Nebius falls 10% as company expects increased 2026 capex spend, issues weak outlook

Nebius delivered the kind of quarter that is supposed to end an earnings sell-off: revenue of $980.8 million, up 571% year over year, beat the roughly $907.5 million consensus; adjusted EBITDA reached $465.8 million; and adjusted net profit was $90.9 million against an expected $271.4 million loss.

The reason was not the quarter. It was the price of the next one. Nebius said full-year 2026 revenue would land at the low end of its $3 billion-$3.4 billion range, below the roughly $3.35 billion-$3.37 billion analysts expected. It also raised 2026 capital expenditure guidance to $27 billion, above the prior outlook and the $23.85 billion consensus, then forecast $40 billion-$45 billion of capital spending for 2027. The shares were down about 10% in the reported trading, with one market snapshot showing a steeper 14% decline at $133.67 after a 15% gain the previous day.

That combination changes the central question for Nebius. The company is not struggling to sell AI infrastructure; it is struggling to make the financing of that infrastructure legible. Its first nine months of capital spending totaled $15.43 billion, while total assets expanded from $12.43 billion at the start of the year to $38.74 billion. Cash stood at $7.43 billion, and non-current liabilities reached $14 billion. A business whose revenue is growing at triple-digit rates can still look fragile if investors cannot tell which future contracts will pay for which data centers.

Nebius is trying to answer that question with customer money rather than only debt or equity. CEO Arkady Volozh said new contracts signed in the third quarter carried longer terms, higher prices and larger upfront payments; advances covered more than 100% of the capital expenditure associated with those contracts, up from 50%-60% in the second quarter. Contract liabilities rose by $7.83 billion during the first nine months. The company expects advances and investment-grade contract-asset-backed financing to cover 75%-85% of 2027 capital expenditure, with advance-payment volume rising from about 10 billion yuan in 2026 to 17 billion-20 billion yuan in 2027.

That is a meaningful financing mechanism, not a cosmetic reassurance. If it works, the $40 billion-$45 billion plan is less a bet on Nebius’s own balance sheet than a bet that a widening roster of customers will prepay for capacity. Databricks, Datadog, Palantir and Snowflake signed a first multi-year agreement, according to the shareholder letter. Longer contracts and higher prices also provide a cushion against the capacity delays that pushed 2026 revenue toward the bottom of guidance.

But the accounting still contains a warning. Nebius reported a GAAP net loss from continuing operations of $37.5 million, or 13 cents per diluted share, despite positive operating profit of $50.4 million. Interest expense of $112.5 million and a one-time $49.5 million fee tied to convertible-bond conversion drove the loss. The company’s $393.3 million nine-month net profit included an $780.6 billion equity-investment revaluation gain; excluding that gain, the underlying result remained a loss. The operating business is improving, but reported earnings are still highly sensitive to financing costs and asset marks.

The market’s reaction also reflects a crowded trade. Nebius had rallied 15% on November 11, helped by strong results from fellow neocloud operators IREN and CoreWeave, and then gave back more than that after its own report. The sell-off spread across the group, suggesting investors are reassessing the sector’s spending appetite rather than singling out one company’s execution. Political resistance to data-center development and uncertainty surrounding Nvidia’s leadership and financial ties to Nebius add further risk, although Volozh said the U.S. election outcome was not expected to affect construction of new and existing sites.

For Nebius, the decisive proof will not be another revenue beat. It will be whether customers keep signing contracts long enough, and at prices high enough, to fund the build before the cash arrives. The company has shown it can sell capacity faster than it can deliver it; now it must show it can finance the gap without giving away the economics of the boom.