저자: Alex · 외국인
요약 보기
엔비디아는 게이밍 중심 기업이 아니라 데이터센터(AI·슈퍼컴퓨터), 자율주행 등으로 사업이 확장된 기업이라고 설명. 2022년 이더리움 채굴 수요 소멸과 팬데믹 특수 종료로 게이밍 매출이 급감했으나 3분기 게이밍 매출 저점 가능성이 제기됨. 데이터센터 매출은 전년 대비 71% 증가한 32.6억 달러로 게이밍의 2배, 데이터센터 시장점유율 약 90% 주장. 자율주행 사업 파이프라인 약 110억 달러. 미국 중국 수출규제 영향은 매출의 3~4% 수준으로 제한적으로 평가.
다듬은 글
whether you put a piece of your paycheck into a retirement account every month or you actively pick your own stocks you should know that Nvidia sits near the top of the S P 500 and the NASDAQ so knowing what they're up to can give you a huge leg up as an investor well Nvidia stock has collapsed by more than 40 percent over the last year making now a great time to decide if it's a good investment so in this episode I want to answer a few simple questions is NVIDIA mostly just a gaming company that sells expensive graphics cards did Nvidia stock just go up because of the pandemic and since interest in gaming cryptocurrencies in the metaverse have all virtually vanished is nvidia's future in serious trouble your time is valuable so let's get right into it if you don't know much about Nvidia a good place to start is understanding why graphics cards are so important to so many different Industries besides video games while video cards have been around since the 1980s things really changed in 1999 when Nvidia released the GeForce 256 this was the first full-fledged Graphics processing you unit or GPU before Nvidia computers relied on central processing units or CPUs to perform the complex but repetitive calculations involved in generating Graphics think about transforming the scene as a viewer moves through it or handling surface and lighting effects and so on this was slow because CPUs process only one or two instructions at a time but gpus are designed to handle many similar tasks in parallel making them great for rendering Graphics the GeForce 256 even Incorporated a transform and lighting engine directly onto the chip itself which resulted in a huge performance boost over previous kinds of cards the GeForce 256 put Nvidia on the map and they've been innovating in the graphics space ever since today Nvidia owns over 85 percent of the GPU Market but while most people still associate gpus mainly with gaming they're used for much more than that from architecture and product design to artificial intelligence and machine learning these Graphics chips power the applications behind our daily lives in fact the reason you're watching this video right now is because of YouTube's video recommendation algorithm which runs on thousands and thousands of gpus but everything changed in the beginning of 2020 nvidia's GPU sales exploded thanks to shutdowns and stimulus checks stuck at home with nowhere to go Gamers and professionals alike were eager to upgrade their machines as more of everyone's time moved online at the same time when the lockdowns first started they halted everything from production lines to delivery services and of all the industries hit by these supply chain issues computer chips were hit the hardest that year the Auto industry Alone lost over 100 billion dollars because they couldn't get enough chips to put inside their cars the second issue for gpus was that crypto miners found that graphics cards were great for mining ethereum which was rocketing up in price ethereum is the most popular blockchain for nfts and the home to some of the biggest metaverse projects on the planet both of which got a lot of hype during the pandemic as a result ethereum saw massive growth from a high of 730 dollars in 2020 to a whopping 4 800.
in 2021 so demand skyrocketed as Gamers professionals and crypto miners all scrambled to buy new gpus lower Supply Plus higher demand means massive price inflation for nvidia's cards and as a result nvidia's revenue from their gaming segment which is the one that covers all of their consumer GPU sales shot up to over 3 billion dollars in a single quarter that was an 85 percent increase from just one year before but in 2022 all those massive Tailwinds turned into headwinds for example as the lockdowns ended people spent far less time in front of their screens at home also more apps services and workloads moved on to the cloud which means they can be accessed by mobile devices and laptops instead of needing a beefy desktop computer on top of that almost all the demand for ethereum mining dried up completely not just because ethereum's price crashed by over 70 percent which makes it far less profitable to mine but also because ethereum changed how it got mined altogether back in September of last year and gpus are no longer the best solution as a result nvidia's gaming division reported just 1.57 billion dollars in Revenue in the third quarter of 2022 that's half of what it was at its peak we'll see if things get better or worse for nvidia's gaming division when they report their final numbers for 2022 at their next earnings call which is in just a few weeks either way that's the context for white gaming which was once nvidia's top earning business unit is seeing historically low numbers today now with that context out of the way let's start answering some of those questions from the start of the episode a few months ago and video launched their RTX 40 series of graphics cards nvidia's chips are built by tsmc but Nvidia doesn't just order chips from tsmc like we do from ubereats instead the two companies work together to create a special version of tsmc's four nanometer chip fabrication process that's optimized specifically for gpus that process lets tsmc pack a whopping 76 billion transistors and over 18 000 cores onto nvidia's chips which is about 70 percent more than their previous obvious generation which was their RTX 30 series graphics cards and videos management was pretty optimistic about the future of gpus during their latest earnings call their CFO said that retailers were sold out of the RTX 4090 within minutes even with that Hefty sixteen hundred dollar price tag when the RTX 40 series cards first came out I made a video saying that I think they're overpriced especially given the amazing performance of the 30 series cards and the performance per dollar of competing gpus from AMD I actually still stand by that opinion but you should take it with a grain of salt because nvidia's sales numbers don't lie they shipped over a hundred and fifty thousand units in the first two weeks alone as a result nvidia's CFO thinks that the third quarter of 2022 which is their latest quarter of financial results at the time of this recording marked a bottom for nvidia's gaming segment and they expect the segment to start growing again in quarter four and well into 2023.
Stacey Raskin who is the senior semiconductor Analyst at Bernstein research is also optimistic about nvidia's gaming segment here's what he had to say to CNBC about nvidia's most recent numbers okay is it is it possible for you to say right in the hero now that the worst is in that the worst is behind this company I don't know yeah I mean it does seem like hopefully numbers have hit a bottom like I said the gaming members are pretty good I think one of the big questions into the guide next quarter is going to be the sequentials by by business line um the guide for the total revenues were kind of flatish into Q4 so the big question is like is anything growing is gaming growing is data center growing like almost being equal I think you'd rather have data center grow sequential into Q4 versus gaming so that'll be one of the key questions on the call I think yeah and we overall the num the number is relative to expectations look okay yeah so after the perfect storm of shutdowns ending people spending more time on mobile devices and demand for ethereum miners going to zero it looks like nvidia's GPU Revenue could have found its bottom but that begs the question should investors only focus on gaming what about nvidia's other business units in the third quarter of 2022 nvidia's Data Center business brought in 3.26 billion dollars in Revenue that's double the revenue of their gaming segment and a massive 71 increase from the previous year this is why I spent some time covering the differences between CPUs and gpus it turns out that gpus are great for machine learning even though that doesn't sound anything like rendering Graphics both kinds of workloads take in large amounts of data and solve similar types of math problems in parallel Nvidia has a few different data center chips that are all worth talking about let's start with their a100 tensor core gpus these chips are specifically designed for data centers focused on machine learning data analytics and high performance Computing which is obviously a very hot market right now thanks to AI models like chat GPT and Dolly 2 nvidia's newer h100 data center chips are a massive step up from the a100s in terms of performance promising up to a 9x speed up in AI training like the kinds used for self-driving cars they also offer a whopping 30X speed up for inference models like the massive natural language processing models behind gpt3 which Powers chat GPT or the image processing models that generate and modify images for Dolly too and then there's Grace Grace is nvidia's First Data Center CPU and it goes on sale male in the first half of this year according to Nvidia Grace can deliver 10 times the performance of today's fastest chips when it comes to high performance Computing and AI workloads one thing that's always impressed me about nvidia's designs is that they can connect together to form more powerful systems at virtually any scale for example eight of these h100 chips can be connected to form a dgx h100 server system and if you combine nine of these dgx h100 service systems you get a dgx pod which is the reference design that can be scaled up to offer AI services at Enterprise levels and Beyond but it doesn't stop there you can connect 32 dgx pods together to create a super pod which provides around one exo-fop of computing performance and Nvidia is linking 18 of these super pods together to build a super computer called EOS which will be about four times faster than fugaku in Japan which is currently the world's fastest supercomputer but Nvidia doesn't just build super computers for themselves their Hardware can actually be found in over 70 of the 500 most powerful supercomputers around the world supercomputers like meta platforms his research supercluster or RSC which is one of the largest customer installations of nvidia's Hardware this super computer is connecting over 6 000 Nvidia gpus to do things like Translate speech between large groups of people each speaking a different language Nvidia also just announced that they'll be providing thousands of gpus and accelerators to power Microsoft's Advanced supercomputer infrastructure which will provide scalable AI training and deployment to other businesses so nvidia's gpus are the platforms Behind These platforms Nvidia announced a similar collaboration with Oracle and their cloud service where their chips will power natural language processing engines behind chat Bots and virtual AI assistance and that's on top of their already long-standing partnership with Amazon web services which utilizes nvidia's server gpus for their virtual machines as well Nvidia dominates the data center Market even more than the Consumer one with a whopping 90 market share share for Enterprise gpus in my opinion as the world becomes more and more data driven demand for nvidia's data center Hardware will keep climbing across nearly every industry but it's not all sunshine and rainbows for nvidia's data center business several months ago the US government ordered Nvidia and AMD to stop selling AI chips and systems to China Nvidia said that this ban could cost them around 400 million dollars worth of data center Revenue per year which is about three or four percent of their total but just a few weeks ago Nvidia announced a chip called the a800 as an alternative product for consumers in China Nvidia designed this chip specifically to get around this ban most of the other specs of the chip are identical to the a100 but the big difference is that the interconnects in the a800 run at 400 gigabytes a second instead of 600 to get under the performance threshold set by this chip ban so between how small the impact of these bands are on nvidia's revenues and how quickly they can change their chip designs to get around them I'm not really too worried about how these trade Wars affect nvidia's bottom line but while Nvidia is making big strides in data centers and their gaming segment seems ready for a rebound Nvidia has another division that's growing even faster even though Automotive is only bringing in around four percent of nvidia's overall Revenue right now it grew by 80 year over year they have about 11 billion dollars worth of Automotive business in their pipeline over the next five years and I expect that number to grow substantially as autonomous driving and in-car infotainment both become more mainstream nvidia's Automotive offerings can be broken out into two parts hardware and software nvidia's self-driving Hardware platform is called Hyperion and it supports a wide array of sensors Hyperion 8 can achieve full self-driving with a 360 degree camera radar lidar and ultrasonic sensor Suite hyperion8 will ship in Mercedes-Benz cars starting in 2024 followed by Jaguar Land Rover in 2025.
this is how Nvidia will provide self-driving as a platform Mercedes-Benz Jaguar Land Rover and recently Lucid Motors announced their electric self-driving SUV called the gravity which will use nvidia's Hyperion system and connect to their Drive software platform which also has two parts first there's Drive map which is a crowdsourced real-time digital twin of every road that they get data for eventually Nvidia expects to map every major highway in North America Western Europe and Asia that's hundreds of thousands of miles of physical roads that'll have digital twins if Nvidia knows where a car is in this digital twin they'll also know where it is in real life then the Hyperion sensors in the car can double check what they're seeing against that digital twin then there's Drive Sim which can take in recorded Drive data turn everything in the scene into an interactive object and store those objects for further reuse Drive Sim also lets Nvidia change anything in the recording and rerun a bunch of simulations in that environment or they can change the environment altogether by adding different lighting effects and weather conditions and road hazards and so on Nvidia software can also simulate what different sensors would see in a given scenario so for example they can simulate a city street show what a lidar would see on that City street and then calculate the depth map that it would return that means that they can train their self-driving AI on lidar data even though they didn't collect it in a real City street all this to say nvidia's Automotive business is one of their most promising business units even if it's the smallest one by Revenue today so whether you put a piece of your paycheck into a retirement account every month or you actively pick your own stocks you should know that Nvidia sits near the top of the S P 500 and the NASDAQ and knowing what they're up to gives you a huge Advantage as an investor and now you know that they're much more than just a gaming company that their hardware and software help power some of the biggest cloud services and super computers on the planet they've built a suite of tools for self-driving vehicles that will hit the road as early as next year and we haven't even scratched this surface on their other multi-billion dollar efforts like Earth 2 which is a digital twin of the entire world including its water and weather systems or nvidia's Omniverse which could end up powering a big chunk of the metaverse if it ever becomes a reality or their Cutting Edge work in robotics the list goes on and on but before you go investing your hard-earned money there's another risk you really need to consider and if you want to know what it is yo
댓글 0