My Nebius Q2 2026 Notes

[Iamnzane]

· Revenue surged 514% YoY to $575 million, while ARR climbed 56% sequentially to $3 billion.

· Customer prepayments cover 50% to 70% of project CapEx, supporting a massive $20 billion to $25 billion 2026 expansion. Payoff of capex is now 1 yr 10 months down from 2-3 years. Arkady says it is a top priority to reduce the time for added capacity. It needs to be nearly ready by the time a contract is executed.

· Margins in Nebius AI reached 50%, while the power contract target was raised to 5 GW without compromising the guidance on ARR of $7-$9 billion

· 50% adjusted EBITDA margin, while group EBITDA was $236 million and the margin was 41%. High depreciation undermines the headline 50% adjusted EBITDA margin.

· $2.3B in positive operating cash flow, NBIS’s capex outflows ($5.7B in Q2) and ongoing equity/debt is reflected in negative free cash flow.

· Average Contract Value (ACV) Q1-$12M, Q2-$20M, Q3-$40M. 4X growth FY 2026

Q&A mostly CEO Arkady:

· Holding a small portion of capacity for short term, high margin contracts (e.g. 6 months): 30-50% margins.

o Customers want a controlled, dependable AI service with specialized models and a quick turnaround.

o Allows pricing basis to be collected to feed back into larger contract negotiations.

· Annual contract value (ACV – a new metric?) reached roughly $40 million per MW

· Annual guidance did not change. The management kept $3-$3.4 billion revenue, $7-$9 billion ARR and roughly 40% EBITDA margins. Raised contracted power to 5 GW. (not active power)

· During Q2, Nebius sold 12.7 million Class A shares through its ATM program at an average price of $223.6, raising approximately $2.8 billion. Another 12.3 million shares remained available under the program at quarter-end. In July, Nebius also raised $775 million through its first secured asset-backed facility.

Short float remains at ~25% today. Yesterday’s shorts ‘took it in the shorts’ after yesterday’s big 35% price jump. This volatility shows you the Wall Street bipolar brain on AI stocks. They think the AI crash is coming sooner than you think. Michael Burry has publicized a large short for $212. Says describing the opportunity as a bit like “shooting fish in a barrel.” This is good news for those of us that believe the AI boom is structural and long lasting (well at least through 2027). This gives us volatility that we can play. But sometimes my brain hurts. The thing to watch is any slowdown or plateau in hyperscalar capex spending. Then it may be time to lighten up. The bears talk about the high depreciation and the lack of visibility into a durable profit feed. And talk about the circular flywheel invested by NVDIA in the neoclouds. GM has offered a finance division for years to loan money to customers to buy their cars. Is this any different? On a side note, I heard yesterday that CoreWeave said there is still high demand for its 5 year old H100 GPUs. Now fully depreciated, so pure cash profit now.

I have belief in CEO Arkady, a tough, cloud experienced, old Russian survivor. He talks straight and tough with high exposure. And I like the NBIS AI diversification into Europe which badly wants AI services. Arkady knows how to navigate these regulators. My only neocloud hold is Nebius. After yesterday’s runup, it remains my top holding.

-zane

[jonathan1]

100% agree. Nebius smashed it on every metric and had a truly great earnings report. Needless to say it remains my top holding too (after the 35% rise on ER day it now sits at 26% of my portfolio). I loved what Arkady said about customer prepayments and the Capex cycle. I also loved reading about how they are selling compute to enterprises and neolabs and how they continue to sell all that they have but they are choosing to keep some back for shorter term contracts that are at least double the revenue in some cases of the longer term hyper scale contracts.

It was not a good day to be short or a bear and I continue to be Long Nebius. I can easily see another double from here and I think it may soon be 400 dollars even as early as next year.

Jonathan

[coldmountain]

A couple of notes of my own: There was a brief discussion during the call (by the CEO as I recall) of their ‘data center as a service’ software model – this is my term, not Nebius’. From what I gather, it is something along the lines of them offering their very substantial software stack to other DC operators. I’m sure it is not the same as a company contracting directly with NBIS on nbis’ own dcs, but it likely is a huge step up to all the terawulf etc NeoClouds, that are probably finding customer requirements are quickly evolving beyond just racks of GPUs. THere was no discussion of how significant the income would be.

Another item that was not discussed in detail in the call, but I have found other analysts dealing with it pretty extensively, is the geographic reach of NBIS, and the fact that it is European based. The EU has put up pretty substantial barriers to the transfer of data, particularly sensitive data, which could be anywhere from corporate IP to health data. Even if the data security issues were not at play, the latency inherent in sending data to Texas and back would likely give pause to a company. I don’t know if it is in fact ‘illegal’ for a hyperscaler to handle some workloads, such as health related things, but at any rate the hoops that they must jump through are substantial. I suppose they include certifying in some way that the data never crosses EU-ex EU borders and that it is never stored. I can imagine this would be a real pain in the neck for a company that has little experience with the EU.

Additionally, NBIS has in general (or at least often) an asset light approach to its infrastructure, which fits well according to several analyses I have read with European regulatory frameworks and investment patterns. Apparently large European institutional investors and regulatory bodies are positively inclined to having the hard infrastructure funded by pension funds etc as it provides a ‘safe’ (I guess it is so thought) parking spot for funds and a steady ‘guaranteed’ long term income stream. For NBIS it broadens the base of entities that have a stake in the success of NBIS.

None of this is meant to imply that NBIS will own Europe, only that it has some structural advantages that are significant. The CEO noted several times that the market is growing far faster than NBIS, and they are not concerned with other companies feeling threatened by NBIS’ success (I guess this was an indirect reference to hyperscaler, and whether they might start going after NBIS in some way).