JPMorgan double upgraded a data center stock that struck a deal with Nvidia earlier this year

JPMorgan double upgraded IREN to overweight from underweight and lifted its price target to $65 from $46, implying 48% upside from Friday's close. The bank views IREN's more than $3B, five-year Nvidia AI-cloud agreement and improving neocloud pricing of $15-$20+ per watt as support for returns on its planned $25B-$30B FY2027 capex program. IREN shares have fallen nearly 27% over the past three months amid concerns over AI infrastructure spending, but JPMorgan expects higher-priced contracts for its roughly 0.5 GW 2027 expansion to be materially accretive.

Analysis

IREN’s valuation hinge is shifting from installed GPU capacity to the spread between contracted revenue per MW and the cost of power, GPUs, and construction financing. If higher-rate contracts are independently confirmed, incremental capacity should drive disproportionate EBITDA growth because site power infrastructure is largely fixed; conversely, a delay in energization or customer acceptance turns the same buildout into a negative free-cash-flow and dilution event. The relevant comparison is less BTC miners and more capital-intensive neocloud peers such as CRWV and NBIS, where contract quality, counterparty concentration, and debt terms determine the multiple.

The near-term upgrade is unlikely by itself to resolve the market’s core concern: funding a multi-year buildout before cash receipts fully ramp. Over the next 1-3 months, disclosures on signed capacity, upfront deposits, GPU procurement commitments, project-finance pricing, and power-delivery milestones matter more than another target-price increase. A widening in IREN’s credit spreads, equity issuance, or capex guidance rising faster than contracted MW would falsify the improving-return thesis; these are early indicators that nominal contract values are not translating to equity value.

Consensus optimism creates an asymmetric setup: verified contracts priced above the market’s underwriting assumptions could rerate IREN toward the premium AI-infrastructure cohort over 6-18 months, but the stock remains exposed to a sharp de-rating if AI cloud pricing normalizes as hyperscalers internalize supply. The non-obvious loser is lower-quality GPU hosting capacity without secured low-cost power, including parts of the former mining universe; IREN’s expansion would increase competition for transformers, electrical gear, and GPU financing, potentially compressing weaker operators’ returns. NVDA is a modest indirect beneficiary through accelerated deployment, but its earnings sensitivity is immaterial relative to IREN’s financing and execution sensitivity.

Key Decisions for Investors

• Maintain IREN as a tactical long only after confirmation of incremental contracted MW, customer deposits, and a funding plan that does not require near-term common-equity issuance. Use a 1-3 month catalyst window; target a rerating toward AI-hosting peers, but cut exposure if capex commitments increase without matching signed revenue or if project financing costs materially exceed management assumptions.
• Express the quality spread through long IREN / short a higher-leverage, less-contracted digital-infrastructure or mining proxy such as WULF, sized beta-neutral. Hold 3-6 months; the thesis is that contracted AI capacity receives a premium while speculative capacity remains capital-constrained. Exit if the long-term contract pipeline fails to convert into disclosed revenue backlog.
• Do not chase NVDA on this development. Instead, monitor IREN deployment progress as a marginal read-through for GPU demand; NVDA upside requires broad evidence that neocloud financing is expanding rather than merely reallocating demand from CRWV, NBIS, and hyperscaler channels.
• Set an alert around the next earnings release for three underwriting variables: contracted versus energized capacity, revenue per MW, and net-debt/funding commitments. A positive trade requires all three to improve together; strong contract headlines paired with deteriorating leverage should be treated as a sell-the-news signal.