Nebius Stock: A Rotterdam Megafab, a Prearranged Share Sale and the Capital Question in Between

Nebius signed a letter of intent for up to 200 MW at a planned Rotterdam AI campus, while its shares fell 5.6% amid valuation and data center cost worries.

Nebius has lined up a potential anchor role at a sprawling AI campus on Rotterdam's Maasvlakte — and simultaneously found itself on the receiving end of a 5.6% single-day share price decline. The two developments point in opposite directions, yet they share a common thread: the gap between announced ambition and the capital required to make it real.

The company's shares came under pressure Thursday, weighed down by valuation concerns and the mounting cost of data center expansion, according to media reports. Worries about a stock sale by COO Ophir Nave, alongside a broadly more cautious market mood, added to the drag. That combination matters because it frames how investors are currently reading Nebius: not as a growth story to be taken on faith, but as a balance sheet and execution question that still needs answering.

A Letter of Intent Is Not a Financing

Under plans reported by Financieele Dagblad, Nebius would become the first customer of an AI factory that VOLT intends to build on the Maasvlakte, with the company having signed a letter of intent to take a substantial share of the planned compute capacity over the long term. Up to 200 MW is earmarked for Nebius, per media reports, with the first capacity expected at the end of 2029 or the start of 2030. Nothing is finalized. Should delays arise, Nebius can walk away under the reported terms.

The first construction phase carries a price tag of EUR 7 billion, split between EUR 2.5 billion for buildings and infrastructure and EUR 4.5 billion for chips and equipment. That breakdown is instructive: this is not merely a real estate project. The bulk of the capital sits in the technical fit-out.

An anchor tenant can ease financing. It does not replace it. A letter of intent documents interest in future capacity — it is not proof that funding, construction and delivery will line up. For investors, the timeline is the crux: the Rotterdam announcement describes a possible extension of future capacity, and its economic value depends entirely on whether plans turn into an operating site. The termination right caps Nebius's exposure to delays, but it does not remove the project's uncertainty.

The Capital Side of Growth

Nebius is already spending heavily. Adjusted EBITDA reached USD 236.2 million. Operating progress, however, does not settle the financing question. At the end of June, the company carried roughly USD 8.55 billion in debt against USD 8.04 billion in liquid funds, with restricted cash on top. Those figures are not a complete funding calculation for the expansion pipeline, but they explain why investors are watching the capital structure alongside demand.

The debate extends well beyond Nebius. CNBC reported Thursday that the CEOs of Intel and Broadcom see access to capital and financing costs as a possible central bottleneck for AI build-out, and that companies must additionally demonstrate returns on their investments. Broadcom's Hock Tan continues to expect demand for compute to outstrip supply, while Intel's Lip-Bu Tan does not anticipate a marked slowdown in the near term. Their caution, then, is not aimed at AI demand itself — it concerns the economic conditions under which that demand can be served.

The Sale That Wasn't a Signal

Against that backdrop, the insider transaction deserves a closer look than the headline reaction suggested. An SEC filing dated October 6 documents Nave's sale of 500,000 Class A shares. The transaction took place on October 5 and amounted to roughly 17% of his allocated equity stake — not a trivial sum. The framework, though, is just as important: Nave acted under a Rule 10b5-1 plan adopted on May 22, and according to the filing no further sales will occur under that plan. Treating the transaction as a spontaneous reaction to recent business developments would therefore be a misreading.

A prearranged plan does not automatically render an insider sale meaningless, but it does limit what can be inferred about the motivation behind the timing. No deterioration in the business, and no negative assessment of individual products, can be derived from the transaction alone. Those who see a sweeping warning signal in it are giving the sale more weight than it carries on its own.

Where the Real Test Lies

On the operational side, Nebius released a product update on October 1 covering changes to its cloud platform and AI cloud, including private cloud interconnect and additional data transfer options. The more meaningful yardstick is whether such extensions hold up in practice. Reporting from October 3 underscores that point: Nebius had resolved an InfiniBand disruption affecting GPU clusters in the eu-west2 region, while a separate issue with interruptible nodes in Managed Kubernetes remained open as of October 4, according to the report. Running workloads were reportedly unaffected — an important caveat, since no widespread outage can be inferred from the disclosure. Even so, concrete operational disruptions are a more relevant checkpoint than speculation about a share seller's intentions, because they bear directly on the technical performance of the offering.

Roughly two weeks ago, Nebius also acquired Inferize, whose inference optimization technology and development team complement Nebius Token Factory. That reads as a coherent expansion of technical capabilities — not yet proof of commercial success.

Weighing it all up, a blanket negative reading of the insider sale does not hold up, and product updates alone do not justify an uncritical positive stance either. What should tip the scales for Nebius is operational delivery: technical progress, reliably running services and careful handling of disruptions. The Rotterdam megafab widens the long-term horizon but delivers no completed financing and construction process. The decisive test, then, is not the next large capacity announcement — it is the demonstrable transition from planned sites to funded, timely and economically viable data centers. The share sale is a signal worth noting, but not a reliable verdict on the company.