AI infrastructure and GPU cloud

IREN

IREN

The USD 14.41 November 19 regular close is 19.4% below the USD 17.87 price used in the latest completed assessment. The lower entry price raises attractiveness to 6.4/10 with unchanged scenarios; business strength remains 6.1/10 and financing and dilution remain major risks.

Analysis and ratings: ChatGPT·Business analysis 11/20/2026·Ratings & valuation 11/20/2026
THESIS STATUSWatch — Q1 confirms rapid AI Cloud growth, but H2–4 were delayed
CONFIDENCEMediumHigher for current demand and revenue; lower for long-term FCF per share.
HORIZON2026–2029
WHAT MATTERS MOSTH2–4 acceptance, collections after prepayments, construction financing, GPU replacement and normalized cash per diluted share.

Business strength & price attractiveness

How AI assesses them

Business strength

Medium confidence
6.1/ 10

Previously 6.1 ·
Unchanged · Assessed

Business strength, independent of the stock price.

No new operating evidence was added after the Q1 review. Company strength remains 6.1/10; the share-price move does not affect it.

Five areas & supporting evidence

Price Attractiveness

Low confidence
6.4/ 10

Previously 5.2 ·
↑ +1.2 · Assessed

At 14.41 USD · close
Three-year horizon, through .

**Price attractiveness: 6.4/10 at USD 14.41.** Nasdaq IREN common stock, regular close November 19, 2026 at 4 p.m. EST / 10 p.m. CET; assessed November 20, 2026, methodology 1.2, through November 20, 2029. [History](https://stockanalysis.com/stocks/iren/history/): open 15.97, high 16.66, low 14.18, close USD 14.41, daily decline 5.45% and volume about 121.8m shares. Operating assumptions, the financial bridge and ranges below remain unchanged: base price **USD 24–45**, adverse **USD 3–12**, both with USD 0 dividends. Returns from the new price are **+67% to +212%** (about 19–46% annually) and **−79% to −17%** (about −41% to −6% annually). A 19.4% decline from the latest assessment's USD 17.87 explains the rise from 5.8 to 6.4/10; adverse-case and financing risks prevent a 7–8 score. Strength of 6.1/10 and the five dimension scores 7.0 / 8.2 / 5.8 / 5.3 / 4.0 remain unchanged. At 12.5× normalized flow, 525m diluted shares and an 18% discount rate, USD 14.41 requires about **USD 0.99bn annual 2029 flow**, below the USD 1.2–1.6bn base range but above the USD 0.25–0.55bn adverse range. Each USD 0.5bn permanent flow change moves today's value around USD 7.25 per share. Confidence is low: contract-level collections, advance refunds, repayments and replacement are not detailed. H2–4 acceptance, funding without excessive dilution and two quarters of per-share flow after replacement and debt service could raise the score; further delays, lost contracts or advances, emergency financing and issuance without per-share value growth could lower it. Older prices and returns below are historical, not current.

Scenarios & supporting evidence

What changed in the latest update

New information: The USD 14.41 November 19 regular close is 19.4% below the USD 17.87 price used in the latest completed assessment. The lower entry price raises attractiveness to 6.4/10 with unchanged scenarios; business strength remains 6.1/10 and financing and dilution remain major risks.

Strength 6.1 → 6.1Attractiveness 5.2 → 6.4Scores changed

The numerical comparison does not by itself confirm or weaken the investment thesis.

Dated prices & model comparison
Input / modelPreviousCurrent
Reference price21.64 USD14.41 USD
Base modelled return+59.4 %+139.4 %
Adverse modelled return-65.3 %-48.0 %

Quote dates: 11/6/2026 → 11/19/2026. Model horizons end: 11/7/2029 → 11/20/2029. Cumulative returns include modelled dividends; they are scenarios, not guarantees.

Terminal-price and cumulative-dividend assumptions are unchanged; reference-price inputs differ as shown above.

View rating history →

Assessed · Methodology 1.2. Analytical judgments by ChatGPT; price changes do not automatically update these scores.

MARKET OVERVIEW

IREN · Price chart

Delayed data · English controls.TradingView
Analysis contents · 13 sections
01

Company Snapshot

20. 11. 2026 · ChatGPT

The USD 14.41 November 19 regular close is 19.4% below the USD 17.87 price used in the latest completed assessment. The lower entry price raises attractiveness to 6.4/10 with unchanged scenarios; business strength remains 6.1/10 and financing and dilution remain major risks.

Previous dated analysis and history

AreaReview as of November 7, 2026 — partial coverage
Company and tickerIREN Limited • NASDAQ: IREN
SectorAI infrastructure, data centers and GPU cloud
Thesis statusWatch — Q1 confirms rapid AI Cloud growth, but H2–4 were delayed and per-share returns remain conditional
ConfidenceMedium; higher for current demand and revenue, lower for long-term FCF per share
HorizonWorking horizon 2026–2029; outlook through 2030
Current viewWatch. Q1 improved evidence on demand and operating economics but did not close financing for the full expansion or three-year per-share value
Company strength6.1/10 — methodology 1.2; dimensions 7.0 / 8.2 / 5.8 / 5.3 / 4.0; assessed November 7, 2026
Price attractivenessUnassessed at $18.03, the November 5, 2026 regular-session close; assessed November 6. A defensible bridge from project economics to three-year per-share value is missing
Next quarterly resultsDate not yet researched because fewer than two calendar months have elapsed since Q1 FY27 results on November 5.
Evidence baseQ1 FY27 results, presentation, 10-Q and call; FY26 materials and attached project files. A complete inventory of project discussions is not documented.

What matters most: H2–4 acceptance, per-share economics, AI demand quality, financing for the full plan and renewal pricing. Management moved more than $4 billion of operating ARR into January and potentially late February 2027 because of a customer-requested design change and supplier timing.

02

Investment Thesis

Q1 FY27 update

The thesis is supported by rapid AI Cloud revenue growth, backlog and pricing, but remains conditional on delivery, collections, financing and per-share economics. H2–4 acceptance slipped from the original 2026 target to January and potentially late February 2027.

IREN can monetize scarce land, grid connections and construction capabilities by delivering operational AI compute. Owned data centers, GPUs and supporting services can improve control of costs, delivery and customer relationships. Microsoft's acceptance of Horizon 1 and a broader customer base are major progress from the original future-AI opportunity.

Investment condition: after operations, capital and replacement costs, contracts must earn enough per fully diluted share. Large backlog and ARR alone do not meet this condition.

Thesis pillarCurrent assessmentDecisive next evidence
Demand for delivered capacityStrengtheningRepeat payments, contract renewals and customer economics
Execution capabilityPartly confirmed by Horizon 1; Canadian retrofit reliability under reviewAcceptance of remaining phases and documented uptime/SLA by location
Shareholder valueNot yet demonstratedCompleted-project returns after all costs and FCF growth per share

IREN serves both training and inference, so it need not wait for every AI application to become profitable immediately. Over a longer horizon, however, compute investment must be supported by end-user revenue or demonstrated productivity. Financing AI customers can postpone this test.

Strongest counterthesis: AI succeeds and IREN grows substantially, but competition, GPU renewal, debt and new shares leave existing owners with weaker returns than company growth suggests.

03

The Business

Q1 FY27 update

IREN combines power sites, data centers and GPU cloud capacity. Q1 AI Cloud revenue reached $200 million. H1 entered operation; H2–4 commissioning continues, but customer-requested design changes and supplier timing delayed acceptance.

IREN connects three layers: data centers and power infrastructure → GPUs, networking and storage → cloud services and environment management. For the last layer, plans must be distinguished from operations: in an interview published September 25, 2026, the CCO said managed services were not yet live despite signed contracts. Mirantis is intended to expand this layer; its margin contribution remains unproven. CCO interview, 49:07–52:22

FY2026 management lists Microsoft, NVIDIA, Together AI, Fireworks AI, Perplexity, Cohere, Figure AI, other developers and an unnamed frontier AI lab. Renewed and expanded Together AI and Fireworks AI relationships are better evidence than new-logo counts alone.

Frontier AI customer identity — unverified report: Blockspace, October 12, 2026, cites The Information as naming Anthropic. The official results and available IREN announcements reviewed do not confirm the name; the purported deal's value, capacity and timing are missing. FY2026 results already include an unnamed frontier lab. Analytical treatment: until an additional contract is documented, do not add further ARR, backlog or an extra customer. The report is a secondary citation, not a documented new company contract.

Bitcoin mining is transitional. August management expected its effective end by December 2026. The exit temporarily reduces existing revenue while AI capacity awaits commissioning.

Four different measures

MeasureMeaning
Secured power and pipelineFuture construction opportunity; availability depends on connections and permits
MW capacityDistinguish gross power from IT load, and planned from operating capacity
Contracted or operating ARRAnnualized pace under stated assumptions; always specify date and operating status
GAAP revenue and FCFReported revenue and cash after investment; central to stock returns

June 30, 2026 backlog: USD 5.1 billion ASC 606 RPO and USD 11.4 billion contracted ASC 842 lease value; the company reports roughly USD 16.6 billion total. Rounded components may not sum exactly. These are two accounting regimes. Do not add the total again to RPO, ARR or included contracts. FY2026 10-K, Note 4

04

Competitive Advantages

Q1 FY27 update

Vertical integration across sites, power, data centers and compute remains a meaningful advantage. H1 entered service, while the H2–4 delay shows that time-to-market still carries execution risk.

The main advantage is infrastructure replication time. Power, land, connections, transformers, cooling, networking and operating teams must meet in the right place at the right time. Customers needing compute now may pay for that combination. IREN already has these pieces…

Repeating designs and supply chains across projects can help. Accepted Horizon 1 supports credibility with customers and lenders. Vertical integration reduces reliance on data-center landlords.

The moat is conditional. Rivals also buy GPUs; well-funded hyperscalers own infrastructure and customers. Compute oversupply, pricing pressure and simultaneous hardware renewal and debt service would be especially damaging. Availability, service quality and through-cycle capital returns must establish durability.

05

Growth Drivers

Q1 FY27 update

Growth depends on converting a $20.3 billion backlog into accepted capacity, GAAP revenue and cash collections, while preserving economics through financing and future GPU replacement. Two-thirds of backlog is investment grade according to management.

Financing after the price signal

IREN closed $363 million of Wingspire financing for deployed, revenue-generating AI compute assets. The transaction supports capital recycling, but without a disclosed rate, maturity or precise liquidity addition it does not resolve the entire $25–30 billion investment plan or $13.81 billion of capital commitments.

Near-term: turn signed contracts into operating capacity and reported revenue. In August, the company cited roughly USD 1 billion operating ARR and USD 4 billion contracted ARR for 2026 capacity. Much of December's delivery is expected to affect revenue mainly in the quarter ending March 2027.

Next wave: targeted cumulative capacity roughly 0.3 GW IT at end-2026 and 0.8 GW IT in 2027. About 1.2 GW gross power is a different measure. NVIDIA's contract is expected to add roughly USD 700 million ARR from 2027 and is excluded from the cited 2026 target.

September 8 update: Sweetwater 1 and 2, totaling 2 GW, were conditionally included as Base Load in ERCOT Batch Zero, subject to further approvals. The first 300 MW gross at Sweetwater 1 targets Q4 2027 delivery. Official update

Redbank, October 14, 2026 report: according to DCD, IREN proposes a 900 MW campus near Wagga Wagga in New South Wales, targeting connection from 2029. The project is at an early assessment stage; connection and permits remain conditions. The stated 900 MW cannot be treated as secured power or operating IT capacity. The report places it outside the announced 5 GW secured-power portfolio. This is a development opportunity without a demonstrated contribution to three-year cash flow. The company post and project website are related references; the analysis relies only on DCD's secondary report for details.

Inference is an opportunity whose economics need testing. More tokens could increase utilization, but compute per token, pricing and workload location matter. Some inference depends more on latency and proximity to users.

06

Economics & Financial Profile

20. 11. 2026 · ChatGPT

The financing flywheel is not FCF: the 10-Q clarified that USD 363m of previously announced equipment financing was fully drawn, carries a fixed implicit rate of 9.51% and is repaid in 36 equal monthly installments from September 2026. This demonstrates refinancing of installed equipment but increases fixed capital servicing. GPU loans and customer advances help fund the 0.3 GW program; the same future cash flow cannot simultaneously be counted as project value, an advance and a repayment source. For 0.8 GW and other projects, USD 6bn of planned funding remains market-dependent. NVIDIA's approximately USD 3.4bn five-year cloud contract is a customer contract. Separately, its right to buy up to 30m shares at USD 70, potentially USD 2.1bn, releases subject to delivered GPUs and regulatory conditions and is neither cash nor committed available financing until exercised. NVIDIA and IREN announcement · FY26 10-K. Project financing analysis is an analytical input, not guidance.

Previous dated analysis and history

Reported metricQ1 FY27 to Sep. 30, 2026Q4 FY26
Total revenue$228m$137m
AI Cloud revenue$200m$71m
Adjusted EBITDA$66m$19m
Net result−$263m−$684m
AI Cloud segment gross margin86%86%

AI Cloud revenue grew 182% sequentially and represented $200 million of total $228 million revenue. The 86% segment gross margin excludes depreciation, amortization and part of site costs, and adjusted EBITDA excludes $106 million of stock-based compensation. Operating cash flow was $529 million, but deferred revenue increased $645 million; subtracting that item gives approximately −$116 million before a full working-capital normalization.

At September 30, cash was $3.522 billion, restricted cash $2.256 billion, debt $9.897 billion and capital commitments $18.947 billion. Of $15 billion of remaining FY27 investment, the company identifies $5.8 billion of cash and restricted cash and $3.2 billion of committed funding; another $6 billion remains targeted and market-dependent. The $363 million equipment financing was fully drawn, carries a 9.51% fixed implicit rate and amortizes over 36 monthly payments. These facts support financing access but not yet recurring FCF per share.

07

Key KPIs

Q1 FY27 update

The critical KPIs are H2–4 acceptance, GAAP revenue after acceptance, collections net of prepayments, SLA performance, diluted shares, debt, financing cost and cash after GPU replacement.

KPIBaselineNext test
Horizon acceptanceH1 accepted in AugustH2–4 delivery and acceptance against updated timing
Operating and contracted ARRUSD 1 billion and USD 4 billion in AugustOperating ramp through year-end 2026 and subsequent GAAP revenue
AI Cloud revenueUSD 70.5 million in Q4 FY2026Growth in the September, December and March quarters
Reported and adjusted cash flowQ4 strongly supported by deferred revenueSeparate prepayments and working capital from recurring cash generation
GPU cohort returnsShort history at the new scaleOperating returns after data-center costs, financing and replacement
Shares and potential dilutionMonitor issuance, RSUs and convertiblesShare-count changes alongside revenue and project-value growth
Customer concentration and qualityMore logos and several large buyersLargest buyers' shares, collections and renewals
CapEx and available financingFY2027 plan of USD 25–30 billionCommitted funds, draw conditions and remaining capital needs
New versus renewed pricingStrong new contracts according to the callPrice, duration, prepayments and margins at actual renewal
Uptime, SLA credits and redundancy by locationNew criticism of Canadian retrofits; aggregate operating data missingVerify backup power, incidents, compensation and retention; distinguish Canada from Texas
Managed services versus bare metalManaged-layer operation not yet confirmedLaunch date, paying customers and separately reported revenue and margins

Use actual period-end shares and economic dilution scenarios for per-share comparisons. Loss-period accounting diluted EPS can exclude antidilutive instruments without removing future impact. Distinguish agreed deposits, cash received and ability to use it.

08

Management & Capital Allocation

Q1 FY27 update

Management delivered H1 but delayed more than $4 billion of operating ARR from late 2026 into January or possibly late February 2027. Capital allocation remains dependent on project financing, customer prepayments and additional equity issuance.

Financing after the price signal

IREN closed $363 million of Wingspire financing for deployed, revenue-generating AI compute assets. The transaction supports capital recycling, but without a disclosed rate, maturity or precise liquidity addition it does not resolve the entire $25–30 billion investment plan or $13.81 billion of capital commitments.

Strengths: Daniel and William Roberts build around AI infrastructure's physical constraints. Horizon 1 acceptance, customers and completed GPU financing support credibility. August's call gives more capital-source detail and distinguishes infrastructure investment from each year's deliveries.

Open test: manage a much larger organization, simultaneous construction and Mirantis and Nostrum integration. Ambition matters only with returns above capital costs. Ability to slow further construction becomes important if funding or demand weakens.

Governance: roughly 394.1 million common shares at August 14. July grants to both co-CEOs total about 18.2 million RSUs. Founders' significant voting rights increase independent-board oversight importance. FY2026 10-K

New proxy materials, filed October 13, 2026: the DEF 14A calls an AGM on November 24, 2026. Proposals 2 and 3 seek approval for share settlement of existing prepaid-forward and capped-call transactions; they do not themselves announce a new discretionary repurchase program. The document also confirms that the July co-CEO grants vest over four years with continued service, with another two years of holding for each tranche, and replace further equity incentives through FY2031. This clarifies dilution timing but does not change the number of RSUs already granted. Concerns about the absence of performance conditions in these grants should be distinguished from ordinary compensation, which also includes performance elements.

Ownership supplement, located October 10: the Schedule 13G filed September 30, 2026 reports Kenneth Griffin's indirect beneficial ownership of 18,687,938 shares, or 4.7%, through Citadel-related persons. Related persons' reported holdings overlap and must not be added together; instruments convertible or exercisable into shares may also be included. The footnote dates holdings to the September 30 market open, while percentages use 394,058,648 shares outstanding on August 14. This document therefore does not update the October 10 share count or demonstrate a net directional bet or short-selling motives; it does not independently change the thesis.

August-call assessment: concrete on deliveries, relationships and funding; weaker proof of future per-share returns. Comparing GPU financing to real estate explains contractual cash flow, but rapid technological replacement needs distinct stress scenarios.

09

Risks / Bear Case

Updated bear case

The main downside is a further H2–4 delay, supplier or SLA problems, weaker financing availability, higher capital costs, more dilution and lower residual GPU value. Backlog and ARR remain conditional on delivery and acceptance.

Financing after the price signal

IREN closed $363 million of Wingspire financing for deployed, revenue-generating AI compute assets. The transaction supports capital recycling, but without a disclosed rate, maturity or precise liquidity addition it does not resolve the entire $25–30 billion investment plan or $13.81 billion of capital commitments.

The greatest hidden fragility is simultaneous risk: a customer slows, new contracts price worse, GPUs age faster and funding gets expensive while construction must be completed.

  • AI monetization: customers could finance compute from investment capital longer than end-product economics support.
  • Leverage and dilution: growth needs major investment; available credit alone does not prove common-share attractiveness.
  • Execution and SLA: delays or unreliability may defer revenue and trigger credits or contractual claims. In a public post available October 9, SemiAnalysis criticizes outages and networking or storage problems in Prince George and Mackenzie; these are an external researcher's claims, without published aggregate SLA or revenue-impact data. Earlier management statement: on September 25, the CCO had already acknowledged that Prince George launched without backup power from day one and that commissioning had begun; he said customers knew the limitations. This was not a response issued after the new criticism. Interview, 49:07–52:22. These sources do not establish the scale of problems. Canadian retrofits should not automatically be equated with new Texas construction: Microsoft's August 13 acceptance of Horizon 1 remains documented positive evidence.
  • GPU and data-center renewal: accounting and economic lives may differ; new generations can require modifications.
  • Competition: oversupply, hyperscalers' own chips and more efficient models may change pricing and utilization. In its October 14, 2026 public release, Gartner forecasts that over 80% of neoclouds will cease operating independently by 2030 because of costs and insufficient differentiation. This is an industry forecast, where loss of independence includes consolidation; it does not provide an IREN bankruptcy probability. For the analysis, it reinforces the need to test returns after hardware replacement, new-contract pricing and financing availability. The company may benefit from demand growth while delivering weak returns per share. The underlying paid Gartner model was not available to the source analysis.
  • Connections and geographic expansion: regulation and readiness vary; pipeline is not completed capacity.
  • Concentration and governance: more customers do not necessarily mean balanced revenue shares; founder compensation and issuance affect per-share value. The July co-CEO grants have time-based vesting without performance conditions; they must be distinguished from ordinary compensation with performance elements described in section 8.

Strongest bear case: customers and infrastructure are delivered, but investment, capital service and dilution leave insufficient value per share — even if AI succeeds long term.

10

Valuation

ChatGPT produces the calculations and valuation scenarios. Share-price inputs are dated to the analysis; they are not continuously updated market data.

CHATGPT SCENARIOS · 20. 11. 2026 → 20. 11. 2029

Different outcomes. Different values.

Historical analysis input
14.41 USD · 19. 11. 2026 · 16:00 EST / 22:00 CET
Bear case
3–12 USD
Base case
24–45 USD
Conditional scenarios from the analysis. The dashed line marks the historical share price used in the calculation. Values are neither current intrinsic value nor a share-price forecast.

20. 11. 2026 · ChatGPT

Price attractiveness: 6.4/10 at USD 14.41. Nasdaq IREN common stock, regular close November 19, 2026 at 4 p.m. EST / 10 p.m. CET; assessed November 20, 2026, methodology 1.2, through November 20, 2029. History: open 15.97, high 16.66, low 14.18, close USD 14.41, daily decline 5.45% and volume about 121.8m shares.

Operating assumptions, the financial bridge and ranges below remain unchanged: base price USD 24–45, adverse USD 3–12, both with USD 0 dividends. Returns from the new price are +67% to +212% (about 19–46% annually) and −79% to −17% (about −41% to −6% annually). A 19.4% decline from the latest assessment's USD 17.87 explains the rise from 5.8 to 6.4/10; adverse-case and financing risks prevent a 7–8 score. Strength of 6.1/10 and the five dimension scores 7.0 / 8.2 / 5.8 / 5.3 / 4.0 remain unchanged.

At 12.5× normalized flow, 525m diluted shares and an 18% discount rate, USD 14.41 requires about USD 0.99bn annual 2029 flow, below the USD 1.2–1.6bn base range but above the USD 0.25–0.55bn adverse range. Each USD 0.5bn permanent flow change moves today's value around USD 7.25 per share. Confidence is low: contract-level collections, advance refunds, repayments and replacement are not detailed. H2–4 acceptance, funding without excessive dilution and two quarters of per-share flow after replacement and debt service could raise the score; further delays, lost contracts or advances, emergency financing and issuance without per-share value growth could lower it. Older prices and returns below are historical, not current.

Previous dated analysis and history

Reference price: NASDAQ common stock, $21.64 — November 6, 2026 regular-session close. Assessed November 7. The verified close rose 20.02% from the prior session; intraday and after-hours prices were excluded.

Price attractiveness

5.2/10 at $21.64. Quote date: November 6, 2026 regular-session close. Three-year horizon to November 7, 2029. Methodology 1.2.

ItemBase caseBear case
Business assumptionsH2–4 reach acceptance by February 2027; the first 0.3 GW ramps, funding uses committed sources and customer prepayments, and further growth remains selective. FY29 revenue $8.0bn; EBITDA before SBC $5.60bn; depreciation $2.60bn; net interest $0.75bn; SBC $0.30bn; modeled tax 25%.The first stage completes more slowly, broader expansion is constrained, financing costs rise, and higher GPU replacement and dilution reduce per-share value. FY29 revenue $4.5bn; EBITDA before SBC $2.85bn; depreciation $1.65bn; net interest $0.55bn; SBC $0.25bn; modeled tax 25%.
Replacement and shareholder cash$2.65bn annual GPU replacement and maintenance; normalized cash used for valuation $1.2–1.6bn.$1.60bn annual GPU replacement and maintenance; normalized cash used for valuation $0.25–0.55bn.
Diluted shares and multiple500–550m shares; 11–14× normalized cash. Conversion and debt are not counted twice.470–520m shares; 7–10× normalized cash. Higher capital cost and limited GPU residual value are reflected in cash and the multiple.
Price in three years$24–45$3–12
Three-year dividends$0 working reinvestment assumption, not company guidance.Same $0 working assumption.
Total three-year return from $21.64+11% to +108%, roughly 3% to 28% annually−86% to −45%, roughly −48% to −18% annually

Friday's 20.02% price increase reduced the margin of safety without changing the business assumptions. The low end of the base case offers only about 3% annually, while the bear case still allows a loss of almost half to more than four-fifths of value. Confidence is low because value remains highly sensitive to cash after GPU replacement, financing cost and the fully diluted share count.

11

What Would Change the Assessment?

These are analytical review tests, not automatic buy or sell rules.

Area and checkpointThesis confirmationReason to reassess
Horizon 2–4, Jan.–Feb. 2027Delivery and acceptance with acceptable economicsFurther delays, credits or lost contracted capacity
Conversion of more than $4bn ARROperating ARR becomes GAAP revenue and cash collectionsARR grows while reported revenue and cash repeatedly lag
Financing through FY27Closed sources cover commitments without destructive dilution$6bn of targeted funding does not materialize or capital cost impairs returns
Project economics, 2027–2029Cash return after site costs, renewal and debt exceeds cost of capitalReturns depend on optimistic residual values and more leverage
Renewals and customer qualityPayments and renewals are supported by customer economicsDefaults, departures or unattractive renewal returns
Per-share valueNormalized cash grows per diluted shareIssuance and RSUs absorb the economic benefit
Reliability and redundancyStable uptime and retention documented by siteMaterial SLA credits, incidents or departures
Managed servicesLive paying customers and separately identified marginMarketing without delivered service
AGM, Nov. 24, 2026Clear impact of approved mechanisms on shares and cashApproval confused with a completed buyback

The key questions are project returns after financing and fleet renewal, the recurring portion of operating cash flow, sensitivity to 30% lower renewal pricing, end-customer economics and normalized FCF per fully diluted share.

12

Current View

20. 11. 2026 · ChatGPT

The USD 14.41 November 19 regular close is 19.4% below the USD 17.87 price used in the latest completed assessment. The lower entry price raises attractiveness to 6.4/10 with unchanged scenarios; business strength remains 6.1/10 and financing and dilution remain major risks.

Previous dated analysis and history

Q1 provided the strongest operating confirmation since Horizon 1 and a specific execution warning. AI Cloud revenue reached $200 million, operating ARR $1 billion, backlog $20.3 billion and contract pricing improved. H2–4 were not accepted on the original timetable and now move into January and potentially late February.

The conclusion remains Watch. At the $21.64 close, the unchanged three-year model implies $24–45 in the base case and $3–12 in the bear case. The price rise lowers attractiveness to 5.2/10; it does not change company strength.

Company strength

6.1/10, assessed November 7, 2026 under methodology 1.2.

DimensionScoreMain evidence
Competitive advantage7.0Vertically integrated sites, power, data centers and compute enabled H1 delivery; the H2–4 delay shows execution risk remains.
Demand durability8.2$200m Q1 AI Cloud revenue, $20.3bn backlog, two-thirds investment grade and higher three- and five-year contract pricing.
Management execution5.8H1 entered operation, but more than $4bn of ARR moved into Jan.–Feb. 2027.
Business economics5.386% AI Cloud gross margin and $66m adjusted EBITDA improve the evidence; normalized CFO was about −$116m and FCF per share is unproven.
Financial resilience4.0The 0.3 GW program is described as funded with conditions; debt, commitments, targeted funding and issuance remain large.

No new operating evidence was added after the Q1 review, so all five dimensions remain unchanged. The score reflects company evidence, not the share-price move. Confidence remains medium.

13

Thesis History / Updates

20. 11. 2026 · ChatGPT

The USD 14.41 November 19 regular close is 19.4% below the USD 17.87 price used in the latest completed assessment. The lower entry price raises attractiveness to 6.4/10 with unchanged scenarios; business strength remains 6.1/10 and financing and dilution remain major risks.

Previous dated analysis and history

Review date: November 7, 2026. No new operating evidence was added after the prior assessment. Company strength therefore remains 6.1/10 (7.0 / 8.2 / 5.8 / 5.3 / 4.0). Friday's $21.64 close was 20.02% above the prior session; with the same $24–45 base and $3–12 bear ranges, price attractiveness fell from 5.8 to 5.2/10. Price-score confidence remains low.

DateThesis developmentRating impact
Aug. 2026Microsoft accepted Horizon 1First-phase delivery reduced one execution risk
Aug. 27, 2026$4bn contracted ARR, $1bn operating ARR and FY27 capex of $25–30bnStronger commercial base and a much larger capital plan
Sep. 8, 2026Sweetwater conditionally entered ERCOT Batch ZeroInfrastructure progress; approvals and construction remain conditions
Oct. 16, 2026Methodology 1.2; strength 6.0, price unassessed at $30.39Integrated three-year project model was missing
Oct. 23, 2026Badajoz received a civil-works licenseDevelopment progress without disclosed project economics
Nov. 1, 2026Multi-year Perplexity relationshipCustomer-demand evidence without disclosed economics
Nov. 3, 2026$363m Wingspire financingPositive financing, limited relative to the full capital plan
Nov. 6, 2026Q1 FY27: $200m AI Cloud revenue, $20.3bn backlog and H2–4 delayStrength 6.0→6.1; price attractiveness 5.8 at $18.03
Nov. 7, 2026Friday close $21.64, +20.02% in one sessionStrength unchanged; price attractiveness 5.8→5.2 with unchanged scenarios

Principal sources: Q1 FY27 results, presentation, SEC filings, public call transcript and price history.

Was this analysis useful to you?

Have a suggestion about this analysis?

Suggestions are not displayed publicly. They help inform future updates. No name or email is required.

Sources and scenarios reflect the date stated in the analysis.Back to all companies