The trend toward neoclouds is a high risk/high reward opportunity for investors. Nebius is similar to CoreWeave, dubbing itself as AI native cloud infrastructure, which means the infrastructure was created specifically for AI workloads with architectures built on bare metal servers instead of hypervisor layers, for example, creating an important differentiation from the Big 3 which we’ve previously covered here and here.
Nebius offers an Nvidia-optimized cloud platform for teams that need to adjust compute resources, want high-performance storage and an easy-to-use AI environment yet do not want to manage the infrastructure or AI operations.
Nebius shares surged in September following the up to $19.4 billion deal with Microsoft. Recently, the company expanded its hyperscaler customer list recently with a $3 billion deal with Meta expected to ramp over the next three months. Backed by these two deals ramping through 2026, Nebius now projects reaching $7 billion to $9 billion in annualized run rate revenue (ARR) by the end of next year, up more than 13X from its current ARR of $551 million from Q3.
However, Nebius is a high-risk stock given its success depends on how much capital the company can raise, and the current financials do not support an easy path to ramping capacity and reaching these targets. Capex needs have already moved much higher, and meeting management’s more aggressive capacity targets for the end of 2026 will require significant capital to deploy the necessary GPUs – there is a chance for capex needs to remain at 7-8X of revenue in 2026, making financing a major challenge.
Regardless of how Nebius executes relative to CoreWeave, it remains an AI bubble stock as the company has to hope the stock price goes up to raise cash, which will dilute shareholders (or raise debt). It’s a vicious cycle as during any months/quarters that AI stocks are soft, Nebius stock will carry outsized execution risk.
For a closer look at Nebius’ deal with Microsoft, its vertical integration and custom servers, read more from our Advanced analysis, Nebius: Mega Microsoft Deal, 5x Growth in Power Fueling AI Cloud Hypergrowth, But High Risk Remains.Nebius: Mega Microsoft Deal, 5x Growth in Power Fueling AI Cloud Hypergrowth, But High Risk Remains.
Revenue Growth Decelerates in Q3, but Expected to Reaccelerate
Revenue decelerated to 355% YoY and 39% QoQ in Q3 to $146.1 million in Q3, down from 605% YoY and 90% QoQ in Q2. Core AI Infrastructure was the primary driver as revenue grew 400% YoY and 40% QoQ to ~$131.5 million, or ~90% of total revenue in Q3.
On that note, Nebius tightened its 2025 revenue outlook, from $450-630 million down to $500-550 million, citing timing of capacity as the primary reason. This is also below current estimates for 372% growth to $555 million.
On the other hand, revenue growth is expected to reaccelerate to 549% to $246.1 million in Q4 and further to 749% to $469 million by Q1, with more capacity coming online next quarter to support the Microsoft deal ramping through 2026 and its new $3 billion, five-year deal with Meta, with capacity rolling out over the next three months.
For fiscal 2026, revenue is expected to reaccelerate to 521% to $3.45 billion, with estimates having doubled since Nebius signed the Microsoft deal in September. Nebius said this quarter that it plans to provide full-year guidance for 2026 in the upcoming quarter.
However, reaching these targets will likely require strong/perfect execution as the company must build and deliver substantial capacity for both the Microsoft and Meta deal, as well as additional capacity to meet external demand. Doing so will likely require capex at >7X of revenue next year, an incredibly challenging position to be in as peers are spending far less yet still struggling to find funding.
Ambitious $7-9B ARR Target by End of 2026, up >7X YoY
Nebius provided an ambitious new medium-term annualized run rate revenue (ARR) target in Q3, forecasting reaching $7 billion to $9 billion in ARR by the end of 2026. For comparison, this would represent >7X YoY growth from 2025’s target of $1 billion at midpoint if it materializes.
Nebius’ two hyperscaler deals with Microsoft and Meta would account for more than half of this at ~$4.1 billion in ARR at full scale combined (approx. $3.5B for Microsoft and $0.6B for Meta). However, this would require Nebius to generate ~$3.9 billion in ARR at midpoint from other customers and other capacity, a challenging task given how costly its capacity expansion plans will be.
Additionally, the QoQ decline in incremental ARR, from $181 million in Q2 to ~$121 million in Q3, highlights a major challenge – this growth boils down to timing, and when Nebius can bring capacity online. This could be power bottlenecks, GPU supply bottlenecks, delays in building out physical infrastructure, funding bottlenecks, etc.
Thus, if the company cannot raise enough capital to afford its ambitious plan of reaching 2.5GW of contracted power and 0.8-1.0GW of connected power by the end of next year, these revenue targets may be unattainable.
Capex Forecast Raised by 2.5X to $5 Billion, 9X of Revenue
Capex requirements will be the number one focal point for Nebius considering it does not have the same depth of cash as hyperscalers, yet is competing with them on securing power, GPUs, and AI workloads.
On this note, Nebius is projected to spend ~9X its estimated 2025 revenue on capex after boosting its 2025 capex forecast by 2.5X to $5 billion, versus its prior view for $2 billion. This capex-revenue ratio is far above what peers such as CoreWeave and even Oracle are spending, at 2.5X and 0.75X. It’s also important to note that the latter two are struggling to in the financing department – CoreWeave is having to take on debt at >9% interest rates, while Oracle’s 520% debt-to-equity ratio has pushed its credit default swaps up to the highest level since 2009.
Financing this capex will be the primary challenge, and it is critical considering the bulk of this capex will go to revenue-generating GPUs. To put this in perspective for 2025 and 2026, here’s what CEO Arkady Volozh explained about capex and Nebius’ connected power targets:
And if we look at it from the CapEx point of view, roughly speaking, it breaks into 3 spending blocks. So first stage, securing land and power. It's pretty cheap. It's around — again, it depends on the scale, but it's around 1% of total CapEx for securing those blocks and electricity. The second stage, building the data centers, building connected power is something around, I don't know, 18%, 20%. And the remaining 80%, the main part is for deploying the actual GPUs. This is the main part of CapEx. So if we want to build as much as our capital will allow us, what should we do?1% of total CapEx for securing those blocks and electricity. The second stage, building the data centers, building connected power is something around, I don't know, 18%, 20%. And the remaining 80%, the main part is for deploying the actual GPUs. This is the main part of CapEx. So if we want to build as much as our capital will allow us, what should we do?
First, we should secure as much capacity as we can because the cost [is] immaterial at this scale. Second, we should build as much as our capital allows us. And third, we will fill GPUs in line with contracted or clearly visible demand. We will need this massive 80% spend [that] will come only when we see real demand. That's why we say that in 2026, we will be securing 2.5 gigawatts total contracted capacity. And we are planning to physically build 800 to 1 gigawatt of connected data centers. This will be done by the end of next year.”
For the updated 2025 capex guide, this would assume ~$50 million towards land, and ~$1 billion towards the physical data center shell and necessary equipment to connect power, and the remaining $4 billion for GPUs. This aligns with management’s forecast to have 220MW of connected power (not yet active) and 100MW of active power by year-end. This would be around ~$10 million per MW of active power based on the comments above, slightly below averages around $12 to 14 million.
Looking ahead to 2026, Nebius is planning to have 2.5GW of contracted power and 0.8-1.0GW of connected power (up 2.5X from a prior view for 1GW contracted). Per management, this includes scaling its existing data centers in the UK, US, and Israel, new data centers under development in the US and Europe, and several large sites with up to hundreds of MW under review, with the chance that some go online by the end of 2026.
Building out this capacity pipeline to reach the connected power targets with powered shells would likely require approximately $3 to $4 billion in capex, while GPUs would likely require $20 billion to $24 billion, potentially higher, depending on mix and how much of that connected power Nebius aims to have active. This is anchored by GPU costs for next-gen hardware now running at $25 billion to $30 billion per GW.
Financing Potential >4X Capex Growth, Still 7-8X of Revenue
There is a likelihood that Nebius’ capex needs for 2026 rise at a multiple of >4X YoY to meet these aggressive capacity expansion targets. This would also be ~7-8X current estimated revenue of $3.45 billion, meaning the company will be unable to break free of this extremely elevated capex cycle next year.
Compare this to Nebius’ balance sheet, which currently show $4.79 billion in cash on hand and $4.1 billion in debt following the company’s recent combined debt and equity offering raising $4.2 billion. This is not likely not even 20% of what the company could need to reach its capacity targets, and thus its revenue targets.
CFO Dado Alonso covered the financing aspect, though it is unlikely that these financing avenues could raise the necessary amount needed for these capacity targets without significantly stressing the company’s balance sheet:
“In order to support our aggressive growth plans in 2026 and to maintain this pace of growth in 2027, we will be utilizing at least 3 sources: corporate debt, asset-backed financing and equity. We are in the process of raising asset-backed debt, which we'll be able to secure with attractive terms supported by creditworthiness of our largest customers. Tomorrow, November 12, we will be putting in place an at-the-market equity program for up to 25 million Class A shares and plan to file a prospectus supplement. We will evaluate the program regularly based on our capital needs. The program enables us to access equity funding on an efficient ongoing basis. However, we will remain dilution sensitive as we prepare to finance future growth opportunities.”
Alonso’s comments suggest that the current cash on the balance sheet likely will go towards other data center opportunities outside of its hyperscaler deals, with the asset-backed debt more likely to fund the Microsoft and/or Meta buildouts in similar fashion to some Bitcoin miners raising substantial cash via asset-secured debt.
The at-the-market program could provide around $2.25 to $2.5 billion in capital around current share prices, if exercised in full between $90 to $100/share, while leading to approximately 10% dilution. Considering management’s goal of remaining dilution sensitive, the ATM program may be utilized at higher share prices to raise more capital, or as a second source behind debt. Again, raising only a few billion via 10% dilution to shareholders is still far from enough from meeting estimated capex needs next year.
While Nebius remains in a better position than CoreWeave at the moment in terms of debt-to-equity, at 0.94x versus 4.85x, there is a very high likelihood that this ratio will move rapidly in CoreWeave’s direction through 2026 given the capital intensity of building this capacity at an accelerated pace. There is also the potentiality for Nebius to be unable to raise $20B+ in capex given that it is approximately equivalent to the company’s current valuation and still 7-8X of revenue.
Why Nebius Must Spend Aggressively – Capacity is Sold Out
The reason that this capex growth is necessary, at least from management’s point of view, is because demand continues to far outstrip the capacity that Nebius can offer. Put another way – it is a spend or get left behind market.
Management emphasized numerous times in Q3’s call that capacity is the main bottleneck to revenue growth, and their current main focus is adding capacity to remove this bottleneck. CEO Arkady Volozh explained that demand was very strong in Q3 with Nebius selling out of all available capacity, and each time capacity was brought online, it was sold. Nebius is currently “selling the remnants of Q4, but [also] now preselling new capacity being delivered in future quarters” in 2026, helping lock in future revenue growth. This would also include the company’s trance of GB300s coming online in Finland in Q4.
One of the more important comments this quarter related to the demand pipeline. Management stated that pipeline generation, or customers wanting to buy capacity, expanded 70% QoQ to $4 billion in Q3, yet they “were only able to convert a portion of that given to the constraints of our capacity.” For context, core AI infrastructure revenue was approximately $131 million in Q3, so the pipeline would be ~30X its current run rate. It makes sense why Nebius is aiming to aggressively grow capacity given it witnessed well over $1.5 billion in sequential growth in the pipeline, as meeting a larger portion of this pipeline via more capacity would quickly translate to revenue growth.
To that point, management said that lead times from power connection and start of GPU deployments to revenue generation range “anywhere from 6 to 12 weeks” or potentially faster in existing facilities, again underscoring why they are willing to pursue this rapid capacity growth through 2026 as they can quickly shift to revenue generation from power connection.
However, this has highlighted a major downside to a capacity-constrained model. Nebius stated that it is “learning to say no to customers as we routinely sell out and have to actually let them down lightly and try to convince them to purchase in the future,” but there is an equal chance that these customers will simply go to CoreWeave or a hyperscaler who has capacity, get locked in to that ecosystem and not return to Nebius.
It also sheds light on the puts and takes of Nebius’ deployment strategy at this small scale. By prioritizing hyperscaler deals with Microsoft and now Meta (though this was ultimately constrained by capacity), Nebius is locking in strong future revenue streams over the next few years, but turning away these smaller startup customers by also locking up a larger portion of its near term capacity to the hyperscalers.
Building Core AI Cloud Business
As we discussed in our prior analysis on Nebius for Advanced members, Nebius: Mega Microsoft Deal, 5x Growth in Power Fueling AI Cloud Hypergrowth, But High Risk Remains, the company believes its vertical integration and proprietary cloud serve as a key advantage and the ‘real future opportunity’:
“In the longer term, Nebius believes that its vertical integration with a full stack of AI services will help broaden its customer base, increase platform stickiness and capture higher margin revenue and services. Nebius offers a proprietary cloud platform with managed MLops services, low downtime and high cost efficiency, combined with its inferencing platform AI Studio. With AI Studio, Nebius says it can offer up to 3x token savings with low latency, and up to 4.5x faster time to token versus other competitors in Europe.”offer up to 3x token savings with low latency, and up to 4.5x faster time to token versus other competitors in Europe.”
As such, the company is prioritizing building out its core AI cloud platform and continuously adding new features, with the predominant goal currently being geared towards enterprise adoption. Two main features from Q3 include Nebius’ new enterprise-ready cloud platform, dubbed Aether, as well as Nebius Token Factory.
With Aether, Nebius brings enterprise-grade security and compliance features along with a comprehensive observability suite, developer tools and more, in an effort to make its platform more attractive and accessible for large enterprise customers. Nebius also is focusing on improving reliability of its network with active health checks to reduce maintenance tasks, and boosting performance and storage speeds.
Token Factory builds on Nebius’ AI Studio and embeds enterprise-grade security in a production-scale inference platform, letting customers run open-source AI models from OpenAI, Alibaba, Meta, DeepSeek and others with 99.9% uptime, per the company. Management says Token Factory will help customers “transform open source models into optimized production-ready systems with guaranteed performance and transparent cost per token” and the “best total cost of ownership.” While it is too early to see how Token Factory contributes to monetization, Nebius’ dedication to improve its platform and expand its suite of offerings can serve as a strong differentiation and potentially aid in customer acquisition over rival platforms.
Financials
Revenue to Reaccelerate
Following the deceleration in Q3, Nebius tightened its 2025 revenue outlook, from $450-630 million down to $500-550 million, citing timing of capacity as the primary reason. This is also below current estimates for 372% growth to $555 million.
For fiscal 2026, however, revenue is expected to reaccelerate to 521% to $3.45 billion, with estimates having doubled since Nebius signed the Microsoft deal in September. Nebius said this quarter that it plans to provide full-year guidance for 2026 in the upcoming quarter.
Margins Improving, but Widely Negative
Gross margin was 70.6%, down slightly from 71.3% in the prior quarter and up from 69.2% in the year ago quarter.
Operating margin began to show signs of improvement, coming in at (89.1%), compared to (105.8%) in the prior quarter and (251.1%) in the prior year. At this rate, Nebius could break even in five quarters assuming it can maintain such improvements consistently, though this may be challenging as capacity ramps up rapidly.
Net margin was (81.9%), not comparable to the prior quarter’s 556% on Toloka’s deconsolidation but up from (293.5%) in the year ago quarter. Adjusted net margin was (68.7%), improving from (123.7%) in the year ago quarter. However, it is important to note that net losses has widened (from ($39.7 million) to ($100.4 million) for adjusted net loss) though margins are showing improvement from the rapid revenue ramp.
Adjusted EBITDA improved to just ($5.2 million) or a (4%) margin, up from a (20%) margin last quarter. Nebius noted that its Core AI Infrastructure business continued to generate positive adjusted EBITDA at a nearly 19% margin in Q3, with the metric weighed down by Avride and TripleTen investments.
Nebius lags CoreWeave by a significant degree for adjusted EBITDA, with CoreWeave posting a 61% adjusted EBITDA margin in Q3, down 4 points YoY. This suggests that there is room for substantial expansion over the upcoming quarters as the business scales to a much larger size.
Earnings Remain Far From Profitability
Nebius reported a 23% beat on adjusted EPS in Q3, though the company remains far from profitability and is not expected to reach profitability for quite some time.
Q3 adjusted EPS was ($0.40), beating estimates for ($0.52) but widening slightly from ($0.38) in the prior quarter. Looking ahead to Q4, adjusted EPS is expected to be ($0.58), before widening to ($0.74) in Q1.
For fiscal 2025, adjusted EPS is projected to be ($1.73) before widening to ($2.29) in 2026, likely driven by increasing expenditures to quickly ramp capacity.
Cash Needs Increasing
The challenge for Nebius is very similar to that of CoreWeave, with the neocloud spending significantly on GPUs and raising substantial debt to fund said spending. As a reminder, Nebius spent nearly $1 billion in capex in Q3, up from $510 million in Q2 and representing more than 6X its revenue. Capex is on pace to be >9X of revenue this year and potentially remain at 7-8X of revenue next year.
Operating cash flow was ($80.6 million) for a (55.2%) margin, while free cash flow was ($1.04 billion) for a (709.1%) margin. This widened from free cash flow of ($678.3 million) in the prior quarter as capex surged more than 87% QoQ to $955.5 million.
Cash and equivalents totaled $4.8 billion, while debt was $4.1 billion. As mentioned previously, surging capex this year and the potential for tens of billions next year means Nebius will likely turn to debt markets for significant funding. Debt to equity sat at 0.94X in Q3, though this is likely to worsen significantly moving through 2026 based on estimated capex needs north of $20 billion.
Valuation
Nebius trades at 37x forward PS ratio, slightly above its average of 32x, though data is limited considering its recent launch on the public markets post-Yandex breakup. Shares have traded as low as 8.6x forward PS and as high as 61.5x.
Rapid revenue growth in 2026 is expected to bring forward PS down to 5.9x next year, though this remains a premium to CoreWeave at 3.0x next year’s revenue estimate of $12.07 billion.
Conclusion
Nebius is on a trajectory of high growth-high debt for the foreseeable future, with the company spending nearly $1 billion this quarter on capex alone, in preparation for capacity ramps for Meta and Microsoft occurring in quick succession.
On one hand, revenue growth is expected to accelerate sharply to 750% over the next two quarters and maintain a hypergrowth profile with >380% growth for the next five quarters, while YoY growth is expected to accelerate 148 points to 521% in 2026. However, capex guidance for 2025 was raised to ~9X of revenue, and aggressive capacity expansion targets for 2026 mean capex will likely remain 7-8X of revenue, far above peers and making finding financing a significant challenge.
Ultimately, even if high-beta stocks catch a bid, there are far cleaner and less capital-intensive ways to gain exposure to the AI buildout. Until Nebius can demonstrate that its growth is not being funded by an expanding cash shortfall, we see limited risk-reward and will remain on the sidelines.
Damien Robbins, Equity Analyst at I/O Fund contributed to this analysis.
Please note: The I/O Fund conducts research and draws conclusions for the Fund’s positions. We then share that information with our readers. This is not a guarantee of a stock’s performance. Please consult your personal financial advisor before buying any stock in the companies mentioned in this analysis.
Recommended Reading: