Premium: A deep dive into Nebius

After going deep into neocloud pioneer CoreWeave across November, it's time to finally take a deeper look at its competitor Nebius. Like with CoreWeave, this turned out to be way longer than expected, so it will be a multi-part series to close out 2025. [Sorry, my IREN dive is delayed to next month. Please don't punish me like the market punished CoreWeave on its delay!]

Nebius was born out of the split-up of Yandex, the leading search engine & hyperscaler cloud in Russia, commonly referred to as the "Google of Russia". This wealth of experience from 2+ decades of rising into a tech giant has allowed Nebius to hit the ground running in their new ambition – to rebuild and reemerge as a different kind of hyperscaler cloud.

I find Nebius to be the polar opposite of CoreWeave in many areas of its strategy, such as building up its cloud software, GPU fleet, and customer base first, and getting into whale deals later. I also find they have a slightly more cautious approach in some ways (margin-over-all) and less in others (expanding capacity in advance of commitments in hand).  

The biggest issue in their trajectory against CoreWeave is that they got a later start. As of Q325, Nebius is roughly at the capacity/revenue point where CoreWeave was at in Q124 (~1.5yrs behind), and CoreWeave has been busy moving fast over that year and half to line up its runway for future capacity across 2026-2028 from here. My main criticism of Nebius over 2025 is that they needed to move faster. But as of their latest Q325 report, I think they are drastically improving on this front, after upping their capacity goals 2.5x and their ARR goals 7-9x for 2026 – thanks in large part to two whale deals.

So let's take a long walk through what makes Nebius tick. This will also be of interest to neoclouds (CoreWeave, IREN), AI chipmakers (NVIDIA, AMD), hyperscaler clouds (Amazon, Microsoft, Google, Oracle), and serverless inference competitors (Databricks, Cloudflare).

• Nebius got a jump start on its neocloud ambitions from its prior 20+ years as a global hyperscaler tech giant. Besides cloud software and data center prowess, this included a long-standing relationship with NVIDIA, as well as the expertise to design their own rack and server hardware to house those GPUs.
• They are not just interested in GPUaaS. Nebius's focus from the start has been on building a "true cloud" – a hyperscaler cloud that specializes in only AI workloads, serving a wide variety of customers through an on-demand, self-service platform. Mgmt views whale deals as a way to jump-start the scale of their capacity towards those long-term ambitions. [CoreWeave landed the big whale deals first in order to build up capacity, and is now building up a self-service cloud platform (after the fact) to eventually make use of capacity that will free up in the future.]
• Because of this, Nebius is buying capacity in advance, before contracts are secured, knowing that it will strengthen its internal fleet and advance its "true cloud" ambitions. [CoreWeave only commits to new capacity with a customer commitment in hand, and has pre-existing debt structures to quickly move from there.]
• But this reversed strategic path doesn't feel particularly risky, as the demand signals for these advance purchases are there; their fleet is continually near or at peak utilization, they are pre-selling out new capacity before it comes online, and they are having to say no to customers looking for more capacity.
• So with these ambitions, why then sell off a huge chunk of coming capacity to whale customers like Microsoft and Meta? Beyond helping to fund the future GPU buildup of their long-term vision, I believe this allows Nebius to adopt other financing strategies that CoreWeave has been leveraging to move faster.
• Nebius focuses heavily on margins in every decision (buildout, procurement, customer deals, financing, etc). However, this has seemingly caused them to move more cautiously compared to pioneer CoreWeave, and, given their later start into this market (mid-2024), they remain well behind them in all metrics.
• Mgmt seems to have become much more aggressive in their growth and financing strategy over the past 2 quarters, and have massively upped their footprint/growth ambitions for 2026 and beyond.
• Expect a strong Q425 from a huge boost in capacity in their AI Cloud, and then again in Q126 as Microsoft & Meta revenue fully kick in along with the Finland expansion.
• Looking forward a year, Nebius increased its early guide for FY26 ARR from $1.1B to a whopping $7-9B in Q3, and has already booked half of it thanks to those 2 big whale deals.

Part 1:

• A different journey & approach
• Financial overview
• Customers & whale deals

Part 2:

• Their custom hardware
• Their footprint and buildout efforts thus far (and when it all hits revenue)
• Likely capacity moves from here

Part 3:

• Financing strategies
• Mgmt & GTM & Subsidiaries
• Software Stack
• TTM quarterly review