Nebius just showed what the AI infrastructure boom looks like when demand is no longer the hard part. The harder question is whether the company can fund and build fast enough without turning growth into a balance sheet trap.
Nebius Group's first quarter was not a normal cloud earnings report. Revenue rose to $399.0 million, up 684% from a year earlier, while the core Nebius AI Cloud business reached $389.7 million, up 841%. For a company trying to position itself between Nvidia's GPU supply and the compute needs of AI builders, those numbers matter because they suggest customers are not just testing capacity. They are consuming it.
The Amsterdam-based, Nasdaq-listed company also moved into positive adjusted EBITDA, reporting $129.5 million for the quarter after a loss a year ago. Its AI cloud unit did even better, producing $174.0 million of adjusted EBITDA on a 45% margin. That gives Nebius a cleaner story than many AI infrastructure names: revenue is rising quickly, utilization is strong, and the business is beginning to show operating leverage.
But this is still an expensive race. Adjusted net loss widened to $100.3 million from $83.6 million, and capital expenditures were about $2.5 billion in the quarter, mostly for GPUs, related hardware, and data center expansion. Net income of $621.2 million looks flattering on the surface, but it was helped by a $780.6 million gain from revaluing equity investments. That is useful for the balance sheet. It is not the same thing as recurring operating profit.
The most important update may not have been the revenue line. Nebius said it has secured up to 1.2 gigawatts of power and land for a new owned AI factory in Pennsylvania, with delivery planned in phases beginning in 2027. That follows the company's May 12 groundbreaking at its 1.2 gigawatt Independence, Missouri campus and a 310 megawatt site in Finland.
In his shareholder letter, CEO Arkady Volozh said compute and cloud needs are exceeding available capacity as companies move from AI experiments into real-world applications. That is the entire neocloud pitch in one sentence. Traditional cloud providers are large, but their best capacity is often spoken for. AI startups and enterprise teams need clusters now, not after a multiyear procurement process.
Nebius says contracted power now exceeds 3.5 gigawatts, already above its prior 2026 target of 3 gigawatts, and it has raised year-end guidance to more than 4 gigawatts. Owned capacity represents more than 75% of that total. The company also expects 800 megawatts to 1 gigawatt of connected power by the end of 2026, which is the nearer-term figure investors should watch. Contracted power is promise. Connected power is what customers can actually use.
As Barron's noted after the results, Nebius shares rose sharply because the market saw a company benefiting from scarce AI data center capacity. That scarcity is real. But scarcity alone does not guarantee durable returns. Data centers need power agreements, local approvals, GPUs, cooling, construction crews, network connections, and customers willing to commit before the assets age.
The Meta deal cuts both ways
Nebius's five-year agreement with Meta gives the story more weight. The deal is worth up to about $27 billion, including $12 billion of dedicated capacity scheduled to begin in early 2027 and up to $15 billion of additional capacity that can be sold to Meta on agreed terms or to other AI cloud customers at market rates. It is a strong endorsement from one of the world's biggest AI spenders.
It also creates a familiar infrastructure risk. Large contracts help finance large builds, especially when customers make commitments early. But they can also concentrate revenue, shape capacity allocation, and make a young cloud provider dependent on the investment cycles of a few enormous buyers. If Meta, Microsoft, or another large customer keeps expanding, Nebius has a tailwind. If those customers slow spending or renegotiate future needs, the same fixed assets become less forgiving.
That is why the customer mix matters. Nebius says its pipeline generation rose about 3.5 times quarter over quarter and that growth is coming from model builders, vertical AI companies, and inference workloads. It named healthcare and life sciences customers, including Sword Health, as examples of demand outside the largest platform companies. That is the right direction. A cloud business becomes stronger when it has anchor contracts for financing and a broad base of customers for pricing power.
The next test is execution. Nebius ended the quarter with $9.3 billion in cash, helped by convertible notes and a $2 billion investment from Nvidia through prefunded warrants. It also carried $8.4 billion of non-current debt and said it is evaluating more financing, including asset-backed and corporate-level debt. This is how the AI infrastructure build-out is being funded: customer prepayments, capital markets, chip partnerships, and a belief that demand will still be there when the buildings are ready.
For AI startups, Nebius's growth is good news because it adds another serious supplier of high-performance compute in a market where access still shapes product roadmaps. For investors, the story is more demanding. The company has shown that demand is there. Now it has to prove that power, capital, and customer concentration can be managed at the same speed as revenue growth.
Also read: Alibaba's AI spending has pushed profit to the edge • Americans are telling AI startups to slow down and earn trust • AI data centers are testing rural America's bargain with Big Tech
