Nebius Captures the AI Inference Pivot: Revenue Surges 684% as $25 Billion Infrastructure Bet Takes Shape

Cloud provider Nebius reports $399M revenue, swing to profit, and $25B capex as AI focus shifts from training to inference. Nasdaq-100 inclusion set for June 22.

The narrative driving artificial intelligence infrastructure is shifting. For the past two years, the heavy lifting was focused on training ever-larger models, but the center of gravity is now moving to inference — the real-time execution of AI models for end users. Nebius, the Amsterdam-headquartered cloud provider, has positioned itself squarely in that fast-growing segment, and the market is taking notice.

The company’s first-quarter results delivered a stark confirmation of the thesis. Revenue hit $399 million, a 684% leap from the same period last year. Even more striking, adjusted EBITDA swung from a loss of $53.7 million to a profit of $129.5 million. Management characterized the business as fully booked — demand outstripped available capacity, an unusual starting point in an industry where overcapacity is the typical risk.

That growth is attracting analyst attention. Bank of America has set a price target of $280, while Citigroup sees fair value at $287. Both projections imply significant upside from Friday’s close of €200.60, which still sits roughly 17% below the 52-week high of €242.95. Year to date, the stock has already gained 162%.

Billions Poured Into Capacity

To sustain its expansion, Nebius has raised its capital expenditure budget for the current year to as much as $25 billion. The goal is to scale data center capacity from 170 megawatts to between 800 megawatts and one gigawatt by the end of 2026. In Pennsylvania alone, the company has secured up to 1.2 gigawatts of power and land for a dedicated AI factory. Vertically contracted power capacity is expected to reach 4.0 gigawatts by year-end.

The build-out extends across Europe and the US. Nebius is investing £1.7 billion in new Nvidia-powered facilities in the UK, complementing existing projects in Finland and a massive complex in Missouri. To bolster its software and orchestration capabilities, the company is acquiring the startup Eigen AI for $643 million, paid in cash and shares.

Nasdaq-100 Entry Triggers Passive Inflows

The most immediate near-term catalyst arrives on June 22, when Nebius joins the Nasdaq-100 index. The inclusion forces passive ETFs and index funds to mechanically buy the stock, regardless of fundamental views. While some of that buying pressure may already be priced in, the actual rebalancing flows will hit at the close of trading on that date.

Technically, the stock appears stretched by conventional measures. At €200.60, it trades about 26% above its 50-day moving average and roughly 90% above the 200-day average. The relative strength index sits at 56.7 — neutral territory — following a sharp pullback from the highs. Annualized volatility of 110% reflects the market’s honest assessment of execution risk at this scale.

Guidance Points to Temporary Margin Compression

For the full year 2026, management reaffirmed revenue guidance of $3.0 billion to $3.4 billion, with an adjusted EBITDA margin of roughly 40%. Capital spending is being pulled forward by customer contracts already secured for 2027. The second quarter, however, will show a temporary margin dip due to the timing of investments and backend-loaded capacity. Margins are expected to recover to first-quarter levels in the third quarter and improve further in the fourth.

Structural Logic Behind the Growth

Nebius’s strategy hinges on a simple dynamic: the hyperscalers — Amazon, Google, Microsoft — handle the largest training workloads, but when their own capacity is exhausted, excess demand spills over to third-party providers. Neoclouds like Nebius offer raw GPU capacity built for both training and inference, without the database, storage, or analytics layers that hyperscalers bundle in. Management has reported that multiple customers are competing for every available GPU across all chip generations, and contract durations are lengthening noticeably. Multi-year commitments are replacing spot purchases, transforming revenue from volatile to predictable.

That predictability comes with execution challenges. The planned capacity expansion is aggressive by any standard, and the semiconductor sell-off on June 5, when stocks like AMD dropped double digits in a single day, demonstrated how quickly neoclouds can amplify market moves. Nebius enters the week of June 16 with fully booked capacity, accelerating margins, and the promise of index-driven buying — yet it remains 17% off its own high. The inference wave is real. Whether the current valuation already accounts for it is the only question that matters.