Is It Too Late to Buy IREN Stock After Its Monster Run?

IREN has secured $4.0 billion of contracted annualized revenue run rate for current capacity, including a five-year, $9.7 billion Microsoft GPU-cloud contract expected to generate $1.94 billion of ARR at an 85% EBITDA margin once fully commissioned. The company owns roughly 5GW of grid-connected power capacity and has more than 150,000 GPUs installed or on order, positioning it to capitalize on AI compute shortages. However, IREN plans $25 billion-$30 billion of fiscal 2027 capex, and its shares remain 40% below their 52-week high amid investor concern over AI infrastructure spending, leverage, and GPU pricing durability.

Analysis

IREN’s equity is no longer primarily a Bitcoin-mining valuation; it is underwriting a leveraged conversion of contracted power into depreciating GPU capacity. The key mismatch is that customer ARR becomes visible only as clusters are commissioned, while cash capex, financing costs, and hardware-obsolescence risk arrive first. An 85% EBITDA-margin target excludes the economic burden of rapid GPU refresh cycles, interest expense, and potential dilution, so the relevant debate is free-cash-flow yield on installed compute rather than headline contracted revenue.

The Microsoft relationship materially improves customer-credit quality and project-finance access, but it is immaterial to MSFT’s earnings and gives IREN limited bargaining power at renewal. IREN’s spending plan creates a near-term beneficiary set in NVDA and data-center equipment/power suppliers, while competing GPU-cloud operators such as CRWV face a mixed outcome: higher contract pricing validates scarcity, but IREN adds a well-capitalized source of supply into the 2027-28 window. The second-order risk is that hyperscalers shift from emergency external capacity purchases to owned infrastructure before IREN’s new sites reach full utilization.

Over the next 1-3 months, financing announcements, GPU delivery schedules, and evidence that customer prepayments cover a meaningful share of build costs will drive the stock more than incremental AI-demand headlines. Over 6-18 months, the thesis requires realized commissioning, sustained utilization, and hardware paybacks near management’s estimate; any deterioration in lease rates or utilization would sharply compress equity value because fixed capital costs are so large. Consensus appears to treat grid interconnection as equivalent to monetizable capacity; the missing constraint is transformer availability, cooling-water/build execution, and the cost of converting nominal power rights into live GPU revenue.

Key Decisions for Investors

• Do not chase IREN on contract headlines; maintain a watch position only until the company discloses committed financing, all-in cost of capital, and commissioned GPU capacity versus plan. Upgrade only if customer prepayments plus non-recourse/project financing cover the majority of the next 12 months of capex without a material equity raise.
• For a tactical AI-infrastructure expression, favor long NVDA versus short IREN over the next 3-6 months. NVDA captures equipment demand at shipment, while IREN bears construction, utilization, depreciation, and refinancing risk; close the pair if IREN demonstrates sustained positive free cash flow after interest and maintenance/refresh capex.
• If IREN rallies materially ahead of its next financing or earnings update, consider a defined-risk bearish structure such as 3-6 month put spreads rather than an outright short. The asymmetric downside catalyst is a dilutive capital raise or delayed cluster commissioning; invalidate if contracted ARR converts to revenue on schedule and capex guidance is fully financed on acceptable terms.
• Treat MSFT as a customer-credit beneficiary rather than a material earnings trade. Monitor whether MSFT expands third-party GPU-cloud commitments or instead redirects workloads to internal Azure capacity; the latter would be a negative read-through for IREN and external GPU-cloud peers including CRWV.