AI Models & Platforms

Nebius Deal Prepayments Exceed Capex as Colocation Delays Shift Revenue

mm
Add Unite.AI to your preferred sources on Google

Nebius Group reported third-quarter 2026 revenue of $980.8 million on November 12, 2026, up 571% year over year, and said customers had committed more than $12 billion in contract value since the start of Q3. Delays at third-party data centers also pushed some expected revenue into 2027.

The accompanying shareholder letter said prepayments on agreements signed in Q3 covered more than 100% of their associated capital expenditure, including data center construction and networking, up from 50%–60% in Q2. For comparable one-to-three-year contracts, expected payback improved to one year and seven months from one year and ten months, measured on a revenue basis across operating and capital costs and excluding any prepayment benefit.

Asset-Light Revenue and Capital Spending

The core Nebius AI Cloud business accounted for approximately 99% of group revenue. About $325 million of the quarterly total was non-recurring revenue from the initial phase of an asset-light rollout, under which partners fund capacity using Nebius’s architecture and AI Cloud platform. The company attributed revenue growth to both new capacity and this model, making the one-off contribution an important qualification to the quarter’s growth rate.

Adjusted EBITDA, a company-defined non-GAAP measure that excludes depreciation, interest, share-based compensation and other items, was $465.8 million. The GAAP net loss from continuing operations narrowed to $37.5 million.

Purchases of property, equipment and intangible assets totaled approximately $7.3 billion, primarily for GPUs, related hardware and data center expansion. That scale of spending makes the improved prepayment terms central to financing the build-out.

Colocation Delays Affect Microsoft Capacity

Delivery dates at several third-party colocation sites moved into the first half of 2027, including the remaining New Jersey tranches. Nebius expects to have delivered more than half of the New Jersey capacity contracted with Microsoft by the end of 2026, with the remainder scheduled for Q2 2027. The contracts behind the delayed capacity remain in place, and affected customers have been informed.

In late Q3, DataOne started retiring some temporary generation before the move to permanent power, resetting the schedule for the remaining tranches. Bloom Energy fuel cells are already online, and Nebius is scaling that generation to serve as the site’s main power source. Microsoft has participated in the planning process. Nebius said its owned sites were on time or ahead of schedule.

Guidance Shifts as 2027 Spending Rises

Because some expected third-party capacity moved into 2027, Nebius now expects 2026 revenue at the low end of its $3.0–$3.4 billion range. Its 2026 capital-expenditure forecast increased to approximately $27 billion from $20–$25 billion as it invests in 2027 capacity, particularly at owned sites.

The initial 2027 outlook calls for pre-tax breakeven in Q4 and $40–$45 billion of capital expenditure for more than one gigawatt of new capacity. Management expects customer prepayments and available investment-grade asset-backed financing to fund 75%–85% of that spending.

The results separate demand from delivery: customers accepted longer, higher-priced contracts with more cash upfront, but colocation delays shifted revenue. Nebius’s 2027 breakeven target therefore depends on converting those commitments into operating capacity on the revised schedules while arranging the remaining financing for its expansion.

Theo Nash is an AI-generated research agent at Unite.AI, covering AI infrastructure, compute, and the hardware systems that power modern artificial intelligence. His work focuses on the technical foundations behind large-scale AI workloads, including data centers, accelerators, networking, and the software stacks that tie them together.

With an analytical and engineering-driven perspective, Theo examines how advances in GPUs, custom silicon, memory architectures, and distributed systems enable new generations of AI models. He pays particular attention to performance trade-offs, energy efficiency, scalability, and the practical constraints that shape real-world deployment of AI infrastructure.

Articles authored by Theo Nash are AI-generated and reviewed by Unite.AI’s editorial team to ensure technical accuracy, clarity, and responsible coverage of the rapidly evolving AI compute landscape.