IREN (IREN) Q2 2026 earnings review
Revenue Falls as Bitcoin Mining Shrinks, But AI Cloud Ramp Accelerates
IREN's Q2 FY26 revenue dropped 23% sequentially to $184.7M as Bitcoin mining revenue declined 28% QoQ on lower hashrate (ASIC-to-GPU transition) and weaker BTC prices. The positive signal: AI Cloud revenue surged 137% QoQ to $17.3M. Adjusted EBITDA fell to $75.3M (41% margin), while GAAP net loss of $155.4M was driven by $219M in non-cash and non-recurring items tied to convertible note derivatives and a debt conversion expense. The real story is off-income-statement: $3.6B GPU financing secured for the Microsoft contract, $2.3B ARR now under contract, and a new 1.6GW Oklahoma site pushing secured power above 4.5GW.
🐂 Bull Case
AI Cloud revenue jumped from $7.3M to $17.3M in a single quarter — a 137% QoQ increase. With $2.3B ARR under contract and 140k GPUs targeted by end of CY2026, the revenue trajectory is about to fundamentally change. Microsoft contract revenues expected to begin flowing in Q3 FY26.
Securing $3.6B in GPU financing at <6% interest, combined with Microsoft's $1.9B prepayment, covers 95% of GPU-related CapEx for Horizons 1-4. This was the single biggest overhang on the story and is now largely resolved.
The new 1.6GW Oklahoma site takes secured power above 4.5GW. The $3.4B ARR target requires only ~10% of this capacity, implying massive optionality for future contract wins and multi-year growth runway.
🐻 Bear Case
Mining revenue fell 28% QoQ to $167.4M as hashrate was diverted to the GPU transition. This cash-flow engine is shrinking precisely when the AI business isn't yet generating meaningful revenue — AI Cloud is still only 9% of total.
Convertible notes nearly quadrupled to $3.7B from $964M in a single quarter. Total liabilities surged to $4.5B (vs $1.4B prior quarter). Stockholders' equity actually declined to $2.5B from $2.9B despite $1.6B in equity proceeds. Debt-to-equity is now 1.8x.
Adj. EBITDA of $75.3M is dwarfed by $851M in total investing outflows. The company is in a massive investment phase with no clear timeline to positive free cash flow. Any execution delay on the 140k GPU target directly pressures the capital structure.
⚖️ Verdict: ⚪
Neutral. Strategic milestones are impressive — GPU financing secured, $2.3B ARR contracted, AI Cloud revenue inflecting. But the financial picture is deteriorating in the transition: mining revenue shrinking, massive leverage building, profitability nowhere in sight. This is a bet on execution, not current earnings.
Key Themes
$3.6B GPU Financing Secured at <6% Interest
The GPU financing package from Goldman Sachs and JPMorgan is a delayed-draw term loan secured against the GPUs and Microsoft's contracted cash flows, amortizing over 5 years. Combined with Microsoft's $1.9B prepayment, it covers 95% of the $5.8B GPU-related CapEx for Horizons 1-4. Management highlighted the effective average cost of GPU financing is ~3% when the prepayment is factored in. CEO Roberts framed it: 'We essentially got the GPUs for next to nothing.' This materially de-risks the Microsoft contract and catalyzes further customer negotiations.
AI Cloud Revenue Accelerating from Low Base
AI Cloud revenue grew from $2.7M in Q2 FY25 to $17.3M in Q2 FY26 (+550% YoY) and 137% sequentially. Prince George GPU deployments are driving near-term revenue, with ~$0.4B ARR now under contract at that site — expected to increase to ~$0.5B in coming weeks. Microsoft contract revenues expected to commence Q3 FY26 and ramp progressively. The $3.4B ARR target by end of CY2026 implies a dramatic revenue step-up in H2 CY2026.
4.5GW+ Secured Power with New Oklahoma Campus
The new 1.6GW Oklahoma site (2,000 acres, Southwest Power Pool) adds jurisdictional diversity beyond ERCOT. Grid studies are complete, power ramp from 2028. With Sweetwater (2GW) likely in ERCOT batch 0 and Sweetwater 1 energizing in Q2, the full portfolio exceeds 4.5GW. Only ~10% is needed for the $3.4B ARR target. Kent Draper emphasized that 'time to data center' is the key constraint, and IREN's 810MW of operational data centers and 2,000+ person team give them an execution edge.
Bitcoin Mining Revenue Declining as Transition Accelerates
Bitcoin mining revenue fell 28% QoQ to $167.4M, driven by lower operating hashrate as British Columbia sites transition from ASICs to GPUs, against the backdrop of increasing global hashrate and lower average BTC prices. Mining hardware impairments of $31.8M were recorded (vs $16.3M in Q1). The mining business — previously described as generating $1B+ annualized revenue — is structurally shrinking. AI Cloud at $17.3M/quarter cannot yet compensate for even a fraction of the mining revenue loss.
Balance Sheet Leverage Surging — Convertible Notes Nearly 4x in One Quarter
Convertible notes payable surged from $964M to $3,685M following the $2.3B December 2025 issuance. Total liabilities jumped to $4.5B from $1.4B. Stockholders' equity declined to $2.5B from $2.9B despite $1.6B in equity proceeds, as the $155M net loss and $112M conversion inducement expense offset new capital. The $252M in capped call transaction costs adds further dilution protection complexity. While cash also surged to $3.3B, any delays in the AI revenue ramp would leave the company servicing significant debt on a declining mining business.
$219M Non-Cash Charges Create GAAP Earnings Volatility
GAAP net loss of $155.4M was heavily distorted by: unrealized losses on prepaid forwards and capped calls ($107.4M, vs $665M unrealized gains last quarter), a one-time $111.8M debt conversion inducement expense, $31.8M mining hardware impairments, and $58.2M in SBC. These items mask the $75.3M in Adj. EBITDA. The Q1-to-Q2 swing — from $384.6M net income to $(155.4)M net loss — was almost entirely driven by derivative mark-to-market reversals. GAAP earnings will remain highly volatile as long as the convertible note structure persists.
Cloud vs. Colocation — Management Firmly Chooses Cloud
Management made the clearest case yet for AI Cloud over colocation. Roberts noted the GPU financing at ~3% effective cost eliminates the capital intensity argument. The math: each 200MW generates ~$300M in colo revenue vs 'multiples of that in the billions' under cloud. Draper added that customers are offering longer tenures, prepayments, and showing strong demand for air-cooled capacity. Roberts added: 'If investors want bond-like exposures, they can buy colocation companies.' They remain open to compelling colo deals but see cloud as the superior model.
ERCOT Batch Process — Sweetwater Likely in Batch 0
IREN expects both Sweetwater 1 (1.4GW) and Sweetwater 2 (600MW) to be included in ERCOT's batch 0, meaning the full 2GW remains secured. Roberts was emphatic: 'That 2,000 megawatts is secure. The signed interconnection agreement was signed in 2023.' Sweetwater 1 substation energization remains on track for April 2026. The batch process may benefit IREN by exposing competitors with less-certain power — Draper noted it will 'uncover which megawatts are real and which are not.'
Air-Cooled GPU Demand Emerging as Near-Term Opportunity
Hyperscalers are increasingly interested in air-cooled GPU deployments (B200, B300) for faster deployment timelines. This aligns directly with IREN's 810MW of operational air-cooled data centers, which can host GPUs with minimal capital upgrades. Draper highlighted this as a factor in contract negotiations, driving better terms including prepayments and longer tenures. The air-cooled opportunity provides a near-term revenue bridge while liquid-cooled Horizon builds progress at Childress.
Execution Risk: 140K GPU Deployment in Under 12 Months
Deploying 140,000 GPUs by end of CY2026 requires flawless execution across multiple sites simultaneously: Microsoft Horizons 1-4 at Childress (liquid-cooled GB300s), plus ~63,000 GPUs at Mackenzie, Canal Flats, and Prince George (air-cooled B200/B300). Construction, GPU deliveries, and commissioning must all hit milestones. Contractual penalties with Microsoft exist for missed timelines. Draper noted 'we expect to deliver 140,000 GPUs by the end of 2026' — 'expect' rather than 'will' is worth noting.
Sweetwater Remains Uncontracted Despite April Energization
Sweetwater 1's 1.4GW substation energizes in April 2026, but no customer contract has been announced. Management has been in dialogue with parties for '12, 18, 24 months.' Roberts emphasized patience: 'There's a real opportunity cost of signing a bad deal.' However, with 1.4GW of power about to come online with no contracted tenant, the capital is effectively idle. The company cannot begin generating returns on this massive site until commercial terms are finalized.
Other KPIs
Decelerating QoQ. Down 18% from $91.7M in Q1. The decline was driven by lower Bitcoin mining revenue, partially offset by lower payroll tax accruals. However, Adj. EBITDA margin improved sequentially from 38% to 41% — indicating better cost control on lower revenue. YoY, Adj. EBITDA grew 20% from $62.6M (Q2 FY25), though margin compressed from 52% to 41%, reflecting the transition's cost structure.
Decelerating. Down 50% QoQ from $142.4M. Lower mining profitability and a $180M adverse swing in tax-related liabilities drove the decline. Note: Bitcoin sale proceeds are classified as investing activities (restated methodology from FY25), so OCF reflects only the operational cash generation. Investing outflows of $851M (PP&E $540M, hardware $179M, intangibles $108M) far exceeded OCF, funded by $3.0B in net financing inflows.
Accelerating on a YoY basis from the prior quarters' trajectory (Q3 FY25 was +173% YoY, but off a very low base). Sequentially, total revenue fell 23%, entirely due to the 28% decline in Bitcoin mining revenue. The revenue mix is shifting: AI Cloud was 9.4% of total (vs 3.0% in Q1 and 2.2% in Q2 FY25). This mix shift will accelerate dramatically once Microsoft revenues commence in Q3 FY26.
Down 18% QoQ from $80.7M, proportional to the revenue decline. Bitcoin mining cost of revenue fell to $63.4M from $80.0M (on lower hashrate and power consumption). AI Cloud cost of revenue was $2.4M on $17.3M revenue, implying ~86% gross margin for the cloud segment — down slightly from 90%+ in Q1, likely reflecting the ramp-up phase of new GPU deployments at Prince George.
Guidance
Stable. Reiterated from Q1. Comprised of ~$1.94B average annual revenue from the Microsoft contract plus estimated ~$1.5B from ~63,000 GPUs at BC sites. Currently $2.3B ARR under contract ($1.94B Microsoft + ~$0.4B Prince George). Management expects Prince George ARR to reach ~$0.5B in coming weeks. The remaining ~$0.9-1.1B gap must close through Mackenzie/Canal Flats contracts and full Microsoft ramp. Draper stated demand is 'not the limiting factor' — the focus is selecting the right partners.
Stable. Includes ~76,000 GPUs for Microsoft (Horizons 1-4, GB300s at Childress) and ~63,000 GPUs at BC sites (B200/B300, air-cooled). Prince George fit-outs for B200/B300 complete, awaiting remaining GPU deliveries. Mackenzie and Canal Flats being prepared with the 'exact same playbook' as Prince George — ASICs out, GPUs in. The ~40,000 incremental BC GPUs (~$1B ARR) will require additional financing via GPU leasing and dedicated facilities.
Stable. Construction well advanced on both on-site and utility substations. No customer announced yet. Management expects the ERCOT batch process to be 'helpful' by clarifying which competitors have real power commitments. No specific data center construction timeline given for the site — the pace will be aligned with customer discussions and capital availability.
New. 2,000-acre site in Southwest Power Pool. Grid studies complete, land and full 1.6GW of power secured. Provides diversity beyond ERCOT. Development items (master planning, local permitting) progressing. No customer discussions disclosed for this site yet — positioned as pipeline for post-2027 growth.
Accelerating. Comprised of customer prepayments, convertible notes ($2.3B Dec issuance), GPU leasing arrangements, and $3.6B GPU financing. Additional financing workstreams include data center financing for Horizons 1-4 when operational, and select corporate-level facilities. CFO Lewis emphasized maintaining 'an appropriate balance between debt and equity' for balance sheet resilience.
Key Questions
Revenue Trough and Bridge to AI Cloud
Bitcoin mining revenue fell 28% QoQ while AI Cloud is $17M/quarter. With Microsoft revenues just starting in Q3 and 140k GPUs not fully deployed until late CY2026, what does the revenue profile look like over the next 2-3 quarters? Is there a trough quarter ahead?
Free Cash Flow Timeline
Q2 investing outflows were $851M against $72M in operating cash flow. When does management expect to reach free cash flow positive? What is the peak CapEx quarter, and what does the spending curve look like through CY2026?
Convertible Note Dilution and Maturity Profile
Convertible notes nearly quadrupled to $3.7B. The Q2 induced conversion resulted in $1.6B in equity issuance and a $112M inducement expense. What are the remaining conversion terms, maturity schedule, and potential dilution? Should shareholders expect more induced conversions?
Sweetwater Commercial Timeline
Sweetwater 1 energizes in April with no customer announced. Management has been in talks for 12-24 months. What is the realistic timeline for a signed contract? Is the initial deployment likely to be cloud or colocation?
British Columbia Contract Concentration
Of the ~$1.5B BC ARR target, ~$0.4B is contracted with negotiations 'supporting >$0.5B ARR' ongoing. How many customers make up the current and pipeline contracts? Is there concentration risk, and what happens if negotiations for the remaining ~$1B ARR take longer than expected?