Offshore
Photo
memenodes
Imagine receiving salary in bitcoin and it starts dumping immediately https://t.co/D5VuWzHI1y
tweet
Imagine receiving salary in bitcoin and it starts dumping immediately https://t.co/D5VuWzHI1y
tweet
Offshore
Photo
WealthyReadings
$ADBE is trading at its lowest multiples ever.
$LULU is trading at its lowest multiples ever.
$PYPL is trading at its lowest multiples ever.
$NVO is trading at its lowest multiples ever.
Which other name deserves to join that list? https://t.co/3iqWMj7Hqq
tweet
$ADBE is trading at its lowest multiples ever.
$LULU is trading at its lowest multiples ever.
$PYPL is trading at its lowest multiples ever.
$NVO is trading at its lowest multiples ever.
Which other name deserves to join that list? https://t.co/3iqWMj7Hqq
tweet
Offshore
Photo
Dimitry Nakhla | Babylon Capital®
$NVDA & $SNPS just announced a notable strategic partnership to overhaul how modern electronics are built
$NVDA is investing $2B into $SNPS, signaling that they view $SNPS software as the critical layer needed to design the next generation of chips and physical systems
By rewriting $SNPS engineering tools to run on $NVDA powerful AI chips, they aim to reduce design cycles from years to months
The collaboration focuses on “Agentic AI”—assistants that autonomously complete engineering tasks—and “Digital Twins,” allowing companies to fully test products virtually before building them physically
This deal essentially gives engineers an edge, combining the best AI hardware with the best design software to accelerate innovation across industries like automotive, aerospace, and semiconductors
https://t.co/LxF7fZtnlN
tweet
$NVDA & $SNPS just announced a notable strategic partnership to overhaul how modern electronics are built
$NVDA is investing $2B into $SNPS, signaling that they view $SNPS software as the critical layer needed to design the next generation of chips and physical systems
By rewriting $SNPS engineering tools to run on $NVDA powerful AI chips, they aim to reduce design cycles from years to months
The collaboration focuses on “Agentic AI”—assistants that autonomously complete engineering tasks—and “Digital Twins,” allowing companies to fully test products virtually before building them physically
This deal essentially gives engineers an edge, combining the best AI hardware with the best design software to accelerate innovation across industries like automotive, aerospace, and semiconductors
https://t.co/LxF7fZtnlN
tweet
WealthyReadings
RT @MarkosAAIG: Okay guys. Very interesting conversation about $NBIS and $IREN by @Agrippa_Inv and @aleabitoreddit.I’ve read it with a lot of pleasure, and I’m writing this response also with a big smile. What’s happening in this thread is that two analyses and I would say, decent-level analyses are landing on totally different outcomes because they’re both built on completely different underlying mechanics. Let me explain..
@aleabitoreddit is looking at $NBIS as a full cloud provider, which I will talk about in a bit, and Agrippa is looking at Nebius as a GPU wholesaler to $MSFT, which is also true but misses an essential part. So the irony is that both of you are right inside your own frame, but each frame only captures half of the business. And I will fill in the missing half, where the actual economics sit in my opinion.
So,@aleabitoreddit mostly views the part of Nebius that behaves like a real full-stack cloud, with an enterprise-facing infrastructure that earns higher revenue per GPU-hour because it runs, for example, orchestration, compliance, inference, RBAC which is the full stack that enterprise-level companies pay for. Let’s call this business line A. That part of his analysis is directionally correct, but not for the Microsoft contract. It is how Nebius wants to position themselves for other customers.
Where @aleabitoreddit model goes wrong is when he tries to drag those cloud economics into the Microsoft deal. This contract is GB300 on a per-GPU structure, with a very different risk and margin profile than simply providing full services to another tenant. And like @Agrippa_Inv said, Microsoft is not using Nebius for their full-stack solutions as an addition to Azure.
Public evidence strongly supports the idea that the Microsoft–Nebius contract is intended to expand Azure capacity. The Nebius press release describes the deal as delivering “dedicated capacity to Microsoft,” and industry coverage and analysts all frame it as a way for Microsoft to relieve Azure’s capacity constraints not merely to supply internal training infrastructure. That implicitly means Microsoft is going to run their Azure workloads, including Azure-AI/LLM workloads, on this Nebius capacity.
And adding to that like @Agrippa_Inv pointed out: Nebius’s communication is very vague and not fully transparent, something I really hate also. So I fully agree with him on that part. But in terms of GPUs, I personally think it’s fine to reference that they have access to around 100,000 GB300 GPUs, because according to Bloomberg-sourced reports, and even taking into account that these are external this number is repeated across multiple industry sources. So while it’s not an official Nebius disclosure, I think we can reasonably assume that the scale is in that ballpark.
Moving on, @Agrippa_Inv is correct about the wholesale nature of the Microsoft agreement. It’s a GPU contract, not a megawatt contract, as I just talked about. So you can’t assert cloud-like gross margins on a hyperscaler deal that is structurally set up more like a fixed-price infrastructure supply. @Agrippa_Inv IRR logic, with his emphasis on prepayment and focus on leverage dynamics, all of that makes a lot of sense to me for the wholesale business proposition. And of course, Nebius does this kind of business also, as you can see in the Microsoft contract, so let’s call this business line B.
The thing is: in his critique, he states a correct point about the Microsoft contract so wholesale (business line B). But then he tries to turn it into a generalized statement about Nebius’s whole business, which I just assessed as being business line A and business line B. That’s not how this company is structured. So in my opinion, that’s where he is wrong. Read further below.👇
This is incredibly poorly researched.
At least you don't charge money for you work, because this is bottom of the barrel stuff.
1) You don't even get the $NBIS GPU models right that are contracted to $MSFT. It'[...]
RT @MarkosAAIG: Okay guys. Very interesting conversation about $NBIS and $IREN by @Agrippa_Inv and @aleabitoreddit.I’ve read it with a lot of pleasure, and I’m writing this response also with a big smile. What’s happening in this thread is that two analyses and I would say, decent-level analyses are landing on totally different outcomes because they’re both built on completely different underlying mechanics. Let me explain..
@aleabitoreddit is looking at $NBIS as a full cloud provider, which I will talk about in a bit, and Agrippa is looking at Nebius as a GPU wholesaler to $MSFT, which is also true but misses an essential part. So the irony is that both of you are right inside your own frame, but each frame only captures half of the business. And I will fill in the missing half, where the actual economics sit in my opinion.
So,@aleabitoreddit mostly views the part of Nebius that behaves like a real full-stack cloud, with an enterprise-facing infrastructure that earns higher revenue per GPU-hour because it runs, for example, orchestration, compliance, inference, RBAC which is the full stack that enterprise-level companies pay for. Let’s call this business line A. That part of his analysis is directionally correct, but not for the Microsoft contract. It is how Nebius wants to position themselves for other customers.
Where @aleabitoreddit model goes wrong is when he tries to drag those cloud economics into the Microsoft deal. This contract is GB300 on a per-GPU structure, with a very different risk and margin profile than simply providing full services to another tenant. And like @Agrippa_Inv said, Microsoft is not using Nebius for their full-stack solutions as an addition to Azure.
Public evidence strongly supports the idea that the Microsoft–Nebius contract is intended to expand Azure capacity. The Nebius press release describes the deal as delivering “dedicated capacity to Microsoft,” and industry coverage and analysts all frame it as a way for Microsoft to relieve Azure’s capacity constraints not merely to supply internal training infrastructure. That implicitly means Microsoft is going to run their Azure workloads, including Azure-AI/LLM workloads, on this Nebius capacity.
And adding to that like @Agrippa_Inv pointed out: Nebius’s communication is very vague and not fully transparent, something I really hate also. So I fully agree with him on that part. But in terms of GPUs, I personally think it’s fine to reference that they have access to around 100,000 GB300 GPUs, because according to Bloomberg-sourced reports, and even taking into account that these are external this number is repeated across multiple industry sources. So while it’s not an official Nebius disclosure, I think we can reasonably assume that the scale is in that ballpark.
Moving on, @Agrippa_Inv is correct about the wholesale nature of the Microsoft agreement. It’s a GPU contract, not a megawatt contract, as I just talked about. So you can’t assert cloud-like gross margins on a hyperscaler deal that is structurally set up more like a fixed-price infrastructure supply. @Agrippa_Inv IRR logic, with his emphasis on prepayment and focus on leverage dynamics, all of that makes a lot of sense to me for the wholesale business proposition. And of course, Nebius does this kind of business also, as you can see in the Microsoft contract, so let’s call this business line B.
The thing is: in his critique, he states a correct point about the Microsoft contract so wholesale (business line B). But then he tries to turn it into a generalized statement about Nebius’s whole business, which I just assessed as being business line A and business line B. That’s not how this company is structured. So in my opinion, that’s where he is wrong. Read further below.👇
This is incredibly poorly researched.
At least you don't charge money for you work, because this is bottom of the barrel stuff.
1) You don't even get the $NBIS GPU models right that are contracted to $MSFT. It'[...]
Offshore
WealthyReadings RT @MarkosAAIG: Okay guys. Very interesting conversation about $NBIS and $IREN by @Agrippa_Inv and @aleabitoreddit.I’ve read it with a lot of pleasure, and I’m writing this response also with a big smile. What’s happening in this thread is…
s GB300s, not H200s. That's the most basic stuff & almost impossible to get wrong.
2) You got no clue about $NBIS's CAPEX spend regarding the $MSFT deal, because management NEVER publicized it. Management didn't even disclose how many GB300s are contracted for this deal.
As a result, your GPU Model, GPU Cost / Unit, GPUs / MW, GPU CaPex / MW data is all off when it comes to $NBIS.
Even your GPU depreciation estimate of "4 years" is totally off. For god's sake, both $IREN & $NBIS just signed FIVE year deals with $MSFT, how can you assume 4 years of depreciation... make it make sense.
It's also not correct to compare the deals based on revenue per MW.
Microsoft is NOT contracting MWs, this is not a colocation deal. This is a CLOUD deal. They're exclusively paying for cloud capacity, i.e. for the numbers of GPUs contracted.
So, the only reasonable yardstick to compare revenue across both deals is on a per GPU basis.
As I just pointed out, $NBIS never gave us any data on GPU quantity (lack of transparency is a big red flag), so we can only speculate how many GPUs Nebius must purchase for their agreement.
We know that $IREN fits 19k GB300s at its 50 MW facilities. But I'm also aware that $IREN left white-space capacity in their 50 MW Horizon data centers, for reasons I went into in my recent Deep Dive on the deal.
So, we can't use $IREN to extrapolate on Nebius's GPU quantity.
The best case study to extrapolate from is the Nscale x Microsoft deal (also GB300s). Nscale disclosed that it's fitting ~22-23k GB300s across 50 MW.
Using that same GPU / MW range for Nebius, we get to a total quantity of 132-138k GB300s for 300 MW IT.
On a revenue per GPU that leads us to:
$IREN = $127.6k
$NBIS = $126.1k - $131.8k
In other words, $NBIS & $IREN get paid the same for their deals with Microsoft.
The only difference is that $IREN got a massive 20% prepayment upfront, while $NBIS did not.
$IREN will be able to use that prepayment to lower their CAPEX burden and decrease financing costs.
Meanwhile, $NBIS got nothing, or else I'm sure they would have advertised it to shareholders.
On top of that, Nebiushas to pay DataOne a big cheque for using its facilities, while $IREN is effectively their own landlord due to vertical integration (owning their self-built data centers).
The irony here is that $IREN clearly got the much better deal, WITHOUT any "software" stack...
I got news to all of you $NBIS investors:
Microsoft isn't paying 1 cent for Nebius's software. NOWHERE in the 6k filings does $NBIS even state that they are charging Microsoft for software, it's very clearly outlined as a hardware (i.e., bare metal) deal.
Come on... do some work once in your life instead of relying on X info. Go on the Nebius IR page & download the Microsoft deal 6k filing and read it for yourself or run it through your favorite LLM and see for yourself.
All the $NBIS shillers on X are doing you a disservice for claiming otherwise.
It's also ridiculous to think otherwise. Think about it for 1 moment. Do you seriously think Microsoft, an S-tier SOFTWARE provider running one of the largest cloud services on the planet (Azure), needs help from $NBIS in the software department?
Come on...
I had discussions with many software engineers who work at hyperscalers, all of them laugh at the idea of gullible $NBIS investors seriously believing that Microsoft is outsourcing software from them.
The irony here is that, $NBIS management could easily disprove any doubt, by publishing hard data on the Microsoft deal.
But guess what?
They haven't done so.
Why could that be? Why is a company like $IREN so transparent with their respective deal, but $NBIS doesn't even disclose the number of GPUs it must purchase to fulfill their end of the bargain?
It's pretty obvious why that's the case.
If you calculate how much $ $NBIS stands to make on their deal (using the 132-138k GPU estimates), they are basically just breaking even on the ha[...]
2) You got no clue about $NBIS's CAPEX spend regarding the $MSFT deal, because management NEVER publicized it. Management didn't even disclose how many GB300s are contracted for this deal.
As a result, your GPU Model, GPU Cost / Unit, GPUs / MW, GPU CaPex / MW data is all off when it comes to $NBIS.
Even your GPU depreciation estimate of "4 years" is totally off. For god's sake, both $IREN & $NBIS just signed FIVE year deals with $MSFT, how can you assume 4 years of depreciation... make it make sense.
It's also not correct to compare the deals based on revenue per MW.
Microsoft is NOT contracting MWs, this is not a colocation deal. This is a CLOUD deal. They're exclusively paying for cloud capacity, i.e. for the numbers of GPUs contracted.
So, the only reasonable yardstick to compare revenue across both deals is on a per GPU basis.
As I just pointed out, $NBIS never gave us any data on GPU quantity (lack of transparency is a big red flag), so we can only speculate how many GPUs Nebius must purchase for their agreement.
We know that $IREN fits 19k GB300s at its 50 MW facilities. But I'm also aware that $IREN left white-space capacity in their 50 MW Horizon data centers, for reasons I went into in my recent Deep Dive on the deal.
So, we can't use $IREN to extrapolate on Nebius's GPU quantity.
The best case study to extrapolate from is the Nscale x Microsoft deal (also GB300s). Nscale disclosed that it's fitting ~22-23k GB300s across 50 MW.
Using that same GPU / MW range for Nebius, we get to a total quantity of 132-138k GB300s for 300 MW IT.
On a revenue per GPU that leads us to:
$IREN = $127.6k
$NBIS = $126.1k - $131.8k
In other words, $NBIS & $IREN get paid the same for their deals with Microsoft.
The only difference is that $IREN got a massive 20% prepayment upfront, while $NBIS did not.
$IREN will be able to use that prepayment to lower their CAPEX burden and decrease financing costs.
Meanwhile, $NBIS got nothing, or else I'm sure they would have advertised it to shareholders.
On top of that, Nebiushas to pay DataOne a big cheque for using its facilities, while $IREN is effectively their own landlord due to vertical integration (owning their self-built data centers).
The irony here is that $IREN clearly got the much better deal, WITHOUT any "software" stack...
I got news to all of you $NBIS investors:
Microsoft isn't paying 1 cent for Nebius's software. NOWHERE in the 6k filings does $NBIS even state that they are charging Microsoft for software, it's very clearly outlined as a hardware (i.e., bare metal) deal.
Come on... do some work once in your life instead of relying on X info. Go on the Nebius IR page & download the Microsoft deal 6k filing and read it for yourself or run it through your favorite LLM and see for yourself.
All the $NBIS shillers on X are doing you a disservice for claiming otherwise.
It's also ridiculous to think otherwise. Think about it for 1 moment. Do you seriously think Microsoft, an S-tier SOFTWARE provider running one of the largest cloud services on the planet (Azure), needs help from $NBIS in the software department?
Come on...
I had discussions with many software engineers who work at hyperscalers, all of them laugh at the idea of gullible $NBIS investors seriously believing that Microsoft is outsourcing software from them.
The irony here is that, $NBIS management could easily disprove any doubt, by publishing hard data on the Microsoft deal.
But guess what?
They haven't done so.
Why could that be? Why is a company like $IREN so transparent with their respective deal, but $NBIS doesn't even disclose the number of GPUs it must purchase to fulfill their end of the bargain?
It's pretty obvious why that's the case.
If you calculate how much $ $NBIS stands to make on their deal (using the 132-138k GPU estimates), they are basically just breaking even on the ha[...]