Arkady Volozh's Nebius seeks to raise up to $4.6 billion for AI data centers and GPU purchases

Nebius, led by Yandex co-founder Arkady Volozhov, plans to raise up to $4.6 billion through convertible bonds. The company intends to direct the funds toward building data centers for AI and purchasing GPUs — essentially a sharp expansion of computational infrastructure following a major contract with Meta.

Where the Money Will Go

This is not about cosmetic business expansion, but rather an attempt to rapidly increase capacity in one of the most scarce market segments. Companies that provide access to AI cloud compete not only on software and pricing, but also on actual computing power: server racks, energy, cooling, networks, and accelerator supply. If a provider doesn't have available GPUs, it simply cannot keep up with demand from developers and corporate clients.

The convertible bonds format is also telling in this context. It allows the company to attract a large volume of capital now while giving investors an option to convert debt into shares later if the company continues to grow. For an infrastructure AI business, this is a convenient tool: the money is needed upfront because data centers and accelerator purchases require substantial investments long before new capacity starts generating revenue.

• New data centers for model training and inference
• GPU purchases to expand cloud capacity
• Preparation of facilities with power, cooling, and network connectivity
• Launch of capacity under already signed and future corporate contracts

Betting on the Contract

The catalyst for this fundraising was a record-breaking contract with Meta. According to published information, its total value could reach $27 billion — a sum that fundamentally transforms Nebius's business scale in a single step. Even if the final revenue turns out to be below the upper estimate, the deal itself demonstrates that the company is viewed as a serious AI infrastructure supplier, not merely a notable player in an overheated market.

For Meta, this is a way to more quickly secure access to external computing resources amid persistent GPU scarcity. For Nebius, it's a test of the ability to rapidly convert a major contract into functioning infrastructure. Signing an agreement is not enough: the company must guarantee hardware delivery, meet facility deployment timelines, ensure service stability, and scale without quality degradation.

This is precisely why, after a major contract, questions about financing inevitably arise alongside strategy. In this sense, Nebius is attempting what nearly all ambitious AI providers do today: secure their place in the supply chain before the largest cloud platforms definitively divide the market. The earlier a company builds facilities, reserves accelerators, and demonstrates readiness to rapidly deploy new capacity, the higher the chances of retaining a client not for a quarter, but for years ahead.

The Race for Capacity

The AI infrastructure market currently operates under fairly harsh economics. Demand for computing grows faster than new data centers appear, and the cost of error is very high: underutilization impacts profitability, while delays in deploying capacity hurt client trust. This is why companies that take on billions in borrowing for GPUs and data centers are essentially betting that the generative AI boom will persist and corporate customers will continue to shift more tasks to specialized AI clouds.

This strategy has its downsides. Infrastructure growth requires not only servers but also access to electricity, engineering teams, communication channels, maintenance, accelerator fleet upgrades, and long-term planning horizons. However, if utilization remains high, these assets become a solid business foundation: clients find it harder to migrate, and revenue becomes more predictable than for companies that survive only on one-off projects or pilots.

What This Means

If Nebius truly raises up to $4.6 billion and quickly converts that capital into operational capacity, the company will solidify its position among the major players in the AI cloud market. For the entire industry, this sends another clear signal: the biggest bets in AI today are not being made on flashy demos, but on data centers, electricity, supply logistics, and GPUs.