TypyCom @TypyCom

Elon Musk you fucking genius 🫡 Shoutout to the entire SpaceX team 💯 That’s a 17 floor building dispersing satellites the size of a school bus in space btw. (The satellites open up to the size of an airplane) We are living in the craziest of times $SPCX @elonmusk

$HOOD says 24/7 stock trading is coming soon including weekends. Public markets slowly turning into a casino that never closes.

keyboard_arrow_left Previous keyboard_arrow_right Next

$AMD is misunderstood w/ $META Muse & $GOOGL $AMZN competition on Agentic Consumers🧵 The Inference Queen will dominate server CPUs and Agentic AI for years to come! Not FInancial Advice! DYOR! The market still prices @AMD as a #2 GPU company that also sells CPUs. That is the wrong stack ranking for 2026–2030. Venice is the first HPC product in volume on TSMC N2, and the best server CPU offerings vs all. Agent sandboxes, GPU host nodes, and general-purpose cloud boxes all want the same thing: cores, threads, memory bandwidth, and PCIe 6.0 in an x86 box that already runs Linux, Chromium, compilers, and enterprise software. AMD is the only merchant vendor that leads that product and is getting the fastest rise in TSMC advanced packaging share to actually ship it. Flagship EPYC 9996: 256 Zen 6c cores, 512 threads, 1.6 TB/s memory, first PCIe Gen 6 on a CPU, up to 1 GB L3, 600 W, TSMC 2 nm GAA. Intel’s current top Xeon 6980P is 128 cores on an older node. NVIDIA Vera is an 88-core Arm companion chip for Rubin racks, not a general-purpose sandbox CPU. Graviton5 and Axion are captive cloud parts. Arm is nowhere near to compete. The SKU split is the tell that Su saw this early. Venice HF for GPU host nodes (high frequency, high I/O). Dense 256 core Venice for agent sandboxes. 128 core parts for enterprise dollar per watt. EPYC Verano next, with LPDDR, for even cheaper scale out. Nobody else has a merchant family mapped that cleanly across host node, sandbox, and general purpose like AMD. Hyperscaler Arm may not work in the next 3-5 years on efficiency inside one cloud. It loses when the job is “give this agent a real computer.” That is AMD’s socket. @Muse does not stay a Meta only event. It forces Google and @amazon to ship their own always on consumer agent or watch Meta sit between the user and the checkout. @Google already has Gemini Spark as a 24/7 cloud agent inside Gmail, Docs, and Chrome. Amazon already has Buy for Me and Alexa for Shopping, and it blocked Muse from Amazon within two weeks of launch. That is a platform defending the customer relationship. The next move is obvious: each of them will offer a Free or paid version of “do whatever I want” agent with a persistent computer, a browser, credentials, and background work after the app closes. Free and $20–$100 tiers turn that into millions, then tens of millions, of reserved VMs. Tokens still hit shared GPUs. The box the agent lives in is CPU, RAM, and disk. When three hyperscalers race to give every user a personal cloud PC, sandbox density, not training clusters, CPU becomes the volume/revenue driver. That is the 50-100m units demand for Venice within 3 years of intense Agentic Consumers competition. EPYC Venice was built for: 256 cores, 512 threads, x86 compatibility, and more agents per watt than a captive Arm part that cannot drop a real Linux computer under Chromium and a compiler. Google and Amazon will still run a lot of Graviton and Axion inside their own clouds. They cannot serve the whole agent fleet that way once enterprises, ISVs, and multi-cloud agents need a merchant socket. Every Muse clone raises the CPU bill. AMD is the the best CPU supplier with the product and the TSMC wafers to take the incremental sockets. UBS this week raised AMD CoWoS demand another 11%, lifted 2027 Venice shipments from 4 million to 5 million units, left Instinct at 1.9 million on supply constraints, and said AMD’s CoWoS share goes from 6% in 2025 to 18% in 2027, the largest increase of any GPU or ASIC vendor. Morgan Stanley also projected AMD to sell 6.75M Venice units, or more than everyone in 2027. That is not a rounding error. That is TSMC reallocating the scarce wafers toward AMD CPUs because agentic demand showed up. And AMD is well on track to be #2 largest customer and #1 for TSMC. Agentic consumer products pay off the 2022 bet. Dr. Su kept funding the unfashionable x86 socket while the market only wanted GPUs. The result is the merchant CPU with the highest core count, the best process node among shipping server parts, the only broad x86 software story, and the steepest TSMC packaging share gain into 2027-2030. NVIDIA still sells more GPU dollars. Intel and custom Arm still take a % share of general cloud cores. AMD is the vendor that kept the best high-core x86 line through the GPU years, then sold Meta both that CPU and the GPU in the same rack and TSMC is now feeding that CPU the wafers. If OpenAI and Anthropic want to compete in this space, that will only 20-30x more CPU orders for AMD. Conclusion: Dr. Su kept funding the socket the market had already written off. In 2022 the consensus was GPUs only or "the world only runs on GPUs". She still called the datacenter AMD’s largest growth opportunity, shipped EPYC into Meta’s Open Compute fleet, and built Venice as a family host node, dense sandbox, general purpose. AMD stock got punished for that. But It was the right map for what agents actually are: reserved Linux computers that only sometimes wake a GPU. Muse made the architecture visible. Google and Amazon now have to answer with their own paid/free "Muse", always on agents or surrender the customer or pay the tax to Meta like Apple store. That race multiplies CPU demand, not just token demand. AMD already has the best merchant part on TSMC 2 nm, the steepest CoWoS share gain into 2027, and a Meta rack that buys both Instinct and EPYC. Server CPU revenue is compounding at rates the street used to reserve for accelerators. Demand for Venice is stronger than any prior generation. The CPU line “far exceeded” her own forecast because she underwrote the unfashionable layer before the workload had a consumer name. The bet was that when agents went from millions to billions of users, the company that still owned the high-core x86 socket would get paid 3-5x more in demand. And Dr. Su was right. Not FInancial Advice! DYOR!

Mike @MikeLongTerm

3 weeks ago

$AMD| $META Muse is pushing 4-8 CPU : 1 GPU ratio 🧵 1. “Make CPU Great Again” @Muse is the clearest example. Meta does not just run a chatbot on shared inference clusters. Each user gets an isolated Linux VM with its own browser, storage, CPU, and memory so the agent can

14 7 92 91K 23

PGM Niewiadów targets 180,000 155 mm rounds a year as PLN 250 million ammunition expansion moves towards production defence-industry.eu/pgm-niewiadow-…

We’ve come a long way. Tesla IPO value was a thousandth of its current value!

X Freeze @XFreeze

a month ago

Tesla is now worth MORE than all other publicly traded automakers combined Tesla: $1.511 Trillion Everyone else combined: $1.446 Trillion Toyota, BYD, GM, Hyundai, Mercedes, BMW, Ford, Ferrari, Volkswagen, Porsche...all of them combined And Tesla still stands taller... nothing

236 431 3K 3.8M 369

Revolut is now Poland’s third-largest bank by the number of Polish zloty current accounts, with 5.6M. That puts it just 270,000 behind Bank Pekao (5.87M) and well ahead of Erste. PKO Bank Polski remains far out in front on 9.93M. It is adding about 1M customers a year, or 19% YoY - several times faster than the high-street banks - even though it still has no live Polish branch. It operates on a Lithuanian licence; a Kraków branch was registered in April 2026 but is not yet open. Local PL account numbers are provided by a partner bank, now UniCredit.

@JonErlichman I know this sounds crazy right now, but I think both SpaceX and Tesla will exceed these estimates

Amazon is destroying tons of rare books to feed its AI training. 🫢 This practice was first reported by 404 Media, which placed a tracking device in a rare book that ultimately arrived at an Amazon facility in Las Vegas. Companies like Amazon need unfathomably large amounts of text to train their LLMs, which have already ingested what they can from the internet, according to TechCrunch.

@PeterDiamandis Few realize this

$AMD just shared this video of the Cerebras $CBRS CEO talking about how the two companies work together

A coffee shop is using AI to track every barista and customer in real time. Through their existing cameras. How many cups each person made. How long every customer has been sitting. Who is fast. Who is falling behind. Movement patterns. Wait times. All live on a dashboard. No new hardware. Regular CCTV turned into a performance tracker. This used to be enterprise-only tech. Now a single coffee shop owner can run it.

.@elonmusk: "You don't need college learning." • You can learn almost anything you want for free. • The internet has made knowledge widely accessible. • College can still have value as a way to demonstrate discipline and the ability to complete difficult assignments. • It also gives young people a chance to spend time with others their age before entering the workforce. "Colleges are basically for fun and to prove you can do your chores, but they're not for learning."

That was 10 years ago Imagine 10 years from now

テスラが目指してる未来 #テスラ #ロボ

ELON JUST DROPPED THE MOST BULLISH LINE OF THE YEAR on the SpaceX earnings call: “The limiting factor currently is memory. Memory output is increasing by around 20% a year. Demand is increasing by 200% a year, maybe higher. Economics 101: prices go up, not down.” Right now the names he’s effectively pointing at: • $SNDK $1,427 • $MU $892 • $SKHY $154 • $DRAM $54 While half the market is still debating if AI has peaked… the smart money is loading memory. Question for you: Are you still waiting for a dip, or are you already in? (Not financial advice. DYOR.)

keyboard_arrow_left Previous keyboard_arrow_right Next

BREAKING $AMD Dr. @LisaSu on CNBC Full🆕🚀 After Q2 2026 Triple Beat ER Revenue: $11.5B vs $11.28B est ✅ EPS: $1.66 vs $1.61 est ✅ Q3 Guide: $13.3B vs $12.51B est ✅ Helios starts shipping from Sept 2026✅ EPYC Venice significant Ramp from Q4 ✅ VERY BULLISH H2 2026 and 2027! ~Our data center busienss is accelerating! ~Growing well over 100% as we go over 2027 as the base gets larger ~In the last 6-7 months, Agentic AI is changing the CPU:GPU Ratio. We are the leader in CPUs, we just announced Venice, leadership in every aspect. Venice is the best CPU ~Enterprises are finding models that fit them at the right cost. We will have all kind of models from largest frontier models to small to medium models ~ We see a broad ecosystem for AI Models. The most advanced models will be extremely important and we will also have open-source models that fit different needs for enterprises. We need all AI Models, they will co-exist! ~ You need more AI Compute to service all AI Models. It is all about the value AI Models bring. ~ Dr. Su sees $SPCX as an incredible business and we are proud to work with SpaceX. U need the right compute for the right workloads. We have a very strong roadmap, we are ready to help. ~ We have a broad portfolio CPU, GPU, FPGA... we do business with every large businesses out there. We are looking forward to longer term partnership. ~ OpenAI, Meta, Anthropic are very long term partnerships, we will update as we get through the quarters ~ CPU : GPU ratio at 1:1. Dr. Su sees the shift in ratio, better than 1 to 1. We will see higher CPU ratio in the next 3-5 years as AI scales more into Agentic AI. ~ Data Center buildout are long term and we are planning with customers 12-24 months in advance. We will be significantly higher than 100% grwoth ~ Customers are giving us more longer term to plan for more supply with entire ecosystem. So that we can have much higher supply to customers. We are very good at increasing supply. ~ We have to keep raising the AI TAM. We building all kind of AI with a broad portfolio ~ This is the inflection moment for AI and AMD ~ It is time to be bullish but we are very disciplined Source & Credit: CNBC & Dr. Su

Mike @MikeLongTerm

2 months ago

$AMD| Why Analysts will upgrade AMD tomorrow and stock back to green! Or this AH is stupid!✅🚀 TLDR: In years I covered @AMD, first time Dr. Su told analysts their numbers are garbage low. And Data Center to grow significantly higher than 100% in 2027. 1. Now I want to

3 13 151 112K 39

$AMD Q2 capex was $808M vs. $298.6M expected which is absolutely insane. This is why the stock is down.

Elon Musk on why the smartest people drop out of college: "You don't need college to learn. Learn stuff. Everything is available basically for free. You can learn anything you want for free. It is not a question of learning." Musk explains what college actually provides: "There is a value that colleges have, which is seeing whether somebody can work hard at something, including a bunch of annoying homework assignments, and still do their homework, and kind of soldier through and get it done. That's the main value of college. And also, you probably want to hang around with a bunch of people your own age for a while instead of going right into the workforce. So I think colleges are basically for fun and to prove you can do your chores. But they're not for learning." On hiring at his companies: "There is a requirement of evidence of exceptional ability. I don't consider going to college evidence of exceptional ability. In fact, ideally you dropped out and did something. Obviously, Gates is a pretty smart guy, he dropped out. Jobs was pretty smart, he dropped out. Larry Ellison, smart guy, he dropped out. Obviously not needed." Musk shares how education should work: "Generally, you want education to be as close to a video game as possible. Like a good video game. You do not need to tell your kid to play video games; they will play video games on autopilot all day. If you can make it interactive and engaging, you can make education far more compelling and far easier to do." He challenges the current system: "You really want to disconnect the whole 'grade level' thing from the subjects. Allow people to progress at the fastest pace that they can, or are interested in, in each subject. It seems like a really obvious thing." Musk criticizes traditional teaching: "Most teaching today is a lot like vaudeville. Somebody's standing up there lecturing to you. They've done the same lecture several years in a row. They're not necessarily all that engaged. That lack of enthusiasm is conveyed to the students; they're not very excited about it. They don't know why they're there. 'Why are we learning this stuff?' We don't even know why. A lot of things people learn, probably there's no point in learning them, because they never use them in the future." On whether university is necessary: "A university education is often unnecessary. That's not to say it's unnecessary for all people. But I think you learn about as much, the vast majority of what you're going to learn there, in the first two years. And most of it is from your classmates. If the goal is to start a company, I would say no point in finishing college." Musk started his own school for his kids: "I created a little school. It's small, only 14 kids now, and it'll have 20 in September. It's called Ad Astra, which means 'to the stars.'" He explains what makes it different: "There aren't any grades. There's no grade one, grade two, grade three. Not making all the children go in the same grade at the same time, like an assembly line. People are not objects on an assembly line. That's a ridiculous notion. Some people love English or languages. Some people love math. Some people love music. Different abilities at different times. It makes more sense to cater the education to match their aptitudes and abilities." Musk shares a key principle: "It's important to teach problem-solving, or teach to the problem, not to the tools. Let's say you're trying to teach people about how engines work. A more traditional approach would be: 'We're going to teach you all about screwdrivers and wrenches. You're going to have a course on screwdrivers, a course on wrenches.' This is a very difficult way to do it." He offers a better approach: "A much better way would be: 'Here's the engine. Now let's take it apart. How are we going to take it apart? Oh, you need a screwdriver, that's what the screwdriver is for. You need a wrench, that's what the wrench is for.' And then a very important thing happens: the relevance of the tools becomes clear." The result: "It seems to be going pretty well. The kids really love going to school. I think that's a good sign. I hated going to school when I was a kid; it was torture. The fact that they actually think vacations are too long, they want to go back to school. Weird, I know." Musk reframes what education really is: "If you think about it, what is education? You're basically downloading data and algorithms into your brain. And it's actually amazingly bad in conventional education. It shouldn't be this huge chore. The more you can gamify the process of learning, the better."

Jaynit @jaynitx

5 months ago

x.com/i/article/2019…

48 486 3K 2.0M 7K