Summary
- Nebius proposes $4.5bn of convertible senior notes split between 2030 and 2034 maturities.
- Proceeds are intended to support data centre construction, build-outs, AI cloud expansion, and purchases of infrastructure including GPUs.
- The financing illustrates the capital intensity of scaling European AI infrastructure as operators commit before new facilities are fully delivered.
Nebius is seeking to raise $4.5bn through a proposed convertible debt offering as the Amsterdam-headquartered AI infrastructure provider finances further data centre construction, cloud expansion, and hardware purchases.
Nebius plans to offer $2.75bn of convertible senior notes due in 2030 and a further $1.75bn due in 2034 through a private placement to qualified institutional buyers.
The company also expects to give initial purchasers options to acquire up to an additional $375m of the 2030 notes and $300m of the 2034 notes.
Nebius said the proceeds would support its continuing growth, including construction and build-out of data centres, expansion of its full-stack AI cloud platform and geographic footprint, and purchases of key infrastructure components including graphics processing units.
General corporate purposes are also included in the proposed use of funds.
The financing puts another large number against the capital requirements emerging around AI infrastructure. Building compute capacity at current scale involves spending on more than GPUs: sites require grid connections, substations, switchgear, cooling plant, networking, buildings, and the supporting systems needed to operate high-density equipment reliably.
Those costs arrive on different timelines. Long-lead electrical equipment and property commitments may need to be secured well before IT hardware is installed, while new data centre developments can take years to move from land acquisition and power reservation to commissioning.
Convertible debt provides Nebius with another route to finance that expansion while deferring the point at which some potential equity dilution occurs. The company said it may also enter separate privately negotiated exchange agreements involving existing convertible notes due in 2029 and 2031.
The scale of the transaction is notable in a European market where AI infrastructure providers are competing simultaneously for capital, electrical capacity, chips, suitable buildings, and engineering resources.
Large hyperscalers can fund much of their data centre expansion from very large operating cash flows, while newer AI cloud providers frequently need external debt or equity to secure capacity quickly enough to keep pace with customer demand.
That creates a different risk profile. The commercial case for a new data centre or GPU deployment depends not only on demand forecasts but on utilisation, hardware economics, energy prices, financing costs, and the ability to bring facilities online on schedule.
It also means capital availability can become an infrastructure constraint in its own right. Securing a grid connection does not produce usable capacity if the operator cannot finance the building and IT equipment behind it.
Conversely, raising billions of dollars does not eliminate the physical bottlenecks associated with grid queues, construction lead times, and equipment procurement. Money can secure options, advance orders, and development rights, but it cannot compress every part of the delivery chain.
That relationship between capital and buildability is becoming more visible as AI infrastructure providers scale across Europe. DataCentral has previously examined how institutional money is following AI infrastructure into data centre development, reflecting the increasingly direct link between compute growth and long-duration infrastructure investment.
Nebius’s proposed offering is another version of the same pressure. Its stated use of proceeds reaches from physical data centre construction through to GPUs, placing the building, electrical systems, and compute layer within one capital programme.
The company will still need to convert that financing into commissioned capacity. In a European market where suitable powered sites remain difficult to reproduce quickly, the operational value of the raise will depend on how efficiently capital can be translated into connected, cooled, and occupied infrastructure.
