Conquering markets instead of empires. 🏰📈 Buying stocks like they're Persian territories, HODLing like it's 323 BC!
Joined October 2023
- Tweets9.7K
- Following228
- Followers8.3K
- Likes14.9K
$NUAI Great stuff!
The "AAIG Technical Series: The Road To NUAI’s TCDC Build" article is now fee to read for all subscribers. I develop and build data centers for a living and went through the whole process of what is needed to build TCDC. Hope you all enjoy, feedback is very welcome and appreciated 👍
$NUAI $NBIS $IREN
aaig.substack.com/p/aaig-tec…
$NBIS First win for the Physical AI workbench?
🚨 $NBIS $NVDA: NEBIUS AND NVIDIA ARE BRINGING AI-TRAINED ROBOTS INTO ZALANDO'S WAREHOUSES.
Sereact just announced a collaboration with NVIDIA, Nebius, CEVA Logistics and Zalando to improve how robots handle returned products in warehouses.
The initial deployment targets Zalando's returns operations at selected CEVA Logistics fulfillment centers.
Processing returned clothing is actually a difficult job for robots.
One customer sends back a neatly folded shirt. Another returns the same shirt crumpled inside a plastic bag. Some items have tags, others don't, and the products keep changing every season. A robot needs to recognize the item, pick it up, check its condition and sort it correctly.
Now here's where NVIDIA and Nebius come in.
Sereact already has its Cortex robot brain and robotic systems working with real warehouse data. When those robots make mistakes, they record what happened.
Previously, engineers had to collect that data, retrain the models, test them and repeat the process. That could take weeks.
The new system is designed to bring that process down to days.
$NVDA provides Isaac Sim, Isaac Lab and Cosmos world models to create simulated versions of real warehouse situations. $NBIS provides the Physical AI Workbench, storage and GPU infrastructure needed to run the simulations, train the models and evaluate the results.
Sereact provides Cortex, its vision language action model that controls the robots. CEVA Logistics operates the fulfillment centers, while Zalando provides the actual returns-handling use case.
Now imagine a robot struggles to pick up a crumpled jacket. Instead of waiting for engineers to manually collect hundreds of similar examples, the system can take that failure and use NVIDIA's simulation tools to generate thousands of variations.
$NBIS Just like the Volt project in Rotterdam Harbour here Nebius is also the anchor customer, which means a big discount on the colocation lease rate and as first PIF backed @HUMAIN commercial customer a new entry into sovereign compute for Saudi Arabia!
$NBIS BREAKING:
Nebius signed a preliminary framework with HUMAIN(One of Saudi Arabia's largest sovereign wealth funds) and DataVolt to be the anchor customer at NEOM's Oxagon AI campus.
The details:
• Up to 50MW, half of the first 100MW now under construction
• Phase 1 is planned at 360MW total
• Capacity comes online in phases starting Q4 2027
• Nebius is HUMAIN's first customer
• Signed at NVIDIA GTC Berlin
It's still preliminary, so not a done deal yet. But Saudi Arabia's sovereign fund is picking Nebius as its first customer.
$NBIS Probably also linked to the strategic Palantir partnership, which also included deployment of modular datacenters!
BREAKING: NVIDIA introduces forward-deployable, containerized modular AI data centers.
The company announced DSX POD, a factory-manufactured modular data center design intended to reduce deployment times from six months to as little as 14 days at prepared sites. The systems feature pre-integrated computing, networking, cabling, and liquid cooling.
Key partners include @armada_ai and @anduriltech, among others, with initial deliveries expected in early 2027.
ALT Source: NVIDIA
Alexander retweeted
This part is so underestimated by the market... and it's a reason why Nebius trades much higher than its peers. Last week we spoke with a GPU cluster founder who has been deploying compute at scale since the mining days, over 150,000 GPUs. He walked us through why this gap is so hard to close for the smaller players. A lot of them have a background in crypto mining, where you never had a customer to serve. You just do it for yourself. If it's down, it's down... you fix it and you're operational again.
So no high-level automation, and every new customer means an engineer configuring the switch, rebooting the servers and installing the drivers by hand. That's a huge process they have to go through when they pivot from mining into building infrastructure for customers. Now, some do it better than others. You don't have real customer feedback either, because if you have one or two big contracts, the client already knows what it wants and teaches you nothing. For a big part, you work off their expertise. Of course, you learn from it, but it's different than handling tens or even hundreds of customers. So what they do is outsource big parts of the stack, fight over whatever talent is left, and end up with the B, C and D grade contracts.
The A grade contracts go to the ones that own the full stack, have automated it, and actually know how to service a customer. And that's the big moat behind $NBIS they've been handling this since the Yandex days.
Nebius and CoreWeave fall in that leader group. Iren $IREN is making steps to get there I'll give them time. WhiteFiber $WYFI same but they also need more to prove. lot of other smaller players I have strong doubt if they will survive.
“A couple of hyperscalers have actually asked us about our experience (with the new pricing mechanisms). We’re getting a lot of interest and attention across the board.”
— Marc Boroditsky, CRO of $NBIS
$ASTS Again this is an AI connectivity platform from space!
Telecom networks are becoming platforms for AI.
We’re working with NVIDIA to advance AI-powered radio processing and autonomous network operations, while operators expand AI computing from data centers to the network edge.
Read the latest from NVIDIA: nvidianews.nvidia.com/news/n…
#AI #AIRAN #6G
$ASTS This gives us the entire year of 2027 and Q1 2028 to get the constellation up. Seems reasonable!
Mobile World Live: SpaceX targets early 2028 for own D2D service launch mobileworldlive.com/starlink… @mobileworldlive
Alexander retweeted
“A couple of hyperscalers have actually asked us about our experience (with the new pricing mechanisms). We’re getting a lot of interest and attention across the board.”
— Marc Boroditsky, CRO of $NBIS
Alexander retweeted
“People have been forecasting that supply will eventually catch up with demand. That’s just not happening.
We’re hearing from customers about what they’ll need in late 2027 and early 2028. We don’t have enough capacity to support them.”
— Marc Boroditsky, CRO of $NBIS
$IREN Recognition from NVIDIA!
"NVIDIA DSX is full stack AI factory reference architecture, co-innovated with the world’s leading infrastructure companies."
IREN sites in Jensen Huang's #NVIDIAGTC Berlin keynote. Watch the full presentation: youtube.com/watch?v=PvFjbPYE…
$NBIS $IREN Anyone except SpaceX, Nebius and IREN already sold their soul (2027/2028 compute) to frontier labs and hyperscalers. Elon just validated the winning strategy! Shitco’s like $CRWV , Nscale are almost completely sold out for 2027 and sold theirs for bargain pricing. Congratz also to @danroberts0101 who say this very early as well and almost sold no 2027 capacity yet!
Sharing this answer from @elonmusk at $SPCX conference call last night:
Q: "When you think about a framework for applying that compute to the external opportunity that presents itself from cloud services relative to the internal needs of that compute for scaling Cursor and Grok and Grok Bot. How has your view evolved about striking the right balance in that mix, and how to think through the output from a monetization standpoint of what that mix might produce for the business in the next 12 to 18 months?"
Elon: "Our belief is that the value of compute per watt will increase over time, not decrease. The reason for that, I think it should be obvious to people really, is that as you improve the AI software, as the models get better and better, you get more useful intelligence per watt.
The value of that compute is proportionate to the useful intelligence per watt. That is why you see things like even old computers, like the H100s, renting for more than they rented when they were new. I think that is likely to continue. I think we are going to face a shortage of super intelligence compute, and that shortage is going to remain for a long time. I think the smart move is to do short-term contracts and essentially to place the bet on useful intelligence per watt increasing rapidly, which obviously that is what has happened so far. Also if you look at, say, how are our short-term contracts renewing? Are they renewing at higher dollars per watt or lower? We see them renewing at higher dollars per watt, not lower dollars per watt.
We are also, frankly, hounded relentlessly by every hyperscaler and by dozens of smaller companies to rent more of our compute. We could rent out our compute several times over if we said yes to everyone. That is just the situation here. That is why I posted that video meme of the Monty Python Cheese Shop sketch, where the guy goes into a cheese shop to get some cheese, and they actually have no cheese.
He goes to like, "What about this kind of cheese? What about that kind of cheese?" They really just do not have any cheese. We have actually tried to offload some of our inference compute to other cloud providers so that we can use more compute for training. We get the cheese shop sketch. We will ask them about, what about some GB300s? "Nope, do not have any of those." What about some GB200s?
Nope, do not have any of those." What about some TPUs? "Nope." What about some Trainiums? "Nope." Okay, how about some H100s? They are like, "Well, we might have some overseas somewhere." Then they will check and come back and say, "Nope, we do not have that either." It is ridiculous. I do not think we are going to have a demand issue here. I am not saying this will happen, it is not impossible that we will have a six-month payback period for compute next year."
$NBIS Good recap👇
GTC week, @nebiusai edition 👇
• $10M for Europe’s AI builders
• Arkady live with @GavinSBaker
• @NVIDIA × Nebius Academy courses
• Build Week: workshops, hackathons + Reichstag rooftop
• Token Factory × Tavily + Physical AI demos
• 3 GTC talks: AI cloud, @Shopify + security
• Nebius LIVE with @dee_bosa
• Open Source AI Salon with @Kimi_Moonshot
• Nebius House: breakfasts, watch party + afterparty
• Vera Rubin technical briefing
• Live GPU spot pricing
• PyTorch Foundation Platinum membership
Berlin is busy 🟢 #NVIDIAGTC
Alexander retweeted
@GavinSBaker promotes @nebiusai spot compute auction in the GTC live:)
$NBIS Jensen shoutout!
“Startup companies like Nebius are doing incredibly well.” - @JensenHuang, @nvidia CEO at #NVIDIAGTC Berlin keynote
$ASTS $SPCX Every Elon shill and SpaceX fanboy point out the difference between Starshield and Starlink. That’s the novice take.
Starshield is owned by US DOD including Mission Control, HOWEVER Starshield is operated via a SpaceX satellite operations center via a contractor relationship with the US DOD.
It are SpaceX employees that processed the speed limit in Ukraine via the SpaceX SOC. Elon is their boss and can give those engineers orders as well.
The question is did Elon act unilaterally and break his contract and later tried to fix that with the White House or did the White House give the order.
Regardless Elon and SpaceX are involved one way or the other.
US Secretary of State Marco Rubio said that decisions regarding Ukraine's access to Starlink were up to SpaceX and Elon Musk, not the White House, comments that suggested Musk may have restricted Kyiv's access to the service reut.rs/4yJ7cUM
