of the last four quarters (TIKR) [31].
    • Adjusted EBITDA margin fell from 38–41% to 14% in Q4 [8].
    • FY26 operating loss was −$1.05B [1].
    • Q1 FY27 cash SG&A is guided up another $40–50M sequentially [22].
    • The shares fell 12.5% after the Q4 print [40].
  • Why it matters: Q1 FY27 (~2026-11-05, unconfirmed) is likely to show a further EBITDA trough before Microsoft revenue lands. Consensus for the September quarter is only ~$65M EBITDA [31].
  • 🛡️ Counter-argument: mining revenue is being switched off by design, so the trough is a transition effect. Consensus EBITDA rises to ~$193M in the December quarter and ~$683M in the March quarter [31].
  • Severity: Medium Low — mostly a timing and credibility issue until March 2027.

🐻 Bear #9Two customers hold ~79% of contracted value

  • The point: concentration is extreme, and both customers have alternatives.
  • Evidence:
    • Microsoft plus Nvidia are ~79% of the $16.6B of contracted value; Microsoft alone is ~58% (my arithmetic from [1]).
    • Microsoft holds a change-of-control termination right if Amazon, Oracle or Alphabet gains control of IREN (from report #1's reading of the SOW).
    • Jones Trading's Hold (2026-09-28) cites customer creditworthiness among the smaller AI labs [38].
  • Why it matters: renewal terms in 2031 and any renegotiation depend on two buyers. The change-of-control clause narrows the list of possible acquirers.
  • 🛡️ Counter-argument:
    • The contracts are long and prepaid, and the customer list has grown to 13+ named customers [8,12].
  • Severity: Medium

🐻 Bear #10Texas froze data-centre approvals with no end date

  • The point: the same policy that protects incumbents may delay IREN's next Texas phases.
  • Evidence:
    • The 2026-09-21 directive halts TCEQ permits and all state data-centre approvals until the audit is reviewed. There is no end date and no listed exemptions; TCEQ reports to the governor by 2026-10-19 [42,43].
    • The PUCT must finish its review of how large-load transmission costs are allocated by 2026-12-31 under SB6 [1].
    • The legislature meets in January 2027 [44].
    • The 10-K says the August good-cause exceptions "could delay energization" of IREN projects [1].
    • IREN has not quantified its exposure (not found).
  • Why it matters: the Sweetwater 1 300 MW phase (Q4 2027) and the Childress air-cooled conversions underpin 2027–28 growth.
  • 🛡️ Counter-argument: Childress and the Sweetwater 1 substation are already energised. Projects that are purely behind-the-meter or already connected "may not be subject" to the pause [44].
  • Severity: Medium Low for 2026–27; higher for 2028.

🐻 Bear #11Anthropic's Australian anchor went to a rival developer

  • The point: the most visible new growth lead, Anthropic's Australian tender, went elsewhere.
  • Evidence:
    • In July the AFR reported IREN was shortlisted for Anthropic's ≥1.4 GW Australian tender; the stock rose ~13% [48].
    • On 2026-09-16 reports said Anthropic signed its first Australian lease as anchor tenant of Zerra DC's 2.16 GW Western Downs Digital Park in Queensland (Macquarie, Dexus's ADC), subject to FIRB and council approval [49,50].
    • IREN's 800 MW Bundey site in South Australia was its pitch [48].
  • Why it matters: Bundey loses its most obvious anchor, and the land bank's option value outside Texas looks further away.
  • 🛡️ Counter-argument:
    • The tender was expected to be split into four or five contracts, so other awards may remain [48].
    • Neither Anthropic nor Zerra has confirmed the lease [49].
  • Severity: Low this was never in consensus.

🐻 Bear #12Founders granted ~$320M each, with no insider buying

  • The point: the Roberts brothers control the vote and received very large awards while outside holders were diluted.
  • Evidence:
    • Each co-CEO received 9,099,328 RSUs on 2026-07-01: ~$317M each at $34.80 (my arithmetic). They vest over 4 years with a 2-year hold [14,16].
    • The independent directors defended the award in a 2026-07-09 letter [15].
    • There were no open-market insider purchases or sales in calendar 2026 through 2026-10-12 [16,33].
    • Class B shares carry 15 votes per ordinary share held (report #1).
  • Why it matters: capital allocation is the decisive skill here, and outside holders have little say over it.
  • 🛡️ Counter-argument: each founder holds ~23.6M shares [16], and holding periods run to ~FY2033. There are no further grants until FY2031 [14].
  • Severity: Low Medium.

Tailwinds NOT in the forecasts

Baseline: FY27 (Jul-26 to Jun-27) consensus revenue is $2.8B (16 analysts) and FY28 ~$7.3B (17) [31,32]. FY29 is ~$14B per TIKR (10 analysts) [31], but another source shows $8.4B from 5 analysts (unverified snippet). FY27 EPS consensus is about −$3.42 [32], so I use EBITDA per share. FY28 consensus EBITDA was not found. My estimate is ~$4.4B ($7.3B × ~60% margin), or **$11.17 per share on 394.06M shares**. All per-share impacts below use 394.06M shares unless stated.

  • A binding ≥300 MW (IT) Sweetwater anchor contract — 40% chance within 36 months

    • Impact on revenue: +$2.25B (+16%) vs FY29 consensus ~$14B (TIKR).

    • Impact on EPS: +$4.28 EBITDA per share in FY29 (+~25% vs ~$17 of FY29 EBITDA per share at a 50% margin, estimate).

    • How we got the number (estimate): 300 MW × $15M/MW (below the quoted $20M) = $4.5B, of which I assume half is already in the higher consensus → $2.25B × 75% incremental margin ÷ 394M ≈ +$4.28.

    • Expected value: 0.40 × $4.28 ≈ +$1.71 EBITDA per share.

    • What it is: a hyperscaler, Nvidia (under the non-binding 5 GW pact) or a frontier lab contracts a large block of Sweetwater 1 for a 2028 start.

    • Why it's not priced in: only the high-end analysts model Sweetwater, with no named customer. The 5 GW Nvidia partnership is explicitly non-binding [1].

    • Trigger to watch: an 8-K naming a Sweetwater customer; Nvidia's $70 right starting to vest; Batch Zero results.

    • Why this probability: energised power in a frozen state is exactly what buyers want. But Anthropic chose a developer in Australia, and a GW-scale deal needs financing IREN hasn't yet done, so it is below a coin flip.

  • 2027 air-cooled capacity contracted at ≥$20M/MW instead of the ~$9M/MW consensus implies — 30% chance by end-2027

    • Impact on revenue: +$2.0B (+27%) vs FY28 consensus ~$7.3B.

    • Impact on EPS: +$4.31 EBITDA per share in FY28 (+39% vs the ~$11.17 estimate).

    • How we got the number (estimate): half of ~0.5 GW of new 2027 capacity = 250 MW × ($20M − $12M assumed modelled price) = $2.0B × 85% (price uplift carries little cost) ÷ 394M ≈ +$4.31.

    • Expected value: 0.30 × $4.31 ≈ +$1.29.

    • What it is: IREN signs the ~0.5 GW of 2027 capacity it calls "late-stage" at the prices it quotes for new 3-year deals.

    • Why it's not priced in: FY28 consensus ÷ ~0.8 GW ≈ $9.1M/MW, close to the Microsoft rate. Analysts appear not to model the new pricing. The consensus price per MW is not published, so this is an inference.

    • Trigger to watch: contract announcements with $/MW or prepayment disclosures; the Q1 FY27 call.

    • Why this probability: the pricing direction is supported, but the BC "rock bottom" pricing and Redburn's deflation thesis suggest not all new capacity gets $20M.

  • A large non-dilutive data-centre financing (≥$3B) replaces equity — 55% chance within 12 months

    • Impact on revenue: none.

    • Impact on EPS: +$2.00 FY28 EBITDA per share (+22%) vs an equity-funded alternative. FCF per share is lower by ~$0.69 for interest.

    • How we got the number (estimate): $3B ÷ $34.80 ≈ 86M shares avoided; $4.4B ÷ 394M = $11.17 vs ÷ 480M = $9.17 → +$2.00. Interest $3B × 9% = $270M ÷ 394M ≈ $0.69.

    • Expected value: 0.55 × $2.00 ≈ +$1.10.

    • What it is: a loan or sale-leaseback on the unencumbered Childress and Sweetwater halls.

    • Why it's not priced in: the market is pricing equity funding: the stock fell >20% in ten sessions with ATM dilution named in bear notes [27,28]. Consensus targets use today's share count, though.

    • Trigger to watch: an 8-K Item 1.01 for a data-centre facility; the Q1 call (~2026-11-05, unconfirmed).

    • Why this probability: peers have raised this kind of debt against contracted halls and management says it is "not an equity number" [23]. But six weeks have passed with nothing filed, and lenders prefer long leases over 3–5-year GPU contracts.

  • Nvidia exercises a meaningful part of its $70 right — 10% chance within 36 months

    • Impact on revenue: none.

    • Impact on EPS: about +$0.35 FCF per share from interest avoided, before the 7.6% share increase (roughly offsetting). Net: not meaningfully quantifiable.

    • How we got the number (estimate): $2.1B × ~7.1% ≈ $149M ÷ 424M shares ≈ +$0.35.

    • Expected value: ≈ +$0.04.

    • What it is / why not priced / trigger: a $2.1B equity injection at about twice today's price, dependent on GPU volume. Watch for vesting disclosure in a 10-Q.

    • Why this probability: it needs both the share price above $70 and very large Nvidia-supplied deployments.

Headwinds NOT in the forecasts

  • GPU rental prices fall >30% before the 2027 capacity is contracted — 35% chance by end-2027

    • Impact on revenue: −$1.37B (−19%) vs FY28 consensus ~$7.3B.

    • Impact on EPS: −$3.13 FY28 EBITDA per share (−28% vs the ~$11.17 estimate).

    • How we got the number (estimate): 0.5 GW × $9.1M/MW × 30% = $1.37B × 90% (price cut falls to EBITDA) ÷ 394M ≈ −$3.13.

    • Expected value: 0.35 × −$3.13 ≈ −$1.10.

    • What it is: Rubin supply plus CoreWeave's $35–39B capex and the hyperscaler builds flood the market.

    • Why it's not priced in: consensus and management's "~$25M/MW talks" both assume firm prices.

    • Trigger to watch: the SemiAnalysis and Silicon Data indices; IREN prepayments falling below 45%; peer price cuts.

    • Why this probability: every Nvidia generation has cut the last one's rent. But 2027 hyperscaler capex near $1–1.2T [55] and server price rises argue against a fast collapse.

  • Credit markets close to non-investment-grade GPU debt for 12 months — 20% chance

    • Impact on revenue: −$2.28B (−31%) vs FY28 consensus.

    • Impact on EPS: −$4.91 FY28 EBITDA per share (−44%).

    • How we got the number (estimate): ~0.25 GW of 2027 capacity delayed × $9.1M/MW = $2.28B × 85% ÷ 394M ≈ −$4.91.

    • Expected value: 0.20 × −$4.91 ≈ −$0.98.

    • What it is: an AI sentiment shock or a peer default stalls the ~$8B of "targeted" GPU financing.

    • Why it's not priced in: the funding plan assumes continued access [22].

    • Trigger to watch: CoreWeave's 9.6–9.75% notes [53]; the cost of the next Mackenzie-style loan; any failed raise.

    • Why this probability: IREN raised $2.4B at 9% in August, but the sector's debt stacks make a pause plausible within a year.

  • Heavy further dilution: more than 60M new shares within 12 months — 55% chance

    • Impact on revenue: none.

    • Impact on EPS: −$1.48 FY28 EBITDA per share (−13%).

    • How we got the number (estimate): $4.4B ÷ 394M = $11.17 vs ÷ 454M = $9.69 → −$1.48.

    • Expected value: 0.55 × −$1.48 ≈ −$0.81.

    • What it is: ATM sales, a direct offering or an equity-linked deal to close the ~$2.6B commitment gap and the $3–8B FY27 gap.

    • Why it's not priced in: consensus per-share targets use ~394M shares.

    • Trigger to watch: share count on the Q1 FY27 10-Q cover; 424B5 supplements.

    • Why this probability: ~$3.5B of ATM is registered and 137M shares were issued in 14 months. Only a data-centre loan avoids it, and that is a near coin flip (55%).

  • Texas freeze delays Sweetwater 1's 300 MW phase by two quarters — 20% chance

    • Impact on revenue: −$1.05B (−14%) vs FY28 consensus.

    • Impact on EPS: −$2.27 FY28 EBITDA per share (−20%).

    • How we got the number (estimate): ~230 MW IT × $9.1M × 0.5 year = $1.05B × 85% ÷ 394M ≈ −$2.27.

    • Expected value: 0.20 × −$2.27 ≈ −$0.45.

    • What it is: the TCEQ halt, the audit or the transmission-cost review holds up permits for the Q4 2027 phase.

    • Why it's not priced in: the delivery date is unchanged and IREN says its practices "align" [21].

    • Trigger to watch: the TCEQ update due 2026-10-19; the PUCT decision by 2026-12-31; IREN commentary.

    • Why this probability: the substation is already energised, which lowers the risk. But the halt has no end date and covers "all regulatory approvals".

  • Horizons 2–4 slip a quarter beyond the December quarter — 25% chance

    • Impact on revenue: −$363M (−13%) vs FY27 consensus ~$2.8B.

    • Impact on EPS: −$0.78 FY27 EBITDA per share. A % is not computable because FY27 EBITDA consensus was not found.

    • How we got the number (estimate): Microsoft $1.94B/yr × 3/4 (Horizons 2–4) × 1/4 year = $363M × 85% ÷ 394M ≈ −$0.78.

    • Expected value: 0.25 × −$0.78 ≈ −$0.20. I put termination of at least one tranche at ~6%.

    • What it is: commissioning or Nvidia component delays push acceptance into the June quarter, with delay credits.

    • Why it's not priced in: consensus March-quarter EBITDA of ~$683M assumes the ramp [31].

    • Trigger to watch: acceptance 8-Ks like the one on 2026-08-13; wording on the Q1 call.

    • Why this probability: Horizon 1 slipped on components, and three buildings at once is harder. But Horizon 2 was already commissioning in August.

  • BC reliability costs credits, churn or price cuts — 30% chance within 12 months

    • Impact on revenue: −$100M (−1.4%) vs FY28 consensus.

    • Impact on EPS: −$0.22 FY28 EBITDA per share (−2%).

    • How we got the number (estimate): 20% × ~$0.5B Prince George contracted ARR = $100M × 85% ÷ 394M ≈ −$0.22.

    • Expected value: 0.30 × −$0.22 ≈ −$0.07.

    • What it is: customers claim credits, fail to renew or demand discounts. Mackenzie acceptance could also slow the 9% loan draws.

    • Why it's not priced in: guidance includes the BC capacity and IREN has not addressed the claims.

    • Trigger to watch: an IREN response; BC ARR on the Q1 call; any ClusterMAX upgrade.

    • Why this probability: user reports are consistent across sources, but the exposed revenue is small.

Investment Risk Score: 69/100 (high)

# Factor Weight Measured value Score Why
1 Liquidity & funding runway 15% Unrestricted cash $5.9B; FY26 FCF −$2.23B (−$4.08B ex-prepayments); 12-month commitments $13.6B vs ~$11.0B identified sources (cash + $2.7B MSFT DDTL/USPP + $2.4B Mackenzie); FY27 capex $25–30B with $3–8B unnamed; no going-concern language 66 Effectively 12–36 months of runway if financing keeps coming. The build is committed, but the cash pile and conditional facilities keep it out of the 81+ band.
2 Leverage 10% Debt $7.71B (+$270M leases); net debt ~$1.8B ≈ 7.4x FY26 adj. EBITDA ($245.7M), Q4 run-rate far worse; effective 7.1% / 9% on GPU loans; DSCR covenant 1.05x 72 Above 5x on trailing EBITDA. Kept below 81 because most debt is 0–3.5% converts maturing 2029–33 and forward EBITDA consensus rises sharply.
3 Dilution (last 12 months + overhang) 10% Shares +44.9% y/y (271.98M → 394.06M); overhang ~201.6M (51%): converts 127.2M, Nvidia 30M, RSUs/options 44.4M; ATM ~$3.5B ≈ 100M sh 88 Well over +25% realised, plus a large overhang. Most converts are out of the money, so not 95+.
4 Profitability 10% FY26 FCF margin −316%; operating margin −148%; adj. EBITDA margin 35% FY26 / 14% Q4; AI cloud direct cash cost 13% of revenue 76 Deep GAAP and FCF losses, improving as contracted revenue arrives; adjusted EBITDA is positive.
5 Revenue quality & concentration 10% Microsoft ~58% and Microsoft + Nvidia ~79% of $16.6B contracted; mining (spot) still 49% of Q4 revenue, ending Dec 2026; contracts take-or-pay, ~4-yr term 72 Over 35% from one customer, offset partly by long prepaid contracts with top credits.
6 Execution & track record 10% Horizon 1 accepted; Sweetwater energised on schedule; ARR target raised 4 times; ">$500M by Q1 2026" ~1 quarter late; 4 straight EBITDA misses vs consensus; FY24 restatement (remediated); BC reliability claims 58 Mixed: physical milestones mostly met, financial forecasts mostly missed, with a big step-up ahead.
7 Competitive position & disruption 10% ClusterMAX "Underperforming" (lowest tier) vs CoreWeave/Nebius Platinum; 4 GW+ of power, energised in Texas; ARR target $4B vs CoreWeave $18.5–19.5B, Nebius $7–9B 62 Contested, with a power edge but a weak software and service reputation. Losing on quality at BC, winning on power.
8 Cyclicality & commodity/macro sensitivity 10% Fully tied to the AI capex cycle; GPU rents fell ~33% at hyperscalers in 2025 [51]; beta ~4.3; 2022 crash caused a default on $103M of non-recourse debt [56,57] 68 Highly cyclical end-market, partly cushioned by 3–5-year take-or-pay contracts.
9 Regulatory, legal & geopolitical 5% Texas permit halt (no end date); BC Hydro caps; US Customs ~$100M origin dispute; CRA ~$27.6M; securities appeal pending (3rd Cir.); Section 232 chips with data-centre exemption 55 Material but contained; no single ruling is existential.
10 Valuation & expectations 10% EV ~$14–15.5B ≈ 5x FY27 / 2x FY28 consensus revenue; price $34.80 vs avg target ~$78; short interest ~21.8% of float, 2.3 days to cover; −20% in 10 sessions 66 Not expensive on FY28, but heavy short interest and a story under public attack make expectations fragile.

Overall: weighted average = 0.15×66 + 0.10×72 + 0.10×88 + 0.10×76 + 0.10×72 + 0.10×58 + 0.10×62 + 0.10×68 + 0.05×55 + 0.10×66 = 9.90 + 7.20 + 8.80 + 7.60 + 7.20 + 5.80 + 6.20 + 6.80 + 2.75 + 6.60 = 68.85 → 69.

Change vs last: +2 vs 67 (old 6-area method).

  • Most of the change is the method: v3 scores dilution, leverage and profitability separately, and all three sit at the bad end.
  • New evidence cuts both ways:
    • Worse: the 12-month commitment gap is now measured (~$2.6B); effective loan rates are ~7.1% rather than 6%; ClusterMAX's lowest tier is confirmed in SemiAnalysis's own text; Anthropic went to Zerra.
    • Better: the Microsoft contract has no termination for convenience and narrow delay remedies.

Ranking check: IREN sits between IPX (67) and WYFI (70). Its $25–30B build, $7.7B of debt and +45% share count carry more balance-sheet risk than IPX's debt-light pre-revenue plant. But it is safer than WYFI, whose own 10-Q says resources "may not be sufficient", because IREN has $5.9B of cash and investment-grade-financed Microsoft revenue.

Conclusion

  • What you are really betting on: that IREN can turn energised Texas power into accepted, financed GPU capacity faster than its share count grows. And that the next 0.5 GW is signed at the new ~$20M/MW pricing, not the ~$10M/MW of its anchor deals.
  • The 2–3 things that will decide the outcome:
    1. Acceptance of Horizons 2–4 and the BC capacity by ~March 2027. This unlocks $2.7B of Microsoft financing, the Mackenzie draws, and the jump from $1B to $4B operating ARR.
    2. How the ~$2.6B near-term and $3–8B FY27 funding gaps are filled: data-centre debt or shares at ~$35.
    3. Contract pricing and prepayments on 2027 capacity and Sweetwater, against a possible GPU rent downturn as Rubin ships.
  • Signposts:
    • Confirming the bull case:
      • Acceptance 8-Ks for Horizons 2, 3 and 4 by 2027-03-31.
      • A ≥$3B data-centre financing with no share issuance.
      • New contracts disclosed at ≥$20M/MW with ≥45% prepayments.
      • Share count ≤ ~410M on the next two 10-Q covers.
    • Breaking it:
      • Any Microsoft tranche terminated or past the delay window.
      • Shares above ~455M by mid-2027 without new contracts.
      • BC customer credits or churn disclosed, or IREN's silence on SemiAnalysis continuing past the Q1 call.
      • The next GPU loan priced above ~10%, or prepayments below 30%.
  • New since the last report:
    • The full Microsoft SOW shows no termination for convenience. Late delivery means credits or termination of one tranche. A non-breach termination makes Microsoft owe the contract value. The termination risk is smaller than we wrote, but the parent remarketing guarantee is now documented.
    • Measured funding gap: $13.6B of 12-month commitments against ~$11.0B of identified sources. The "6%" Microsoft financing costs ~7.1% effective. The anchor contracts price at ~$9.7–11.3M/MW, about half the quoted new-deal rate.
    • SemiAnalysis's own ClusterMAX 3.0 text confirms the lowest tier and alleges "rock bottom" BC pricing; IREN has not responded. Anthropic's Australian anchor went to Zerra DC. Texas escalated from an audit to a permit halt.

Open questions & data gaps

  • Microsoft redactions (credit rate, delay window, liability cap, tranche dates): still redacted [6]. High importance for sizing a Horizon slip.
  • Capex split and uncovered funding: the GPU vs data-centre split of the $25–30B is not in primary documents. Secondary sources say GPUs are about two-thirds [23]. High importance.
  • BC exposure: no site-level ARR split, and no customer credits disclosed or ruled out. Whether BC Hydro's competition rules cover a mining-to-AI "change in end use" at IREN's existing capacity is not found [46]. Medium.
  • Consensus EBITDA/EPS for FY28–29: not found. Aggregators conflict (FY28 revenue $5.7B vs $7.3B; FY27 EPS −$3.42 vs +$0.23). My FY28 EBITDA base of ~$4.4B is an estimate, so all EBITDA-per-share impacts carry that uncertainty. Medium.
  • Capped-call cap prices on the converts: not found. This affects the true dilution overhang. Low-Medium.
  • Call transcripts: the Q4 FY26 Q&A was read only through a summarised third-party copy. The Q1/Q2 FY26 transcripts were again not read in full.
  • Mackenzie loan documents: not filed as an exhibit yet, and no 8-K Item 1.01 found. Cross-default terms are unverified.
  • Price data: the 2026-10-12 price ($34.80) is from an aggregator and may be intraday.
Sources
  1. IREN Ltd Form 10-K, fiscal year ended 2026-06-30 — SEC EDGAR — 2026-08-27 — https://www.sec.gov/Archives/edgar/data/1878848/000187884826000052/iren-20260630.htm
  2. IREN Ltd Form 10-K, fiscal year ended 2025-06-30 — SEC EDGAR — 2025-08-28 — https://www.sec.gov/Archives/edgar/data/1878848/000187884825000063/iren-20250630.htm
  3. Form 10-Q, quarter ended 2025-09-30 — SEC EDGAR — 2025-11-06 — https://www.sec.gov/Archives/edgar/data/1878848/000187884825000081/iren-20250930.htm
  4. Form 10-Q, quarter ended 2025-12-31 — SEC EDGAR — 2026-02-05 — https://www.sec.gov/Archives/edgar/data/1878848/000187884826000015/iren-20251231.htm
  5. Form 10-Q, quarter ended 2026-03-31 — SEC EDGAR — 2026-05-08 — https://www.sec.gov/Archives/edgar/data/1878848/000187884826000026/iren-20260331.htm
  6. Microsoft–IE US Hardware 3 Partner Statement of Work (10-Q Ex. 10.3, redacted) — SEC EDGAR — dated 2025-11-02, filed 2026-02-05 — https://www.sec.gov/Archives/edgar/data/1878848/000187884826000015/exhibit103-partnerstatem.htm
  7. Common Terms Agreement, Microsoft GPU financing (10-K Ex. 10.37) — SEC EDGAR — dated 2026-05-29, filed 2026-08-27 — https://www.sec.gov/Archives/edgar/data/1878848/000187884826000052/a1037commontermsagreemen.htm
  8. IREN Reports FY26 Results (8-K Ex. 99.1) — SEC EDGAR — 2026-08-27 — https://www.sec.gov/Archives/edgar/data/1878848/000187884826000051/irenreportsfy26results.htm
  9. IREN Business Update and Q3 FY26 Results (8-K Ex. 99.1) — SEC EDGAR — 2026-05-07 — https://www.sec.gov/Archives/edgar/data/1878848/000187884826000025/irenreportsq3fy26results.htm
  10. IREN Reports Q2 FY26 Results (8-K Ex. 99.1) — SEC EDGAR — 2026-02-05 — https://www.sec.gov/Archives/edgar/data/1878848/000187884826000014/a991irenreportsq2fy26res.htm
  11. IREN Reports Q1 FY26 Results (8-K Ex. 99.1) — SEC EDGAR — 2025-11-06 — https://www.sec.gov/Archives/edgar/data/1878848/000187884825000079/q1fy26resultspressreleas.htm
  12. IREN Signs $2.8bn in New Customer Contracts, Raises 2026 ARR Target to over $4bn (8-K Ex. 99.1) — SEC EDGAR — 2026-07-20 — https://www.sec.gov/Archives/edgar/data/1878848/000114036126028871/ef20078253_ex99-1.htm
  13. IREN Delivers Horizon 1 to Microsoft (8-K Ex. 99.1) — SEC EDGAR — 2026-08-13 — https://www.sec.gov/Archives/edgar/data/1878848/000114036126032638/ef20080141_ex99-1.htm
  14. Form 8-K, Item 5.02 (co-CEO RSU grants) — SEC EDGAR — 2026-07-01 — https://www.sec.gov/Archives/edgar/data/1878848/000114036126027202/ef20077190_8k.htm
  15. DEFA14A, independent directors' letter on the 2026 award — SEC EDGAR — 2026-07-09 — https://www.sec.gov/Archives/edgar/data/1878848/000114036126028014/ef20077507_defa14a.htm
  16. Forms 4 (co-CEOs, CFO, directors) — SEC EDGAR — 2026-07-01 — https://www.sec.gov/Archives/edgar/data/1878848/000095010326009958/ownership.xml
  17. Schedule 13G (Citadel, 18,687,938 shares, 4.7%) — SEC EDGAR — 2026-09-30 — https://www.sec.gov/Archives/edgar/data/1878848/000110465926112411/primary_doc.xml
  18. Schedule 13G (Goldman Sachs, 33.6M shares) — SEC EDGAR — 2026-08-12 — https://www.sec.gov/Archives/edgar/data/1878848/000088698226000449/primary_doc.xml
  19. Form 8-K Item 4.01, change of auditor (RCGT → KPMG) — SEC EDGAR — 2025-11-28 — https://www.sec.gov/Archives/edgar/data/1878848/000114036125043599/ef20060170_8k.htm
  20. EDGAR submissions feed, CIK 0001878848 (latest filing 2026-09-30) — SEC — fetched 2026-10-12 — https://data.sec.gov/submissions/CIK0001878848.json
  21. IREN Q4 FY26 Earnings Call Transcript — Motley Fool (third-party copy, read via summary) — call 2026-08-27, published 2026-08-31 — https://www.fool.com/earnings/call-transcripts/2026/08/31/iren-iren-q4-2026-earnings-call-transcript/
  22. Transcript: IREN Q4 Earnings Call Highlights — MarketBeat (summary) — 2026-08-27 — https://www.marketbeat.com/instant-alerts/transcript-iren-q4-earnings-call-highlights-2026-08-27/
  23. IREN FY27 capex plan is not an equity requirement — Blockspace (search summary only) / Nasdaq — 2026-08 — https://blockspace.media/insight/iren-fy27-capex-plan-not-equity-requirement/
  24. ClusterMAX 3.0: The Industry Standard GPU Cloud Rating System Returns — SemiAnalysis — 2026-09-23 — https://newsletter.semianalysis.com/p/clustermax-30-the-industry-standard
  25. Silicon Valley's top AI scrutineer says Firmus is "up for the challenge" but IREN runs "#1 worst site" — Capital Brief — 2026-09-24 — https://www.capitalbrief.com/article/silicon-valleys-top-ai-scrutineer-says-firmus-is-up-for-the-challenge-but-iren-runs-1-worst-site-74ce1b6f-95fb-482d-a5de-6f3752cc758e/
  26. Iren shares fall as SemiAnalysis highlights data center reliability issues — Investing.com — 2026-10-07 — https://www.investing.com/news/stock-market-news/iren-shares-fall-as-semianalysis-highlights-data-center-reliability-issues-4937184
  27. IREN stock falls after report flags infrastructure and disclosure concerns — Investing.com — 2026-10-08 — https://www.investing.com/news/stock-market-news/iren-stock-falls-after-report-flags-infrastructure-and-disclosure-concerns-93CH-4939589
  28. IREN shares sink 20% in ten days — Startup Fortune — 2026-10-10 — https://startupfortune.com/iren-shares-sink-20-in-ten-days-as-a-scathing-infrastructure-report-collides-with-wall-streets-ai-bull-case/
  29. IREN Drops 4% as SemiAnalysis Flags Its Canadian Data Centers — 24/7 Wall St. — 2026-10-08 — https://247wallst.com/investing/2026/10/08/iren-drops-4-as-semianalysis-flags-its-canadian-data-centers-core-scientific-pulls-back-2/
  30. IREN Leans On NVIDIA AI Validation As SemiAnalysis Questions Its Infrastructure Claims — Yahoo Finance/Stocktwits — ~2026-10-10 — https://finance.yahoo.com/technology/ai/articles/iren-leans-nvidia-ai-validation-095223955.html
  31. IREN Stock Fell 6% in a Day After SemiAnalysis Flagged Its Canadian Sites (consensus revenue/EBITDA) — TIKR — 2026-10 (data to 10-08) — https://www.tikr.com/blog/iren-stock-fell-6-in-a-day-after-semianalysis-flagged-its-canadian-sites-heres-where-the-stock-could-go
  32. IREN Forecast (consensus, price targets, ratings) — StockAnalysis — updated 2026-09-29, price 2026-10-12 — https://stockanalysis.com/stocks/iren/forecast/
  33. IREN data page (short interest, insider activity, market cap) — MarketBeat — page dated 2026-10-12 — https://www.marketbeat.com/stocks/NASDAQ/IREN/news
  34. IREN Stock Slips, But A Top Wall Street Bank Is Bullish (JPMorgan upgrade to OW, $65) — Stocktwits — 2026-09-14 — https://stocktwits.com/news-articles/markets/equity/iren-stock-slips-but-a-top-wall-street-bank-is-bullish/cZtmjA6RBTB
  35. Northland stuft IREN mit Outperform ein, Kursziel 99 Dollar — goldesel.de — 2026-09-18 — https://goldesel.de/aktien/news/northland-stuft-iren-mit-outperform-ein-und-setzt-kursziel-auf-99-dollar
  36. CoreWeave (CRWV) or IREN (Redburn initiation, IREN Neutral $40) — TipRanks — 2026-09-21 — https://www.tipranks.com/news/coreweave-crwv-or-iren
  37. Rothschild Redburn downgrades Nebius, CoreWeave to Sell on lower GPU prices — 24/7 Wall St. — 2026-09-21 — https://247wallst.com/investing/2026/09/21/rothschild-redburn-downgrades-nebius-coreweave-to-sell-on-lower-gpu-prices/
  38. IREN Aktie: Governance-Zweifel schlagen zu (Jones Trading Hold) — Boerse-Global — 2026-09-28 — https://www.boerse-global.de/iren-aktie-governance-zweifel-schlagen-zu/829480
  39. IREN Aktie: JPMorgan bleibt skeptisch bei 46 Dollar — Boerse-Express — 2026-05-13 — https://www.boerse-express.com/news/articles/iren-aktie-jpmorgan-bleibt-skeptisch-bei-46-dollar-904227
  40. IREN Stock Faces 3-Month Losing Streak After Q4 Miss (HC Wainwright Buy $90) — Stocktwits — 2026-08-28 — https://stocktwits.com/news-articles/markets/equity/iren-stock-q4-miss-august-gains-analyst-buying-opportunity/cZYyoMFRJs3
  41. IREN granted conditional Base Load status for 2 GW Texas campus (rewrite of IREN release of 2026-09-08) — w.media — 2026-09-09 — https://w.media/iren-granted-conditional-base-load-status-for-2-gw-texas-campus/
  42. Texas Governor Directs a Halt on Data Center Permitting With State Agency Audits — Greenberg Traurig — 2026-09 — https://www.gtlaw.com/ar/insights/2026/9/texas-governor-directs-a-halt-on-data-center-permitting-with-state-agency-audits
  43. Texas halts data center permitting pending state audits — BakerHostetler — 2026-09-23 — https://www.bakerlaw.com/insights/texas-halts-data-center-permitting-pending-state-audits-what-developers-investors-and-energy-users-need-to-know/
  44. Governor Abbott pauses Texas data center interconnections and calls for verification and audit — Foley & Lardner — 2026-08 — https://www.foley.com/insights/publications/2026/08/governor-abbott-pauses-texas-data-center-interconnections-and-calls-for-verification-and-audit-what-data-center-developers-need-to-know-now/
  45. Texas PUC adopts data center interconnection rules — Utility Dive — 2026-09-21 — https://utilitydive.com/news/texas-puc-adopts-data-center-interconnection-rules/830899
  46. Data centre regulation in British Columbia: competing for a limited supply — BLG — 2026-07 — https://www.blg.com/en/insights/2026/07/data-centre-regulation-in-british-columbia-competing-for-a-limited-supply
  47. Stitt signs bill to prevent higher utility costs from data centers into law (Oklahoma HB 2992) — KGOU — 2026-05-13 — https://www.kgou.org/energy/2026-05-13/stitt-signs-bill-to-prevent-higher-utility-costs-from-data-centers-into-law
  48. IREN Stock Jumped 13% in a Day on an Anthropic Australia Report (AFR tender, shortlist) — TIKR — 2026-07 — https://www.tikr.com/blog/iren-stock-jumped-13-in-a-day-on-an-anthropic-australia-report-heres-where-the-stock-could-go-in-2026
  49. Anthropic signs first Australian data center lease (Zerra DC, Western Downs) — W.Media — 2026-09-16 — https://w.media/anthropic-signs-first-australian-data-center-lease/
  50. Macquarie partners with Zerra and ADC to build Australia's largest data centre campus for Anthropic — IPE Real Assets — 2026-09 — https://realassets.ipe.com/news/macquarie-partners-with-zerra-and-adc-to-build-australias-largest-data-centre-campus-for-anthropic/10138513.article
  51. H100 rental price over time — Silicon Data — data through 2025-12 — https://www.silicondata.com/blog/h100-rental-price-over-time
  52. The Great GPU Shortage (H100 1-year rental index) — SemiAnalysis — 2026 — https://newsletter.semianalysis.com/p/the-great-gpu-shortage-rental-capacity
  53. CoreWeave Announces Pricing of $1.25B and €2B Senior Notes — CoreWeave IR — 2026-06-11 — https://investors.coreweave.com/news/news-details/2026/CoreWeave-Announces-Pricing-of-1-25-Billion-of-Senior-Notes-and-2-Billion-of-Senior-Notes/default.aspx
  54. NVIDIA confirms Vera Rubin launch for Q3 as Blackwell demand keeps climbing (incl. Bloomberg report on >15% server price rises) — Digital Citizen (secondary) — 2026 — https://www.digitalcitizen.life/nvidia-confirms-vera-rubin-launch-for-q3-as-blackwell-demand-keeps-climbing/
  55. Moody's hyperscaler capex forecasts marked up by $85bn to close in on $1trn by 2027 — DatacenterDynamics — 2026 — https://datacenterdynamics.com/en/news/moodys-hyperscaler-capex-forecasts-marked-up-by-85bn-to-close-in-on-1trn-by-2027/
  56. Iris Energy 6-K, quarter ended 2022-12-31 (impairments, SPV defaults) — SEC EDGAR — 2023 — https://www.sec.gov/Archives/edgar/data/1878848/000114036123007261/brhc10048129_ex99-4.htm
  57. Iris Energy receives default notice on $103 million — The Block — 2022-11 — https://www.theblock.co/post/183799/iris-energy-receives-default-notice-on-103-million
All IREN Bulls vs Bears reports
Disclaimer. AI-generated research and podcast (Claude by Anthropic, voices by Google Gemini) for educational purposes — not financial advice and not a recommendation to buy or sell. Willem (MR.INVEST) owns this stock. Check the company’s own filings before making any decision.
← All Bulls vs Bears reports
IREN Bulls vs Bears — Stock Analysis — MR.INVEST