Nebius: Wall Street Is Watching The Wrong Number

Nebius is targeting $3.0-$3.4 billion of 2026 revenue and a $7-$9 billion exit run rate as it expands AI computing capacity. Connected power is projected to reach 800 MW-1 GW by year-end, while contracted power commitments have risen to 5 GW. Customer prepayments could exceed $9 billion in 2026, and recent contracts are expected to cover roughly 50%-60% of related infrastructure CapEx, supporting the funding of its capacity buildout.

Analysis

The equity should be underwritten as a leveraged infrastructure-financing story rather than a conventional cloud-software multiple. The implied monetization of energized capacity requires sustained high GPU utilization and premium pricing; any slippage in customer deployment schedules would hit revenue disproportionately because depreciation, power commitments, and data-center operating costs are largely fixed. Prepayments improve working-capital optics, but partial CapEx coverage still leaves NBIS exposed to equity issuance, secured debt, or lease financing if build costs rise or capital markets tighten.

Near term, the key catalyst is evidence that contracted demand converts into disclosed customer names, binding take-or-pay terms, and funded construction milestones. Over 6-18 months, scarce interconnection and power availability could make NBIS strategically valuable versus capacity-constrained GPU-cloud peers such as CRWV, IREN and CIFR; however, that scarcity premium disappears if hyperscalers pull forward proprietary builds or NVIDIA supply loosens enough to compress rental rates. The less appreciated bottleneck is electrical equipment and construction execution: VRT, ETN, GEV and PWR can capture incremental spend, while transformer/switchgear delays would defer NBIS revenue without necessarily reducing its cash burn.

Consensus may be assigning too much value to headline contracted power rather than deliverable, financed, GPU-equipped capacity. The critical operating proof point is revenue per live MW and utilization, not aggregate pipeline: a material discount to the implied run-rate economics would force a lower terminal multiple and raise dilution risk. Conversely, verified long-duration customer commitments with capex reimbursement above current levels would materially de-risk the model and justify rerating.

Key Decisions for Investors

• Establish only a starter long NBIS position (maximum 1% NAV risk budget) after the next filing or investor update confirms named counterparties, contract duration and take-or-pay protections. Treat as a high-volatility 12-18 month infrastructure option: potential upside is substantial on financing de-risking, but a 40-50% drawdown is plausible if funding needs exceed expectations.
• Do not add on capacity announcements alone. Add only if reported revenue per energized MW and utilization validate the implied exit economics for two consecutive quarters; failure to provide these KPIs or a downward 2026 revenue revision is the thesis falsifier.
• Use a relative watch trade rather than an immediate pair: long NBIS / short CRWV only if NBIS secures more customer-funded build commitments while CRWV's financing spreads widen. Without comparable contract disclosure and valuation data, the pair is not yet actionable.
• Accumulate VRT or ETN on broad industrial pullbacks as a lower-volatility second-order exposure to AI data-center electrical intensity. This thesis weakens if AI infrastructure operators begin materially delaying energized-capacity timelines or if order/backlog commentary signals normalization.