Nebius Group stock came roaring back on Tuesday, jumping roughly 9% to around $253.25 and reversing Monday’s 4.22% decline as investors piled back into one of the market’s most aggressive artificial-intelligence infrastructure bets. The rebound pushed NBIS through the closely watched $250 area after repeated selling pressure around that level, but the violence of the two-day move also captures exactly what makes the Nebius stock forecast 2026 so difficult: investors are trying to value a company whose revenue is exploding, whose AI computing capacity is effectively selling as fast as it can be deployed, and whose infrastructure ambitions require tens of billions of dollars of investment.
The 9% rally did not arrive in isolation. AI-linked technology stocks were broadly strong Tuesday as enthusiasm around infrastructure spending pushed the Nasdaq deeper into record territory, even while elevated Treasury yields continued to challenge richly valued growth companies. That backdrop matters enormously for Nebius because few stocks are more directly exposed to the market’s belief that AI spending will remain extraordinary for years. Nebius is building an Nvidia-powered AI cloud across the United States and Europe, and its growth targets have become almost difficult to comprehend: second-quarter group revenue surged 454% year over year to $582.3 million, annualized run-rate revenue reached $3 billion, and management continues to target $7 billion to $9 billion of run-rate revenue by year-end.
Those numbers explain why buyers rushed back into NBIS after Monday’s drop. They also explain why the stock can fall 4% one day and jump 9% the next. Nebius is no longer being valued on what the business looks like today. Investors are betting on what it might become several years from now—and the amount of money required to build that future is enormous.
Monday’s Selloff and Tuesday’s Surge Expose the Real NBIS Debate
Nebius closed Monday around $232.57, down 4.22%, before Tuesday’s rally carried the shares back above $250. On the surface, that might look like little more than another volatile session for a high-beta AI stock. Yet the reversal illustrates a deeper battle taking place between investors who see Nebius becoming a major independent AI-cloud provider and skeptics who believe the market is capitalizing future growth far too aggressively before the infrastructure required to produce that growth has even been fully built.
The bullish argument starts with demand. Nebius said in its second-quarter update that AI-cloud revenue reached $574.9 million, representing roughly 98% of group revenue and rising 514% from the previous year. Annualized run-rate revenue increased from $1.9 billion at the end of March to $3 billion at the end of June, while Nebius AI Cloud generated an adjusted EBITDA margin of approximately 50%. Management said it was once again selling out available capacity essentially as quickly as that capacity could be brought online.
Related articles
That is the kind of problem every infrastructure company wants to have. Nebius isn’t currently struggling to convince customers to rent its GPUs. It is struggling to install enough computing power to satisfy customers already asking for it.
The bearish question is what happens after billions of dollars have been committed to solving that shortage.
Nebius Revenue Is Growing at a Speed Normally Reserved for Tiny Startups
A company generating hundreds of millions of dollars per quarter normally does not grow revenue by more than 400% year over year. Nebius is doing exactly that because its starting infrastructure base was relatively small and new GPU capacity is coming online rapidly into an exceptionally tight AI-compute market.
Second-quarter group revenue of $582.3 million increased 46% sequentially and 454% year over year, while AI-cloud revenue increased even faster. Adjusted EBITDA swung from a $21 million loss in the year-earlier period to a $236.2 million profit, producing a group margin of roughly 41%. Nebius AI Cloud itself generated approximately $285.7 million of adjusted EBITDA, equivalent to a margin close to 50%.
Those margins are particularly important because one of the biggest criticisms of the emerging “neocloud” industry has been that companies are spending huge amounts on Nvidia GPUs only to rent them out through businesses that may eventually become commoditized. Nebius‘ current results point in the opposite direction. High utilization, strong pricing and infrastructure efficiency are creating significant operating leverage as new capacity becomes productive.
Management still expects $3 billion to $3.4 billion of group revenue for 2026, an adjusted EBITDA margin of approximately 40%, and year-end annualized run-rate revenue of $7 billion to $9 billion. If Nebius reaches the midpoint of that ARR target, its exit rate would be more than six times the $1.25 billion level reported at the end of 2025.
That kind of acceleration is why investors keep returning to NBIS after selloffs.
But revenue growth isn’t the most spectacular number in the Nebius story anymore.
Meta’s $27 Billion Deal Changed the Scale of the Nebius Story
The clearest validation of Nebius‘ infrastructure strategy came in March, when the company announced a massive five-year agreement with Meta. Under the arrangement, Nebius will provide $12 billion of dedicated capacity across multiple locations using Nvidia’s Vera Rubin platform, with deployments beginning in early 2027. Meta also agreed to purchase certain capacity that Nebius is unable to sell to other AI-cloud customers, creating an additional potential commitment of up to $15 billion. The combined contract value could therefore reach approximately $27 billion.
That second component is especially interesting because Nebius still intends to sell the relevant compute to third-party customers wherever possible. Meta effectively provides a powerful backstop for unsold capacity, substantially reducing part of the utilization risk associated with building enormous GPU clusters before every individual customer has been identified.
Nebius already had Microsoft as a major customer. Its multi-year Microsoft agreement, announced in September 2025, involves dedicated AI infrastructure from the company’s Vineland, New Jersey data center. Nebius subsequently said it delivered the first tranche of Microsoft capacity on schedule in November 2025.
The significance of those customer names is difficult to overstate. Microsoft and Meta are not speculative AI startups trying to rent a few GPUs. They are two of the world’s largest technology companies and among the biggest AI infrastructure spenders on the planet.
For Nebius, those contracts provide more than revenue.
They provide credibility.
Nebius Is Trying to Turn Customer Contracts Into a Financing Machine
Building the infrastructure required to fulfill enormous AI contracts creates a second challenge: someone has to pay for the GPUs, power systems and data centers before customers generate years of revenue from them.
Nebius is trying to solve that problem by using contracted cash flows themselves as financing assets. In July, the company secured approximately $775 million in senior secured debt, backed by deployed GPU infrastructure and cash flows from an investment-grade customer. The facility is priced at SOFR plus 2.50% and, according to Nebius, together with the customer cash flows covers more than 100% of the capital expenditure needed to deploy the underlying GPU infrastructure.
That financing structure could become enormously important. Nebius said the model provides a repeatable framework for more than $40 billion of additional customer commitments, while its second-quarter presentation showed that 50%-60% of capital expenditure on certain new contracts can be self-financed through customer prepayments.
This is the bull case in its most sophisticated form. Nebius signs long-term agreements with investment-grade technology companies, uses those contracts and customer prepayments to finance infrastructure, deploys the GPUs, generates cash flow and then repeats the process at increasingly larger scale.
If that machine works, Nebius can expand much faster than a company relying entirely on its own balance sheet.
If it breaks, the capital requirements become much more intimidating.
And those requirements are already enormous.
The $20 Billion-to-$25 Billion Capex Bill Is Why NBIS Can Never Be a Simple Growth Story
Nebius spent approximately $5.7 billion on capital expenditures in the second quarter alone, primarily on GPUs, GPU-related equipment and data-center expansion. For the full year, management expects capital expenditures between $20 billion and $25 billion.
That number puts the entire NBIS thesis into perspective. The company expects $3 billion to $3.4 billion of 2026 revenue while potentially investing more than six times that amount in infrastructure during the year. Such spending can make perfect economic sense if the resulting assets generate years of highly profitable contracted revenue, but it means investors cannot evaluate Nebius like an ordinary software company whose incremental customers require almost no physical capital.
Every new wave of Nebius growth requires GPUs, networking equipment, buildings, cooling systems and enormous quantities of electricity.
Management ended the second quarter with roughly $8 billion in cash and has been expanding its financing options, but the company will need to keep balancing customer prepayments, asset-backed borrowing, cash generation and potentially other financing instruments as the buildout accelerates.
The market therefore faces an unusual equation: breathtaking revenue growth on one side and breathtaking capital requirements on the other.
Whichever grows faster will eventually determine whether today’s valuation was justified.
Power May Be More Important Than GPUs for Nebius’s Next Phase
Buying Nvidia chips is difficult, but obtaining enough electricity to operate them may be an even larger bottleneck.
Nebius has been aggressively securing power across multiple regions and raised its year-end contracted-power target to 5 gigawatts after its second-quarter results. The company still expects approximately 800 megawatts to 1 gigawatt of connected power by the end of 2026 and plans to deploy more than 1 GW of additional capacity annually beginning in 2027.
Its pipeline includes enormous new sites. In May, Nebius announced plans for an owned AI factory in Pennsylvania with up to 1.2 GW of power, adding to another planned 1.2 GW-scale site in Missouri and a major expansion in Finland. The Pennsylvania facility is expected to be delivered in phases beginning in 2027.
Those projects reveal how quickly the AI-cloud industry is changing. A company that only recently emerged in its current form is planning power infrastructure measured in gigawatts because hyperscalers and AI developers are consuming compute faster than suppliers can build it.
Nebius says its pipeline supports that aggression. In Q1, customer pipeline generation increased approximately 3.5 times quarter over quarter, while the second quarter produced four major deals averaging more than $1 billion each.
Demand therefore does not appear to be the immediate problem.
Execution is.
The Biggest NBIS Risk Is That AI Infrastructure Becomes Less Scarce
Nebius currently operates in an extraordinary market where demand for advanced AI computing exceeds available supply. That scarcity supports high utilization, attractive pricing and strong margins, but investors should be careful about assuming today’s economics will last forever.
Nebius is not the only company building.
CoreWeave, hyperscalers and other specialized cloud providers are racing to add GPU capacity, while Meta itself has considered monetizing spare computing infrastructure. Every major technology company understands that AI compute is scarce, and billions of dollars are being directed toward eliminating that scarcity.
GPU technology also moves extremely quickly. Nebius has already deployed Nvidia B300 and GB300 systems and plans to be among the first AI-cloud providers deploying Vera Rubin systems. Staying competitive therefore requires continuously purchasing newer hardware while existing GPUs depreciate economically as more powerful generations arrive.
Higher interest rates add another complication. The broader AI rally is continuing even with long-term Treasury yields at historically elevated levels, but companies requiring massive infrastructure financing are inherently more sensitive to the cost of capital than asset-light software businesses.
The nightmare scenario for NBIS investors is therefore straightforward: Nebius builds enormous amounts of expensive infrastructure just as industrywide capacity catches up with demand, utilization falls and pricing weakens.
The bullish scenario is almost the mirror image.
Demand remains so strong that Nebius continues selling capacity before it is finished.
So far, that is the scenario management says it is experiencing.
Nebius Stock’s 9% Rally Matters Less Than What Happens to Its $7 Billion-to-$9 Billion Target
Tuesday’s rebound above $250 will attract traders because it erased Monday’s selloff and pushed NBIS through an area where the stock had repeatedly encountered resistance. But investors considering the Nebius stock forecast 2026 should probably care much more about another number: $7 billion to $9 billion.
That is management’s year-end target for annualized run-rate revenue.
Nebius reached $3 billion at the end of June, meaning the company must still bring a huge amount of additional capacity online during the second half of the year to reach its goal. Management expects infrastructure deployed late in the second quarter to begin generating revenue during Q3, while additional capacity is scheduled to arrive as the year progresses.
If Nebius approaches the upper end of its target while maintaining strong AI-cloud margins, investors will gain powerful evidence that the company can translate billions of dollars of capital spending into revenue at the speed management promises. The Microsoft and Meta contracts would then look less like isolated wins and more like the foundations of a rapidly scaling AI-infrastructure platform.
If capacity deployment slips or ARR growth fails to accelerate sufficiently, the market may suddenly focus much more intensely on that $20 billion-to-$25 billion capital-expenditure bill.
That is why the violent stock swings make sense.
Investors aren’t debating a mature company’s next quarter. They are trying to price a massive infrastructure network while it is still being constructed.
The Nebius Stock Forecast 2026 Comes Down to One Brutal Question
Nebius has already accomplished several things that would have seemed extraordinary only a year ago. Quarterly revenue has increased more than fivefold from the prior year. AI Cloud has reached approximately 50% adjusted EBITDA margins. Microsoft and Meta have signed enormous infrastructure agreements. Contracted power has expanded toward 5 GW, and Nebius says demand is strong enough that new capacity continues selling rapidly.
That explains why a 4% selloff can be followed immediately by a 9% rally. Every decline attracts investors who see the next independent AI-cloud giant being built in real time, while every surge raises fresh questions about how much future success has already been capitalized into NBIS stock.
The company’s next phase will therefore be less about proving that demand for AI computing exists. Microsoft, Meta and Nebius‘ broader customer pipeline have already provided considerable evidence of that.
The challenge is converting demand into infrastructure, infrastructure into revenue and revenue into enough cash flow to finance the next round of expansion without destroying shareholder returns.
With billions committed to GPUs and data centers, the stakes are enormous. If AI compute remains scarce and Nebius continues selling capacity almost as quickly as it can install it, today’s explosive revenue trajectory could continue surprising investors.
If the supply-demand equation normalizes before Nebius earns an adequate return on its massive infrastructure buildout, today’s valuation could look very different.
Tuesday’s 9% surge tells investors that the market still believes.
Now Nebius has to prove that a company spending up to $25 billion in one year can build its AI cloud faster than the AI boom can cool down.
Disclaimer
This article is for informational purposes only and does not constitute financial or investment advice. Readers should conduct their own research and, where appropriate, consult a qualified financial advisor before making investment decisions. This article was researched and drafted with the support of AI, then reviewed, fact-checked, and edited by the editorial team before publication.
