Got $1,000? 2 Growth Stocks Securing Every AI Data Center.

Iren is scaling AI data-center capacity following a five-year, $9.7 billion Microsoft contract for 200MW, equivalent to roughly $1.94 billion of annual recurring revenue, and targets $4 billion in ARR exiting 2026. Its 5.8GW development portfolio includes Sweetwater, where the first 300MW is targeted for Q4 2027; at current pricing above $20 million per MW annually, that capacity could generate more than $6 billion of annual revenue. Micron reported fiscal Q4 2026 revenue of $54.2 billion, up 479% year over year and 30.8% sequentially, beat guidance, and indicated fiscal 2027 will be even stronger amid AI-driven memory demand and constrained supply.

Analysis

IREN is transitioning from a commodity-like power/hosting valuation toward a contracted AI-infrastructure model, but the equity remains highly sensitive to execution rather than simply AI demand. The key underwriting variable is not headline GW capacity; it is the spread between contracted revenue per MW and all-in delivered capital cost, including grid interconnection, GPUs/networking where applicable, financing, and construction overruns. Over the next 1-3 months, confirmation of customer prepayments, binding take-or-pay terms, and non-recourse project financing would justify multiple expansion; any reliance on corporate equity issuance would likely overwhelm the positive demand narrative.

The second-order beneficiary of incremental AI-site commissioning is MSFT, which can monetize scarce compute capacity without carrying all development risk, while electrical-equipment vendors such as VRT, ETN, PWR and GEV may capture more reliable economics than merchant data-center developers. Conversely, IREN competes for power, transformers, construction labor and debt capacity with CoreWeave and hyperscaler-backed projects; those bottlenecks can raise capex faster than quoted rack economics. The principal 6-18 month risk is that AI customers secure capacity aggressively but defer utilization, leaving developers with high fixed charges and contract renegotiation risk.

MU's apparent valuation discount should not be treated as structurally durable: memory earnings are cyclical, and a low forward multiple can reflect a peak HBM/DRAM pricing environment rather than mispricing. The more important catalyst is whether HBM supply remains allocation-constrained into the next contract-reset cycle, allowing MU to convert revenue growth into sustained gross-margin strength despite capacity additions from Samsung and SK Hynix. Contrarian view: AI capex beneficiaries with contracted power may be more vulnerable to financing conditions than semiconductor suppliers; a rise in long rates or tighter project-finance spreads can impair IREN's equity value even if AI demand remains intact.

Key Decisions for Investors

• Maintain a tactical long MU position into the next earnings and memory-contract pricing update, preferably funded by a short SOXX basket or a partial NVDA hedge. Target a 10-15% upside on continued HBM margin expansion over 1-3 months; exit if management signals weaker next-quarter DRAM/NAND pricing or gross-margin guidance misses consensus by more than 300bp.
• Treat IREN as an event-driven watchlist name rather than a core long until binding customer terms, capital commitments, and project-level return metrics are disclosed. Initiate only after evidence that incremental capacity is funded without material equity dilution; use a 15-20% downside stop or reduce if financing spreads widen materially.
• For cleaner AI-infrastructure exposure, favor a pair of long VRT or ETN versus short IREN on a relative basis over 6-12 months. This expresses continued data-center buildout while reducing exposure to IREN's power-delivery, construction, and refinancing risk.
• Set alerts for AI data-center lease cancellations, utility interconnection delays, and HBM capacity-expansion announcements from Samsung or SK Hynix. These are the highest-probability thesis breakers for IREN and MU, respectively.