Bernstein reiterates Outperform on IREN stock after Nvidia deal

Bernstein SocGen reiterated an Outperform rating and $100 price target on IREN, citing its partnership with Nvidia and a pipeline of roughly 5 GW of secured power across campuses. Nvidia is starting with a 60 MW AI cloud contract, while IREN retains 450 MW at Childress and remains on track to deliver 200 MW of Horizon data centers for Microsoft. However, the stock is already up 725% over the past year and the latest Q3 FY2026 results missed badly, with EPS of -$0.30 versus -$0.2161 expected and revenue of $144.8M versus $219.87M expected.

Analysis

This is less a pure IREN story than a validation event for the AI infrastructure supply chain. The market should treat NVDA as the economic center of gravity: when it starts underwriting multi-year capacity with select operators, it effectively compresses perceived execution risk for the entire buildout stack and widens the moat around preferred infrastructure partners. That should lift sentiment for GPU-adjacent capex beneficiaries, but the bigger second-order effect is that scarce power + permitted land + retrofitted cooling becomes the real bottleneck, not chips.

For MSFT, the implication is more subtle: outsourcing incremental capacity to a constrained partner can improve time-to-deploy, but it also signals that hyperscaler self-build is not enough to satisfy near-term demand. That creates a tactical bid for outsourced AI campuses and electrical contractors, while pressuring less differentiated colo names that lack secured power or GPU allocation. The risk is that the equity market extrapolates the partnership economics too aggressively before utilization, financing, and margin conversion are proven.

The main tail risk is execution dilution over the next 6-18 months. If growth is funded with heavy equity issuance before cash flow inflects, the narrative can shift from "strategic scarcity asset" to "expensive capital sink," especially if AI demand pauses or GPU cycles move faster than site commissioning. The recent earnings miss also matters because it gives skeptics an anchor to argue that headline partnerships are ahead of operating reality.

Consensus is likely underestimating how much of the upside is already tied to financing structure rather than operating performance. If NVDA’s implied commitment becomes a template, the winners are the few operators with power already locked; everyone else may face a rising cost of capital and longer payback periods. The move is therefore bullish for the category, but not uniformly so—stock selection should favor balance-sheet strength and contracted megawatts over promotional growth stories.

Key Decisions for Investors

• Long NVDA vs. basket of unprofitable AI infrastructure names over 3-6 months; thesis is that NVDA captures the most durable economic rent while preserving optionality across the buildout.
• Avoid chasing IREN after the partnership headline; if initiating, use a small starter only on a post-event pullback and size it as a financing-risk trade, not a core long.
• Pair trade: long MSFT / short weaker AI campus or colo peers with limited power visibility for 3-9 months; thesis is that hyperscale demand persists, but scarce capacity accrues to the best-capitalized operators.
• For higher-conviction tactical exposure, buy NVDA call spreads 2-4 months out to express continued AI capex upside with defined downside; risk/reward is favorable if the market keeps rewarding strategic supplier lock-in.
• Watch for any announced equity issuance or revised capex timeline at IREN over the next 1-2 quarters; that is the likely catalyst for multiple compression and a good opportunity to fade the rally.