Revenue Soars 684% at Nebius, But the Real Story Is Its Bid to Escape the GPU Rental Trap
12.06.2026 - 12:23:29 | boerse-global.de
The numbers alone would make most growth-equity investors salivate. Nebius generated $399 million in revenue during the first quarter of its fiscal 2026 — a 684% surge from the same period last year — and guided the full year to between $3.0 billion and $3.4 billion. Its contracted backlog stands at roughly $46 billion, anchored by a $27 billion deal with Meta and a $17 billion agreement with Microsoft. Yet for all the top-line firepower, the market has priced in something more complicated: a stock that has rocketed 165% since January and 365% over twelve months, but still sits 17% below its 52-week high of €242.95.
That gap between the euphoric trajectory and the recent pullback captures the central tension surrounding Nebius. The company is simultaneously a hypergrowth Neocloud that rents NVIDIA GPUs to desperate hyperscalers and a self-described builder of something more durable — a vertically integrated inference platform it calls Token Factory.
From Capacity Arbitrage to Software Economics
The Neocloud thesis is brutally simple and brutally fragile. AI compute demand is expanding faster than the balance sheets of even Microsoft and Meta, so excess capacity gets outsourced. Margins in this segment are fat because supply is scarce; customers pay hefty premiums for guaranteed GPU access. But once the supply crunch eases, those margins will compress. For Nebius, the real long-term wager is whether it can look any different in a world awash with chips.
Management’s answer is Token Factory. Through the acquisition of Eigen AI and a licensing deal for Clarifai’s intellectual property, Nebius is layering software control onto what was once a pure hardware business. Tavily roots agentic reasoning in real-time data, Eigen optimises the model, and Clarifai orchestrates the whole stack. The product is no longer raw compute capacity but managed inference endpoints, where service-level agreements are defined by token throughput, latency, and cost. The economics of AI are shifting decisively from training to inference, and Nebius is betting that the firm which controls the full optimisation chain — from silicon to software endpoint — will win the cost-per-token battle.
Should investors sell immediately? Or is it worth buying Nebius?
The Revolut Test and a Crowded Pipeline
The strongest evidence for this shift came from an unlikely customer. Revolut, the London fintech with more than 75 million clients across 40 markets, has rebuilt its AI stack on Token Factory. The platform now powers FinCrime agents that detect financial crime in real time and a chat orchestrator handling over one million customer-service tickets every month. A regulated financial institution running production agentic workloads on a Neocloud changes the sales conversation: it is no longer about GPU availability but about whether Nebius can run your AI business.
That pipeline is already thickening. Nebius reported a 3.5-times sequential increase in prospective deals during Q1 when hyperscaler contracts are excluded. The demand is so intense that for every newly deployed cluster, four or five potential customers are competing for capacity. The entire current fleet was fully booked in Q1, and most upcoming capacity over the next few quarters has already been allocated to clients.
Spending Billions to Meet Demand — and Hedge Risk
To relieve that bottleneck, Nebius is investing at a staggering pace. Capital expenditures for fiscal 2026 are planned between $20 billion and $25 billion. In the US, the company is building a 300-megawatt data centre in Vineland, New Jersey alongside DataOne, and has announced a gigawatt-scale “AI Factory” campus in Missouri. A partnership with Rowan University, launched on June 11, will train students in data science and cloud computing — a clear nod to the growing workforce needed to run these sites.
In the UK, €2.3 billion is being poured into three facilities that will collectively deliver 65 megawatts of capacity by 2027, all equipped with Nvidia Blackwell Ultra chips. And a global partner programme, built on a distribution pact with TD Synnex, is set to launch in July or August, offering resellers double-digit margins on Nebius cloud services.
Yet the company’s customer structure also contains a paradox. Microsoft and Meta are simultaneously its largest buyers and its most dangerous long-term competitors. Management insists that strategic partnerships with hyperscalers complement a broader base of AI-native startups, independent software vendors, and enterprise clients. The revenue numbers suggest that diversification is underway, but the market’s annualised 30-day volatility of more than 110% reflects just how much execution risk remains — not in generating demand, but in delivering on construction timelines and chip supply.
Nebius at a turning point? This analysis reveals what investors need to know now.
The Structural Tailwind
Nebius is not betting alone. Bank of America forecasts the server CPU market will exceed $170 billion by 2030, expanding at a 37% compound annual rate, driven by the spread of agentic AI systems that plan and execute tasks autonomously. The Nebius backlog and revenue trajectory align with that thesis.
The stock has rallied 433% from its 52-week low, and the EBITDA margin of 32% in Q1 suggests that growth is not yet destroying profitability. But the ultimate test will come when GPU scarcity gives way to abundance. At that point, the value of Token Factory’s cost-per-token economics — not the size of its hardware inventory — will determine whether Nebius is a high-margin infrastructure platform or merely an expensive bet on a cycle that will eventually turn.
Ad
Nebius Stock: New Analysis - 12 June
Fresh Nebius information released. What's the impact for investors? Our latest independent report examines recent figures and market trends.
