Why Nebius Remains My Highest Conviction AI Bet

Nebius revenue surged 684% year over year to $399 million, while AI cloud EBITDA margins expanded to 45%, highlighting sharply improving operating leverage. Nvidia's $2 billion strategic investment strengthens Nebius's position as an AI infrastructure partner amid GPU supply shortages. A 328 MW Bloom Energy agreement could unlock an estimated $3.3 billion to $4.9 billion in annual AI infrastructure revenue opportunities for Nebius.

Analysis

NBIS is morphing from a growth story into a scarcity asset: in a world where GPU access is the bottleneck, the market is likely to pay more for secured capacity than for raw revenue growth. The second-order effect is that NVIDIA’s capital stake is not just endorsement; it reduces customer acquisition risk for Nebius and may crowd in adjacent hyperscalers/enterprise buyers who want a “safe” counterparty with preferential supply. That makes the equity re-rate potentially faster than the operating model alone would justify, because the market tends to capitalize supply security before it fully underwrites terminal margins.

The bigger takeaway is that infrastructure scarcity shifts pricing power downstream. If Nebius can keep utilization tight, the implied economics suggest a multi-year step-up in the value of every incremental MW, which should pressure smaller AI cloud competitors with weaker financing and less direct access to GPUs. Bloom Energy is the indirect beneficiary: its role becomes less about power hardware and more about being an enabling constraint-relief trade, but that also means the stock may be overowned as a clean “AI power” proxy if actual conversion of MW agreements into deployed capacity slips.

The main risk is execution lag, not demand. These deals can look highly accretive on paper months before revenue hits, and any slippage in permits, interconnects, rack deployment, or chip delivery can compress the narrative quickly. A second tail risk is competitive response: if large cloud providers or other GPU-backed infra players secure incremental supply faster, NBIS’s premium multiple could normalize despite strong fundamentals.

Consensus may be underestimating how reflexive NVIDIA’s investment is for the rest of the stack. If the market interprets the stake as a signal that supply will remain structurally tight, that supports both NBIS’s capacity monetization and BE’s power backlog, but it also lowers the probability that pricing on compute meaningfully normalizes over the next 12-18 months. In that regime, the right trade is to own the scarce-capacity enablers and avoid the “aspirational” AI names that still depend on abundant, cheap inference capacity.

Key Decisions for Investors

• Long NBIS into near-term weakness; use 3-6 month horizon. Risk/reward favors upside continuation as scarcity premium expands, but size modestly because execution and financing risk can still cause 15-20% drawdowns.
• Pair trade: long NBIS / short a basket of undercapitalized AI infrastructure peers with no strategic GPU partner. Thesis: NBIS deserves a premium for supply security and customer confidence; expect relative outperformance over 1-2 quarters.
• Long BE on a 6-12 month basis, but treat it as a conversion-rate trade rather than a pure headline trade. Best entry is on pullbacks if the market doubts project rollout; upside comes from backlog monetization, downside is delay risk.
• Buy NVDA as a low-volatility express on persistent AI infra scarcity, but prefer call spreads over outright stock if positioning is crowded. The investment signal supports multiple expansion, yet much of the strategic value may already be partially discounted.
• Avoid chasing smaller AI cloud names until there is proof of shipped capacity and contracted utilization. If you want optionality, use long-dated calls only after confirmation of deployment milestones.