Bernstein maintains Outperform on IREN as its AI-cloud buildout accelerates

Overview

Bernstein updates its IREN model after recent operational and financing developments. It maintains Outperform and a $100 price target, arguing that contracted cloud demand, expanding capacity and financing execution support IREN's shift into an AI-cloud platform.

Core views

Bernstein argues that IREN has moved from a legacy Bitcoin-mining business toward an AI-cloud platform. Operating cloud ARR had reached $1 billion as of August, including $0.5 billion from Horizon 1 delivered to Microsoft and $0.5 billion from enterprise and AI-lab contracts. The report says 2026E capacity is fully sold at $4 billion of ARR, and expects Bitcoin mining to be wound down by CY26E as the Childress site is converted to cloud deployments. The revenue case rests on contract visibility and a blended customer strategy. Of $4.7 billion in contracted ARR, Microsoft and NVIDIA account for $2.6 billion under five-year contracts with annual revenue yields of $10-15 million per IT MW. Enterprises and AI labs account for about $2 billion, with recent deals carrying annual revenue yields of $20 million per IT MW and average terms of roughly three years. Bernstein expects Horizon 2-4 and Mackenzie deliveries in Q4 2026 to allow the contracted ARR to be fully realized by Q2 CY27E, after allowing for a two- to three-quarter lag between commissioning, testing, customer acceptance and utilization. Microsoft represents less than 50% of contracted ARR, which the report views as evidence of greater customer diversification. Capacity expansion is the other central driver. Bernstein expects IREN to reach about 2 GW of gross cloud power by CY29E, with Childress and British Columbia transitioned fully into cloud by CY27E and Sweetwater 1 scaled to about a 1 GW cloud campus by CY29E. The report cites a 5.8 GW power portfolio, including 350 MW at Childress contracted to Microsoft, 130 MW in British Columbia, construction and retrofit activity at Childress, initial 300 MW liquid-cooled construction at Sweetwater 1, and further planning in Oklahoma, Australia and Spain. Its forecasts show AI-cloud revenue rising from $0.6 billion in CY26E to $5.0 billion in CY27E, $10.9 billion in CY28E, $15.5 billion in CY29E and $16.8 billion in CY30E; adjusted cloud EBITDA rises from $0.2 billion in CY26E to $3.4 billion, $8.7 billion, $12.4 billion and $13.5 billion over those respective later years. Bernstein characterizes the CY30E outcome as roughly $17 billion of cloud revenue and $13 billion of adjusted EBITDA, implying about an 80% EBITDA margin. Financing is a key assumption behind that buildout. Bernstein estimates incremental cloud capex of $65 billion through CY30E, or about $50 million per IT MW, and notes company guidance for $25-30 billion of FY27 capex. IREN secured $19 billion of funding over the prior 12 months, including $16 billion across customer prepayments, GPU financing and convertibles, plus $3 billion through equity ATM; around $14 billion was cash or undrawn capacity. The report expects another $8 billion from new GPU financing and prepayments, with operating cash flow, data-center financing and other corporate sources supplying the remaining $3-8 billion. It highlights $3.65 billion of Microsoft-backed GPU financing at about 6%, $2.4 billion of non-investment-grade financing at 9%, customer prepayments covering 45-55% of GPU capex, and unencumbered data-center assets as funding flexibility. Bernstein changes its valuation approach from a prior sum-of-the-parts framework to 8.5x EV/EBITDA for 2028E as IREN progresses toward a neocloud model. Using $8.7 billion of 2028E adjusted EBITDA, a 9x one-year-forward EBITDA presentation yields a $73.9 billion target enterprise value; after $27.0 billion of net debt, including deferred revenue, the report derives a $46.8 billion target market capitalization and a $100 target price on 461 million shares. Bernstein therefore reiterates Outperform.

Analysis framework

Bernstein updates forecasts using IREN's announced capacity delivery, contracted ARR, customer mix, site-development plans and funding arrangements. It models the timing lag from commissioned capacity to recognized revenue, projects cloud revenue and EBITDA through CY30E, and applies an EV/EBITDA valuation multiple to 2028E EBITDA after deducting net debt.

Methodology notes

• Valuation methodsEV/EBITDA valuation

  EV/EBITDA valuation

  Bernstein values IREN using an 8.5x multiple of projected 2028E EBITDA, then deducts net debt to calculate equity value and the $100 target price.

• Industry AnalysisVolume-price decomposition

  Capacity, contracted ARR and revenue-per-IT-MW analysis

  The report links cloud capacity buildout and customer contract yields to future ARR, revenue and EBITDA, while allowing for a two- to three-quarter revenue-recognition lag.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• IREN Ltd (IREN)

  Primary covered company; Bernstein expects its AI-cloud capacity expansion and contracted customer base to drive revenue and EBITDA growth.

  Strengths

  $1 billion of operating cloud ARR, $4.7 billion of contracted ARR, a diversified customer base, unencumbered data-center assets and substantial committed funding.

  Weaknesses

  The transition requires substantial capex and the Bitcoin-mining business is expected to wind down by CY26E.

  Comparison

  Bernstein replaces its previous sum-of-the-parts valuation approach with a neocloud-oriented EV/EBITDA framework.

  Risks

  Financing access, timely data-center development and securing additional customer contracts are explicit risks.

Key data

• Operating cloud ARR$1 billionAs of August; includes $0.5 billion from Horizon 1 delivered to Microsoft and $0.5 billion from enterprise and AI-lab contracts.
• Contracted ARR$4.7 billionExpected to be fully realized by Q2 CY27E after Horizon 2-4 and Mackenzie deliveries.
• Projected cloud capacity~2 GWBernstein's projected gross cloud power by CY29E.
• CY30E cloud revenue$16.8 billionForecast AI-cloud revenue; described in the report as roughly $17 billion.
• CY30E adjusted EBITDA$13.5 billionForecast AI-cloud adjusted EBITDA, implying approximately 80% EBITDA margin.
• Incremental cloud capex$65 billionEstimated requirement through CY30E, or roughly $50 million per IT MW.
• Target price$100Based on 8.5x 2028E EV/EBITDA.

Impact & implications

Bernstein's thesis is that IREN's contracted cloud demand, diversified customer mix and financing access can support a rapid conversion from Bitcoin mining to AI cloud. Its valuation depends on the company delivering capacity, realizing contracted ARR on schedule and funding a large capital program.

Risks

• IREN's AI-cloud business is capital intensive and depends on access to flexible, low-cost financing.
• Long-term capacity expansion depends on timely land and equipment procurement and successful data-center development; supply-chain or execution delays could impair scaling.
• The on-demand cloud business depends on winning long- and short-term contracts from multiple customers.

What to watch

• Delivery of Horizon 2-4 and the Mackenzie site in Q4 2026, and realization of contracted ARR by Q2 CY27E.
• Progress in financing the $25-30 billion FY27 capex program, including GPU financing and customer prepayments.
• Cloud capacity conversion at Childress and British Columbia and Sweetwater 1's ramp toward a roughly 1 GW cloud campus.
• Further customer contracts for 2027 and 2028 capacity.