저자: Alex · 외국인
요약 보기
엔비디아 실적 발표(2024년 8월)를 분석. 매출 300억달러(전분기비 15%, 전년비 122%), EPS 68달러, 데이터센터 매출 263억달러(전년비 154%)로 AI 수요가 강하다고 설명. 블랙웰 칩은 양산 지연 우려가 있었으나 설계 결함이 아닌 TSMC 수율 문제로 보고, 4분기부터 램프업 및 수십억달러 매출 예상. 매출총이익률은 75.1%로 하락했으나 개선 조치 완료 주장. 경쟁 위협은 제한적이며 주요 리스크는 공급 제약으로 제시. 자사주 매입 500억달러 승인, 네트워킹 매출 증가 언급.
다듬은 글
Nvidia just had their most important earnings call ever whether you're a longtime Nvidia shareholder or looking to buy Nvidia stock for the first time you should know that this isn't just about revenues and profit margins nvidia's earnings decide the speed and the scale of the entire AI Revolution so in this video I'll break down everything you need to know about nvidia's latest earnings call and what it means for the entire stock market your time is valuable so let's get right into it to say this was one of the most anticipated earnings calls of all time would be a massive understatement this revolution is just starting but it all starts with Nvidia in my opinion it is the most important earnings not just of the Year potentially in many years because of what this represents in terms of attack let me explain the reason for all this hype ever since open AI released chat GPT in November of 2022 the stock markets returns have been largely driven by the Magnificent 7 that's because companies like Amazon meta platform forms and even Tesla already have huge data center infrastructures they have Global products and services that benefit from Ai and they have very Deep Pockets that puts them in the perfect position to invest aggressively in Ai and outspend the startups and smaller businesses that compete with them especially at a time where interest rates and inflation are making everything more expensive boy do I feel that so that's where investors put their money and as a result the Magnificent 7 outperformed the other 493 companies in the S&P 500 by a massive margin but even among those seven one company far outshine the rest and that's Nvidia since they're supplying all the AI chips that everyone else is investing in in a gold rush you want to be the company that sells picks and shovels if nvidia's Revenue drops below expectations that means companies are investing less into AI infrastructure than we thought and the AI Gold Rush is slowing down for the rest of the market if nvidia's Reven grow that means the opposite this AI revolution could be bigger than we thought or at least happen faster which drives up forward estimates for the rest of the market so in a very real way Nvidia has been driving the returns for the rest of the S&P 500 that's one reason why their earnings are always so important but I said this earnings call specifically was the most important ever and that's because of Blackwell nvidia's Next Generation AI chip I believe the Blackwell ecosystem will be the primary picks and shovels in this entire AI Gold Rush so its speed and scale will largely be determined by how fast envidia can get black Wells into Data Centers and at what cost so that's what I'll cover in this video I'll walk you through nvidia's earnings results focusing on data centers I'll explain what's going on with Blackwell they're hugely important Next Generation AI chip I'll talk about nvidia's biggest risks right now like their supply constraints and their competition and of course what's in store for Nvidia stock in the short medium and long term there's a lot to talk about so let's Dive Right In to nvidia's most important earnings call Nvidia posted record revenues of $30 billion for the quarter which is a 15% quarter over quarter and a whopping 122% year-over-year Nvidia also posted earnings per share of 68 Which is 171% higher than a year ago after accounting for their 10 for one stock split things get even crazier when we focus on nvidia's data center revenues which now account for about 88% of their total revenues today that's why I'm not covering their other business units nvidia's data center revenues came in at$ 26.3 billion up 16% quarter over quarter and an insane 154% year-over-year just to be clear that means nvidia's data center revenues grew by 2 and 1 half times in one year after they already became a trillion doll company and they're on Pace to make over a $100 billion in AI accelerator Revenue this year alone compare that to nvidia's two biggest quote unquote competitors AMD isn't even close to Nvidia data center sales and Intel is well as a result Nvidia holds over a 90% share of the data center GPU market and that's before their Blackwell GPU sales ramp up that huge market share is why Nvidia has so much pricing power even as a hard Ware company they reported gross margins of 75.1% putting nvidia's profit margins on par with most software companies and that's huge since according to Market us the global artificial intelligence Market is expected to almost 12x over the next 8 years which is a compound annual growth rate of 36.8% but many of the companies building Next Generation AI applications are not publicly traded think about the 9s and early 2000s companies like Amazon and Google went public very early in their growth cycle but today companies are waiting an average of 10 years longer to go public that means investors like us can miss out on most of the returns from the next Amazon the next Google the next Nvidia so I spent a lot of time digging into this and The fundrise Innovation fund is a great way to invest in some of the best tech companies before they go public venture capital is usually only for the ultra wealthy but fund Rises Innovation fund gives regular investors access to some of the top private pre-ipo companies on Earth without breaking the bank The fundrise Innovation fund has an impressive track record already investing over $100 million into some of the largest most in demand Ai and data infrastructure companies so if you want access to some of the best late stage AI companies before they IPO check out the fundrise Innovation fund using my link below today all right so Nvidia reported record revenues and earnings per share both of which also beat analyst EXP expectations on top of that they issued stronger than expected guidance for the current quarter so why did Nvidia stock drop 8% after such good earnings well even though their 75.1% gross margins are actually up 5 Points year-over-year they're down 3.3 points from the previous quarter and this is where things get really interesting nvidia's margins dropped because of something in Blackwell's designs that I've actually been calling out for a few quarters now something that's really important for investors to understand this is Blackwell the successor to nvidia's current generation of gpus called Hopper and compared to Hopper these new Blackwell chips perform a whopping four times better at AI training and an insane 30 times better at AI inference that's why every company wants these things since this huge leap in performance means data centers get a lot more compute power from a Blackwell system of the same cost size weight power or whatever is limiting their current data center footprint but Blackwell gpus are actually two separate dieses connected by an ultra high-speed 10 TB pers second link this connection is so fast that it actually tricks the two dieses into thinking there a single chip it's a really clever design that gets around a lot of the physical limitations of the machines that build these chips a lot but not all of them it turns out that the rumored 3-month delay for Blackwell isn't because of a design flaw at all but due to the accuracy limitations of some of the machines at tsmc the company that builds and video chips the placement of these Ultra high-speed chipto chip links needs to be insanely precise to hit those 10 terb pers second speeds if the link placement is off even by a tiny amount the whole chip could fail as its different metal layers and materials warp and expand when they heat up during normal operations as a result the percentage of working Blackwell chips that tsmc could produce or the yield went down fewer chips for the same amount of production time and materials means lower profit margins causing nvidia's margins to drop to 75.1% for the quarter and as for that rumored 3-month delay Nvidia says it already made the necessary changes to improve yields and Blackwell production is scheduled to ramp up in the fourth quarter and into next year Nvidia expects to ship several billion dollars worth of Blackwell chips in Q4 of this year and one big reason that data centers really care about Blackwell in particular is direct to chip liquid cooling today around 90% of all server RS are air cooled that means that a lot of data centers including hyperscalers like Amazon Google and Microsoft are making massive infrastructure changes to support liquid cooling for their Blackwell systems specifically industry estimates suggest that up to 80% of cooling will become direct to chip liquid cooling over time so this is a huge shift in how data centers operate around the world that's why Blackwell is such an important chip and this was such an important earnings call all right all right now that we walk through nvidia's data center earnings what's going on with Blackwell and why it's so important to Data Centers let's talk about the risks 's talk about nvidia's risks starting with their competition Nvidia has two kinds of competition and Ironically neither of them actually compete with Nvidia here let me explain the first kind of competition comes from chipmakers like Intel and AMD but nvidia's revenue from data centers is five times bigger than intels and amds put together and beyond that a lot of Intel and amd's data center revenues actually come from selling CPUs not gpus or other kinds of accelerators so they don't really compete with Nvidia as much as serve the part of the data center Market that doesn't run on nvidia's ecosystems given nvidia's 90 plus% market share for gpus I'd say this is a pretty small piece of the market the second kind of competition comes from hyperscalers that make their own accelerators mainly Microsoft Google and Amazon but there are two key reasons they don't compete with Nvidia either the first and most obvious reason is they don't sell their chips to anyone else Microsoft's Azure Maya Amazon's tranium and inferentia chips and Google's tensor processing units only exist within their own clouds to help their own customers with AI workloads specific to their use cases nvidia's gpus are much more general purpose and over 55% of nvidia's data center revenues come from companies outside these big three Cloud providers and that number is only growing over time but the second reason that Microsoft Amazon and Google don't really compete with Nvidia is that's just not how data centers work in the first place let me give you an example most people only have one phone or one car at a time so every sale for one company is a missed customer for their competition every car Tesla sells is one GM doesn't every phone Google sells is one that Samsung doesn't but data centers don't work like that especially when it comes to AI when Google buys a GPU from AMD or uses their homegrown tensor processing units that's not one less sale for NVIDIA since data centers handle a wide variety of workloads for many different kinds of businesses data centers are portfolios of different hardware and software Solutions and as demand for a specific kind of workload grows data center operators will buy more of the right Hardware to support it but in the end their goal is always to meet the changing needs of their customers and optimize their costs only a small portion of all workloads around the world involve AI today so Blackwell would be total Overkill that's why data centers buy chips from AMD or Intel for certain tasks they build their own chips for others and they use nvidia's gpus for heavy duty training and inference and no matter what Intel or AMD might say inference is becoming a heavyduty AI application as the industry moves from predicting the next word in a sentence to generating images and entire videos so while companies might make chips that Benchmark well against Nvidia for large language models their performance remains to be seen for text to video models like open AI Sora or complex protein structure and interaction models like Alpha fold 3 by Google deepmind my point is nvidia's biggest risk isn't competition at all it's complexity Nvidia didn't change Blackwell's design to compete with AMD or Amazon they changed it because even the best chip fabrication company on the planet had trouble placing their Ultra high-speed chipto chip connections and that's just one piece of one part of the ecosystem investors may not realize this but when Nvidia comes out with a new chip architecture like Hopper Blackwell or Reuben they're not designing one chip they're designing five separate chips that make up an entire computing platform a new GPU an nvlink switch chip that connects multiple gpus together a network interface card and two separate Smart Switch chips to connect multiple racks together one for ethernet and one for infiniband if any one of those chips has an issue nvidia's entire data center ecosystem has an issue that's why I spend so much time understanding nvidia's products not just their profits and why I focus on the science behind this stock but credit where credit is due look how fast Nvidia was able to address this issue with Blackwell and how little it dropped their overall margins from 78 to 75% for a single quarter so the only other risk worth mentioning is nvidia's Supply constraints look every company on Earth is constrained by one of two things they're either Supply constrained or demand constrained and as long as Nvidia keeps making massive leaps in compute power they're going to be Supply constrained but Jensen was crystal clear that Nvidia is ramping up Supply to meet the insane demand for Blackwell and backfill demand for Hopper it seems like a very clearly this was a production issue and not a fundamental design issue with Blackwell but the deployment in the real world what does that look like tangibly and is there a sort of delay in the timeline of that deployment and thus revenue from that product I I um let's see that's just the fact that I was so clear and it wasn't clear enough uh kind of tripped me up there right away and so so let's see we uh we made a mass change to improve the yield functionality of Blackwell is wonderful we're sampling Blackwell all over the world today uh we have started volume production uh volume production will ship in Q4 Q4 we will have billions of dollars of Blackwell revenues and um we will ramp from there we will ramp from there either way I'd rather see Nvidia run out of supply then run out of demand and as a shareholder I trust in video's leadership to keep innovating expanding their total addressable market and work closely with tsmc and their other partners to meet as much demand as they can speaking of which let's talk about what's in store for Invidia stock over the short medium and long term because there's plenty for shareholders to look forward to in addition to Blackwell sales ramping up over the next few quarters they're also ramping up their ethernet based networking Solutions like their Spectrum Force switch and their Bluefield data processing units both of these are chips that do calculations to spread AI workloads out across multiple gpus and then bring the final outputs back together $3.7 billion of nvidia's Revenue came from networking products this past quarter which is actually 30% more than all of amd's data center revenues for the quarter and Beyond the networking sales themselves Nvidia is expanding its total addressable Market big time by supporting ethernet as well as infiniband and that means even more data centers can integrate Blackwell and Hopper systems down the road over the medium term Nvidia is joining the likes of Apple Google and meta platforms when it comes to share BuyBacks when a company buys back some of its shares the value of each outstanding share goes up since it represents a slightly bigger piece of the company nvidia's board of directors approved another $50 billion in share Buybacks in addition to the $7.5 billion of shares they're still authorized to buy back from the last repurchase approval I like share BuyBacks way more than dividends because they show more long-term confidence in the business and those shares can be sold again at better valuations while dividends never make it back to the company for future growth besides BuyBacks Nvidia is gearing up for the aipc market with new GeForce RTX graphics cards and Nims at a high level nvidia's Nim microservices stitch together different functions to create a higher level service for example nvidia's Avatar cloud engine or Ace is built on microservices to do things like Translate speech to text and text to speech match a characters lips and facial expressions to what they're saying follow a specific set of rules context and guard rails and so on I think nvidia's Nims will be a real Revenue multiplier for NVIDIA over the long term since companies pay a per GPU subscription for these services and in the long term Jensen actually announced three different gpus at computex 2024 the Blackwell Ultra chips which ship in 2025 the architecture after Blackwell which is called Reuben which will ship in 2026 and then Reuben Ultras which will ship in 20127 all of these gpus are architecturally compatible which means that a hopper compute tray can be replaced with Blackwell or a Reuben tray down the road so when a data center buys chips from Nvidia today they're already investing in the infrastructure that supports new Chips making future sales much more likely for NVIDIA that also means that older chips will benefit whenever Nvidia or one of their customers writes a new acceleration library or software application so the whole Nvidia Hardware ecosystem gets better over time which leads to more adoption which leads to more developers and so the cycle repeats full circle how I picked some of the highest performing tech stocks before they made massive moves earlier this year check out this video next or if you want to see all my Nvidia covers including interviews with their Executives this playlist is for you le yo
Comments 0