When extraordinary growth is not enough
Nebius Group's November 12, 2026, earnings report put a sharp contradiction at the center of the AI cloud investment case. The company reported third-quarter revenue of $980.8 million, an increase of 571% year over year, yet its shares sold off heavily as management acknowledged delays in getting some third-party data-center capacity online and gave an even larger capital-spending outlook for 2027. In capital-intensive infrastructure, demand growth and shareholder returns are not the same variable. Customers can want more computing capacity while financing costs, delivery schedules and dilution concerns weigh on the equity.
The information cutoff is November 12, 2026, at 7:15 p.m. ET; the latest complete U.S. stock-market cash session was Thursday, November 12. Nebius released its unaudited financial results that morning for the three and nine months ended September 30. Reuters reported a decline of roughly 16% in Nebius shares during Thursday's session. That is a reported cash-session move, not an extended-hours price extrapolation. The release is a new earnings event, not a replay of earlier announcements about AI partnerships or the prior day's discussion of another cloud company's results.
What Nebius actually earned
Nebius reported consolidated revenue of $980.8 million in the third quarter, up from $146.1 million in the comparable quarter of 2025. Group adjusted EBITDA was $465.8 million, compared with an adjusted EBITDA loss of $5.2 million a year ago. Continuing operations recorded a GAAP net loss of $37.5 million, an improvement from a $119.6 million loss the year before. The company also reported positive adjusted net income of $90.9 million. These measures are not interchangeable: adjusted EBITDA excludes costs that remain economically important to a business building expensive infrastructure.
The figures include the core Nebius AI cloud business alongside other consolidated activities, including autonomous-vehicle and education businesses. Consequently, total revenue cannot be treated as a pure one-to-one measure of GPU rental revenue. The company and its management have also described asset-light partner arrangements that can create revenue at different stages of deployment. Investors must distinguish recurring capacity utilization from upfront or nonrecurring contributions when estimating steady-state margins.
The new guidance exposed a timing problem
The biggest negative surprise was not the prior quarter's sales; it was the timetable for future capacity. Management said some third-party facilities expected to come online in 2026 had shifted into the first half of 2027, while the underlying contracts remained in place. It expects 2026 sales of approximately $3.0 billion to $3.2 billion, compared with its earlier $3.0 billion to $3.4 billion range. That change trims the upper end rather than erasing all of the forecast revenue. Nevertheless, for an equity priced on rapid capacity ramp-up, the difference between revenue recognized in December and revenue recognized in a later quarter can be material.
Management also said a December annualized revenue milestone was now expected in February. This is a timing expectation, not an event that had already occurred by November 12. The company emphasized that its own facilities under development were on schedule or ahead, with delays concentrated at certain third-party sites. The distinction is relevant because partner-controlled construction projects can expose an operator to execution risks it cannot directly resolve. Contracts may survive, yet the carrying cost of capital can continue while revenue is postponed.
The scale of the capital requirement
Nebius increased expected 2026 capital expenditure to around $27 billion from an earlier $20 billion to $25 billion plan, according to Reuters' coverage of management's update. Its initial 2027 outlook calls for $40 billion to $45 billion in capital expenditures and $12 billion to $13 billion in revenue. The company expects to reach pre-tax break-even by the end of 2027, a forward-looking objective rather than demonstrated profitability. There is a striking asymmetry between the level of planned investment and the revenue projected for a single year; that does not by itself prove the spending is irrational, because assets may generate returns over multiple years.
Management says customer prepayments and financing secured against contracts with large customers could fund roughly 75% to 85% of next year's spending. That is an important potential protection against shareholder dilution, but it should not be misunderstood as cash already collected for every planned asset. Funding availability, borrowing terms, asset collateral and project completion all matter. Leverage may help shareholders if capacity fills quickly and contracted pricing holds; it can also magnify losses if commissioning falls behind or equipment becomes obsolete faster than assumed.
Why the stock reaction matters beyond Nebius
Nebius is part of a category often called neoclouds: specialized providers building AI computing clusters and selling capacity to other businesses. Their model helps hyperscalers and software developers obtain computing power without constructing every facility themselves. It also raises a difficult question of financial architecture. Heavy capital expenditure arrives early, while contracted revenue and operating cash flows arrive over time. Large customers can reduce demand uncertainty, but customers' bargaining power, dependence on chip suppliers and concentration remain genuine risks.
The broader market was also navigating Nvidia's officially confirmed November 12 CEO transition. That created additional uncertainty around the strategic relationships and sentiment that support AI infrastructure. But a simultaneous selloff does not establish that Nvidia's management change caused Nebius' capacity delay. The third-party construction timing was an operational disclosure in Nebius' own earnings materials. Keeping company-specific execution separate from sector-wide sentiment is essential for any responsible analysis.
Debate: durable demand versus an expensive financing bridge
The bullish argument starts with the extraordinary revenue growth, improved adjusted operating performance and management's comments about customers committing to longer-duration agreements at higher prices. Reuters reported that Nebius expects demand for capacity well beyond the present year. A large contracted pipeline can support asset-backed funding and provide visibility unavailable to a purely speculative infrastructure buildout. If facilities open and achieve strong utilization, capital-intensive projects may produce attractive operating cash flows over their lives.
The bearish case focuses on the distance between accounting earnings and the cash needed to build the network. Financing commitments are not necessarily unconditional, rising debt raises interest expense, and a construction delay can push back several years of projected returns. Investors also face potential equity dilution, power availability constraints, contractor performance risk, export controls and competition. Adjusted EBITDA can look strong while future depreciation and financing obligations remain large. It is therefore hazardous to value the stock simply by applying a software-company revenue multiple to its reported sales growth.
What to watch after November 12
The next useful operational milestones are management's expected February 2027 annualized revenue level and the rollout of third-party capacity now shifted into the first half of 2027. These are guidance periods, not confirmed dates of completion. Track recognized revenue against the revised $3.0 billion–$3.2 billion 2026 range, the development of free cash flow, interest costs, customer prepayments and the balance of secured funding versus common-equity issuance. The company's $12 billion–$13 billion 2027 revenue projection and $40 billion–$45 billion capital-spending expectation should be revised as actual facilities come online. At the cutoff, no exact date for the next quarterly results announcement had been verified.
Conclusion: capacity is valuable only when it becomes cash flow
Nebius demonstrated that AI cloud demand can generate spectacular year-over-year revenue growth. It also demonstrated why market confidence remains fragile when the business requires immense upfront investment. Capacity delays, even on contracts that remain intact, change the timing of cash receipts and can increase financing pressure. The key test is not whether management can repeat that demand is strong; it is whether commissioned infrastructure, recurring revenue and capital costs converge into sustainable returns. November 12 supplied both the evidence of scale and the warning that execution timing matters.