Nebius Plans $40 Billion to $45 Billion in 2027 Spending as AI Data Centre Race Intensifies

Nebius Group is preparing for one of the largest infrastructure investment programmes in the artificial intelligence industry, forecasting capital expenditure of $40 billion to $45 billion in 2027. The Amsterdam-based AI cloud provider plans to use customer prepayments and financing backed by long-term contracts to fund much of the expansion. The strategy reflects growing demand for computing capacity from AI developers and enterprises, but it also exposes the company to risks involving construction delays, power availability and the financial commitments required to build data centres at scale.

The spending outlook was announced alongside Nebius’s third-quarter 2026 results on November 12. The company reported revenue of $981 million, up 571% year-on-year, and forecast revenue of $12 billion to $13 billion for 2027. However, delays at third-party data centre projects have forced it to revise its near-term revenue expectations. The situation highlights a central question facing the emerging neocloud industry: can specialised AI cloud providers build infrastructure quickly enough to meet demand while keeping their financing commitments manageable?

Nebius Plans Up to $45 Billion in Capital Spending

Nebius expects to spend between $40 billion and $45 billion on capital expenditure in 2027, a substantial increase from its current investment programme.

The spending will primarily support the expansion of AI data centres, GPU infrastructure, networking equipment and computing capacity.

The company also raised its expected 2026 capital expenditure to approximately $27 billion, compared with its earlier forecast of $20 billion to $25 billion.

The investment reflects the enormous infrastructure requirements of modern AI systems. Training and operating large models require specialised processors, high-speed networking, cooling systems and reliable electricity supplies.

Nebius wants to expand its infrastructure footprint to serve AI laboratories, software developers and large enterprise customers.

Its expansion strategy includes building and operating its own data centres while also working with third-party infrastructure partners.

The company expects to bring online more than one gigawatt of additional capacity annually as its infrastructure network expands.

How Will Nebius Finance the $40 Billion Expansion?

One of the most important aspects of Nebius’s strategy is its reliance on customer prepayments and asset-backed financing.

Rather than funding the entire infrastructure programme through conventional borrowing or equity issuance, the company intends to use money committed by customers under long-term contracts.

Nebius expects customer prepayments and asset-backed financing of investment-grade contracts to cover approximately 75% to 85% of its planned 2027 capital expenditure.

The company also expects customer prepayments to increase from approximately $10 billion in 2026 to between $17 billion and $20 billion in 2027.

Its financing strategy includes:

• Customer prepayments: Customers pay a portion of their contractual commitments before the associated computing capacity is delivered.
• Asset-backed financing: Nebius can borrow against long-term customer contracts and the revenue they are expected to generate.
• Cash reserves: The company ended the third quarter with approximately $7.4 billion in cash and cash equivalents.
• Operating cash flow: Growing revenue and margins could increasingly contribute to infrastructure investment.

Nebius said prepayments attached to deals signed during Q3 covered more than 100% of the associated capital expenditure, including data centre construction, hardware and networking costs.

This is a significant improvement from the 50% to 60% coverage reported in the previous quarter.

The strategy can reduce the amount of external capital required, but it also makes timely infrastructure delivery particularly important. If a project is delayed, the company may face additional costs before it can begin generating the revenue needed to fulfil its customer commitments.

Microsoft Contract Highlights the Risk of Data Centre Delays

Nebius’s expansion plans are facing a major execution challenge at a data centre project in Vineland, New Jersey.

The facility is operated by third-party infrastructure provider DataOne and is intended to supply computing capacity under Nebius’s agreement with Microsoft.

The Microsoft contract is valued at up to $19.4 billion.

However, the New Jersey project has experienced delays, pushing some planned capacity deliveries into 2027.

Nebius said only slightly more than half of the relevant Microsoft capacity is now expected to be delivered by the end of 2026, with the remaining portion scheduled for the second quarter of 2027.

The delay is linked to changes in the site’s power arrangements.

DataOne had been using temporary natural gas generators while transitioning towards Bloom Energy fuel cells. According to Nebius executives, the temporary generation was wound down faster than initially expected.

The project has also faced regulatory and local concerns. New Jersey regulators fined the facility $1.1 million over violations involving the operation of generators, while nearby residents have raised noise complaints through legal proceedings.

These developments illustrate how power availability, permitting and local opposition can complicate the construction of AI data centres.

Nebius said Microsoft had been involved in the planning process for the revised delivery schedule. Microsoft had not publicly announced the cancellation of the agreement at the time of the reports.

Why Microsoft and Other Hyperscalers Are Becoming More Cautious

The New Jersey delays come amid reports that Microsoft is increasing oversight of infrastructure projects operated by third-party AI cloud providers.

Large technology companies are investing heavily in computing capacity, but they depend on external partners to deliver data centres, electricity connections and specialised equipment on schedule.

If infrastructure is delayed, customers may have to postpone AI product launches or find alternative computing resources.

Microsoft has reportedly been reassessing some third-party infrastructure arrangements, including projects involving Nebius, IREN and Nscale.

The issue is important for neocloud providers because large contracts often underpin their financing arrangements.

A long-term agreement with a major technology company can help a provider secure loans and raise money for construction.

However, missed delivery deadlines can create financial penalties and, under certain contractual conditions, termination rights.

Nebius has said that delays have primarily affected third-party sites it does not directly control. It maintains that data centres it is building itself are progressing on schedule or ahead of schedule.

The company expects its owned facilities to provide most of its additional capacity from 2027 onwards.

Nebius Reports 571% Revenue Growth in Q3

Despite infrastructure delays, Nebius reported substantial revenue growth during the September quarter.

Revenue reached $981 million, an increase of 571% from the corresponding period last year.

Its Nebius AI Cloud business generated approximately $974 million, accounting for nearly 99% of group revenue.

The company also reported adjusted EBITDA of $465.8 million, compared with an adjusted EBITDA loss of $5.2 million in Q3 FY25.

Financial indicator	Q3 2025	Q3 2026
Total revenue	—	$981 million
Year-on-year revenue growth	—	571%
Adjusted EBITDA	-$5.2 million	$465.8 million
AI cloud adjusted EBITDA	—	$518.2 million
Cash and cash equivalents	—	$7.4 billion
Quarterly capital expenditure	—	$7.3 billion

The company also reported approximately $38.7 billion in remaining performance obligations at the end of September, representing contracted future revenue that has not yet been recognised.

However, a portion of the reported quarterly revenue was linked to a one-time payment associated with a new partnership model. This means investors should distinguish between recurring cloud revenue and revenue arising from specific contractual arrangements.

Nebius Lowers Its 2026 Revenue Expectations

While quarterly revenue exceeded analyst expectations, Nebius reduced its expectations for full-year 2026 revenue.

In its shareholder letter, Nebius said 2026 revenue would be near the low end of its $3 billion to $3.4 billion range. On the earnings call, CFO Dado Alonso narrowed guidance to $3 billion to $3.2 billion. Both statements are forecasts, not reported full-year results.

The revision reflects delays in bringing third-party computing capacity online.

The company also expects to reach its target of more than $7 billion in annualised run-rate revenue approximately two months later than originally planned.

The change illustrates the relationship between infrastructure availability and revenue growth in the AI cloud business.

Even when customers have signed contracts, a provider cannot fully recognise the associated revenue until the required computing capacity becomes operational and services are delivered.

Nebius Forecasts $12 Billion to $13 Billion Revenue in 2027

Despite the near-term delays, Nebius expects its business to expand substantially next year.

The company has forecast revenue of $12 billion to $13 billion for 2027 and expects to reach pre-tax break-even by the fourth quarter.

The forecast depends on several factors, including:

• Bringing new data centre capacity online.
• Delivering contracted computing services on schedule.
• Expanding relationships with major AI customers.
• Maintaining strong demand for GPU infrastructure.
• Managing construction and financing costs.

Nebius has also reported growing interest from AI developers and enterprise customers seeking access to advanced computing resources.

The company said customers had committed more than $12 billion in contract value since the beginning of the third quarter.

Its customer base includes major technology companies, AI laboratories and software developers requiring dedicated computing infrastructure.

What Is a Neocloud and Why Is the Market Growing?

A neocloud is a specialised cloud computing provider focused primarily on supplying GPU infrastructure and computing services for artificial intelligence workloads.

Traditional hyperscalers such as Amazon Web Services, Microsoft Azure and Google Cloud offer broad cloud platforms covering storage, databases, networking, software and computing.

Neocloud companies concentrate more heavily on the specialised infrastructure required for AI model training, inference and other compute-intensive workloads.

Companies such as Nebius and <Entity value=”CoreWeave” category=”company”/> are competing to provide access to high-performance GPUs and dedicated AI computing environments.

Their growth is linked to increasing demand for AI applications, larger models and agentic systems that require substantial computing resources.

However, the business requires significant upfront investment. Data centres need land, electricity, cooling infrastructure, networking equipment and expensive processors before they can generate meaningful revenue.

This creates a challenging financial model in which companies must secure customer commitments and funding while managing construction timelines.

The Bigger Picture

Nebius’s planned $40 billion to $45 billion investment programme demonstrates the scale of infrastructure spending required to support the global AI industry. The company is attempting to finance much of its expansion through customer prepayments and loans backed by long-term contracts, reducing its dependence on traditional equity funding. Its rapid revenue growth and improving adjusted EBITDA show the commercial opportunity in specialised AI cloud infrastructure.

However, the delays at the New Jersey facility reveal the risks behind this strategy. AI data centre construction depends on electricity, permits, equipment availability and third-party execution, all of which can affect delivery schedules. If capacity is delayed, revenue recognition can slip even when customer demand remains strong. Nebius’s ability to deliver projects on time, maintain customer confidence and manage its large capital commitments will determine whether its aggressive expansion plan translates into sustainable growth.

Looking Ahead

Nebius will focus on bringing delayed third-party capacity online while expanding its own data centre infrastructure across the United States and Europe. Investors will monitor progress on the Microsoft agreement, the pace of customer prepayments and whether the company can keep its construction programme within its financial targets. The projected 75% to 85% financing coverage for 2027 capital expenditure will also be an important measure of how effectively the company can fund its expansion.

The broader neocloud market is likely to remain closely linked to the global AI investment cycle. Demand for computing capacity could create significant opportunities, but the companies supplying that capacity must manage enormous infrastructure costs and execution risks. Nebius’s 2027 revenue and profitability targets provide an ambitious benchmark, and the coming year will test whether its financing model can support growth at this scale.

Correction, November 13, 2026: The earlier guidance range was $3 billion to $3.4 billion, not $3 billion to $3.2 billion. The latter is the narrowed range given on the earnings call. Sources: Nebius shareholder letter filed with the SEC and the November 12 earnings-call transcript.