Where Will Nebius Stock Be in 3 Years?

Nebius has secured more than $40 billion in multiyear AI-computing commitments, including up to $27 billion from Meta and $17.4 billion from Microsoft, while Q2 revenue surged 454% year over year to $582.3 million and adjusted EBITDA reached $236.2 million. Management targets a $7 billion-$9 billion AI-cloud revenue run rate exiting 2026 and plans to add more than 1GW of capacity annually from 2027. However, the capital-intensive expansion required $8.1 billion of first-half 2026 capex and $5.75 billion of August convertible-note funding, potentially adding about 18 million shares on top of existing dilution. The article estimates a wide 2029 valuation range of roughly $125-$835 per share and argues the current $244 share price lacks sufficient margin of safety given risks to per-megawatt pricing and future dilution.

Analysis

NBIS is evolving from an AI-cloud scarcity beneficiary into a leveraged infrastructure underwriter. The relevant valuation variable is not headline backlog but the spread between contracted revenue per MW, GPU/power financing costs, and utilization during deployment ramps; a modest decline in renewal pricing can create a disproportionate equity impact because fixed power, land, and depreciation costs are committed before revenue is recognized. Prepayments improve construction liquidity but do not eliminate counterparty concentration, delivery-acceptance, or residual-value risk.

PLTR gains strategically if a sovereign-AI channel expands its addressable infrastructure footprint, but the economic benefit is likely indirect and slower than the share-price narrative: monetization depends on sovereign procurement cycles, security certifications, and whether PLTR captures software attach rates rather than merely naming a preferred compute partner. META and MSFT are better viewed as capacity buyers with increased supplier optionality; NBIS scaling can marginally improve their negotiating leverage versus hyperscale and GPU-cloud alternatives, potentially pressuring industry AI-compute pricing in 2027-28.

The market appears to be capitalizing contracted demand as though it were software-like recurring revenue. The more likely inflection is when substantial new industry supply tests realized $/MW and when NBIS must fund the next build tranche. Over the next 1-3 months, additional sovereign-AI customer disclosures or evidence of high prepayment conversion can support momentum; over 6-18 months, delayed energization, lower realized pricing, or another equity-linked financing would force a capital-intensity rerating. The contrarian upside is that scarce energized power plus advance-funded contracts could make NBIS a strategic acquisition or financing partner before supply catches up, but that outcome requires execution without incremental dilution.

Key Decisions for Investors

• Do not chase NBIS following the partnership-driven move. Maintain a watchlist entry only after quarterly disclosure separates contracted power from live, revenue-producing MW and confirms realized annual contract value above $15M/MW; a break below that threshold would support a bearish valuation reset.
• For existing NBIS exposure, use the $313-$325 convertible conversion range as an overhead supply/dilution checkpoint over the next 6-12 months. Reduce if share-price strength is not matched by live-capacity growth, EBITDA-to-free-cash-flow conversion, and limited incremental equity-linked issuance.
• Express the sovereign-AI software angle through a modest long PLTR / short NBIS pair over 3-6 months only if PLTR discloses booked sovereign workload or software revenue tied to the partnership. This isolates potential software attach economics from NBIS construction and financing risk; exit if PLTR fails to quantify monetization or NBIS reports sustained realized pricing above $20M/MW.
• Monitor META and MSFT AI infrastructure commentary as a read-through for buyer power rather than take a directional position. Any emphasis on lower inference/training unit costs, multi-sourcing, or reduced external-compute needs would be an early warning that NBIS backlog economics may be less durable than nominal commitments imply.