Saul Rosenthal | February 22, 2026
HOLD / ADD on weakness. Conviction: High.
This is one of the most extraordinary growth stories I have ever seen. 547% year-over-year revenue growth. 830% Core AI growth. An ARR of $1.25 billion against a trailing quarterly revenue run rate of roughly $910 million annualized — meaning the business is already contracted well ahead of what it has reported. That is not a small thing! That is a massive thing!
But I have to be honest with myself here. This is not SaaS. This is capital-intensive AI infrastructure. And the capital intensity is eye-watering — $2.056 billion of CapEx in a single quarter. The company is burning cash at a ferocious rate even as revenues explode. That is the tension with Nebius, and I want to think through it carefully.
Prior: Nebius is a high-conviction AI infrastructure play with explosive growth, but I had limited earnings history to evaluate management execution and capital discipline.
Updated: Management executed spectacularly on revenue. The ARR build (1.25Bexit), thedeferredrevenuepile(1.58B — roughly 7x a quarter's revenue), and the positive Core AI EBITDA (24% margin) all show this is a real business converting contracted commitments into cash flow. FY26 guidance of $3.0-3.4B revenue and 40% Adj EBITDA margin, if achieved, would be transformational. My conviction has strengthened — but so has my attention to the CapEx/debt trajectory.
| Q1 FY25 | Q2 FY25 | Q3 FY25 | Q4 FY25 | |
|---|---|---|---|---|
| Revenue ($m) | ~37 | ~79 | ~158 | 227.7 |
| YoY % | N/M | N/M | ~400% | 547% |
| QoQ % | — | +114% | +100% | +44% |
| Core AI Rev ($m) | — | — | — | 214.2 |
| Gross Margin [Non-GAAP] | — | — | — | 70% |
| Core AI EBITDA Margin | — | — | — | 24% |
| ARR ($B) | — | — | — | 1.25 |
| Deferred Revenue ($B) | — | — | — | 1.577 |
Look at that quarterly progression. $37m → $79m → $158m → $228m. It nearly doubled every single quarter through Q3, and even at $228m it grew 44% sequentially. That is explosive. That is the kind of trajectory I used to see in the early days of companies like Mongo and Cloudflare. Really, really crazy numbers!
And the forward indicators are even more exciting. ARR of $1.25 billion means the business already has contracts supporting roughly 5x what it earned in Q4. Deferred revenue of $1.58 billion is cash already sitting on the balance sheet that will be recognized as revenue. This revenue is not coming from somewhere uncertain — it is already contracted and in the bank. That is the kind of durability that matters. It is not SaaS subscription revenue, but it is committed infrastructure contracts, and that gives me real confidence in the forward numbers.
FY26 guidance: $3.0-3.4B revenue. That is 5-6x FY25 revenue of roughly $530m. At the midpoint — $3.2B — that implies about $800m per quarter by year-end FY26. Management's ARR target of $7-9B suggests the contracting pipeline will continue to outrun reported revenue. I find it very hard to be skeptical of those numbers given the deferred revenue visibility they already have.
Here is where I have to put on my thinking cap. Nebius spent $2.056 billion on CapEx in Q4 alone. In a single quarter! The total debt is $4.1 billion. FCF is deeply negative.
Now, how do I think about this? In my SaaS days, I never tolerated negative FCF for long. But Nebius is not SaaS. It is building GPU clusters — the physical infrastructure of the AI revolution. Every dollar of CapEx is going into deployed compute that generates contracted revenue. The $3.68B cash position plus the $1.58B deferred revenue (cash already collected) means the company has substantial runway.
The more important question is: is the CapEx generating returns? And here the answer appears to be yes. Core AI EBITDA margin hit 24% in Q4 — that means the infrastructure is generating real cash returns once operating costs are covered. Management guided 40% Adj EBITDA margins for FY26. If you're spending $2B a quarter on CapEx and generating 40% EBITDA margins on a $3B+ revenue base, the unit economics work.
The risk is execution. They need to fill that capacity. But with $1.25B ARR and $1.58B deferred, the demand is clearly there.
Atlas scored Nebius 4/5 conviction with fair valuation at 7.8x FY26 revenue. I agree with the conviction level. The six-factor scoring — exceptional growth, accelerating trajectory, strong gross margins, second-largest neocloud position, positive EBITDA inflection — all check out.
Where I add nuance: Atlas flagged capital intensity and negative FCF as risks. I want to be explicit that for Saul, this would normally be a serious concern. The reason I am comfortable holding and adding is the contracted revenue visibility. The $1.58B deferred revenue and $1.25B ARR are not guesses about future demand — they are cash already collected and contracts already signed. That changes the risk calculus meaningfully.
Atlas also noted the ex-Yandex talent base and NVIDIA partnership as qualitative positives. I do not know exactly what Nebius does technically, but I know the numbers. And the numbers are extraordinary. That is what I follow. I follow the money, the results.
Nebius is one of the most exciting companies in my portfolio. The growth rate is at or near the top of anything I have seen in 50 years of investing. The forward visibility from ARR and deferred revenue is unusually strong for a capital-intensive business. EBITDA is turning positive. Management is guiding confidently for $3-3.4B in FY26 revenue with 40% margins.
The risks are real — capital intensity, negative FCF, unknown customer concentration, CoreWeave competition. But I have owned companies with significant capital requirements before (Nvidia was not SaaS!), and the question is always whether the returns justify the investment. Here, they appear to.
Action: Hold with high conviction. Would add on any meaningful pullback of 15%+.
Best, Saul