저자: JoseNajarro · 외국인
요약 보기
테슬라 로보택시 가치평가(ARK 2029년 주당 2,600달러 전망)를 근거로 AI 칩 수요가 지속될 것이라 봄. 휴머노이드 로봇 대량 상용화는 5년 내 어렵다고 평가. MLPerf 벤치마크에서 엔비디아가 여전히 선두이며 H100 소프트웨어 최적화로 성능이 계속 개선된다고 언급. 브로드컴은 AI 실리콘 매출 증가로 실적 호조, 브로드컴 CEO의 엔비디아 경쟁 관련 발언을 인용하며 엔비디아 우위를 강조.
다듬은 글
welcome back to another episode today we're going to take a closer look at Nvidia and we have a lot and a lot of great updates today we have some Tesla price points that somehow I'm going to bring it back to Nvidia we also have some new benchmarks that investors should know and Broad come A huge Semiconductor Company just reported earnings and stock price is up dramatically they also said something interesting about Nvidia so let's take a closer look in ul the first thing I want to take a closer look at is for those that are not familiar today Arc may maybe you could care less about Arc but I just what they mention here I believe can help kind of understand where maybe the market might be in the future so Arc expects their value for Tesla in 2029 to be 2, $600 per share you might be like Jose why are you talking about this this is crazy this is insane but I just want to kind of see why they value it so much and one of the reasons that they value it so much is they estimate that 90% of Tesla's Enterprise Value and earnings will be attributed to the robo taxi business in 2029 and the reason I bring this up for NVIDIA is maybe Nvidia uh well Nvidia obviously they're using all their um Tesla is using nvidia's chips to do things like training um and if things like Robo taxis need to go out there I do believe FSD and all these kind of autonomous driving solutions that Tesla needs has needs to be a little bit better even though I believe FSD 12.3 the one that I have is insane I use it almost on a daily basis I still believe it has ways to go for it to be a full R bolt taxi so for them to do that they obviously need to train more of their models and to train more of their models they're probably going to need a lot more Nvidia AI chips so that got me pretty excited if they believe that this is where the future will be within the next few years then Robo taxi Tesla's not going to be the only one doing that right now some things that they did not mention here or they did not expect much is the Optimus so Optimus is their humanoid robotic and they believe while Tesla can grow in this they don't see this kind of ramping up into the commercial Market within the next 5 years so this is something that's pretty interesting because right now we are getting a lot of excitement here in the humanoid robots right this is one of the things that we saw during GTC events but Arc and again you might disagree or agree but to some extent maybe they might have some knowledge in the tech Market maybe you might have some different input on their investing but I would have to agree that right now for us to see a very huge meaningful ramp up in any type of humanoid robotss is not going to happen within the next 5 years so if your bullish case for maybe Invidia right now is robotics will come out in the next three years in forms of crazy crazy volume then I don't think that's the case maybe in forms of training right because before you ramp up that production you need to train those robots for the training for the software solution Nvidia can definitely benefit but if you want from the volume number then that's something that might take a lot longer um but overall I find this if this is to be true and if Robo taxi is expected to be within the next five years and if humanoid robots is something that can happen with after 5 years then we definitely have a lot of needs for nvidia's AI chips so overall I find that bullish before we go any further we are now on the road to 50,000 subscribers by the end of the year thank you all for the crazy support So if you haven't and are enjoying the content make sure to hit the thumbs up and the Subscribe button finally if if you want to support the channel a little bit more make sure to check out my special offer at full.com hose now back to today's episode now the second thing I want to take a closer look at is ml perf so for those that are not familiar there is a benchmark and it happens every few months where all these companies that have ai chips they kind of hey look we test all our chips on these benchmarks to kind of be able to Showcase hey look our chips are better than everybody else at the end of the day no one has beaten Nvidia and no one is coming close so Nidia is still the top ler but Intel did announce that their gouty 2 is doing pretty well right so they kind of showcase gudy 2 and they even made it in form of a large system this large system had over 1,000 2 AI accelerators uh so it kind of showcases that the gudy 2 is scalable so that's actually great news for Intel and being able to Showcase they also mentioned that the gy the sunu be gy 3 accelerator will be in the next Benchmark and right now it continues to show that the only AI accelerator out in the market right now to be able to kind of showcase that they're an alternative to nvidia's h100 is Intel with the 2 and I find that pretty interesting right that right now an AMD with their Mi 300X that's beenin out they still haven't showcased this in their ml performance so that's pretty interesting I really always wonder why they don't show it there um um but the gouty 3 is expected to be uh generally available from oems in Fall of 2024 now Nvidia like I mentioned showcased a lot of great things and Still Remains the top player here in the benchmarks there's no one anywhere better than that so the only one Nvidia is competing with is themselves and they're showcasing a few things first they showcase that in June 2023 they had a system that had about 3,500 h100s about a year later that unit has increased by over three times and the performance of that unit also increased by three times and let me just say that's pretty impressive because when you start to scale up multiple units together sometimes their performance starts to decrease so by having almost a near perfect scaling performance growth that's actually very bullish and that's what they're trying to Showcase here the other thing is inidia showcases that hey look their h100 in model like stable diffusion V2 for training um the same h100 is about 1.8 times faster on that model thanks to kind of all the software Solutions Nvidia has done so it showcases that hey look if I'm buying a product from Nvidia the h100 for example I know that the company's going to continue to improve via software and I can get better performance over time um another one that they showcase is the h100 the h100 in forms of training of a huge model with 175 billion parameters did increase about 27% versus last year and again thanks to things like their software optimization some other things we have here is NVIDIA right now mentions that in this ml performance they train the the their biggest accelerator unit that they showcase was one with 11,000 gpus but they believe in the future in 2024 they're going to have 100 Nvidia Hopper GPU from Frontier models and maybe that'll be able to train and in 2025 they'll have ai factories with 300,000 Nvidia Blackwell gpus so we can kind of already see the huge demand of Nvidia gpus at the moment which can also be very bullish I was actually very excited to see this and we can see the huge dramatic growth of gpus used per data center or per super computer the final thing that we are seeing is here's another example of kind of the improvements of Intel G 2 here they're kind of showcasing that hey look you say you're an alternative to us but Nvidia h100 is 2.8 times better and the h200 is 3.2 better in things like fine-tuning performance now in forms of influencing the improvements is even a lot better where the h200 is about 4.7 times faster so overall some great news there for NVIDIA shareholders and I do believe like I mentioned Nvidia right now is just competing against the itself now some other bullish news we have reports that came out today base that tsmc is accelerating their coas Advanced packaging capacity construction so obviously right now we are on the supply constraint market so if tsmc can improve that capacity that bols well for NVIDIA in the long term of things so just wanted to bring that out now the final thing I wanted to take a closer look at is broadcom so broadcom is up 14% after hours this is not a stock I'm personally in right now um but we can see the stock has been dominating year-to date the stock is up 37% not counting the after hours results and the main reason is this company one announced a 10 for one stock split but they also did beat Revenue I believe and beat earnings and beat earnings per share yes they did but one thing that they mention is that they are seeing huge growth in their AI accelerators so their AI accelerators are custom A6 Solutions which is very different from Nvidia but I'll let the CEO explain it in a bit um but the great thing about this company is they expect for 2024 that AI link chips to be in to be 11 billion in 2024 versus their prior forecast of $10 billion so obviously we can see the huge excitement for their AI chips uh and obviously that definitely is creating some excitement on the stock price now this is what I believe everybody should know about and everybody's here kind of considering hey Jose all these big Tech players are designing their own AI chip and these AI chips are going to eliminate the need for nvidia's chips in general and most of those AI chips that are being built are either from companies like broadcom or Marvel so these big Tech players go to them and say hey look we need help building these chips so one of the analysts hey can you asked Can you explain the difference or the competition between broadcom and Nvidia and this is what the CEO said on AI accelerators I think we are operating on a different to start with scale much a different model it is and on the gpus which are the AI accelerator of choice on Merchant in a merchant environment is something that is extremely powerful as a model and it's something that Nvidia operates in in a very very effective manner we don't even think about competing against them in that space not in the least that's where they're very good at and we know where we stand with respect to that so the CEO mentions that hey look Nvidia is the top dog we're not even competing against them and I'm not even going to lie to you guys that we compete against them right that's not where we at we know where we stand we know exactly where the where inidia stands and we respect that but what they do is they work on very select Ive or selective hyperscalers if there is a scale and this is pretty much if there is a specific particular and that's the thing particular complex AI workloads they're able to create these Asic AI accelerators what Nvidia provides are more gpus which are programmable and can work for any AI workloads but if there is one particular very complex AI workload that that big Tech player does over and over and over again then it makes sense for them to make an AI chip but again we can kind of see from the CEO of Bram a huge company how much respect they have on how well Nvidia does in the GPU market so I thought that's something that we should end this not on today's episode and reading that actually again I don't know how it can but it just showcases how strong of a lead Nvidia is amongst anybody else so brockham I appreciate you to bring uh the CEO of baram I appreciate for bringing the real thought process of how we should see Nvidia and this other Asic players now Nvidia is definitely dominating the space so congratulations to any investor uh let me know your thoughts let me know what you guys think of what brockum CEO said take care have a good day and see you next time
Comments 0