($) Nebius Earnings: A GPU Cloud Masterclass
Today’s post is for our premium members only. If you are reading this in full, thank you for being an Interconnected Premium member! If you aren’t, I hope you become one by scrolling down and tear down that paywall! 😎
Nebius, a GPU cloud platform, reported their Q4 earnings today. It is a special one, because it marks the first full calendar year of operations the company has gone through, after its rather unique “IPO” just 16 months ago.
Readers of this newsletter know that I have written quite a few deep dives on Nebius, from its one of a kind origin story, to how it stacks up with other GPU clouds, especially CoreWeave, to its unique position in benefiting from the “sovereign AI” theme in Europe. (See the bottom of this post for links to my growing “library” of Nebius writings.) I am transparent about holding a rather large position in Nebius in my fund. (See my 2025 annual letter if you missed it.)
But you don’t have to like Nebius to benefit from today’s post. Even though I like it, I can see many reasons why investors or analysts of the AI industry do not like it or at least have skepticisms about it in a crowded and noisy field. However, as long as you are involved in some way in the world of AI and new data center constructions, which are becoming an increasingly large part of GDP growth, you should listen to Nebius’s earnings calls.
The company is more transparent, and more knowledgeable, than most its competitors and peers about what actually goes into building a GPU data center from scratch and why. It is quite a master class of the logistical sequencing, unit economics, and long term profitability of a GPU data center.
Here are three things I picked up from listening to today’s call and reading Nebius’s shareholder letter that you likely wouldn’t find anywhere else. Rich, useful information, just hidden in plain sight.
