Nebius: Assessing Its Position on the AI Infrastructure S-Curve

The investment case for Nebius is rooted in a foundational shift, not a fleeting trend. We are witnessing the dawn of a new technological paradigm, where artificial intelligence is moving from a software application to the very operating system of the digital economy. This transition demands a complete overhaul of the world's computing infrastructure, creating a supercycle of exponential demand for data centers. The scale is staggering: nearly 100 GW of new capacity is projected to come online between 2026 and 2030, effectively doubling global supply. This isn't just growth; it's a fundamental expansion of the planet's compute substrate, representing a $1.2 trillion asset creation opportunity in real estate alone.

This build-out is being driven by a profound change in the nature of AI workloads. In 2025, AI represented about a quarter of total data center demand, with training-the initial model creation-being the primary driver. But the curve is about to steepen dramatically. By 2030, AI is projected to account for half of all workloads. More critically, a pivotal shift is anticipated by 2027, when inference workloads could overtake training as the dominant requirement. Inference-the ongoing, real-time application of trained models-is where AI generates sustained revenue and user engagement. This means the demand for computing power becomes not a one-time spike but a persistent, growing stream, creating a need for widespread, distributed infrastructure to minimize latency.

Nebius is positioned directly on this exponential adoption curve. Its core model is a pure-play bet on this infrastructure layer. The company rents out space in data centers, owns some of its own, and provides AI-optimized computing power, primarily through clusters of Nvidia GPUs. This is a genius business model for the paradigm shift: it offers hyperscalers and other AI developers the flexible, scalable compute they need without the massive upfront capital expenditure of building and powering their own facilities. The company's own growth trajectory illustrates the opportunity. It recently adjusted its power contracting plans, moving from 1 gigawatt to 2.5 gigawatts for 2026 due to overwhelming demand. Its projected annualized revenue run rate for the end of that year is a staggering $7 billion to $9 billion, up from just $551 million in Q3. This isn't just scaling; it's a step-change, directly renting out the fundamental rails of the AI economy.

Execution & Financial Metrics: Scaling the Rails

The foundation for Nebius's explosive growth is now concrete, built on two major hyperscaler contracts worth a combined $20 billion over five years. The deal with Microsoft is valued at $17 billion, while the partnership with Meta adds another $3 billion. These aren't just large orders; they are long-term commitments that de-risk the company's massive capital expenditure plans and validate its technical and operational model. This is the first-mover advantage in action, securing a dominant position on the AI infrastructure S-curve before the next wave of demand hits.

The key metric to watch is capacity. Management expects to have over 2.5 gigawatts of contracted power by the end of 2026, with 800 MW to 1 GW of connected power already deployed. This is the physical manifestation of the $20 billion in contracted revenue. The company is moving from securing demand to delivering the fundamental compute substrate. The scale is immense, and the execution here will determine whether Nebius can meet its own ambitious targets or face bottlenecks in construction, permitting, or power procurement.

The market's high expectations are reflected in the stock's performance. Over the last 120 days, the share price has surged 79.2%, and it has climbed 201% over the last year. This isn't a reaction to quarterly earnings; it's a bet on the entire paradigm shift. The valuation metrics confirm this optimism. The stock trades at a trailing price-to-sales ratio of nearly 97, a forward P/E of 26, and a market cap near $24 billion. These multiples price in flawless execution and sustained exponential growth for years to come.

The bottom line is that Nebius is scaling the rails at an unprecedented pace. The financial health is currently supported by massive contracted revenue and a clear path to capacity deployment. However, the capital intensity of this build-out is extreme. The company must convert these power contracts into physical, connected data centers without missing a beat. The stock's recent pullback of 4% over the last 20 days shows that even with this foundation, the market remains sensitive to any sign of execution risk. The next phase is pure operational grind: turning promises into gigawatts.

The Capital & Cost Curve: Navigating the Build-Out

Scaling the AI infrastructure rails is a capital-intensive sprint, and the cost of the track is rising. Data center construction costs are forecast to increase 6% in 2026 to $11.3 million per megawatt. This inflationary pressure is a direct headwind to margins, as Nebius must deploy its contracted power capacity at a higher per-unit cost. The company is not immune to this trend; it must convert its $20 billion in hyperscaler deals into physical reality while navigating a steeper financial slope.

Nebius's strategic response is a prudent pivot toward profitability. Co-founder Roman Chernin has explicitly stated the company's focus is on margins, not revenue when negotiating its massive deals with Microsoft and Meta. This is a critical balancing act. By prioritizing margin in its largest contracts, Nebius aims to secure the capital and de-risk the build-out while preserving the financial flexibility needed to serve its broader, more diverse client base. This approach acknowledges that the initial hyperscaler wave is a launchpad, not the entire journey.

The operational balancing act is now the core challenge. The company must aggressively deploy its 2.5 gigawatts of contracted power by the end of 2026 to meet its own growth targets. Yet, it also needs to serve its core business of traditional enterprises and emerging AI firms, which are just beginning to tap its specialized GPU infrastructure. This dual mandate creates a tension between speed and selectivity. The company is preparing for a potential market "winter" by developing high-margin services and long-term relationships, but it cannot afford to slow its build-out pace in the face of soaring construction costs.

The bottom line is that Nebius is navigating a classic S-curve inflection point. The demand curve is steepening, but so is the cost curve. Its success will depend on its ability to execute this capital-intensive build-out with disciplined financial management, ensuring that the exponential growth in capacity does not come at the expense of the exponential growth in profitability.

Catalysts, Risks, and What to Watch

The investment thesis for Nebius now hinges on a series of near-term execution milestones and market dynamics that will validate its position on the AI infrastructure S-curve. The path forward is clear, but the company must navigate a steep climb.

The most immediate catalyst is the deployment of capacity for its landmark deals. Over the next three months, Nebius plans to deploy the capacity needed for its $3 billion agreement with Meta. This is the first tangible step in converting a multi-billion dollar contract into connected compute. Success here will set the tone for the entire 2026 build-out. The company's goal is to have over 2.5 gigawatts of contracted power by the end of 2026, with 800 MW to 1 GW already connected. The clock is ticking, and the market will be watching for quarterly updates on connected power milestones.

The primary risk is execution risk in delivering this promised capacity on time and within budget. The company is navigating a rising cost curve, with data center construction costs forecast to increase 6% in 2026. Any delay or cost overrun on the Microsoft and Meta projects would directly threaten the financial model built on securing long-term, high-margin deals. The stock's recent pullback shows the market is sensitive to such friction. The company's focus on margins, not revenue in these hyperscaler deals is a prudent hedge, but it must still deliver the physical infrastructure to meet its own ambitious targets.

Beyond execution, investors must watch for shifts in the AI workload mix and signs of demand softening. The paradigm shift is toward inference, which is expected to overtake training as the dominant requirement by 2027. This transition validates Nebius's model for sustained, distributed compute. However, any unexpected slowdown in hyperscaler demand for these foundational workloads could impact the growth trajectory. The company is preparing for a potential market "winter" by developing high-margin services, but its core growth is still tied to the relentless expansion of AI compute. The watchlist includes quarterly connected power figures, margin trends in the hyperscaler deals, and any commentary on the inference vs. training workload balance from management.

The bottom line is that Nebius has secured its place on the S-curve. The next phase is pure operational grind, turning promises into gigawatts while managing a rising cost curve. The catalysts are clear, the risks are defined, and the market's patience is being tested.