NVIDIA to invest $2 billion in Nebius and help deploy more than 5 GW of AI capacity by 2030

NVIDIA will invest $2 billion in Nebius, an Amsterdam-based AI cloud company that grew out of Yandex's international assets. This is not just a funding round: the partners plan to scale Nebius's AI infrastructure to more than 5 GW of NVIDIA-based capacity by the end of 2030.

How the deal is structured

The deal is a private placement. NVIDIA bought pre-funded warrants that can be converted into 21,065,936 Class A shares of Nebius at a nominal exercise price of $0.0001 per share. For Nebius, it is a way to raise about $2 billion at once without waiting for a classic public offering; for NVIDIA, it is a way to enter the capital as a major shareholder through a more flexible structure. If the warrants are converted, the stake would correspond to about 8.3%, and both the instrument and the future shares are subject to a six-month lock-up.

But the technical side matters more than the money. NVIDIA and Nebius agreed to work across the entire AI stack, from data center design and hardware commissioning to inference software and management of huge GPU clusters. In today's market, it is not enough to reduce this to "buy accelerators and switch them on." The winners are those who build AI factories faster, get new platforms earlier and can keep clusters stable under real customer workloads.

• Designing and launching AI factories with NVIDIA's involvement
• Developing a stack for inference and agentic AI
• Early access to Rubin, Vera CPU and BlueField
• GPU health monitoring and server fleet management

Why Nebius

Nebius is not a typical startup out of nowhere. The company emerged after the split of Yandex N.V. amid Russia's war against Ukraine and a subsequent two-year restructuring. The Russian business was sold to a consortium of local investors, while the international assets tied to AI and infrastructure stayed in the new structure. That is how Nebius appeared: a company headquartered in Amsterdam, listed on Nasdaq, and betting on a full-stack AI cloud rather than on search, taxis or advertising.

NVIDIA's bet looks even more logical given the pace of construction. Nebius is already deploying several gigawatt-scale AI factories in the US, including a site in Kansas City. The company also recently received approval to build a new 1.2 GW site near Independence, Missouri; power delivery there is expected in the second half of 2026. In other words, this is not a slide-deck presentation but a provider that is actually trying to lock in electricity, land and hardware for future demand.

"Nebius was built for AI from day one, not retrofitted from a regular cloud."

The company's financial picture is typical of the infrastructure race: fast growth and huge upfront spending. In the fourth quarter of 2025, Nebius reported $227.7 million in revenue, up 547% year over year, but also posted a net loss of $249.6 million. Capital expenditure for the quarter reached $2.1 billion as the company bought H200 and Blackwell GPUs and prepared new capacity. At the end of 2025, annualized recurring revenue stood at $1.25 billion, and the target for 2026 is already $7–9 billion in ARR.

What this means

For the market, this is another confirmation that the AI infrastructure shortage will be addressed not only by selling chips but also by directly financing cloud operators. NVIDIA has already made a similar bet on CoreWeave, and now it is strengthening Nebius. Yes, there is a circular loop here: the manufacturer invests in customers who then buy its own hardware. But such alliances, it seems, will determine who manages to skim the cream off the agentic AI boom.