Nebius’s $27 Billion Bet: Why Meta’s “Excess” Capacity Rumor Is Just Noise in a Supply-Starved Market

Posted on by

(SeaPRwire) –   By: Reginald Vance

Nebius Group is facing a classic infrastructure paradox. The market is panicking over a rumor while ignoring the hard physics of the deal. Stock dropped 16% in five days. That is a violent correction for a name up 158% year-to-date. Investors are spooked by Bloomberg reports that Meta Platforms might sell its own excess computing capacity. They see a competitor eating their lunch. This view misses the fundamental nature of the relationship. Meta is not a rival. It is a anchor tenant. The fear is misplaced. The reality is far more structural.

The core contradiction lies in the scale of commitment versus the noise of speculation. Nebius signed a $27 billion deal with Meta. This is not a pilot project. It is a massive, long-term binding agreement. Meta is backing around 300 MW of AI capacity through this partnership. That is hundreds of megawatts of dedicated power and cooling infrastructure. When a company like Meta explores selling excess capacity, it usually refers to idle, non-contracted resources. It does not mean they will cancel the $27 billion contract with Nebius. The two actions are logically distinct. One is operational triage. The other is strategic infrastructure investment. Confusing the two is a costly error for retail traders.

Look at the revenue trajectory. It is explosive and undeniable. Q2 2025 revenue was just $105 million. By Q4, the annualized run rate hit $1.25 billion. Q1 2026 delivered 684% year-over-year growth. Management targets over $3 billion for the full year. This pace suggests doubling again in 2027. Such growth requires physical assets. Land. Power. Chips. Nebius has secured 1.2 GW of power and land for a new AI factory in Pennsylvania. Contracted power capacity guidance jumped from 1 GW to over 4 GW. This is not software. This is heavy industry. The barriers to entry are measured in gigawatts, not lines of code.

Wall Street is split because the valuation demands perfection. NBIS trades around $215. Northland analyst Nehal Choksi sees a Buy at $248. He points to the shift toward higher-margin AI-native customers. He cites the Tavily acquisition as a value-add. Morgan Stanley’s Josh Baer sees a Hold at $144. He argues near-term targets are aggressive. Profitability is unproven. Large net new bookings are required. The average price target is $237.38. This implies only 10% upside. The 52-week range is $43.89 to $299.86. The volatility is inherent to the business model. It is capital intensive. It is execution heavy.

The real risk is not Meta. The risk is execution. Can Nebius actually build 4 GW of contracted power capacity? Can they deploy the silicon fast enough to meet the $3 billion revenue target? CoreWeave is active in the same space. Any slowdown in AI infrastructure spending would hurt NBIS harder than the broader tech sector. Jensen Huang’s involvement adds credibility. Nvidia is connecting AI-native companies with Nebius. This is a validation of the tech stack. It is not a guarantee of margin stability.

Investors need to stop reading headlines about Meta’s internal resource allocation. They need to look at the balance sheet. The $27 billion deal is a floor, not a ceiling. It provides visibility. It provides cash flow certainty. The pullback is a reaction to fear, not fundamentals. The supply of AI compute is scarce. The demand is insatiable. Nebius is building the pipes. Meta is buying the water. They are partners in scarcity.

The endgame is consolidation. Only players with massive capital access and power grid connections will survive. Nebius is positioning itself as one of those few. The stock drop is a test of conviction. Those who bought the $105 million revenue story should hold. Those betting on easy money will get shaken out. The infrastructure play is long. It is hard. It is real.

Author bio: Reginald Vance, a venture partner specializing in semiconductor valuation and advanced materials, focusing on the intersection of capital markets and physical AI infrastructure constraints.