IREN posts $707m FY26 revenue as AI Cloud surges eightfold

IREN's AI Cloud Services revenue grew nearly 8x to $129m in FY26, with $4bn contracted ARR targeted by year-end as it pivots from

IREN posts $707m FY26 revenue as AI Cloud surges eightfold

IREN Limited (NASDAQ: IREN) has reported total revenue of $707m for the fiscal year ended 30 June 2026, up from $501m a year earlier, as the company accelerates a wholesale shift from Bitcoin mining to AI Cloud Services. AI Cloud revenue climbed to $128.8m, roughly eight times the $16.4m recorded in FY25, while Bitcoin mining contributed $578.2m, a segment the company is actively running down as sites are converted.

The headline financials carry a significant accounting charge: IREN recorded a net loss of $702.6m for the full year, driven primarily by $638.8m in non-cash impairments linked to the decommissioning of Bitcoin mining hardware. Adjusted EBITDA, which strips out those charges alongside stock-based compensation and other non-recurring items, came in at $245.7m, down from $269.7m in FY25. The adjusted EBITDA margin contracted to 35% from 54%, reflecting a deliberate increase in headcount (up nearly threefold in FY26) and platform investment ahead of the AI Cloud revenue ramp.

Contracted ARR and customer expansion

The more consequential metrics for investors are forward-looking. The company reports $1bn in operating annualised run-rate revenue as of 26 August 2026, and says contracted ARR for 2026 capacity stands at $4bn, with its current capacity largely sold out. IREN did not name the frontier AI lab with which it signed a new multi-year contract, but its existing disclosed customer roster includes Cohere, Perplexity, Figure AI and Higgsfield AI alongside hyperscaler exposure via the Microsoft relationship.

On the infrastructure side, IREN delivered Horizon 1, the first of four 50MW (IT) liquid-cooled deployments at its Childress, Texas site, to Microsoft during August. Horizon 2 is commissioning and Horizons 3 and 4 are in late-stage construction, targeting Q4 2026 delivery. The company targets cumulative deployed capacity of approximately 0.3GW (IT) by end of 2026 and 0.8GW by end of 2027. Recent three-year contracts have been priced at above $20m revenue per megawatt of IT load, with active discussions ongoing at around $25m per megawatt, implying an estimated two-year payback on GPU capital expenditure.

GPU financing has been structured at scale: a $3.6bn investment-grade facility at 6.0% supports the Microsoft contract, covering 96% of associated GPU capex together with customer prepayments. A further $2.8bn in financing, including a $2.4bn facility led by Blue Owl and PIMCO at a 9.0% fixed rate, covers the Mackenzie air-cooled expansion. Customer prepayments across recent deals represent 45-55% of GPU capex, a structural feature IREN says provides funding in excess of GPU capital requirements.

Market context and competitive positioning

IREN sits in a rapidly expanding category of vertically integrated GPU cloud operators, competing with CoreWeave, Lambda Labs and a clutch of well-capitalised peers for long-term AI compute contracts. The distinguishing characteristic IREN emphasises is its ownership of power and land in renewable-rich regions: sites in Texas, Oklahoma, British Columbia, Australia and, following the acquisition of Spanish firm Nostrum Group, Spain. The Mirantis acquisition adds managed Kubernetes and software capabilities, broadening the platform beyond raw GPU rental toward managed services, where margins are typically higher.

The structural supply-demand imbalance in GPU compute capacity has enabled IREN and its peers to push contract pricing upward and demand larger customer prepayments, effectively shifting financing risk to buyers. That dynamic may moderate as hyperscaler data centre programmes and new entrants add capacity through 2027 and 2028. IREN acknowledged in its results that discussions for 2027 capacity are at a late stage and 2028 financing and customer conversations are running in parallel, suggesting the contracting pipeline remains active but is not yet locked.

Co-founder and co-CEO Daniel Roberts said the founding thesis, that AI infrastructure cannot scale as quickly as digital demand, "became tangible" in FY26 and described compute scarcity as a deepening structural shortage. The company's next milestones are Horizons 3 and 4 delivery in Q4 2026, the 2027 capacity contracting cycle, and demonstrating revenue recognition that closes the gap between the $4bn contracted ARR figure and reported GAAP revenue.