원문 보기
AMD just had their most important earnings call ever whether you're a longtime AMD shareholder or you're just wondering if AMD can keep up with Nvidia you should know that this isn't just about revenues and profit margins amd's latest earnings show us their place in amd's latest earnings call one key Insight that changed the way I see the company and what that means for the future of the stock here's everything I'll cover in this video amd's latest earnings their next Generation Instinct gpus the biggest risks for AMD stock like competing with Nvidia and of course where all of this puts AMD pretty well year to date let me start by saying that this isn't a video about AMD being a bad company or a bad investment in fact I've made video after video talking about how AMD is the only company on Earth that even has a shot at competing with Nvidia but the further we get into the AI era The Wider the gap between AMD and Nvidia seems to get and I think I figured out why let's start with amd's latest earnings call AMD posted record revenues of $6.8 billion for the quarter which is up 17% quarter over quarter and 18% year over-year AMD has four major business units data centers client PCS gaming and embedded chips if you've been watching this channel for a while you know that I focus on investing in technology platforms that other businesses pay to build on top of so for this video I'm just going to focus on amd's data center chips which is their biggest and fastest growing business unit anyway AMD reported $3.5 billion in data center revenues which is up a whopping 122% year-over-year that sounds great at first but let's compare that to Nvidia nvidia's data center revenues clocked in at $26.3 billion last quarter which was up by and insane 154% year-over-year and represents 88% of their overall Revenue remember that number nvidia's data center revenues are 7 and a half times bigger than amds and they're growing 25% faster but why is that well nvidia's Hopper gpus are the current industry standard for both AI training and inference which is why companies like Google Amazon Microsoft and meta platforms are buying hundreds of thousands of h100s at a time in fact Elon Musk and the team behind xai set up a super cluster of 100,000 Nvidia h100 gpus earlier this month with plans for another 50,000 h20s in the near future and as a side note it's pretty amazing that elan's team built the entire super cluster in just 19 days when it would normally take 3 or 4 years are we already at the age of clusters of 200 and 300,000 gpus um the answer is yes and then the um uh first first of all uh acknowledgement of of achievement where it's deserved I mean Elon is singular in this understanding of engineering and and construction and large systems and from the moment that we decided to get to go um the planning of with our engineering team our networking team or our infrastructure Computing team the software Team all of the preparation Advance all of the infrastructure all of the logistics and the amount of technology and equipment that came in on that day Nidia nvidia's infrastructure and Computing infrastructure and all that technology to training 19 days just you know you don't want did anybody sleep that's huge for AI investors because it could mean even faster AI adoption for Enterprises and even larger GPU orders now that the entire industry knows how fast these super clusters can come online and start training on mountains of data and speaking of data or emails lately they might be selling yours too that's why I join 've everything they've done I just 's by going to join delet me.com symbol2 so here's the big Insight nvidia's Data Center business accounts for 88% of their total revenue and that revenue is tied almost exclusively to AI on the other hand amd's data center revenues only account for half of their total with their client side CPUs gaming chips and embedded processors making up the rest on top of that though amd's data center revenues are actually split between their Instinct gpus and their epic CPUs which obviously compete with Intel and not Nvidia so if we compare Apple's to Apples nvidia's revenue from AI accelerators is actually around 15 times bigger than amds and the reason that Nvidia is also growing faster is 88% of their total revenues are directly tied to their AI ecosystem compared to just 25 to 35% of amd's Revenue depending on how much amd's epic CPUs are benefiting from this AI boom said another way for every dollar I put into Nvidia stock 88 cents are directly exposed to AI but for every dollar I put into AMD stock it's more like 25 to 35 cents that's a big difference in terms of exposing my portfolio to the overall AI market and just to be clear AI is one of the only markets that matters for AMD right now the PC market actually shrank during the pandemic and has been growing by low single digits ever since 35% of amd's total revenues come from PC CPUs gpus and gaming chips and those revenues are still lower today than they were back in 2022 compare that to the global artificial intelligence Market which is expected to almost 11x over the next 9 years that's a compound annual growth rate of over 30% and the market for data center accelerators is expected to more than 5x by 2032 or grow 24% per year for the next 8 years that's another reason I'm so focused on data centers and AI infrastructure earlier this year I said that AMD could become a serious competitor to Nvidia for two key reasons first nvidia's chips were seriously Supply constrained with weit times of up to a year or longer giving AMD a chance to take meaningful market share and second nvidia's data center infrastructure ran on infiniband which meant that most ethernet-based data centers weren't making big investments into nvidia's Hardware ecosystem at the time but over the last few quarters Nvidia fixed both of those issues by seriously ramping up hopper and Blackwell production as well as coming out with ethernet-based networking Solutions in fact 3.7 billion of nvidia's Revenue came from AI networking products last quarter which is still more than all of amd's data center revenues today now that AMD doesn't have those two big Tailwinds it just comes down to how their gpus stack up against nvidias so let's talk about that next the Mi 300X is amd's AI accelerator that directly competes with nvidia's h100s it's designed for more memory intensive AI workloads like running large language models with huge context windows or generating video clips that's why companies like Microsoft and meta platforms buy these chips to run AI models like GPT 4 and llama 3.1 and when AMD released the Mi 300X they claimed it performed around 30% better than nvidia's h100s but here's the problem AMD released these chips more than a year after Nvidia released the h100 which means the biggest cloud service providers and supercomputers were already reserving nvidia's h200 gpus by then and just a few weeks ago AMD CEO Lisa Sue announced the Mi 325x today I'm very excited to launch Mi 325x our next Generation Instinct accelerator with leadership generative AI performance Mi 325 again leads the industry with 256 gbt of ultra fast hbm 3E memory and 6 terabytes per second of bandwidth and when you look at that across some of the key models we're delivering between 20 and 40% better inference performance and latency on things like llama mistra and mraw the hpu version features 2 terabyt of hbm 3E memory and 48 terabytes per second of aggregate memory band with enabling our customers to run more models as well as larger models on a single Mi 325 server when you put all of that capability together what you see is that Mi 325 platform delivers up to 40% more inferencing performance than the h200 on llama 3.1 and a lot of people also are doing training and when you look at training performance we've made very significant progress on optimizing our software stack for training on a growing number of customers custers and what you see with Mi 325 we have excellent training performance that's very competitive compared to the competition so amd's mi 325x competes with nvidia's h200 gpus and just like last time amd's chips appear to outperform nvidias by 20 to 40% for inference and offer at least the same performance when it comes to training at least according to amd's early benchmarks that means if AMD can be Nvidia on price they're offering some serious performance performance per dollar just like with the Mi 300X it's also worth pointing out that AMD was making $0 on these data center gpus just one year ago so building a multi-billion Dollar business that competes directly with Nvidia in four quarters is an incredible achievement that's what I was betting on when I put AMD on my list the challenge is that nvidia's Blackwell gpus are coming out in a few weeks and remember Blackwell isn't just a GPU it's an accelerated Computing platform platform chips based on the Blackwell architecture will go into self-driving cars and robots not just data centers and supercomputers Blackwell is also four times better at AI training and a whopping 30 times better at AI inference than Hopper that massive leap in performance gives Nvidia a lot of pricing power which is why they have 76% gross margins while amd's margins are at 50% that's a pretty big difference and by the time that AMD has an answer to the B100 Nvidia will be close to launching their already announced b200 chips we know from nvidia's investor presentations that they've moved to a one-year release cycle between architectures meaning Hopper to Blackwell and then Blackwell to the Reuben architecture that Jensen already announced at computex earlier this year so one of the biggest risks for AMD is that Nvidia releases new Chips almost twice as fast as them right now another big risk for AMD is spreading themselves too thin even though this video focuses on data centers AMD makes CPUs and gpus for gaming PCs custom chips for video game consoles and has a big embedded chip business thanks to their acquisition of xyl links together those business units account for almost 50% of amd's Revenue remember for every dollar I put into Nvidia stock 88 cents are directly exposed to AI but for every dollar I put into AMD stock it's more like 25 to 35 that leads me to the third risk for AMD which which is the innovators dilemma AMD recently partnered with Intel to form an industry Advisory Group around the x86 CPU ecosystem I'm not saying that x86 is going away anytime soon but most major chip designers are leveraging the arm architecture more and more nvidia's grace CPU is arm-based so are Google's tensor processing units and Microsoft's Cobalt chips and aws's graviton processors and while AMD makes arguably the best 86 sh on Earth I'm worried that big Partnerships around x86 are their way of doubling down on today instead of pivoting to the architectures of tomorrow and as a shareholder I want to see AMD skate to where the puck is going diversify their product lines and make some arm chips in the future now that I've covered amd's latest earnings their Data Center gpus and the risks of competing with Nvidia while being spread across many different verticals let's talk about what that all means for AMD make let's talk about where AMD belong one of the big reasons I put AMD on this list at the start of 20124 was to own the entire Enterprise AI accelerator Market that's one way this entire list of stocks works together and why I think it performed so well but as time went on it became clear that Nvidia can solve their biggest challenges that faster then AMD can take advantage of them on top of that AMD has spread across many more low growth markets like PCS game consoles and embedded chips only about a third of amd's Revenue comes from AI accelerators today and nvidia's data center revenue is eight times bigger and growing faster once I realize that I can get around three times more direct AI exposure for every dollar I spend on Nvidia stock than I do from AMD it really changed the way I value both companies s a result I'm moving AMD down to spots on the list that doesn't make AMD a bad company or a bad investment it's just not a pure play on AI and I don't need to bet on every single company that makes AI data center hardware and if you want to see how I pick this list of stocks at the start of the year make sure to check out this episode
Comments 0