Demand for computing infrastructure dedicated to artificial intelligence is extending well beyond the traditional planning horizon, with customers already seeking capacity for 2028. Nebius CEO Arkady Volozh says the company is facing substantially more demand than it can physically serve, highlighting the persistent gap between the rapid adoption of AI and the pace at which the infrastructure required to support it can be built.

AI Compute Demand Is Extending Into 2028

Volozh said Nebius is already receiving requests for capacity in the first and second quarters of 2028, with some customers seeking tens of thousands of GPUs and vGPUs. The company had previously indicated that it could sell its entire planned 2027 capacity if it chose to do so, but has maintained some capacity for shorter-term requirements and opportunities to capture higher pricing.

The gap between demand and supply is also being driven by physical constraints. Data centers, power connections, advanced chips and computing infrastructure cannot be expanded as quickly as demand for AI is growing. For AI cloud providers such as Nebius, that means future capacity is becoming a strategic asset even before it becomes operational.

Nebius Growth Reflects the Scale of AI Infrastructure Demand

Nebius’ financial performance illustrates the acceleration taking place across the market. The company generated $582.3 million in revenue during the second quarter of 2026, an increase of 454% from the same quarter a year earlier. AI cloud revenue reached $575 million, up 514% year over year.

During the quarter, Nebius signed four significant contracts with AI customers, with the average value of each agreement exceeding $1 billion. The company said these contracts generate returns of $20 million to $25 million per MW, while upfront payments cover 50% to 60% of the capital expenditures associated with building the capacity.

Nebius raised its target for contracted capacity at the end of 2026 to 5GW, compared with a previous target of more than 4GW. Starting in 2027, the company plans to deploy more than 1GW of capacity each year.

Limited Capacity Is Supporting Higher Pricing

The shortage of computing capacity is not simply a supply issue. It is also changing pricing dynamics across the market. Nebius said that in short-term deals, typically covering periods of up to six months, customers are willing to pay a significant premium for capacity that is available immediately. The company is managing deals in the range of $40 million to $50 million per MW, and sometimes higher.

The environment indicates that AI customers are willing to pay more to secure access to computing resources, particularly when delays in obtaining GPUs could postpone model training or product launches. At the same time, higher pricing can support the economics of building new infrastructure while raising the level of investment required from AI cloud providers.

Through the remainder of 2026 and into 2027, investors will be watching Nebius’ pace of capacity expansion, new contracts, GPU availability and the costs of financing and infrastructure development. If demand continues to extend into 2028 as the company currently reports, the gap between AI compute demand and supply could remain a defining feature of the infrastructure market. However, faster data center construction, improvements in power availability and increased chip supply could eventually ease the shortage and reshape competitive and pricing dynamics.