Neocloud Boom: Comparing Nebius, CoreWeave, & IREN

AI hyperscalers Microsoft, Alphabet, Meta, Amazon and Oracle are expected to spend about $700 billion on AI infrastructure capex in 2026 and $729 billion in 2027, supporting demand for GPU-focused neocloud providers. Nebius is positioned as the strongest of Nebius, IREN and CoreWeave, with estimated 2026 revenue growth of 530.09%, $8.04 billion of cash, and a 2.3% debt rate, versus 346% growth and $5.9 billion cash for IREN and 151% growth, $5.52 billion cash, and an 8.3% debt rate for CoreWeave. The sector remains capital-intensive and likely unprofitable in the near term, with debt costs and equity dilution key investment risks.

Analysis

The investable distinction within GPU lessors is not headline revenue growth but the conversion of contracted capacity into unlevered cash flow after GPU depreciation, power, and financing costs. NBIS appears structurally advantaged if customer prepayments and lower-cost funding genuinely reduce construction-period cash burn; that can support a higher EV/revenue multiple than CRWV even before profitability. The key second-order beneficiary is NVDA: third-party clouds broaden the buyer base and reduce hyperscaler bargaining power, but they also create a future secondary-GPU supply overhang if lease demand disappoints.

CRWV is the weakest balance-sheet expression of the theme. A high fixed funding cost makes its equity disproportionately sensitive to utilization, GPU pricing, and refinancing spreads: a modest shortfall in realized revenue per GPU-hour can erase equity value because interest and depreciation do not flex. Conversely, IREN's lower cash interest is not automatically superior economics if convert dilution becomes the recurring source of expansion capital; its valuation should be assessed on fully diluted FCF per deployed MW, not nominal debt cost.

Over the next 1-3 months, the catalyst path is contract disclosure quality—duration, termination rights, customer concentration, prepayment, and whether revenue is take-or-pay—as well as quarterly capex-to-backlog conversion. Over 6-18 months, power availability is the binding constraint, favoring operators with secured energized capacity rather than merely GPU purchase commitments. The consensus is likely extrapolating a supply shortage without pricing the eventual risk that hyperscalers internalize workloads once utilization is proven; that would compress neocloud rental rates before hardware demand necessarily weakens.

The article is promotional and its growth estimates are insufficient to establish equity value. Falsify a NBIS-over-CRWV thesis if NBIS's incremental contracted backlog lacks prepayment or if its cash capex per MW rises materially, while CRWV secures durable refinancing at a substantially lower spread and demonstrates stable realized GPU utilization through a full quarter.

Key Decisions for Investors

• Initiate a 3-6 month pair: long NBIS / short CRWV, sized beta-neutral. The trade expresses lower financing and execution risk at NBIS against CRWV's operating leverage; target a 15-25% relative return, with exit/review if CRWV reports lower funding costs or NBIS's net cash burn exceeds management's buildout plan.
• Maintain NVDA as the higher-quality AI-infrastructure exposure rather than adding broad neocloud beta. Add on sector-led pullbacks only; the relevant confirmation is sustained data-center revenue guidance and supply discipline, while a material reduction in hyperscaler capex intentions is the stop signal.
• Do not underwrite IREN as a standalone long until fully diluted share count, conversion terms, contracted power capacity, and realized GPU utilization are reconciled. Establish an event-driven watch ahead of earnings; favor a long only if incremental capacity is funded without materially worsening per-share economics.
• Monitor GPU rental-rate data, power interconnection timelines, and high-yield spreads weekly. A 10%+ decline in realized rental rates or a meaningful widening in AI-infrastructure credit spreads would warrant reducing the NBIS/CRWV gross exposure, because both equities remain long-duration capex vehicles.