Nebius Is Becoming An AI Infrastructure Giant

Nebius expanded contracted AI capacity above 3.5GW and is targeting over 4GW by 2026 while owning 75% of the infrastructure. AI cloud ARR surged 674% year over year to $1.9 billion, with pipeline generation accelerating 3.5x sequentially. The business is also showing strong operating leverage, with adjusted EBITDA margins at 45% and cost of revenue falling from 49% to 26%.

Analysis

NBIS is moving from a “story stock” to an infrastructure compounding story, and that changes the competitive map. Owning a large share of the asset base means the company can keep expanding capacity without handing away economics to third-party landlords, so the margin inflection is likely to persist longer than the market typically prices in for AI infrastructure providers. The second-order winner is the broader AI hardware and networking stack: sustained utilization growth should keep demand tight for accelerators, high-speed interconnect, power gear, and liquid cooling across the ecosystem.

The key read-through is that supply, not demand, is now the bottleneck that matters. If contracted capacity is already above current deployment and pipeline is still accelerating, pricing power can stay intact for several quarters, which is how the market gets surprised on both revenue durability and EBITDA leverage. The losers are less obvious: smaller AI cloud entrants and traditional colocation names that depend on leasing economics may face a widening cost disadvantage if NBIS keeps internalizing more of the infrastructure stack.

Risk is mostly execution and capital intensity over the next 6-18 months. The bullish case breaks if power delivery, permitting, or GPU sourcing constrains the buildout, or if customer concentration and pre-commitments prove less sticky than implied by the pipeline acceleration. The market may also be underestimating how fast competition responds: hyperscalers can compress the economics of standalone AI clouds if they decide to subsidize capacity to defend strategic accounts.

Consensus may be too focused on ARR growth and not enough on the quality of the asset base behind it. If the company is truly locking in returns on owned infrastructure, the right framework is not revenue multiple alone but FCF durability and reinvestment efficiency, which could justify a premium re-rating even after the recent move. But if the market is already pricing perfection, the next leg up likely needs confirmation from sustained margin hold and sequential pipeline conversion, not just headline capacity announcements.

Key Decisions for Investors

• Long NBIS on 3-6 month horizon into any post-news consolidation; use pullbacks as entry because the setup is now margin-and-capacity led, not just narrative-driven. Risk: execution slippage on buildout; reward: continued multiple expansion if EBITDA margins stay near current levels.
• Pair trade: long NBIS / short a premium-valued colocation or leased-capacity peer over the next 1-2 quarters. The relative thesis is that owning more of the stack should preserve economics better than asset-light competitors if AI demand stays tight.
• Buy NBIS call spreads 3-6 months out rather than outright calls to express upside while capping premium decay. This works best if the market needs one more confirmation print on pipeline conversion and margin durability.
• Watch for any signs of power, GPU, or delivery bottlenecks over the next 60-90 days; if they surface, trim 25-40% of longs because the multiple can compress quickly when growth is intact but execution slips.
• If NBIS continues to outperform on both growth and margins, consider rotating exposure from software-only AI beneficiaries into infrastructure enablers, as the market may be underweight the picks-and-shovels leg of the AI capex cycle.