Nebius shares sink 12% as customer-funded AI buildout faces a test

Nebius shares slid further on Friday afternoon, turning a difficult week for companies building artificial-intelligence data centers into a direct test of the company’s funding model. At 2:19 p.m. EDT on October 16, TECHi’s Yahoo Finance-sourced quote put NBIS at $166.00, down $22.41, or 11.89%, from Thursday’s $188.41 close. The source did not state the exact delay; this is a dated market observation, not a closing price. Morning coverage had the stock down about 5.9% at $177.29 around 10:30 a.m. EDT. The selling had plainly intensified by early afternoon, even though we found no new Nebius earnings release or contract announcement behind that later leg down.

That distinction matters. A falling share price is a reportable event; it does not, by itself, reveal which holder sold or why. The broader AI-infrastructure trade was weak, and a Gartner warning published Wednesday has put renewed attention on whether specialized cloud providers can earn enough from costly hardware. But the sharper question for Nebius is more specific: how much of its construction bill is truly funded by customers, and how much still has to come from its own balance sheet or investors?

The answer is not visible in a single stock chart. It requires separating revenue, cash received in advance, capital spending and financing. Nebius’s June-quarter results and interim report provide a useful starting point. They show a business gaining customers rapidly, but also one buying infrastructure far faster than current operating cash can pay for it. Friday’s price move does not prove that strategy has failed. It shows how little room investors are leaving for delays.

A market move, not a fresh company disclosure

By the 2:19 p.m. quote, Nebius had given up nearly one dollar in eight of Thursday’s closing value. The intraday observation was still changing as the U.S. session continued, so any percentage in this article is a dated snapshot rather than a prediction of the closing move. The source showed more than 23 million shares traded at that snapshot, versus an average of about 17 million across 30 completed sessions.

Earlier Friday reports linked the decline to sector pressure, including the Gartner forecast and a reassessment of AI data-center spending. That is a plausible context, not proof that Gartner alone caused Nebius’s afternoon losses. CoreWeave, IREN and Applied Digital were also down heavily in a mid-afternoon market snapshot. Those companies differ in contracts, financing and assets, and a common selloff should not erase those differences. The useful comparison is the question investors are asking each company: when will contracted demand become dependable cash after power, GPUs, construction and financing are paid for?

Gartner’s October 14 release warned that more than $3.5 trillion of AI-infrastructure spending could be exposed to impairment by 2030. It also forecast that more than 80% of neocloud providers could cease operating independently by then. Those are industry forecasts, not a finding that Nebius has impaired an asset or will cease operating. Applying the headline percentage to one company as though it were a measured probability would be misleading. Nebius’s own contract terms and cash conversion are more informative.

The $3.41 billion gap inside a fast-growing quarter

Nebius reported $582.3 million of group revenue for the three months ended June 30, 2026, up 454% from the comparable quarter a year earlier. Its non-GAAP adjusted EBITDA was $236.2 million. Those are material improvements. Yet the same earnings release also reported a $190.4 million loss from continuing operations. A healthy adjusted measure and a statutory loss can coexist because they include different costs; neither figure alone answers whether building the next wave of capacity is self-funding.

The cash-flow table is more revealing. Continuing operations provided $2.2461 billion of cash in the June quarter. Purchases of property, equipment and intangible assets used $5.6574 billion. Subtracting the former from the latter leaves approximately $3.4113 billion that those two cash-flow lines did not cover. This is TECHi’s arithmetic using Nebius’s reported figures, not a company-defined free-cash-flow measure. It excludes acquisitions, other investing cash flows, financing inflows and movements in the cash balance. It should be read as a narrow measure of how far operating cash fell short of infrastructure purchases during that quarter.

The same calculation over the first six months is slightly larger: $8.1303 billion of property, equipment and intangible purchases less $4.5041 billion of operating cash flow equals about $3.6262 billion. That helps explain why a business can post dramatic sales growth, carry substantial cash and still be sensitive to funding conditions. Large server orders and data-center construction are paid before all the associated service revenue is recognized.

Those figures need one more adjustment in the reader’s mind. Nebius said the six-month rise in operating cash was driven primarily by customer advances. Its interim report shows deferred revenue increased by $4.395 billion during the period. Deferred revenue is an obligation to deliver services later, not revenue already earned. The increase is almost as large as the $4.5041 billion of six-month operating cash flow, although other cash movements also affected that total. Calling the full operating-cash number recurring cash profit would overstate what the filing establishes.

Why advance payments are both a strength and a test

Customer prepayments are a valuable feature of Nebius’s model. A cloud client that commits cash before capacity comes online can reduce the amount Nebius must raise in debt or equity. Management’s August shareholder letter said roughly 70% of deals closed in the second quarter included prepayments and that these could cover 50% to 60% of the associated capital expenditure. That was management’s description of deals signed in that period, not a promise that every future project will carry the same terms.

The distinction is central to the investment case. A dollar collected upfront improves cash today, but Nebius then has to build and operate the promised capacity. The benefit to shareholders depends on the remaining construction cost, the timing of delivery, the duration and pricing of the contract, hardware depreciation, electricity and support costs, and whether the customer renews. A customer-funded build can be attractive; it is not the same thing as receiving an unconditional grant.

The strength of Nebius’s proposition is that it can negotiate some of those terms before committing all the hardware expense. The vulnerability is that the expense arrives in large, concentrated chunks. If a site’s power connection, construction schedule or GPU delivery slips, reported revenue can lag even while cash has already been deployed. If contract terms on new deals weaken, the company may have to fund more of each project itself. Neither possibility has been established by Friday’s share-price decline. Both are reasons the next operating update matters more than a one-day percentage move.

Management has emphasized the economics of its recent wins, including a company-estimated payback period of one year and ten months for deals signed in the June quarter. The letter’s footnotes make clear that this is an estimate based on forecast costs and contracted future capacity, including capacity not yet built. Investors should compare the estimate with actual cash spending and revenue as projects come online. The number is useful as a statement of management’s underwriting, not as a completed return on invested capital.

Revenue growth is real, but scale makes timing expensive

Nebius is not a pre-revenue data-center concept. The company’s June-quarter revenue of $582.3 million was more than five times the year-earlier figure, and the group’s adjusted EBITDA moved from a loss to a positive $236.2 million. Its cloud operation accounts for most of group sales. These are meaningful signs that customers are using and paying for capacity, even as reported earnings absorb depreciation, financing and other expenses.

A fast-growing base can nevertheless coexist with an even faster-growing spending program. The $5.6574 billion of June-quarter infrastructure purchases was roughly 9.7 times that quarter’s $582.3 million of recognized group revenue. That ratio is not a valuation multiple or a standard margin. It simply illustrates how much cash Nebius was committing to hardware and facilities ahead of recognized sales at that point in the buildout. Future quarters could look different if installed capacity produces contracted revenue; they could also disappoint if delivery is slower or returns are thinner than modeled.

The interim report says property and equipment purchases primarily supported GPUs, related hardware and data-center expansion. The company also identified access to financing, power, sites, equipment and timely construction as risks to its plans. These are not generic footnotes for a business buying billions of dollars of physical assets. They describe the steps between a signed cloud agreement and revenue that can be earned repeatedly.

Nebius ended June with a cash position of about $8 billion, according to its shareholder letter, and later announced a convertible-note financing of approximately $5.75 billion in gross proceeds in August. Those two figures are from different dates and cannot simply be added to infer today’s cash balance. The financing expanded the company’s resources, while also adding future claims and potential dilution to the capital structure. Friday’s move should be read against both sides of that ledger.

What the selloff does and does not say

It is tempting to treat a 12% drop as a verdict on the entire AI-cloud business. It is not. Market prices can move on interest-rate expectations, crowded positioning, risk limits and news about adjacent companies. We found no Nebius disclosure on Friday announcing a canceled customer contract, a construction halt or a revised forecast. Nothing in the cited filing proves one occurred. The latest verified company operating figures remain the June-quarter disclosures.

The bear case is concrete without inventing bad news. Nebius must execute a capital-intensive expansion, and operating cash flow boosted by advances does not cover the full current pace of asset purchases. Capital markets and customer contract terms therefore remain important. If a larger share of new contracts lacks upfront funding, or if project delivery slips while expenses continue, outside funding requirements could rise. Gartner’s broader concern is that the industry may build more capacity than end demand can support profitably. Its forecast is a scenario, not a company-specific impairment test.

The bull case is also concrete. Nebius has reported strong revenue growth and significant customer prepayments. Those advances suggest that at least some buyers are willing to commit cash against future capacity rather than waiting for spot supply. If new sites arrive on schedule and contracts convert into earned revenue, today’s high spending could support a larger, higher-utilization business. The June quarter’s positive adjusted EBITDA and management’s account of deal terms lend substance to that possibility, although they do not eliminate construction or financing risk.

Neither case can be settled by comparing Nebius with CoreWeave on stock-price performance alone. Their debt structures and customer agreements differ. TECHi’s separate CoreWeave report focuses on bond-market signals; the distinct Nebius question is whether customer advances continue to reduce the amount of construction Nebius must fund itself. Both companies operate in the same stressed trade, but investors need different evidence to judge each one.

The next numbers that would change the picture

The first item to watch is the next cash-flow statement: operating cash, changes in deferred revenue and purchases of equipment and facilities. Read together, they will show whether customer advances keep pace with capital commitments. A larger deferred-revenue balance is encouraging only if Nebius also delivers services and recognizes revenue at acceptable margins. Cash collected ahead of work is powerful financing; cash generated after delivering the work is a stronger measure of lasting economics.

Second, compare contracted capacity with capacity actually energized and earning revenue. A gigawatt of future contracted power is not equivalent to a gigawatt of operating data-center load, and neither automatically indicates GPU utilization. The company’s plans can be substantial while the timing of revenue still matters. The useful question is whether the buildout is arriving when customers need it and whether the realized price justifies the entire cost of the cluster, including networking, power and support.

Third, watch the mix of funding on new deals. The 50% to 60% prepayment coverage described for selected June-quarter contracts would materially change the company’s capital needs if repeated at scale. It would be a mistake to assume those exact terms apply to every future contract. More granular disclosure about how many deals include prepayments, their duration and the costs they cover would let readers judge the model more accurately.

Finally, separate management’s adjusted profitability from the full cost of ownership. Depreciation, interest and potential share dilution affect what ultimately accrues to investors. The June-quarter statutory loss is not evidence that customer demand is absent; the positive adjusted EBITDA is not proof that the entire expansion has paid for itself. Both are parts of the same developing business.

At 2:19 p.m. EDT Friday, the market was marking down Nebius sharply. The freshest verified development was that price, not a newly disclosed company failure. The more durable story is the arithmetic behind the buildout: $5.66 billion of infrastructure purchases against $2.25 billion of operating cash in the last reported quarter, with customer advances doing much of the work of funding growth. The next filing can show whether that balance is improving as capacity comes online. Until then, the drop measures investors’ concern more clearly than it measures the outcome.