Nebius Q4: Building an AI Empire

The Q4 earnings for Nebius are finally out, and while the headline numbers and capex guidance might have caused a bit of a stir in the pre-market, looking beneath the surface reveals a business that is growing at a scale that is genuinely hard to wrap your head around. We are watching the birth of a potential AI hyperscaler in real-time, but as with any empire-building project, it comes with a massive bill and a unique set of risks that every shareholder needs to understand.

Read on below or watch my video.

The Revenue Illusion: Why the Miss Doesn’t Matter

On paper, the Q4 results might look like a slight disappointment. Revenue came in at $227 million, which was a bit below the $240–$250 million the market was looking for. For the full year, the $529 million total was also shy of the $550 million estimate. However, in a business like this, focusing on the backward-looking revenue miss is missing the forest for the trees.

The reality is that Nebius is selling out every ounce of capacity as soon as it comes online. If revenue was lower than expected in Q4, it’s simply because capacity came online a few weeks later than modeled. That revenue doesn’t disappear; it just slides into Q1 2026. What matters is the trajectory of Annualized Run Rate (ARR) and forward guidance.

The market expected Nebius to end 2025 with about a billion dollars in ARR. Instead, they blew past that, hitting $1.25 billion. More importantly, they reaffirmed a 2026 target of $7 billion to $9 billion in ARR. We are looking at a 6x to 7x growth rate in a single year.

The Magic of Cash Flows

One of the most interesting parts of the Nebius story right now is how they are actually funding this growth. Despite a revenue number at $227M, Nebius reported positive operating cash flow of $834 million in Q4.

How does a company with $227 million in revenue generate $834 million in operating cash? The answer lies in their deals with giants like Microsoft and Meta. These contracts are structured with significant upfront payments. This is deferred revenue, money that Nebius gets today to build out the capacity that will be delivered over the next year or two but will show up in the revenue when capacity is deployed.

This is a beautiful setup for an infrastructure play. It means the customers are essentially helping to finance the very data centers they will eventually use. While the income statement shows low revenue and an adjusted net loss, the cash flow statement tells a story of a company that is successfully leveraging its relationships with the world’s largest tech companies to offset its massive capital expenditures.

The $20 Billion Question: Capex and the Funding Gap

Now we get to the part that actually worried the market: the capex guidance. Nebius announced they plan to spend $20 billion on capital expenditures in 2026. To put that in perspective, they spent about $4 billion in 2025. This is a massive step up.

When you run the math on their current $3.7 billion cash pile and the operating cash flow the business is expected to generate, they have about 60% of that capex covered meaning there is a clear funding gap of about $8 billion. This is the growing pain of building an AI empire.

As a shareholder, this is where the risk of dilution becomes real. Nebius has a few ways to bridge this gap:

1. Monetizing Equity Stakes: They hold valuable stakes in companies like ClickHouse and AVRide plus others, which could be worth anywhere from $7 billion to $10 billion.
2. Debt and Convertible Debt: They have a history of using these instruments successfully.
3. At-The-Money (ATM) Equity Program: This is the most direct path to dilution, and it’s what the market is currently bracing for.

At a $22 billion market cap, any sort of capital raise is significant especially if it’s in the billions. If they choose to raise that through equity, current shareholders will face substantial dilution. This is the primary reason why, despite my long-term bullishness, I’m cautious about adding new capital at these specific price levels.

The Physical Foundation: Power and Locations

You can’t build an AI cloud out of thin air. You need two things: power and physical space. Nebius is getting ahead of the game on both fronts.

They have already secured 2 gigawatts of contracted power and have raised their year-end 2026 guidance to 3 gigawatts. They aren’t going to deploy all of that immediately, but in a world where every tech giant is currently hunting for power, banking that capacity now is a massive competitive advantage. It future-proofs the business for 2027 and 2028.

On the location front, they are expanding from seven sites to a planned 16 by the end of 2026. This includes strategic company-owned sites in Missouri and Alabama, alongside co-locations in the UK, France, Israel, and other parts of the US like Oklahoma and New Jersey.

The strategy here is to build sovereign AI clouds, infrastructure that sits within specific borders to satisfy local regulatory and latency requirements. However, this expansion isn’t without risk. We are seeing a growing Not In My Backyard (NIMBY) sentiment regarding data centers due to their impact on local power grids. If state or local governments start pushing back on these builds, it could threaten Nebius’s ability to hit its aggressive 2026 and 2027 targets.

Beyond Infrastructure: The Full-Stack Vision

If Nebius were just a GPU rental shop, a 42x sales multiple might be seen as insane but the forward numbers are a lot lower given their significant growth. On top of that, they are building something much more integrated. The recent acquisition of Tavili for $250 million is a clear signal of their intent to move up the stack.

Tavili focuses on agentic search and agentic development. In a world where AI is moving from chatbots that answer questions to agents that execute tasks, having this capability in-house is critical. Nebius is building a platform that allows AI-native companies to not just rent chips, but to build, train, and deploy agents easily on a single, optimized stack.

The launch of Token Factory and the AI Cloud 3.1 Aether platform further differentiates them from players like Coreweave. They are aiming to be the AWS of AI, a verticalized hyperscaler that owns everything from the power contract to the inference API.

Valuation: The Road to $100 Billion

Where does this end? If Nebius can achieve its goal of 20% to 30% EBIT margins as it matures, and those start flowing down to 15% to 20% free cash flow margins, the returns could be staggering.

In some models, I’ve looked at the path to a $100 billion valuation. If they hit $20 billion in revenue by 2029 (which analyst estimates suggest is possible) and maintain healthy margins, a $100 billion market cap isn’t just a dream, it’s a mathematical likelihood.

But, and this is a big but, the journey to 2029 is going to be bumpy. We have to account for the dilution that will likely happen over the next two years. If the company hits a $50 billion valuation but has increased its share count 30% to get there, the return for an individual shareholder might be far less impressive than the company’s overall growth.

Final Thoughts: Holding, Not Adding

I’ve been a believer in this management team since the Yandex days. They have the talent, the vision, and now they have the multi-billion dollar contracts with Microsoft and Meta to prove their model works.

However, as an investor, you have to separate the company’s success from the stock’s performance. At current multiples, the market is pricing in a lot of great execution which I believe is realistic. Given the $8 billion funding gap and the certainty of future capex-driven volatility, I’m perfectly happy holding my current stake, which has a very low cost basis.

If the stock sees a significant pullback, perhaps into the $60 or $70 range, as the market digests the earnings and potential of dilution, I would consider adding more. But for now, I want to see exactly how they bridge that $8 billion gap in 2026.

Nebius is building something special, but empires aren’t built for free. We are paying for the future today, and as long as you understand the cost of that seat at the table, the long-term view remains incredibly exciting.

Disclosure : I am long NBIS, MSFT, GOOGL, AMZN and may be long stocks discussed in this article in the near future. I am not a financial advisor and do stock analysis as a hobby. Opinions are my own. This is not financial advice. Consult with a financial advisor before making any investments as stock investing comes with risk of loss.