GET IN EARLY! I'm Investing In This HUGE AI Chip Breakthrough

Video Statistics and Information

Video
Captions Word Cloud
Reddit Comments
Captions
I just got back from GTC and I'm convinced that Wall Street does not understand Nvidia that's because the key to finding great stocks is understanding a company's products not just their profits and I'm convinced that Nvidia will be bigger than Apple bigger than Microsoft and Nvidia stock will sit on top of the indexes as the biggest company on Earth let me show you why your time is valuable so let's get right into it first things first I'm not here to hold you hostage here's everything I'm going to talk about in this video I'll cover nvidia's insane new Blackwell gpus I'll talk about a few things that investors need to know about Blackwell's design costs and margins the networking technologies that Nvidia built to scale Blackwell data centers to mindboggling sizes and of course whether I'm buying Nvidia stock as a result and where I'd put it on the list that I made at the start of the year to get rich without getting lucky but let me point out a few quick things before I dive into nvidia's insane new Blackwell gpus first Nvidia is not just an AI hardware company they're an endtoend AI Computing platform and Nvidia showcased a lot of software at GTC I spent my time at the conference learning as much as I can about every part of this business I interviewed nvidia's Executives I attended close to a dozen expert panels and sessions I got my hands on new products and service demos everything that I think the mainstream media and Wall Street analysts are going to ignore until it's too late and as you can imagine there's way too much to cram into a single video from self-driving cars and supercomputers to humanoid robots and digital twins for heavy Industries and even foundational AI models that other businesses can build on top of so I'm going to spend the next few videos breaking it all down for you why because the world is going through a major technology transformation right now that's my second Point big businesses are changing everything from the hardware they run on to the kinds of tools and processes they build on top of it to the services and products that they sell on top of that and like I've been saying for a couple of years now nvidia's data center Hardware is at the center of this transformation and third that's why I think Nvidia will become the biggest company on the planet just like the iPhone becoming the platform of the mobile Computing era made Apple the biggest company in the world Blackwell and Hopper will be the platforms of the generative AI era and if you still think that big companies can't get bigger remember this Warren Buffett didn't invest in Apple until 2016 10 years after the iPhone came out Apple was the biggest company in the world by market cap for 5 years when Warren Buffett bought it and Apple stock is still 50% of his portfolio today I don't want to wait 10 years to find the next iPhone or hold the next Apple stock that's why I'm making these videos now all right one last point and then I'll dive right into nvidia's Hardware I get asked all the time how I make these videos and I won't lie it can be a real Mental Health Battle I like to do everything myself so my main challenge is burnout I get this brain fog that drains my energy and crushes my productivity for days and I didn't think that anyone could relate so instead of talking to someone I would just ignore it or try to power through it which would just make me fall even further behind that's why I'm so excited to tell you about betterhelp the sponsor of this video whether you're stressed out about balancing work and family or you're dealing with depression or anxiety you are not alone and getting help is nothing to be ashamed of and honestly it's never been easier it can be fully remote and online and after answering a few questions better help can usually match you with a professional therapist within 48 hours that's huge because they make it easy to switch to a new therapist at no additional cost without worrying about insurance or who's in your network that way you can focus on finding the right therapist for you between running my own business and preparing for my wedding next year my mental health is something I'm investing in more and more so if you want better help with your mental health check out the link in the description or go to betterhelp.com slts yoou to save 10% off your first month and a big thank you to betterhelp and to you for supporting the Channel all right I can't wait anymore let's talk about Blackwell the rate at which we're advancing Computing is insane and it's still not fast enough so we built another chip Hopper is fantastic but we need bigger gpus and so ladies and gentlemen I would like to introduce you to a very very big GPU this is Blackwell the successor to nvidia's Hopper gpus and just like Hopper Blackwell has two variants the B100 and the b200 the B100 is a dropin replacement for the h100 which means that data centers and supercomputers can swap out an h100 tray and L Al put a B100 tray in its place that's important for investors because it means that Nvidia won't have a hard time ramping up Blackwell sales since the infrastructure to support the b100s is already deployed anywhere that there's an h100 today the B100 is a whopping 80% faster than the h100 while drawing the same 700 watts of power which means the entire Blackwell architecture scales much better than hopper in data centers and supercomputers that's why they're going to want to phase in the B1 00s over time on the other hand the b200 is nvidia's new state-of-the-art chip it's anywhere from 10 to 25% faster than the b00 depending on the workload while drawing around 1,000 Watts instead of 700 compared to Hopper these new Blackwell chips perform a whopping four times better at training AI models and an insane 30 times better for AI inference but how Nvidia made it happened is really important for investors to understand this is the most advanced GPU in the world in production today this is Hopper this is hopper Hopper changed the world this is Blackwell it's okay Hopper 28 billion transistors and so so you could see you I can see that there's a small line between two dieses this is the first time two dies have abutted like this together in such a way that the two CH the two guys think it's one chip there's 10 terabytes of data between it 10 terabytes per second so that these two these two sides of the Blackwell Chip have no clue which side they're on there's no memory locality issues no cash issues it's just one giant chip when we were told that Blackwell's Ambitions were beyond the limits of physics uh the engineer said so what and so this is what what happened so there are two important points here first this blackwall GPU actually combines two dieses into a single chip and connects them with a 10 terb per second link so that they act as one GPU this is a really clever chip design but two dieses an ultra high-speed link and a bigger total area for advanced packaging means that Blackwell costs over twice as much to make as Hopper and higher unit costs means lower margins personally I think the trade-off is worth it in the short term because Blackwell's insane performance gives Nvidia a lot of pricing power compared to other GPU makers and that flexibility is what lets them get even more businesses onto the Cuda platform at a time where every Data Center and cloud service provider is scrambling to get as many accelerators as they can during a gold rush getting everyone on your pick and shovels is what matters most second if you've been watching this channel for a while you know that nvidia's chips are made at tsmc on a process node called n4p which was custom made for NVIDIA so as tsmc's chip fabrication process keeps improving shareholders should want Nvidia to offer a single die version of these chips or find another packaging Solution that's higher margin than the current architecture not necessarily higher performance just lower manufacturing costs while still meeting Market demands and speaking of Market demands let's see how Blackwell systems scale to create full data center and even super computer sized Solutions because that's where the magic really happens this is a fully functional board and I just be careful here this right here is I don't know $10 billion two Blackwell chips and four Blackwell dyes connected to a Grace CPU the grce CPU has a super fast chipto chip link what's amazing is this computer is the first of its kind where this much computation fits into this small of a place so the gb200 super chip pairs two b200 gpus with one Grace server CPU they're connected via NV link which is a chipto chip connection with 900 GB per second of bandwidth that's fast enough to move 150 featurelength 4K movies between chips every single second this gb200 super chip is what Jensen is holding in his left hand while he casually holds the10 billion initial prototype board in his right this right here is I don't know 10 billion yeah I'm never going to be that cool two of those gb200 Grace Blackwell Super Chips go into a Blackwell compute node which is one tray in an AI data center rack the 2GB 200 Super Chips in each tray are also connected via nvlink at a high level nvlink is for chipto chip connections while infin band fiber optics connect different compute nodes and Trays at the server level Nvidia also makes these blue field dpus or data processing units here's an easy way to think about dpus CPUs have a low amount of cores that run one operation at a time gpus have more cores and can run more operations in parallel dpus have tons of cores and can handle highly parallel workloads like coordinating securing analyzing and transferring data between many different chips in a data center so dpus end up handling most of the networking work workloads so that CPUs and gpus don't have to all right so 18 of these compute trays go into one gb200 nvl 72 system remember the gb200 super chip has two Blackwell gpus on it connected via NV link there are two Super Chips per compute tray which are also connected via NV link and there are 18 trays per system 2 * 2 * 18 is 72 Blackwell gpus connected by NV link that's why it's called the gb200 nvl 72 all right so we already have a good handle on the gb200 but what about this NV link well it turns out that NV link is way more than just a chipto chip connection the rate at which we're advancing Computing is insane and it's still not fast enough so we built another chip this chip is just an incredible chip we call it the Envy link switch it's 50 bill transistors it's almost the size of Hopper all by itself this switch chip has four MV links in it each 1.8 terabytes per second what is this chip for if we were to build such a chip we can have every single GPU talk to every other GPU at full speed at the same time that's insane by the way and vlink infiniband and these Bluefield dpus all came out of nvidia's $7 billion acquisition of melanox back in 2019 which will probably go down as one of the best Acquisitions in Silicon Valley history all right so sandwiched between the compute notes that we just talked about are nine EnV link switch trays Each of which have two of these EnV link chips in them and like we saw in the compute trays each Envy link chip connects four Blackwell gpus four ports per nvlink chip times two chips per tray times nine trays is the 72 ports supporting all to all GPU Communications at a mindboggling bandwidth of 130 terabytes per second that's the back that's the that's the back the dgx MV link spine 130 terabytes per second goes through the back of that chassis that is more than the aggregate bandwidth of the internet we could basically send everything to everybody within a second more bandwidth than the entire internet don't worry I'll give you a second to let your brain stop melting all right the 18 compute trays and the nine Envy link switch trays come together to form this gb200 nvl 72 system which is just one Rack in an AI data center this one rack has 1.4 xoff flops of AI training performance in it just to be clear there are only a few exaflop supercomputers on Earth today and Nvidia just built one in a single 120 KW rack that they sell off the shelf and at the top of each of these single rack supercomputers exists a single infin band switch tray to connect them to even more single rack supercomputers and the resulting system can be connected together to make one 32,000 GPU 645 exaflop AI Factory for the new Industrial Revolution one Computing platform for generative AI one giant GPU and while your brain finishes melting again let me ask you a simple question why would Nvidia build something like this and does Humanity really need gpus this powerful 600,000 Parts somebody used to say you know you guys make gpus and we do but this is what a GPU looks like to me when somebody says GPU I see this now let's see what it looks like in operation if you were to train a GPT model 1.8 trillion parameter model it took about 3 to 5 months or so uh with 25,000 amp uh if we were to do it with hopper it would probably take something like 8,000 gpus and it would consume 15 megawatts 8,000 gpus and 15 megawatts it would take 90 days about three months if you were to use Blackwell to do this it would only take 2,000 gpus 2000 gpus same 90 days but this is the amazing part only 4 megawatt of power power so from 15 yeah that's right and that's and that's our goal our goal is to continuously drive down the cost and the energy they're directly proportional to each other cost and energy associated with the Computing so that we can continue to expand and scale up the computation that we have to do to train the Next Generation models so you can do the same amount of work with one quarter of the gpus in one quarter of the data center footprint using one quar of the power for Big Data Centers cloud service providers and future supercomputers Blackwell lets them do a lot more in the same amount of space and power and on the other side startups and smaller companies can now buy a lot more compute for the same amount of money and power budget that's why Blackwell is such a big deal all right I know this video is a little long but I want it to be thorough since Blackwell is such an important piece of technology to wrap your head around including its double die design and what that means for unit costs and margins as well as how nvidia's networking Solutions can scale Blackwell data centers to do anything Humanity can throw at it now with all of that context let's talk about whether I would buy Nvidia stock at today's prices and if you feel I've earned it consider hitting the like button and subscribing to the channel that really helps me out and it lets me know to put out more deep Dives like this thanks and with that out of the way let's talk about Nvidia stock I think that Blackwell is about to move the needle in a big way for NVIDIA black Blackwell is not just a new data center GPU it's an entire compute platform chips based on the Blackwell architecture will also go into self-driving Cars Smart factories and even humanoid robots all of which I'll cover over the next few videos because I really believe that Nvidia stock will sit at the top of the indexes and Pension funds and everyone's portfolios for years to come so with everything I've learned from Nvidia GTC I'm moving Nvidia up one spot on my list of stocks to get rich with without getting lucky that's a pretty big move considering how high it was already and the fact that I just can't afford to buy as much at $900 per share I also ended up moving Amazon stock up one spot since AWS is the biggest cloud service provider on the planet and I think that their Bedrock platform for generative AI is about to benefit a lot from nvidia's Blackwell gpus this is why it's so important to understand the science behind the stocks and if you want to see why I pick these stocks here's the video where I walk through it step by step either way thanks for watching and until next time this is ticker symol U my name is Alex reminding you that the best investment you can make is in you
Info
Channel: Ticker Symbol: YOU
Views: 465,121
Rating: undefined out of 5
Keywords: nvidia gtc 2024, nvidia keynote, jensen huang keynote, nvidia robots, jensen huang robots, nvidia blackwell gpu, blackwell, nvidia, nvda, nvidia stock, nvda stock, jensen huang, openai, chatgpt, gpt4, msft, microsoft stock, msft stock, googl, google stock, artificial intelligence stocks, nvidia stock news, semiconductor stocks, gpt-4, nvidia news, ai copilot, omniverse, ai stocks, best ai stocks, amd stock, intc stock, intel stock, amd vs nvidia, nvidia vs amd
Id: g5wf4yY52S0
Channel Id: undefined
Length: 17min 42sec (1062 seconds)
Published: Fri Mar 29 2024
Related Videos
Note
Please note that this website is currently a work in progress! Lots of interesting data and statistics to come.