
IREN Limited
Monetize energized, grid-connected, interconnect-queued power two ways: (1) legacy Bitcoin self-mining (volatile, hash-price-linked revenue at ~$3-5M/MW/yr) and (2) a pivot to long-dated, contracted AI revenue. Unlike shell-leasing peers (WULF, CIFR, CORZ, APLD) that lease datacenter shells to a neocloud, IREN OWNS the NVIDIA GPUs and rents compute directly (AI Cloud) -- capturing more of the value chain but taking on GPU depreciation, technology-obsolescence, and heavier capex/financing intensity. The scarce asset is the site's power, not its ASICs; the whole equity story is the 480MW->1,210MW power ramp converting into contracted AI revenue.
Sources — 27 figures with citations
- Q3 FY2026 revenue by segmentfiled2026-03-31Total $144.795M (Bitcoin mining $111.160M + AI Cloud Services $33.635M) vs $144.823M in Q3 FY2025; 9M FY26 $569.782M vs $313.731Msec.gov — Form 10-Q for the quarter ended Mar 31, 2026 (filed 2026-05-08), Condensed Consolidated Statements of Operations. AI Cloud growth derived: 33.635/3.581 - 1 = +839% YoY. 9M growth derived: 569.782/313.731 - 1 = +81.6%.
- Q3 FY2026 operating loss and net lossfiled2026-03-31Operating loss -$233.523M (-161.3% margin); net loss -$247.827M (-171% margin); 9M FY26 net loss -$18.622Msec.gov — 10-Q. Components: SG&A -$81.750M, D&A -$121.245M, impairment of assets -$140.411M. The 9M loss of only -$18.6M reflects a $533.942M unrealized GAIN on financial instruments offset by $111.799M of debt-conversion inducement expense. Margins derived: -233.523/144.795 = -161.3%; -247.827/144.795 = -171.1%.
- Q3 FY2026 gross margin ex-D&A, by segmentderived2026-03-3172.4% blended; AI Cloud 86.3%; Bitcoin mining 68.2%sec.gov — Revenue and cost-of-revenue lines are filed (total cost of revenue $39.933M: mining $35.336M, AI Cloud $4.597M) and are stated exclusive of D&A. Arithmetic: (144.795-39.933)/144.795 = 72.4%; (33.635-4.597)/33.635 = 86.3%; (111.160-35.336)/111.160 = 68.2%. Adding the $121.245M of D&A makes the quarter loss-making at the gross level.
- Q3 FY2026 adjusted EBITDAfiled2026-03-31$59.5M, 41% margin (vs $75.3M and 41% in Q2 FY26)sec.gov — 8-K Ex-99.1 dated May 7, 2026, non-GAAP reconciliation. Bridge from the -$247.8M net loss adds back D&A $121.2M, impairment $140.4M, stock comp $31.5M, unrealized loss $23.7M and others - i.e. the 41% margin exists only after excluding the fleet's depreciation and impairment.
- Balance sheet at Mar 31, 2026filed2026-03-31Cash $2,213.274M; convertible notes payable $3,687.832M; finance lease liabilities $274.256M ($122.168M current + $152.088M non-current); total liabilities $4,600.369M; total stockholders' equity $2,664.528M; 340,979,966 ordinary shares issued and outstandingsec.gov — 10-Q Condensed Consolidated Balance Sheets. Net debt derived: 3,687.832 + 274.256 - 2,213.274 = $1,748.8M. Cash rose from $564.526M at Jun 30, 2025.
- 9M FY2026 capex and free cash flowderived2026-03-31Capex $2,354.934M ($1,669.157M PP&E ex computer hardware + $685.777M computer hardware); operating cash flow +$289.326M; implied FCF ~-$2,065.6M; financing inflow $3,968.769Msec.gov — Cash-flow lines filed in the 10-Q (net investing -$2,608.931M incl. $107.573M intangibles and $159.059M prepayments/deposits). Derived: 2,354.934/569.782 = 413% capex intensity; 289.326 - 2,354.934 = -2,065.608 FCF = -362.5% of 9M revenue.
- Jul 20, 2026 business update - ARR raise, contracting and prepaymentsfiled2026-07-20Year-end CY2026 AI Cloud ARR target raised from $3.7bn to >$4bn, ~85% under contract; $2.8bn of new multi-year cloud contract value; customer prepayments ~45% of associated GPU capex on contracts signed since Jun 1, 2026; weighted average contract term ~4 years; ~$7.6bn cash at Jun 30, 2026 incl. $1.7bn restrictedsec.gov — 8-K Item 2.02 Ex-99.1 (furnished, not 'filed' for Section 18 purposes). The cash figure is expressly unaudited preliminary and the restricted portion relates to the Microsoft Horizon 1-4 GPU financing. ARR is disclaimed by the company as a non-GAAP operating metric, not a substitute for revenue, contingent on commissioning/testing/customer acceptance. Customer list: Microsoft, NVIDIA, Perplexity, Figure AI, Together AI, Fluidstack, Fireworks AI, Fal AI, Hume AI + one unnamed.
- NVIDIA contract, 5GW partnership and share purchase rightfiled2026-05-07$3.4bn 5-year air-cooled Blackwell AI Cloud contract (60MW at Childress, ramp from early 2027); 5GW strategic partnership; NVIDIA granted a 5-year right to purchase 30,000,000 ordinary shares at $70 (~$2.1bn gross if fully exercised)sec.gov — 8-K Ex-99.1 May 7, 2026. The $70 strike is ~76% above the $39.75 Aug 3, 2026 close, so the proceeds are currently far out of the money. Capacity roadmap in the same release: 480MW in 2026, 1,210MW in build for 2027, 5GW secured power for 2028+, plus Nostrum (490MW Spain) and Mirantis acquisitions.
- GPU-level financing arranged post-quarterfiled2026-05-29~$3.6bn at IE US Hardware 3 LLC (May 29, 2026): ~$1.5bn delayed-draw term loan at term SOFR + 2.25% (Goldman Sachs / JPMorgan joint lead arrangers, 0.40% commitment fee on undrawn) plus $2.1bn of 5.96% senior notes due Dec 31, 2031; drawable in tranches until May 29, 2027sec.gov — 8-K Item 1.01. Proceeds partially fund GPU infrastructure for the Nov 2, 2025 Microsoft contract at Childress. Maturity is Dec 31, 2031 OR the earlier date on which the final service fee under all Microsoft tranches is paid in full - i.e. the debt is structurally tied to Microsoft contract performance. Parent provided limited guarantees.
- GPU procurement commitmentfiled2026-05-19~$1.6bn Dell purchase agreement for GPUs and ancillary products, payable in installments within 30 days of each tranche shipping, deployed at Childress for the $3.4bn managed-services contract; parent-guaranteedsec.gov — 8-K Item 1.01. Counterparty is Dell Marketing L.P.; the buying entity is IE US Hardware 4 Inc., with IREN Limited unconditionally guaranteeing its obligations.
- Share price (close)market2026-08-03$39.75query1.finance.yahoo.com — Regular-session CLOSE for Mon Aug 3, 2026 (regularMarketTime 2026-08-03 20:00 UTC = 16:00 ET), NasdaqGS, USD. 52-week range $15.49-$76.87 - the close is ~48% below the high. Cross-checked against api.nasdaq.com pre-market Aug 4 quote of $40.00. Not an intraday high.
- Market capitalizationderived2026-08-03~$14.2Bsec.gov — 357,378,674 ordinary shares outstanding as of Apr 30, 2026 (10-Q cover page / dei EntityCommonStockSharesOutstanding) x $39.75 Aug 3, 2026 close = $14,206M. Cross-check: stockanalysis.com reported a $10.47B cap on the same 357.38M share count at the Jul 29 close of $29.31 ($10.47B / $29.31 = 357.3M), which confirms the share count. Excludes NVIDIA's 30M-share right at $70 and the 18.2M co-CEO RSUs granted Jun 30, 2026.
- Q3 FY2026 consensus miss and market reactionmarket2026-05-07Revenue $144.8M vs ~$219.87M consensus (-34.1%); EPS -$0.30 vs ~-$0.2161 consensus (-38.8%); stock -6.77% in regular trading, -2.42% after hours to ~$59.50investing.com — Sell-side consensus is a third-party aggregate, not a filed figure. The ~$59.50 post-print level versus the $39.75 Aug 3 close implies a further ~33% decline over the subsequent three months despite the Jul 20 ARR raise.
- Co-CEO equity grants (dilution)filed2026-06-309,099,328 RSUs granted to EACH of the two Co-CEOs on Jun 30, 2026 (~18.2M shares total), with a combined six-year vesting and holding period; no further equity grants to either until FY2031sec.gov — 8-K Item 5.02. ~5.1% of the Apr-30 share count, unanimously approved by independent directors with an independent compensation consultant.
- Share price (close)market2026-11-03$24.46finance.yahoo.com — Regular-session close for Tue Nov 3 2026, NasdaqGS, USD. 52-week range $21.86-$76.87; the close is ~68% below the high and ~40% below the $40.53 close of Aug 27 2026, the day FY26 results were released after the close.
- Market capitalizationderived2026-11-03~$9.6Bsec.gov — 394,058,648 ordinary shares outstanding as of Aug 14, 2026 (FY26 Form 10-K cover) x $24.46 Nov 3, 2026 close = $9,639M. Excludes NVIDIA's 30M-share right at $70 and the 18.2M co-CEO RSUs granted Jun 30, 2026.
- Balance sheet at Jun 30, 2026filed2026-06-30Cash $5,895.6M plus $1,723.9M of restricted cash; total debt $7,592.9M net of issuance costs ($6,767.6M recourse, $938.0M non-recourse); finance lease liabilities $243.8M; total liabilities $11,604.4M; stockholders' equity $4,185.6M; 380,193,608 ordinary shares outstandingsec.gov — Form 10-K (filed 2026-08-27), balance sheet and Note 23. Net debt derived on unrestricted cash: 7,592.9 + 243.8 - 5,895.6 = $1,941M.
- Latest print: Q1 FY2027 (quarter ended 2026-09-30, released 2026-11-05)filed2026-11-12revenue $228 million; +66% QoQ from $137 million (the release frames growth sequentially; no YoY figure given); guide: More-than-$4 billion operating ARR now expected January 2027 (possibly February 2027); remaining FY27 capex $15 billion ($9 billion covered, $6 billion targeted from additional sources)finviz.com — Recorded Nov 12 2026; figures as the source states them. The entry's analysis text predates this print unless dated otherwise.
- regulatory, 2026-11-03filed2026-11-03Democrat Gina Hinojosa defeated Gov. Greg Abbott in the Texas governor's race (AP call; preliminary 57% to 41.7%) after criticizing Abbott's handling of AI data-center development; she takes office 2027-01-19.fox5atlanta.com — REPORTED; recorded Nov 12 2026; source dated 2026-11-04
- regulatory, 2026-11-04filed2026-11-04Company-specific read of the Texas result: the election has not itself imposed a new moratorium or suspended IREN's construction at Childress or Sweetwater (Crypto Briefing), while a separate analysis lists IREN's Childress data center among permit applications pending at the Texas Commission on Environmental Quality under Abbott's earlier directive to halt data-center permits pending an ERCOT audit (Blockspace); the Q1 FY27 release notes the 2GW Sweetwater Hub has conditional inclusion in ERCOT's Batch Zero process.cryptobriefing.com — REPORTED; recorded Nov 12 2026; source dated 2026-11-05
- earnings, 2026-11-05filed2026-11-05Q1 FY2027 (quarter ended 2026-09-30): revenue $228 million (+66% QoQ from $137 million) AI Cloud $200 million (Services $135 million plus Lease $65 million; from $71 million) and Bitcoin mining $28 million (from $67 million); net loss $(263) million (from $(684) million); adjusted EBITDA $66 million (29% margin); AI Cloud segment gross margin 86%; operating ARR doubled to $1 billion; the more-than-$4 billion ARR target, 'fully contracted and fully funded', moved from year-end 2026 to January 2027 and 'may extend to February 2027' citing 'a customer-requested change and supplier timing';finviz.com — CONFIRMED; recorded Nov 12 2026; source dated 2026-11-05
- guidance, 2026-11-06filed2026-11-06Earnings-call coverage attributes the ARR slip to customer-requested networking upgrades and equipment delivery delays affecting Microsoft-related deployments, describes about $15 billion of planned capex for the rest of fiscal 2027 against about $9 billion of cash and undrawn commitments ('$6 billion in additional planned spending requiring funding'), and quotes management attributing the spending reduction to financing conditions rather than weaker demand.theglobeandmail.com — REPORTED; recorded Nov 12 2026; source dated 2026-11-06
- Watched signal: AI Cloud revenue conversion vs the company-claimed ARR targetsfiled2026-11-12Q1 FY27 AI Cloud revenue reached $200 million (from $71 million) and operating ARR $1 billion, but the more-than-$4 billion operating ARR milestone slipped from December 2026 to January 2027, possibly February 2027.finviz.com — Outcome of a signal this entry said to watch; recorded Nov 12 2026.
- Watched signal: Childress Horizon 1-4 / Sweetwater energization and GPU-deployment cadencefiled2026-11-12Horizon 1 delivered to Microsoft; Horizon 2-4 in late-stage construction with equipment delivery delays on Microsoft-related deployments (call); Sweetwater 1 construction well underway and the 2GW Sweetwater Hub conditionally included in ERCOT's Batch Zero.finviz.com — Outcome of a signal this entry said to watch; recorded Nov 12 2026.
- Watched signal: Cash burn + GPU-financing terms during the buildfiled2026-11-12Unrestricted cash fell to $3,522 million from $5,896 million and net debt rose to $4,430 million from $216 million; $15 billion of FY27 capex remains with $6 billion still to be sourced; the new five-year contract has 90% of GPU capex prepaid by the customer.finviz.com — Outcome of a signal this entry said to watch; recorded Nov 12 2026.
- Watched signal: GPU depreciation policy / used-GPU residual printsfiled2026-11-12H100 GPUs were recontracted at 'double previous pricing', supporting residual values; no depreciation-policy change was disclosed.finviz.com — Outcome of a signal this entry said to watch; recorded Nov 12 2026.
- Board falsifier check (state-of-ai-cloud)filed2026-11-12The more-than-$4 billion operating ARR milestone the conversion schedule for the Microsoft/NVIDIA-anchored book slipped from December 2026 to January 2027 and may extend to February 2027, citing a customer-requested change and supplier timing (equipment delays on Microsoft-related deployments per the call); a one-to-two-month slip meets the letter of 'fails to convert to ARR on schedule' but is not obviously a material slip of the 480MW to 1,210MW ramp.finviz.com — Bears on a condition the board thesis pre-registered; recorded Nov 12 2026.
The thesis on this name
State of the AI Cloud
A genuine mispricing where power was already energized and interconnect-queued: IREN (direct Microsoft ~$9.7B + Nvidia ~$3.4B anchor, owned-GPU AI Cloud model that captures more than shell-leasing) and the Google-backstopped pair (WULF ~360MW + Google ~14% equity, plus a 20-yr ~401MW/~$19B lease signed directly with Anthropic at Justified; CIFR ~207MW of contracted critical IT load at Barber Lake) convert volatile hash-price revenue into contracted, credit-backstopped 10-15yr leases at far higher multiples. The scarce asset is the SITE'S POWER, not its ASICs.
State of the AI Cloud
The best-anchored miner-pivot — direct Microsoft (~$9.7B) + Nvidia (~$3.4B) book and an owned-GPU AI Cloud model that captures more of the value chain than shell-leasing.
State of the AI Cloud
Direct Microsoft (~$9.7B) + Nvidia (~$3.4B) book + owned-GPU AI Cloud on power it controls; the cleanest energized-power re-rate (non-US domicile flagged).
Earnings, margins, COGS & capex
IREN is mid-transition: Q4 FY26 (Jun-2026 qtr) revenue fell to $137.2M (from $144.8M) as it decommissioned Bitcoin mining hardware ahead of GPU installation, and it posted a GAAP net loss of -$684.0M that was driven by $450.4M of non-cash impairments (FY26: revenue $707.0M, +41%, and a net loss of -$702.6M). The signal under the noise: AI Cloud services revenue grew +110% QoQ to $70.5M (~51% of revenue, past mining), FY26 adj. EBITDA was $245.7M (35% margin) and Q4 adj. EBITDA fell to $19.2M (14%). The whole thesis is forward -- management reports $4bn of AI Cloud annualized run-rate ARR contracted for 2026 capacity, with $1bn operating at Aug 26 2026 (company-claimed, not a booked figure), underpinned by a ~$9.7B, 5-year Microsoft GPU-cloud contract (NVIDIA GB300; Horizon 1 delivered in August 2026) and a ~$3.4B, 5-year NVIDIA AI Cloud contract, built on power IREN controls at Childress and Sweetwater, TX. The bull edge is best-in-class counterparties (Microsoft IG + NVIDIA) and an owned-GPU model that captures more value than shell-leasing; the bear edge is GPU depreciation/obsolescence on a leveraged fleet ($7.6B of debt at Jun 30, 2026) and heavy capex/financing intensity while the ramp is unproven.
Income statement — where each revenue dollar goes
% of revenueOf every $1 of revenue, ~31¢ is cost of goods and ~43¢ operating expense, leaving ~26¢ of operating profit (~703¢ net).
Revenue trend
Margins
falling -- 14% in Q4 from 41% in Q3 ($59.5M) and 54% in FY25 on higher employee costs and platform investment as mining rolls off ahead of the AI Cloud ramp
negative/worsening -- Q4 carried $450.4M of non-cash impairments (FY26 $638.8M) and a $102.1M write-down of mining hardware held for sale on the mining-to-AI transition; FY25 was +$86.9M
up sharply -- AI Cloud +110% QoQ to $70.5M, past mining ($66.7M); the intended mix inversion
negative -- capex for 480MW / ~150k GPUs far exceeds current revenue during the ramp
COGS structure
For the legacy mining segment, COGS is dominated by electricity/power and mining-hardware depreciation. For the AI Cloud segment, the far larger economic cost is GPU depreciation + technology-obsolescence on the OWNED NVIDIA fleet plus power and datacenter operating cost -- the structural difference from shell-leasing peers, who carry no GPU-depreciation on their books. The Q4 FY26 GAAP loss of -$684.0M was amplified by $450.4M of non-cash impairments (mining hardware decommissioned ahead of GPU install) and a $102.1M write-down of hardware held for sale, so adj. EBITDA (14% in Q4; 35% for FY26) sits well above the GAAP net line once depreciation ($417.7M in FY26), impairment, and financing are loaded.
Capex
The defining feature and the whole bear case on the balance sheet. 2026 build targets ~480MW of capacity (Childress Horizons 1-4) with ~150,000 GPUs deployed; 2027 targets ~1,210MW including Childress Horizons 5-6 and Sweetwater 1 (management has referenced a 5GW+ longer-term AI expansion ambition). Funded by cash, operating cash flow, GPU financing, and additional capital initiatives; cash rose from $2.2B (Mar-2026) to $5.9B plus $1.7B of restricted cash (Jun-2026) after $3.0B of convertible notes and $2.1B of share issuance in Q4. Sweetwater 1 substation was energized on schedule and Horizon 1 of the four Childress liquid-cooled deployments was delivered to Microsoft in August 2026, Horizon 2 is in commissioning and Horizons 3-4 are targeted for Q4 2026, in support of the $9.7B Microsoft contract.
Latest earnings
Q4 FY26 revenue of $137.2M was below Q3 ($144.8M) as mining fell to $66.7M while AI Cloud rose to $70.5M; the -$684.0M net loss included $450.4M of impairments and adjusted EBITDA fell to $19.2M from $59.5M.
IREN does not guide GAAP revenue. It reports an AI Cloud ARR operating metric: $4bn contracted ARR for 2026 capacity and $1bn operating at 26 Aug 2026 (raised on Jul 20, 2026 from $3.7bn to more than $4bn). 2026 capacity is largely sold out and 2027 is in late-stage discussions. Horizon 1 of four 50MW (IT) liquid-cooled deployments at Childress was delivered to Microsoft in August 2026, with Horizon 2 in commissioning and Horizons 3-4 targeted for Q4 2026. Capacity roadmap: 480MW gross AI Cloud in 2026 and 1,210MW in build for 2027 (gross basis, from the May 2026 update); the August 2026 release targets cumulative delivery of about 0.3GW (IT) in 2026 and 0.8GW (IT) in 2027 (IT-load basis, so not directly comparable), across a pipeline above 5GW. Company caveats: ARR is GPU/hour pricing for commissioned GPUs at the relevant date x 8,760 hours plus storage/ancillaries, is not a GAAP measure, and is subject to commissioning, testing and customer acceptance.
- Q4 FY26 revenue
- $137.2M (mining $66.7M + AI Cloud $70.5M); -5.2% QoQ. FY26 $707.0M (+41% YoY)
- AI Cloud revenue growth
- +110% QoQ in Q4 FY26 ($70.5M vs $33.6M), ~51% of total revenue; FY26 $128.8M vs $16.4M (~8x)
- Adjusted EBITDA
- $19.2M in Q4 FY26 (14% margin), down from $59.5M (41%) in Q3 FY26; FY26 $245.7M (35%) vs $269.7M (54%) in FY25
- GAAP net loss
- -$684.0M in Q4 FY26 (-498% net margin); FY26 -$702.6M (flattered by a $558.5M unrealized gain on financial instruments, net of $111.8M of debt-conversion inducement expense)
- Impairment
- $450.4M in Q4 FY26 alone; $638.8M for FY26
- AI Cloud ARR
- $4bn contracted for 2026 capacity, $1bn operating at Aug 26, 2026; $4bn targeted to be operational by Dec 31, 2026 (target raised from $3.7bn on Jul 20, 2026)
- New contract wins (Jul 20, 2026)
- $2.8bn total contract value in new multi-year cloud services contracts with leading AI developers; a further new multi-year contract with a leading frontier AI lab was announced on Aug 27, 2026
- Customer base
- Jul 20, 2026 update: Microsoft, NVIDIA, Perplexity, Figure AI, Together AI, Fluidstack, Fireworks AI, Fal AI, Hume AI, plus one unnamed new leading AI developer; Aug 27, 2026 release adds a new frontier AI lab and recent signings including Cohere, Prometheus, Perplexity, Figure AI, Fal AI and Higgsfield AI
- Customer prepayments
- ~45% of associated GPU capex prepaid on contracts executed since Jun 1, 2026 (Jul 20 update) and 45-55% on recent contracts (Aug 27 release) - directly reduces net funding need; $1,842.6M of deferred revenue at Jun 30, 2026
- Weighted average contract term
- ~4 years across the portfolio
- Cash (audited)
- $5,895.6M at Jun 30, 2026 plus $1,723.9M of restricted cash ($7,619.5M in total; GPU-financing note proceeds are held in restricted escrow until release conditions are met); vs $2.213bn at Mar 31, 2026
- NVIDIA relationship
- $3.4bn 5-year air-cooled Blackwell contract (60MW at Childress, ramp from early 2027) + 5GW strategic partnership; NVIDIA holds a 5-year right to buy 30M shares at $70 (~$2.1bn)
- Debt structure
- $7,705.6M of debt principal at Jun 30, 2026 ($7,592.9M net of issuance costs): $6,767.6M recourse (six series of convertible senior notes plus $22.0M other) and $938.0M non-recourse Hardware-3 GPU financing ($413M drawn of a $1,545M DDTL at SOFR+2.25%; $525M issued of $2,100M of 5.96% notes due Dec 2031), plus $243.8M of finance leases; ~$2.4bn of further GPU financing for the Mackenzie expansion signed Aug 25, 2026
- GPU procurement
- ~$1.6bn Dell purchase agreement (May 19, 2026) for GPUs at Childress serving the $3.4bn managed-services contract; parent-guaranteed
- Capacity trajectory
- ~3MW of self-built AI Cloud a year ago to 480MW delivered in 2026; 1,210MW in build for 2027; 5GW secured power pipeline
Growth drivers
- Microsoft anchor — ~$9.7B, 5-year GPU cloud contract (NVIDIA GB300) phased through 2026 at the 750MW Childress, TX campus -- the strongest hyperscaler, investment-grade anchor among the miner-pivots [fact]
- NVIDIA AI Cloud contract — ~$3.4B, 5-year (air-cooled Blackwell) within 60MW at Childress, plus an option for NVIDIA to buy ~30M IREN shares [fact]
- Owned-GPU AI Cloud model captures more of the value chain than pure shell-leasing (IREN rents compute directly rather than only leasing datacenter shells) [fact]
- Power already controlled and energizing on schedule — Sweetwater 1 substation energized; Childress Horizon 1-4 liquid-cooled DCs advancing; the scarce asset (energized MW) is in hand [fact]
- Contracted ARR ramp — $4bn of ARR contracted for 2026 capacity and $1bn operating at Aug 26 2026, targeted to be operational by Dec 31 2026 via ~150k GPUs at Childress + British Columbia [company-claimed]
- AI Cloud revenue inflection — +110% QoQ to $70.5M in Q4 FY26 (~51% of revenue, past mining), the intended mix shift away from volatile mining [fact]
Reported financials — SEC EDGAR
Audited GAAP figures pulled from SEC filings · latest filing 2026-08-27. The audited primary-source spine — not financial advice.
Revenue — annual (GAAP)
Margins & balance sheet — FY’26
Bull & bear
IREN went from a $3.7bn ARR target that was partly aspirational to $4bn of ARR contracted for 2026 capacity with $1bn already operating, a customer list running from Microsoft and NVIDIA to Perplexity, Figure AI and a new frontier AI lab, $7.6bn of cash including $1.7bn restricted, and - the most underrated item - customers prepaying 45-55% of the GPU capex. Demand is explicitly described as exceeding available and planned capacity, with 2026 capacity largely sold out.
- The ARR target went UP and got MORE contracted at the same time: raised from $3.7bn to >$4bn with ~85% under contract as of Jul 20, 2026, on $2.8bn of new total contract value, and by Aug 27, 2026 the company described $4bn as contracted for 2026 capacity with $1bn operating. Rising target plus rising contracted percentage is the opposite of the usual neocloud pattern of a flat target quietly losing coverage.
- Customer prepayments of 45-55% of associated GPU capex on recent contracts (~45% on those signed since Jun 1, 2026) is the single most important economic change: it converts the biggest bear point (owned-GPU capital intensity funded with debt) into partly customer-funded capex and materially reduces net funding need. Management also states contracted pricing 'continues to strengthen.'
- Counterparty quality plus diversification, which is rare in this cohort: Microsoft (~$9.7bn contract) and NVIDIA (~$3.4bn) anchor the book, but the customer list now also includes Perplexity, Figure AI, Together AI, Fluidstack, Fireworks AI, Fal AI and Hume AI (with Cohere, Prometheus and Higgsfield AI added by Aug 27, 2026) - so the thesis no longer rests on two names, and the weighted average contract term is ~4 years.
- NVIDIA is a partner, not just a supplier: a 5GW strategic partnership to deploy NVIDIA-aligned infrastructure across the pipeline, and NVIDIA took a 5-year right to purchase 30M IREN shares at $70 (~$2.1bn). A chip vendor taking equity exposure at that strike is a strong third-party validation of the platform, whatever it says about the current price.
- Funding is committed rather than hoped for: $7.6bn of cash at Jun 30, 2026 (audited; $1.7bn of it restricted) plus ~$3.6bn of GPU-level financing arranged May 29, 2026 (a $1.5bn Goldman/JPMorgan delayed-draw term loan at SOFR+2.25% and $2.1bn of 5.96% notes due Dec 2031, drawable in tranches to May 2027) at rates that imply lenders underwrote the Microsoft contract cash flows, not the equity story; $938M of it was drawn at Jun 30, and ~$2.4bn more was signed on Aug 25, 2026 at a 9.0% fixed rate for the Mackenzie expansion, which the company says funds 90% of associated GPU capex.
- Operating cash flow is positive: +$2,100.4M for FY26, though $1,841.7M of it is customer prepayments booked as deferred revenue (+$258.7M without them). Unlike most of the AI-buildout cohort, customers and the operating business fund part of the build; the cash burn is entirely discretionary capex.
- Power is the real scarce asset and IREN owns a lot of it: 5GW of secured power across North America, Europe and APAC, expanded by the Nostrum acquisition (490MW in Spain plus a GW+ development pipeline) and advancing Australian projects. The Mirantis acquisition adds the software layer for deploying and managing that compute.
- The owned-GPU model captures the full services margin rather than a landlord's rent: FY26 AI Cloud gross margin ex-D&A was 86.9% (87.0% in Q4), versus 64.9% for mining (computed from the filed statements) - so as the mix inverts, unit economics improve structurally, not just in dollars.
- Execution cadence has been fast and specific, not vague: 3MW to 480MW of self-built AI Cloud capacity inside 12 months, with Horizon 1 delivered to Microsoft in August 2026, Horizon 2 in commissioning, Horizons 3-4 targeted for Q4 2026 and 2026 capacity largely sold out.
The last reported quarter produced a -$684.0M GAAP loss on $450.4M of impairments, and the entire thesis now rests on an ARR metric the company itself disclaims as not GAAP revenue. Getting from ~$550M of annualized Q4 GAAP revenue to $4bn of ARR operating by December is roughly a 7x step-up in six months, funded by a debt stack that has gone from ~$1bn to $7.7bn of principal in twelve months (a further ~$2.4bn of financing was signed in August) - and the market has voted, taking the stock ~68% below its 52-week high.
- The gap between the metric and the money is enormous and explicitly disclaimed: ARR is defined as GPU/hour pricing for GPUs commissioned as of Dec 31, 2026 multiplied by 8,760 hours plus storage/ancillaries, is 'an operating metric, not a GAAP measure... not a substitute for revenue,' and depends on 'commissioning, testing and customer acceptance.' Q4 FY26 GAAP AI Cloud revenue was $70.5M - about $0.28bn annualized, or about 7% of a $4bn annualized rate.
- Q3 FY26 was a bad miss, not a rounding error: revenue $144.8M vs ~$219.9M consensus (-34%) and EPS -$0.30 vs ~-$0.22 (-39%). Revenue was flat YoY and down 21.6% QoQ. Whatever the strategic narrative, the delivery-versus-expectation record at this point is poor, and Q4 revenue then fell a further 5.2% QoQ to $137.2M.
- Owned GPUs mean owned depreciation and owned obsolescence, and the charges are already visible: $112.1M of D&A and a $450.4M impairment in Q4 FY26 alone, taking FY26 impairments to $638.8M (mainly decommissioned Bitcoin mining hardware) and D&A to $417.7M. Adjusted EBITDA (14% in Q4, 35% for FY26) exists only because D&A, impairment and stock comp ($205.0M in FY26) are excluded; on a GAAP basis operating margin was -452% in Q4 and -148% for FY26.
- Leverage has exploded and is now structurally senior to the equity: $6.77bn of recourse debt at Jun 30, 2026 (mostly convertible senior notes, $3.0bn of them issued in Q4) plus $938M drawn of ~$3.6bn of new Hardware-3 secured GPU financing (with limited parent guarantees; $2.7bn still to be drawn subject to conditions), where the maturity is tied to full payment of the Microsoft contract service fees. If Microsoft's tranches slip, the financing structure - not just the P&L - is exposed.
- The cash is audited now but largely borrowed or prepaid: the $7.6bn at Jun 30, 2026 includes $1.7bn of restricted cash, and it came from $6.3bn of convertible-note proceeds, $4.7bn of share issuance and $1.84bn of customer prepayments booked as deferred revenue (a liability to deliver service, our reading), against $4.3bn of FY26 capex. Q4 revenue ($137.2M), impairments ($450.4M) and the share count (394.1M at Aug 14, 2026) are known; what is not is how much of the $4bn ARR converts to GAAP revenue.
- Bitcoin mining - 82% of FY26 revenue and 49% of Q4 - is being deliberately dismantled, so near-term revenue gets worse before it gets better. Mining revenue fell to $66.7M in Q4 from $111.2M in Q3 and the decommissioning is what produced the impairments.
- The FY26 net loss of -$702.6M is itself cushioned by a $558.5M unrealized gain on financial instruments (capped call and prepaid forward transactions), against which sat $111.8M of debt-conversion inducement expense; without the gain the loss would have been about -$1.26bn (computed). None of it is operating cash.
- Dilution has been the primary funding tool and continues: shares went from 258.1M at Jun 30, 2025 to 380.2M at Jun 30, 2026 and 394.1M by Aug 14, 2026 (+53%), plus $4.74bn of share-issuance proceeds in FY26, plus NVIDIA's right to 30M more shares, plus 9,099,328 RSUs granted to EACH of the two co-CEOs on Jun 30, 2026 (~18.2M shares, six-year vest).
- NVIDIA's $70 strike is now ~186% above the $24.46 share price, so the ~$2.1bn of potential proceeds that were part of the funding narrative are, at current levels, not coming.
- Concentration remains real despite the new names: the ~$9.7bn Microsoft contract plus the ~$3.4bn NVIDIA contract still dominate ARR under contract, and the $2.8bn of newly signed TCV is spread across AI developers and startups whose own funding durability is unproven over a ~4-year contract term.
- Supplier and single-site concentration: a single ~$1.6bn Dell purchase agreement (parent-guaranteed) supplies the GPUs for the $3.4bn contract, and the 2026 delivery program is concentrated at Childress, Texas - one site, one grid, one interconnection risk.
What it is worth
EV/ARR and power-per-MW sanity check, framed as a medium-conviction long on execution. ~$9.6B equity cap (Nov 3 2026) against ~$137.2M of Q4 FY26 revenue (~$0.55B annualized) prices the equity almost entirely on the company-claimed ARR ramp ($4bn contracted for 2026 capacity, $1bn operating at Aug 26 2026) and the 480MW->1,210MW power build, not on trailing fundamentals; adding ~$1.9B of net debt (computed at Jun 30 2026) gives an enterprise value of about $11.6B, roughly 2.9x the $4bn contracted ARR (computed). The multiple only works if Childress Horizon 1-4 energizes on schedule, the ~150k-GPU fleet deploys, Microsoft/NVIDIA revenue converts, and GPU depreciation on the owned fleet stays contained -- several contestable assumptions stacked. Each case is enterprise value over FY2027 consensus revenue ($2.81B, fiscal year ending Jun 2027; estimate, aggregator), the one consensus metric all seven names share because none has positive EPS and the peers do not report ARR, converted to a share price on $1.94B of net debt (Jun 30 2026, unrestricted cash basis) and 394.06M shares (Aug 14 2026; the $2.4B of financing signed in August and any later share issuance are not included): bear at the two named peers that trade cheaper (CoreWeave 2.9x, Nebius 3.6x), base at the stock's own 4.1x, bull from the stock's own high since the FY26 results (7.5x, level with the 7.7x peer median) to the multiple the average analyst target implies (11.3x).
~$15–21
37% to 16% below the $24.46 close: 2.9–3.6x enterprise value over FY2027 consensus revenue of $2.81B (estimate, aggregator), converted on $1.94B of net debt and 394.06M shares: CoreWeave at 2.9x and Nebius at 3.6x, the two named peers that trade cheaper than IREN's 4.1x (EV on Nov 3 2026 closes over FY2027 consensus revenue; estimates, aggregator). The June bear (the multiple compressing back toward the mining-plus-optionality base) carried no price or multiple. The whole range sits below the 52-week low of $21.86 (Nov 2 2026), which is only 11% under the close: the case is IREN converging to CoreWeave and Nebius, the two GPU-cloud peers that trade cheaper, not a level the stock has traded at. Same driver: an AI-capex air-pocket, a GPU-depreciation re-rating, a financing scare, or a ramp slip that delays ARR conversion [estimate].
~$24
0% vs the $24.46 close: 4.1x enterprise value over FY2027 consensus revenue of $2.81B (estimate, aggregator), the stock's own multiple (2.9x the $4bn contracted ARR, computed). The median of the six named peers is 7.7x (CoreWeave 2.9x, Nebius 3.6x, Core Scientific 6.3x, Applied Digital 9.1x, Cipher 9.7x, TeraWulf 10.0x; EV on Nov 3 2026 closes over FY2027 consensus revenue; estimates, aggregator), which would be $49.91 (+104%); the base does not step up to it because the ramp is unproven and cash is burning. The June base was anchored near the then price of ~$46, which is 88% above the Nov 3 close (computed). Same driver: the anchor contracts are real but the ramp is unproven and cash is burning; range-bound on each energization / ARR-conversion print [estimate].
~$49–75
+98% to +208% vs the $24.46 close: 7.5–11.3x enterprise value over FY2027 consensus revenue of $2.81B (estimate, aggregator), converted on $1.94B of net debt and 394.06M shares. Low end: the stock's own high close since the FY26 results ($48.55 on Sep 22 2026 = 7.5x, level with the 7.7x median of the six named peers); high end: the multiple the average analyst target implies ($75.42, 20 analysts, estimate, aggregator = 11.3x). The June bull carried no price or multiple (a re-rating from a miner to a contracted-AI-infrastructure multiple); the close is 68% below the 52-week high of $76.87. Same driver: the $4bn contracted ARR converts, Childress/Sweetwater energize on schedule, and the equity re-rates from a miner to a contracted-AI-infrastructure multiple [estimate/company-claimed].
Adj. EBITDA (35% for FY26, 14% in Q4 FY26) flatters the picture: it excludes the depreciation, impairment ($450.4M in Q4 FY26; $638.8M for the year), and financing that drove a -$684.0M Q4 GAAP net loss (-$702.6M for FY26). ARR is a company-claimed target, not recognized revenue -- the valuation is a bet that the ramp converts. Owned-GPU model means the fleet depreciates, unlike shell-leasing peers valued on lease streams.
SWOT
Strengths
- Best-anchored miner-pivot — a direct ~$9.7B Microsoft (NVIDIA GB300) + ~$3.4B NVIDIA book -- investment-grade / chipmaker counterparties, not a single neocloud [fact]
- Owned-GPU AI Cloud captures more of the value chain than landlord-only peers that merely lease datacenter shells [fact]
- Power in hand — Sweetwater 1 substation energized on schedule and Childress Horizon 1-4 advancing -- the scarce asset (energized, interconnect-queued MW) is controlled, not queued [fact]
- Adj. EBITDA positive (35% for FY26 — 14% in Q4 FY26) even mid-transition, and AI Cloud revenue inflecting (+110% QoQ in Q4, past mining) [fact]
Weaknesses
- GPU depreciation + technology-obsolescence risk on an owned, leveraged fleet -- a cost landlord-only peers (WULF, CIFR, CORZ, APLD) do not carry on their books [estimate]
- Heavy capex / financing intensity to fund the 480MW->1,210MW build against sub-$1B current revenue; FY26 capex was $4.3B against $707.0M of revenue, funded by $9.7B of financing inflows [fact]
- GAAP loss-making (-$684.0M Q4 FY26 — -$702.6M FY26) with large non-cash impairments ($450.4M in Q4) as mining hardware is decommissioned; revenue fell QoQ during the transition [fact]
- Counterparty and geographic concentration — heavy reliance on Microsoft and on one Texas grid region (Childress/Sweetwater) [fact]
- Non-US (Australia) domicile is a framing/governance nuance for US-first mandates, though listed on Nasdaq with US-heavy assets [fact]
Opportunities
- Convert the $4bn contracted ARR ($1bn operating at Aug 26 2026) into recognized revenue as Childress Horizon 1-4 and the ~150k-GPU fleet come online (company-claimed targets) [company-claimed]
- Longer-term 5GW+ AI expansion ambition across Childress, Sweetwater, and international sites if power and financing keep pace [company-claimed]
- Additional hyperscaler / NVIDIA contracts layering onto the owned-GPU AI Cloud as capacity energizes -- the Microsoft/NVIDIA relationships are a validation beachhead [estimate]
- Re-rating from a Bitcoin-miner multiple to a contracted-AI-infrastructure multiple as AI Cloud revenue, which passed mining in Q4 FY26 ($70.5M against $66.7M), keeps growing [estimate]
Threats
- AI-capex air-pocket — the whole miner-pivot slice is one bet that the AI-datacenter capex supercycle keeps compounding; two hyperscaler capex guide-downs would reprice the equity independent of IREN's execution [estimate]
- GPU/architecture obsolescence (Blackwell -> Rubin -> next) strands capital on the owned, leveraged fleet faster than modeled [estimate]
- Financing risk — a risk-off market that raises the cost of GPU financing / capital throttles the build and pressures the balance sheet [estimate]
- Bitcoin price weakness on the still-material mining segment removes the cash-flow bridge during the transition [fact/estimate]
- Execution on the 480MW->1,210MW ramp and grid/interconnect timing is the entire story -- any slip delays ARR conversion [fact]
Moats, dependencies & bottlenecks
Moats
Hyperscaler + chipmaker anchor (~$9.7B Microsoft GB300 + ~$3.4B NVIDIA) the strongest anchor among miner-pivots (IG + chipmaker vs single-neocloud peers), but contractual, not technological lock-in; renewal/conversion is the risk The single most important edge and the clearest de-risker versus the Fluidstack/CoreWeave-counterparty peers [fact]
not shell-leasing) captures more value per MW than landlording, but the owned fleet depreciates and can be replicated by any well-capitalized operator [estimate] The structural differentiator from WULF/CIFR/CORZ/APLD, at the cost of taking GPU-depreciation onto the balance sheet [fact/estimate]
energized, interconnect-queued MW is the genuinely scarce asset, but concentrated in one grid region The scarce asset is the site's power, not its ASICs -- IREN controls it and is energizing on schedule [fact]
capital-replicable; the moat is execution + counterparty relationships, not a structural barrier Advantage is operational execution and the Microsoft/NVIDIA beachhead, not a defensible technology [estimate]
Dependencies
The dominant AI Cloud counterparty; investment-grade (a strength vs single-neocloud peers), but concentration means a reduction/renegotiation would impair the whole thesis [fact]
NVIDIA (GPU supply + ~$3.4B AI Cloud contract + ~30M-share option) Sole effective supplier of the owned GPU fleet AND a customer/investor (circular financing dynamic) -- allocation, refresh cadence, and the ~$3.4B contract all run through NVIDIA [fact]
The 480MW->1,210MW build leans on cash, operating cash flow, GPU financing, and additional capital; debt reached $7.6B at Jun 30, 2026 with $2.7B of GPU financing still to be drawn -- a risk-off window throttles the ramp [fact/estimate]
Energized MW is the scarce asset; concentrated in one grid region, so grid/interconnect timing gates ARR conversion [fact]
Mining was the majority of FY26 revenue ($578.2M of $707.0M) and 49% of Q4 FY26 ($66.7M of $137.2M); BTC weakness removes the cash-flow bridge during the transition [fact]
Advantages
- Best-anchored miner-pivot — direct ~$9.7B Microsoft (IG) + ~$3.4B NVIDIA book, not a single-neocloud counterparty [fact]
- Owned-GPU AI Cloud captures more of the value chain than shell-leasing peers [fact]
- Energized, controlled power at Childress / Sweetwater, TX — the scarce asset is in hand and energizing on schedule [fact]
- AI Cloud revenue inflecting (+110% QoQ to $70.5M in Q4 FY26, ~51% of revenue) with FY26 adj. EBITDA of $245.7M (35%) [fact]
- US-heavy asset base (Texas) and Nasdaq listing despite Australian domicile -- explicitly not a mainland-China name [fact]
Weaknesses
- GPU depreciation + technology-obsolescence risk on an owned, leveraged fleet that landlord-only peers do not carry [estimate]
- Heavy capex/financing intensity — FY26 capex of $4.3B against $707.0M of revenue funded by $9.7B of financing inflows [fact]
- GAAP loss -$684.0M (Q4 FY26 — -$702.6M for FY26) with $450.4M non-cash impairments in Q4; revenue fell QoQ during the transition [fact]
- ARR figures are company-claimed targets, not recognized revenue — the $4bn run-rate targeted to be operational by end-2026 assumes flawless ramp execution [company-claimed]
- Concentration in Microsoft and one Texas grid region — non-US (Australia) domicile is a US-first framing nuance [fact]
Bottlenecks
- Power + grid interconnect at Childress / Sweetwater — energizing 480MW (2026) then 1,210MW (2027) is gated by substation build and grid interconnect timing [fact]
- Capital + GPU financing — the build cannot outrun access to cash, GPU-collateralized financing, and additional capital; $2.7B of GPU financing was still undrawn at Jun 30, 2026 and FY26 capex was $4.3B [fact]
- NVIDIA GPU allocation -- supply of GB300 / Blackwell caps how fast the ~150k-GPU fleet deploys [fact]
- Execution timing — converting company-claimed ARR to recognized revenue depends on Childress Horizon 1-4 coming online on schedule [fact/company-claimed]
- Skilled operations talent to stand up and run liquid-cooled GPU datacenters at high utilization [estimate]
Top signals & trends
Top signals
10 sourced items since this entry's vintage are in the evidence rows (latest print, guidance, deals, resolved watch items); the analysis text is as of the entry's stated date and predates them unless dated otherwise.
AI Cloud grew +110% QoQ to $70.5M in Q4 FY26 (about $0.28B annualized); the bull case needs the $4bn contracted ARR ($1bn operating at Aug 26 2026) to actually convert to recognized revenue -- watch each print [company-claimed]
Horizon 1 delivered to Microsoft in August 2026, Horizon 2 in commissioning and Horizons 3-4 targeted for Q4 2026 is the execution tell that de-risks the ramp, with the Sweetwater 1 substation energized on schedule [fact]
Debt reached $7.6B at Jun 30, 2026 (cash $5.9B plus $1.7B restricted) after $3.0B of convertible notes in Q4; the size and cost of GPU financing into any risk-off window is the core balance-sheet risk [fact]
Owning the fleet means residual-value and obsolescence disclosures matter -- evidence of <accounting-life economic life would pressure the equity [estimate]
Mining was 82% of FY26 revenue but 49% of Q4 FY26 revenue and remains the cash bridge; BTC weakness or faster decommissioning swings the transition-quarter optics [fact]
IREN is a high-beta expression of the AI-infra trade; a hyperscaler capex guide-down reprices the whole miner-pivot slice at once [estimate]
Trends
Repurposing energized, interconnect-queued power into long-dated AI leases re-rates the equity from a mining to an AI-infra multiple -- IREN is the owned-GPU outlier in the slice [fact]
Drives the Microsoft/NVIDIA anchor demand and the whole ARR ramp; the entire bull thesis rests on this persisting [fact]
Latest-gen GPUs win premium contracts (positive) but accelerate obsolescence of the OWNED in-place fleet (negative for depreciation/residuals) -- sharper for IREN than for shell-leasing peers [estimate]
NVIDIA validation is a strength, but supplier/customer/investor loops invite the same scrutiny as the broader neocloud chain [estimate]
Energized, interconnect-queued MW is the scarce asset -- IREN controlling power at Childress/Sweetwater is a genuine advantage while power gates the industry [fact]
Peers are posting large GAAP losses now (WULF ~$428M Q1'26; APLD ~$101M Q3 FY26); IREN's own -$684.0M Q4 FY26 loss and $7.6B of debt fit the pattern [fact]
Ecosystem & competitor graph
Suppliers feed the company; customers pull from it. Line thickness shows the strength of each tie (supply-chain dependency, customer earnings contribution). Hover to isolate a tie.
Core GPU supplier (GB300 / air-cooled Blackwell) for the owned fleet; also a ~$3.4B AI Cloud contract customer and holds an option to buy ~30M IREN shares -- the dominant single dependency
Server/rack integrator assembling GPU systems for AI datacenter build-outs
GPU server/rack supplier to AI datacenters
Power, cooling, and thermal-management infrastructure for high-density liquid-cooled GPU datacenters
Energized, interconnect-queued power at Childress and Sweetwater, TX -- the genuinely scarce asset
Anchor AI Cloud customer: ~$9.7B, 5-year GPU cloud contract (NVIDIA GB300), phased through 2026 at the 750MW Childress campus -- the strongest hyperscaler anchor among miner-pivots
~$3.4B, 5-year AI Cloud contract (air-cooled Blackwell) within 60MW at Childress -- both supplier and customer (circular)
Largest listed neocloud (owned-GPU rental at far greater scale, ~$99B backlog); the scaled owned-GPU model IREN's AI Cloud is a smaller, power-anchored version of [fact]
Listed pure-play neocloud (Netherlands-domiciled, non-US, flagged; explicitly NOT mainland-China) competing for the same GPU-rental workloads [fact]
Miner-pivot peer, full pivot to AI/HPC via Google-backstopped Fluidstack leases; shell-leasing (no owned GPUs) vs IREN's owned-GPU model [fact]
Miner-pivot peer, entire 300MW Barber Lake TX leased to Google-backstopped Fluidstack; shell-leasing structural twin of WULF [fact]
Largest miner-pivot by contracted backlog (~590MW/>$10B to CoreWeave); stayed independent after shareholders rejected CoreWeave's buyout -- landlord model, not owned-GPU [fact]
Build-to-suit AI-factory landlord (CoreWeave + a separate IG hyperscaler, ~400MW Ellendale ND), pursuing a REIT conversion; shell-leasing, not owned-GPU [fact]
Private pure-play energy-first AI factory (OpenAI/Stargate Abilene TX) -- the closest private analog; NOT publicly tradable [fact]