Nebius: Riding the AI Infrastructure S-Curve to the Next Paradigm

Nebius is positioned squarely on the steep, early-mid phase of the AI compute adoption S-curve. Its explosive growth trajectory is the clearest signal that the market is moving from experimentation to fundamental infrastructure build-out. The company's journey from $90 million in annual recurring revenue (ARR) in 2024 to $1.25 billion by the end of 2025 represents a more than 14x increase-a rate that defines the inflection point of a paradigm shift. This isn't just scaling; it's capturing the initial wave of demand as the world's most ambitious tech firms race to secure the compute power needed to train and run next-generation models.

Management's forward view suggests the steep climb is far from over. The company now anticipates up to $9 billion in ARR by the end of 2026, which would imply another potential 7x growth. This projection, if achieved, would place Nebius at the very forefront of the exponential adoption curve. The sheer magnitude of these numbers underscores a critical shift: the market is no longer debating whether AI compute is essential. It is now about who can build and deliver the physical and software infrastructure at the scale required.

A key pillar of this growth is the foundation of multi-billion dollar, long-term contracts with industry giants. The 5-year, $17.4 billion deal with Microsoft and the $3 billion tie-up with Meta are not just large orders; they are strategic commitments that lock in demand for years. These agreements, which together imply more than $3 billion in annual recurring revenue for 300 megawatts of capacity, validate Nebius's model as a critical infrastructure layer. They provide the financial runway and credibility to pursue the next wave of customers, from established digital firms to the mainstream enterprises that will eventually need to adopt AI at scale. In this context, Nebius isn't just selling data centers; it's building the fundamental rails for the next technological paradigm.

Scaling the Physical and Software Rails: Power, Performance, and Profitability

To ride the AI S-curve to its next plateau, Nebius must scale both its physical and software infrastructure in lockstep. The company's plan to have over 2.5 GW of contracted power by the end of 2026 directly addresses the critical bottleneck of energy supply. This isn't just about building more data centers; it's about securing the massive, reliable power needed to run thousands of GPUs. With 800 MW to 1 GW of that power already connected, Nebius is laying down the physical rails at a pace that matches its explosive ARR growth. This power foundation is the essential first step for any compute provider aiming for the scale of hyperscalers.

On the software side, performance is the key differentiator. Nebius's stack is engineered for the latest hardware, as demonstrated by its strong performance across several configurations of the latest NVIDIA Blackwell systems in the MLPerf® benchmark. This integration isn't incidental; it's core to the value proposition. By optimizing its full-stack cloud platform for NVIDIA's cutting-edge chips, Nebius ensures its customers achieve the fastest possible training and inference speeds. This performance advantage, coupled with partnerships like the one with Anyscale for Ray-based workloads, creates a sticky ecosystem that attracts the most demanding AI innovators.

The ultimate test of scaling is profitability. Here, Nebius has crossed a significant threshold. The company achieved first positive EBITDA of $15 million in Q4 2025. This marks a clear shift from pure growth investment to operational efficiency. While net losses remain high due to other business segments, positive EBITDA signals that the core AI infrastructure business is generating cash from operations. It shows the model can work at scale, turning massive revenue growth into tangible bottom-line progress. This is the financial fuel needed to fund the next leg of the expansion, where power contracts and software performance must continue to outpace demand.

Valuation, Catalysts, and the Execution Imperative

The valuation on Nebius is a direct reflection of its position on the AI S-curve. With a market cap near $28 billion, the stock trades at a price-to-sales ratio of 43. That multiple is not a guess; it is a bet on the company's ability to maintain its projected exponential growth. To justify that price, Nebius must achieve the up to 7x ARR growth in 2026 it has forecast. If the company hits that target, the valuation becomes a function of its future scale. The risk is that any stumble in execution could make today's premium look like a bubble.

The most immediate catalyst is the deployment of capacity for the Meta deal. The company has announced an agreement to deliver AI infrastructure to Meta, valued at ~$3B over five years. The plan is to deploy the capacity needed over the next three months. This is a classic S-curve inflection: a large, multi-year contract is being converted into tangible, near-term revenue and operational momentum. Successfully executing this build-out will be a critical proof point for the company's scaling ability and could serve as a springboard for similar deals with other mainstream enterprises.

Yet the primary risk remains execution at scale. The company is attempting to deploy over 2.5 GW of contracted power by the end of 2026, a massive physical build-out. It must also scale its software stack and operational teams in parallel. The path from securing a $3 billion contract to reliably delivering that capacity without cost overruns or technical failures is fraught. The company's first positive EBITDA of $15 million in Q4 2025 shows the core business can be profitable, but it operates against a backdrop of deep net losses. The infrastructure build-out is capital-intensive, and any significant overrun in costs or delays in deployment would pressure margins and test investor patience.

The bottom line is that Nebius is a pure-play on the AI infrastructure paradigm shift. Its valuation demands flawless execution on the exponential growth curve. The next three months are a key test, as the Meta deal moves from paper to physical reality. For investors, the setup is clear: the potential reward is defined by the company's ability to scale its power, software, and operations to meet the insatiable demand for compute. The risk is that the complexity of building the rails for a new technological era proves harder to manage than the growth itself.