Skip to content

Nebius bets on $40 billion spend via pre-payment: power plant failures expose fragility of alternative AI draw

Share
Nebius bets on $40 billion spend via pre-payment: power plant failures expose fragility of alternative AI draw

Listen to this article

Read by Anchor

Nebius, a cloud infrastructure specialist, disclosed an ambitious capital-spending plan of up to $40 billion for 2027, exceeding Wall Street forecasts, but the expansion relies on a high-risk financing mechanism that depends on upfront payments from customers and loans backed by future compute contracts.Building ultra-high-performance compute capacity is no longer just a chip shortage issue; it has become a harsh battle with power grids and local operating licences.This model, which tries to fund data centres at their virtual value before they are completed, has already shown immediate cracks, reflected in a 17 percent drop in the company’s share price after it cut its 2026 revenue outlook by $200 million to $3 billion.

The operational crisis is evident at Nebius’s data centre in Fineland, New Jersey, where Nebius is tied to an AI-compute contract with Microsoft worth $19.4 billion, but the project has slipped several months behind schedule. The company will be able to deliver only slightly more than half of the agreed capacity by the end of the current year, with the remaining half postponed to the second quarter of 2027. The setback stems from the site’s external operator, Data One, which rushed to cut reliance on temporary generation units faster than expected in preparation for a switch to Bloom Energy fuel cells, after regulators imposed a $1.1 million fine for violating federal law by running 45 natural-gas generators without proper licences, alongside lawsuits from residents over continuous noise complaints from the facility.

This timing slip puts alternative-compute contracts to a real test, as the agreement clauses give Microsoft the right to terminate the affected service after a defined grace period if alternative capacity cannot be provided, and impose financial penalties for delays. It coincides with Microsoft tightening its oversight of external projects run by firms such as Nebius, Aerin and In Scale, and dispatching inspection teams to the New Jersey site after its financial reports showed capital spending falling short of prior estimates due to external developer setbacks, prompting it to back away from a double-up agreement in Texas and withdraw as a major client from a UK site because of power-supply disruptions.

Although Nebius reported $981 million in revenue for the third quarter, a year-over-year increase of 571 percent supported by an exceptional cash infusion from a new partnership model and cash on hand of $7.4 billion, CEO Arkadi Volos acknowledged that delays have become an inevitable part of the industry.The bet on upfront cash flows to cover 85 percent of costs collapses if on-ground operational capacity freezes.The finance team confirms that pre-payments from customers such as Microsoft and Meta fund the build-out of new capacity, relying on wholly owned data centres to generate the bulk of 2027 capability away from external operator hiccups.

This setback sends a direct message to technology leaders and digital procurement heads in the Gulf, Egypt and the Arab Levant: relying on cloud-compute reservations with new-cloud providers without vetting power-infrastructure and generation licences carries the risk of costly operational outages. Organizations in the region seeking to secure processing capacity for large models must scrutinise contract terms, delivery schedules and compensation clauses carefully, recognising that server capacity does not become real unless energy sources and feed-networks are licensed and ready for actual use.

Don't miss the next story

Subscribe for updates