Today is the day I join the Substack family, and I have two main reasons:
1) Substack offers better options for sharing longer and more complex work. I want to share, for example, my Excel files, where I have all my calculations, and that works better on Substack than through X articles
2) Feedback from my followers - I was surprised, but many people already asked me if I wanted to create an account on Substack. I assume it is easier for people to read and share than using X.
Please understand that all my articles, deep dives, valuation models, and further analysis will be free for everyone. I am new to Substack, and I must say the fact that the first thing it helps you with is setting up subscriptions is comical. I have some preset settings for subscriptions, but you don’t need to subscribe, as all posts will be open to everyone!
The only limitation I actually set up myself is that commenting is open only to subscribers (the free tier is sufficient) to avoid annoying people.
Feel free to share this so people can follow. Already uploaded my updated Nebius Thesis, and tomorrow I will release the margin impact analysis regarding Token Factory’s new joiner, Inferize.
Thanks for all the support!
https://t.co/dOoF3g0QkQ
$NBIS Q3 Preview
Instead of celebrating like everyone else (which is hilarious because we are still down 28% last month), I will offer a few words regarding tomorrow’s earnings. These are the things I will be paying the most attention to:
1) 2027 ARR number
The majority of us expect this number to be revealed by management, and if they do so, it will definitely dominate the whole earnings report. Consensus was around $16-$17B last week, and I saw some numbers that were closer to $20B.
I expect that Nebius will guide between $23-$28B, which itself is a big surprise to the upside. The actual number will be much higher, let me tell you that.
2) Asset-light model
Nebius introduced it in July, and I would appreciate if we could finally get more information about potential clients and how all of this is connected to the $PLTR partnership (because it is). I understand that during Q2 earnings, it was way too early to give any numbers/direct expectations, but I definitely look forward to getting more clarity on where this is heading.
3) Token Factory numbers
I am not sure we are getting that this quarter, but if we do, it is immediately the headline numbers for me. This is what differentiates Nebius from others the most, and I want some numbers. Inference players like Baseten, Fireworks, and Together all have crazy valuations in private markets, and Nebius won’t get a significant premium for TF unless they provide some actual numbers. An analyst can tell you their business is well diversified, but without numbers, you can’t do much more.
4) Customer prepayments
I expect Nebius to be at a similar level to $CRWV, but I would appreciate more clarity. CoreWeave included that bullish sentence about getting 100% prepayment on a contract, but that’s all we know, which has literally zero implications for how we can model their balance sheet (it is still bullish, yes!).
5) Connected/Active power
I would really appreciate it if we could get more clarity on what is the current connected/active power. My estimates remain the same for connected power around 1GW by the end of 2027 and ~550- 600 MW of active power. However, it would be nice if Nebius stated what is their active/connected power exiting Q3.
Overall, expecting a double beat and a strong print tomorrow!
(My valuation models will be updated, and I will share with you in case something changes dramatically!)
CoreWeave $CRWV CEO Michael Intrator, asked on CNBC why it isn't buying back bonds yielding ~15%, said:
"I can deploy the capital at a higher rate of return."
Jim Cramer pressed whether it can really earn over 16%.
Intrator also said he canceled his 10b5-1 stock trading plan.
$CRWV x $IREN prepayments
Well, it is very important to stress that they said exactly that: “100% of the contract value upfront.”
We have only this one sentence and nothing more. This means that it could very well be just one contract for some H100 GPUs. Unless we have more information, it is a nice headline sentence, but we have no idea about its implications for their balance sheet.
Not bullish/bearish post, just trying to explain the facts here.
$CRWV Prepayment is Stronger than $IREN's
Reading carefully "100% of the contract value upfront" as pointed by @_Sgr_A_Star means not only GPU capex but DC colocation (CRWV uses mostly colo) value over the duration of contract and the profit margin.
This is much stronger than IREN's 90% of GPU Capex which is around 65% of total contract value.
Hope $IREN steps it up for set of contracts on either the prepayment front or economics.
$NBIS $IREN
I will write my summary of $CRWV call tomorrow. The overall compute market seems to continue in the supply/demand imbalance trajectory, which is definitely bullish.
However, I expect more details from the call and how the think about the increasing CDS yields…
Still not convinced that their model is feasible and I remain bearish $CRWV
$CRWV
A lot of words with not much meaning for the broader market. I don’t think that the market is persuaded that they are wrong about this sell-off based on $CRWV’s results and the call.
Nebius will have to do more. Otherwise, I don’t see any sentiment shift to be fair...
$CRWV
"We now expect full-year 2026 capital expenditures of $37 billion to $39 billion, raising the low end of our prior guidance range of $35 billion to $39 billion. This is as a result of our expectation to deploy infrastructure and deliver it to customers sooner than planned."
So CoreWeave is expecting to spend only $4.3 - $8.3B during Q4? Interesting...