Nebius Stock Up 254% in a Year: Should You Buy, Hold or Sell?

Nebius reported Q2 2026 revenue growth of 454% year over year to $582 million, ARR of $3 billion, and adjusted EBITDA margin expansion to 41% from 32% in the prior quarter. The AI-cloud provider reaffirmed 2026 revenue guidance of $3.0-$3.4 billion and ARR guidance of $7-$9 billion, supported by four AI cloud contracts each worth more than $1 billion. However, $5.7 billion of Q2 capex and expected 2026 capex of $20-$25 billion underscore substantial funding needs and execution risk; after a 253.5% one-year share-price gain, the article advises existing holders to hold while new investors await a better entry point.

Analysis

NBIS should be valued less like software and more like a leveraged power-and-GPU conversion vehicle: incremental contracted capacity is valuable only if it reaches billable utilization before hardware economics deteriorate. Price/book is a weak comparison metric because the balance sheet will be dominated by rapidly depreciating accelerators and project assets; EV/forward EBITDA, capex-to-revenue conversion, net debt/contracted ARR, and customer concentration are the critical diligence items. The reported deal pipeline should not be capitalized at face value until disclosure distinguishes binding take-or-pay commitments, cancellation rights, prepayments, and commencement dates.

Near term, the share price can remain momentum-driven, but the next 1-3 month catalyst path is binary around financing structure, commissioning milestones, and whether utilization stays high as new clusters come online. Equity issuance, GPU-backed debt, or weaker-than-expected free-cash-flow conversion would compress the premium multiple even if revenue continues to grow. A delay at any major power site also creates an adverse double effect: revenue slips while depreciation, interest, and fixed operating costs begin accruing.

The less obvious read-through is favorable for NVDA so long as NBIS converts expansion plans into delivered GPU orders, but unfavorable for CRWV if NBIS demonstrates that an asset-light model can source power and customers without carrying as much balance-sheet risk. Contrarily, the market may be underpricing the scarcity value of commissioned power rather than raw GPUs; if NBIS can sustain auction-clearing economics through 2027, its platform and power-access moat could justify a higher multiple than conventional hosting peers. That thesis is falsified by falling realized pricing, utilization below management targets, or a material upward revision to funding needs.

Key Decisions for Investors

• Do not add outright NBIS after the momentum move; establish a watch-level entry only after the next financing disclosure and commissioning update. Require evidence that incremental capacity is generating revenue within the expected ramp window and that funding does not materially dilute existing holders.
• For a relative-value expression over 3-6 months, consider long NBIS / short CRWV in equal beta-adjusted dollars only if NBIS discloses binding customer commitments and CRWV's leverage or pricing metrics worsen. Target 15-20% relative upside; exit if NBIS announces substantial equity financing or CRWV secures comparable long-duration contracted capacity.
• Maintain NVDA as the cleaner AI-infrastructure exposure rather than using NBIS as a proxy for GPU demand. NBIS execution supports accelerator demand, but a project delay would likely affect NBIS far more than NVDA; reassess if hyperscaler/custom GPU substitution materially weakens Blackwell demand.
• Set alerts for: a meaningful increase in 2026 capex or funding requirements, site-permitting slippage, a sequential decline in realized capacity pricing, or customer concentration above a level that makes one contract central to annualized revenue. Any of these should trigger a downgrade of the NBIS long thesis before earnings.