Nebius Stock: A 9.3% Drop, a Shelved Junk Bond, and the $20–25 Billion Question

The sell-off in Nebius shares this week had little to do with anything Nebius announced. That is precisely what makes it worth examining. On Friday, disappointing quarterly results from Amazon and Intel triggered a broad technology rout, Reuters reported, and Nebius was swept up in the current. On Monday the stock gave up another 9.3%, extending a 30-day decline to -38%. Neither move stemmed from a company disclosure — a distinction that matters, because a weaker sector mood is not by itself proof of a weaker business.

Yet the connection cannot be waved away. If investors are starting to question the economics of AI infrastructure more fundamentally, that doubt lands on Nebius's valuation too. The real dividing line runs between two separate claims: the market is skeptical of the investment boom's profitability, and whether that skepticism is warranted at Nebius must be settled by the company's own numbers. A falling share price proves neither the case for nor against the business model — the same caution applies to a rising one.

What Dolphin Research Actually Flags

A fresh analysis from Dolphin Research, published Monday, sharpens the debate. Its core message: Nebius must deliver more than rapid growth to justify its premium. A premium can be earned — but only through convincing advantages in growth, profitability, or the reliability of its expansion. It is on that last point the research raises a caution flag, noting that a meaningful portion of projected revenue is not yet contractually secured. That is not a forecast of collapsing demand; uncontracted revenue can still be won. For valuation purposes, however, the gap between booked business and business that must still be created is what counts — and it argues against explaining the premium on general AI compute demand alone.

The optimistic counterweight comes from Truist, which considers Nebius's valuation at 4.5 times expected 2028 revenue justified, citing anticipated industry-leading growth from 2026 to 2028. That view names its own condition: the exceptional growth has to materialize. A valuation built on later-year revenue offers thin protection if the buildout or utilization falls short.

Scale of the Bet, in Both Directions

The operating momentum is substantial. According to Dolphin Research, Nebius AI Cloud generated $575 million in revenue in the second quarter of 2026 — 98% of group revenue, which means the company's entire assessment now rests on this one segment. Management has confirmed a target annualized revenue run rate of $7–9 billion by the end of 2026, against expected capital expenditures of $20–25 billion for the year. A run rate is not realized annual revenue, and it cannot be netted directly against capex. The magnitudes still show why growth here should not be viewed in isolation: the buildout consumes enormous capital long before its economic success can be fully judged.

Financing has become its own storyline. Roughly two weeks ago, Nebius arranged a $3.5 billion senior secured credit facility with Citi, carrying a 364-day term, earmarked for data centers, GPU procurement, and general corporate purposes. Media reports also said the company shelved a planned high-yield bond issue, with the collapse of the financing market specifically for neocloud providers cited as background. That reads as a warning about the fundability of expansion, not proof of its failure — but for a capital-intensive buildout, access to financing deserves as much attention as the revenue forecast.

Concrete Steel, Concrete Deadlines

The expansion is not abstract. In Béthune, France, Azur Datacenter secured an additional EUR 250 million for a project being developed for Nebius, according to media reports dated October 26. The first phase is scheduled to open in December 2026 — a tangible project standing against the broader argument over AI spending, though its opening remains an expectation rather than a fact.

On the commercial side, Nebius introduced public real-time spot pricing for GPU capacity on October 21 and expanded its European AI developer program together with NVIDIA. The same announcement cited customer-reported development-time reductions of up to 40%. That is a different lens on the AI boom: not just the technology supplied, but its usefulness in daily work. Customer time savings and provider revenue are separate quantities, and who pays for which benefit — and how much of it reaches the provider — matters more than the blanket question of whether AI has become too expensive. A day earlier, Nebius announced a Europe-wide agreement with TD SYNNEX giving sales partners access to the Nebius AI Cloud for AI training and inference, including reserved capacity, predictable pricing, and technical and commercial support. It points to broader distribution, but no specific revenue volume or earnings contribution can be derived from it.

The November 12 Scorecard

Nebius will report third-quarter 2026 results before market open on November 12, 2026 — the next company-specific test of the returns debate. Until then, weakness among other tech giants should not be read as a verdict on Nebius, nor should secured financing be taken as proof of future earnings. The stock sits squarely in the middle of the AI investment cycle, and the economic result of that cycle is what will ultimately have to convince.