Nebius Eyes 50MW at Saudi NEOM AI Campus as First Customer

HUMAIN, the Saudi PIF company, and the data-centre developer DataVolt announced on 21 October 2026 a preliminary framework agreement that would make the AI cloud company Nebius an anchor customer at their planned AI campus in Oxagon, NEOM. According to the release, Nebius would take “up to 50MW (half of the initial 100 MW currently under construction as part of the first phase’s planned 360 MW.)”

The release says the capacity “is expected to become operational in phases from Q4 2027” and that Nebius is “expected to deploy NVIDIA-powered computing infrastructure under the agreement.” HUMAIN CEO Tareq Amin called Nebius “our first customer.” The agreement was “signed on the sidelines of NVIDIA’s GTC Berlin conference,” the release says.

Business Pill · LEAD TIME

A short Business Engineer explainer on lead time: why the wait for capacity is part of its price. It teaches the general idea only and says nothing about HUMAIN, DataVolt, NEOM or Nebius.

NVIDIA GTC Berlin keynote

NVIDIA’s official stream of the GTC Berlin keynote, 21 October 2026. The release says the Nebius agreement was signed on the sidelines of the conference; the framework is not part of the keynote.

What the Framework Covers

The release describes a proposal, not a final contract. It speaks of a “preliminary framework agreement,” a “proposed framework” and a “contemplated offtake,” and says the framework establishes “an initial commercial framework for the project.” In plain terms, an offtake is capacity a customer commits to use once it is built; here the amount is a ceiling, “up to 50MW.”

The 31 August release calls it DataVolt’s planned AI data center campus. That release said construction had begun on 100MW of a 360MW first phase, and that “the groundbreaking marks the start of construction on the 360MW first phase of its wider planned 1.5GW Oxagon campus.” It said the facility “is expected to integrate renewable energy sources, innovative cooling technologies, and high-performance computing clusters.” The 21 October release says the Nebius framework comes “just six weeks after announcing the commencement of construction.”

What Was Posted on X

HUMAIN’s own post announces the framework: up to 50MW of planned capacity for Nebius within the initial 100MW under construction at Oxagon, with phased operations expected to begin in Q4 2027, announced on the sidelines of GTC Berlin. HUMAIN calls it a preliminary framework; this piece takes its figures from the companies’ release.

What the Companies Said

Amin said in the release: “Advancing our commercial relationship with Nebius as our first customer is a significant milestone for HUMAIN and a clear demonstration of the international demand for the infrastructure we are developing.” DataVolt CEO Rajit Nanda said: “Speed to market and infrastructure cost competitiveness will define the winners in the global AI compute race.”

For Nebius, the release says, the agreement “represents an expansion of the infrastructure capacity available to support the continued growth of its AI cloud business.” HUMAIN’s post on X described “up to 50MW of planned capacity for Nebius within the initial 100MW currently under construction, with phased operations expected to begin in Q4 2027.”

Connecting the Coverage

The framework lands on the same day Nebius’s chief revenue officer told CNBC that compute prices were rising. Our Nebius spot-price piece carries a 21 October update in which Nebius chief revenue officer Marc Boroditsky told CNBC that if Nebius brought its 2027 capacity online it would sell out, and that its auction showed “continued upward mobility in terms of the kind of price per megawatt.”

As we read it, the NEOM framework is Nebius lining up more of that future capacity, with the proposed Nebius allocation expected to become operational in phases from Q4 2027.

Our neocloud shakeout piece reported, citing The Information, that executives expect only seven to 10 of hundreds of neoclouds to survive, with Nvidia supporting them through easier credit terms, site selection and leases. Our GTC Berlin wrap covers the conference where this agreement was signed.

The Structural Read

As we read it, the framework tests whether an AI cloud will commit to a campus that is still being built: the release calls it a “preliminary framework agreement” with a “contemplated offtake,” signed “just six weeks after announcing the commencement of construction.”

As we read it, the pitch is energy and speed: the release cites “the local competitive energy and time-to-market advantages available in NEOM’s Oxagon,” and the 31 August release said the facility is expected to integrate “renewable energy sources.”

As we read it, the stakes for HUMAIN are proof of demand: Amin called Nebius “our first customer” and “a clear demonstration of the international demand for the infrastructure we are developing.”

Three Implications

FOR AI CLOUD PROVIDERS The release says Nebius would get capacity to “support the continued growth of its AI cloud business,” with the allocation expected “in phases from Q4 2027.”

FOR GULF AI INFRASTRUCTURE The release describes the framework as a milestone that demonstrates “the potential to attract international AI customers and deploy capacity at significant scale.”

FOR NVIDIA The release says Nebius is “expected to deploy NVIDIA-powered computing infrastructure under the agreement.”

The Business Engineer Lens

This story maps onto the Business Engineer framework The State of AI Data Centers.

The framework argues that “The structural bottleneck isn’t computing, it’s electrons. Companies that solve the power equation will dominate the AI era; those that don’t will watch their chips sit idle.”

As we read it, the NEOM release pitches exactly that power equation, citing “the local competitive energy and time-to-market advantages available in NEOM’s Oxagon.” The framework was published in December 2025 and does not discuss NEOM, HUMAIN or Nebius.

What Is Not Established

The release gives no price, contract length, GPU count or investment figure for the Nebius capacity, and calls the agreement preliminary. We did not find a Nebius statement on the framework in the Nebius posts on X we read for 21 October. The 1.5GW campus figure comes from the 31 August release and describes a planned total, not capacity under construction. That release also said “the first 100MW anticipated to be available in 2028,” while the 21 October release dates the Nebius allocation “in phases from Q4 2027”; the releases do not explain the difference.

The Bottom Line

HUMAIN and DataVolt say they have a first international customer for their NEOM campus: Nebius, for up to 50MW from Q4 2027, under a preliminary framework signed at GTC Berlin. As we read it, the release’s own pitch is power and speed, and the terms that would make it a commitment are not yet public.