re: Nebius - NBIS - AI Infrastructure Company

We about to slip under 190 here?

I remember being ecstatic when we hit 116 in march.. just tryin to make myself feel better

Recap:
Sold all NBIL and calls at slight gains this morning.

Because I needed a loss in my brokerage, I sold 111 shares of OUST that I bought at $44.

I’m so bullish on the NBIS business model that I bought 21 shares in that account.

Then I bought three $260 strike Nov calls at 40.70.

Just wait until Wall Street digests this strategy…

quote:

---

Just wait until Wall Street digests this strategy…

---

Edited for clarification and spelling:


I have to be honest I'm pretty confused by the strategy - seems contradictory to the vertical integration argument they've made for the year. I understand they have great software.

I'm not sure what their target audience is to be honest. I mean, I know what the press release says, but I don't know why a data infrastructure wouldn't just go bare metal with a hyper scaler or color of a hyper scaler. Again if it's just a side shot at business that's fine but it's dramaticallydifferent than their story.

Again if an incremental side gig - fine, but I am struggling to see how transforming bitcoin miners are going to be able to raise billions of dollars to carry the capex and then use NBIS software... to do what? They are not going to be clouds. 95% of these small AI wannabe's are going to go colo or bare metal - in fact they have to to get financing to build out.

The one exception is Iren which has already tethered to Nvidia to flagship the Nvidia software (RunAI) that comes with the DSX set up so they are not a likely customer.

I guess if a bunch of the small companies want to try to be token factories, it would make sense but they are not getting funding without an investment grade customer - and so many of the investment grade customers do not want anything beyond bare metal as they have their own software.

Not saying this as a bear - I have 1600 shares of NBIS and a few options so I am fully allocated lol but if this is truly a transition (I do not think it is) then I disagree with where the bottleneck is. IMO, the bottleneck is not software with an abundance of data centers and power available just looking to be more efficient. The bottleneck is power and data centers.

Gemini:
Under the new asset-light partnership model Nebius announced, there are actually two completely different types of partners buying this hardware, and they have very different motivations.

THE YIELD SEEKERS (COMMERCIAL INFRASTRUCTURE PARTNERS)
For this group, the answer is no. They are not buying graphics processing units (chips) to train their own models or build their own agents.
These partners are data center developers, regional operators, and infrastructure investors. They have land, power, and capital, but they lack the expertise to design a modern artificial intelligence cluster, secure a direct chip allocation from Nvidia, or write the complex orchestration software needed to run it.

They buy the physical assets purely as a real estate and hardware yield play. Nebius installs its software stack and connects the site to its own global sales pipeline. Nebius then routes its massive enterprise tenants (like Meta, Microsoft, or newly signed startups like Reflection AI) to the partner's data center. The partner gets a highly predictable, high-yield return on their physical investment without having to build a sales team or write a single line of cloud software.

THE SOVEREIGN AND ENTERPRISE FORTRESSES
For this group, the answer is a resounding yes. These are the sovereign nations, national security agencies, and highly regulated industries (like pharma or banking) you pointed out earlier.

These entities cannot send their proprietary data to a shared public cloud. They must own the physical servers, control the local power supply, and ensure the data never leaves their geographic borders or private networks.
However, building a highly efficient artificial intelligence cloud from scratch is incredibly difficult. Instead of trying to reinvent the wheel, they buy the physical infrastructure and license Nebius's proprietary software stack and system designs. This gives them a ready-to-use, top-tier artificial intelligence cloud environment inside their own private, self-funded walls.

So I could see Blue Cross Blue Shield or OLOL having a 5MW campus for themselves whereas Blackrock may go in for 200MW as an investment, where they are renting out Nvidia GPUs to “name that AI Company” procured by Nebius.

quote:

---

The knock on every neocloud has always been the same, capex hungry, debt funded, growth capped by how many data centers you can physically finance and build. This model breaks that ceiling.

Nebius can now scale globally using other people’s money, generating what they describe as high margin revenue with minimal incremental capital.

It’s the hyperscaler playbook flipped. Instead of racing to own every megawatt, Nebius becomes the software and demand layer sitting on top of everyone else’s infrastructure.

The platform becomes the product. That’s a higher margin, higher multiple business than renting GPUs, and it’s exactly the kind of model shift markets re-rate.

And the demand side is already proven. Meta at up to $27B, Microsoft at $17B, Reflection’s $1B+ deal yesterday. Nebius has more customers than capacity.

This model turns that exact problem into the growth engine, every partner site becomes new supply for a sales pipeline that’s already overflowing.

They’ve already signed initial arrangements under the model. This isn’t a concept slide. It’s live.

The market saw a press release. It should have seen a business model upgrade.

---

Loading Twitter/X Embed...
If tweet fails to load, click here.

So how is going to build everything for said school district and procure the power for them?

They would have to outsource every aspect - wouldnt they look for an all in one solution?

Again - my bias is the bottleneck is power and data centres - this strategy assumes otherwise and those will just magically appear and need software.

Those sovereign nations, and companies and governments will need everything done for them. Not that nebius cannot do anything, but that is not this strategy.

quote:

---

Those sovereign nations, and companies and governments will need everything done for them.

---

Those who want it will build it, and Nebius won’t have to worry about procuring power or delays.

In the meantime, they’ll simply continue their current build out.

I’ll be shocked if they don’t name a partner on the earnings PR.

Power is the bottleneck. You’re right.

But Nebius’s demand proves business success. They have more customer requests than they have infrastructure. They turn people away.

What better evidence is there for the level of respect that enterprises have for NBIS software layer than the customers themselves taking on the burden of supplying their own land and power to get access to the benefits of Token Factory?

What does that say about all of the neoclouds who have more infrastructure than they have customers?
Their product isn’t as desirable.

Land and electricity do not spit out elite LLM’s.

What we will see is that people want an elite software stack with virtually no downtime and all the regulatory compliance so bad, that they’ll do anything to get it—buy land, buy GPUs, supply power.

This will be extremely interesting to watch.

I of course hope I am understating and wrong lol

quote:

---

Those who want it will build it, and Nebius won’t have to worry about procuring power or delays.

---

Which they would do before this and after this - I do not see how this changes that dynamic. I do not see someone saying the reason we were not building is we could not get access to software. As we all know it will take more than a few years these days to build if you start now, and I do not think this is what is holding people back.

quote:

---

But Nebius’s demand proves business success. They have more customer requests than they have infrastructure. They turn people away.

---

There is no question on that - although I am not sure who these days has an idle deployed GPU.

quote:

---

What better evidence is there for the level of respect that enterprises have for NBIS software layer than the customers themselves taking on the burden of supplying their own land and power to get access to the benefits of Token Factory?

---

If that happens - yes.

quote:

---

What does that say about all of the neoclouds who have more infrastructure than they have customers?
Their product isn’t as desirable.

---

Who has idle GPUs?

quote:

---

What we will see is that people want an elite software stack with virtually no downtime and all the regulatory compliance so bad, that they’ll do anything to get it—buy land, buy GPUs, supply power.

---

You mention this like it is not the bottleneck issue. Again, if you believe there are significant number of Companies out there that are incredibly anxious to build their own data center and really want to, but the only thing holding them back to date is they do not think Nvidia Run AI software alond would be enough.... I just do not see that. Companies do not build data centers (now or in the original cloud days) for a multitude of reasons and right now the practicality is if someone decided today to build it - it will be years until fruition for the need for software.

Now again, I could be completely wrong and it would not be the first time - but I do not see Companies engaging in multi billion contstruction projects because of software availability.