저자: JensenHuang인터뷰;진행자미상 · 외국인
요약 보기
엔비디아 Blackwell 램프업 성공, 실적·가이던스 상회. 추론형 AI로 연산 수요가 크게 증가(추론 연산량 과거 대비 약 100배). 중기 데이터센터 투자 확대로 올해 성장 기대. 중국 매출 비중은 수출통제 전보다 절반 수준이나 경쟁은 치열. Blackwell 수요는 전 분기보다 더 좋다고 평가. 진행자는 장기 강세 관점 유지, 마진은 일시적 하락 후 회복 전망.
다듬은 글
and I started my interview with Jensen hang CEO of Nvidia talking about the Blackwell ramp which he called extraordinary here's what he has to say we had a fantastic quarter a terrific ramp nothing was easy about it and and a couple quarters ago of course people were worried about uh how successfully we'd be able to ramp something as complex as Blackwell Blackwell people don't you know maybe they just forget that it's not just a chip but it's a whole system that system is a ton and a half has a million and a half components uh you know it's it's uh incredibly hard to build it's it's building 350 plants around around the world and 100,000 different operators Factory operators you know contribute to building this thing and so so I think it's it's uh it was it was logical to to be concerned about the ramping of Blackwell but we have now successfully ramped Blackwell the other concern that people had was uh the uh the hopper uh blackw transition and might create a pocket an air pocket and I I think we're we're now uh well successfully behind the air pocket we're going to have a good quarter this quarter we had a great quarter we're going to have a good quarter next quarter and uh we've got a fairly good pipeline of demand for Blackwell yeah um the guide was also above as you mentioned now earlier today speaking of demand I was talking to Amazon CEO Andy jasse he told me that as of now if he had more AI resources uh to to sell through AWS he could sell more that's kind of the the short-term signal of demand that you talked about on the call tell me more about the midterm signals that investors should be aware of that give you confidence in the continued demand that the scale outs of data centers AI factories relative to what you've historically seen the short-term signal are just our POS and the forecasts and on top of that the things that are not forecasted are new startup companies that are spinning off and some of these are quite famous and you know and at the risk of of of forgetting any of them I won't mention any of them but there's some really really fantastic startups that have come out as a result of uh new reasoning AI capabilities and you know artificial general intelligence capabilities that that they they have breakthroughs in and several of them there's several of them that are related to agentic AI really exciting companies and there's there's several of them related to physical AIS you know there's just handfuls of each one of them and each one of them needs additional compute and that's you know the type of things that Andy talks about because they need to go to AWS and they have urgent need for more compute right away and so that's on top of what we already knew to have POS and forecasts and such the midterm comes from the fact that this year's capital investment for data center is so much greater than last year's and of course we had a very large year last year we had a great year last year it stands to reason that with blackw and with all the new data centers going online we're going to have a fairly great year now long term the thing that's really exciting is we're just at the beginning of the reasoning AI era you know this is the air the time when uh AI is is uh thinking to itself before it answers a question instead of just immediately generating an answer they'll reason about it maybe break it down step by step um it'll do maybe some searching in its own mind before it creates and composes a smart answer for you the amount of computation necessary to do that reasoning process is 100 times more than what we used to do so if you could imagine we thought computation the amount of compute necessary was a lot last year and then all of a sudden reasoning AI deep seek was an example of that chat pt40 is an example of that grock 3 reason an example of that so all of these reasoning AI models now need a lot more compute than what we used to used to used to um let me stop you there because even more load because some people took deep seek to mean actually that you need less compute right because the the initial report was that they were doing more with less but you're saying in fact some of what came out of deep seek was the opposite that there's going to be more compute demanded uh UNP that for me there are three phases in how AI Works how AI is developed largely number one is pre-training it's kind of like us going through high school a lot of basic math basic Language basic everything that basic understanding of human knowledge is essential to do what is the next step which is called Post tring in posttraining you might get uh human feedback you know it's like a teacher showing it to you we call it reinforcement learning human feedback you might practice and do thought experiments uh you're this preparing for a test you're doing a whole lot of practices we call it reinforcement learning AI feedback you could e e either uh also do tests uh and practice and we call it reinforcement learning uh verifiable reward feedback so now basically it's uh AIS teaching AI how to be better AIS that posttraining process is where an enormous amount of innovation is happening right now a lot of it happened uh with these reasoning models and that computation load could be a hundred times more than pre-training and then Here Comes inference the reasoning process instead of just spewing out an answer when prompted it reasons about it it thinks about how best to answer that question breaks it down step by step might even reflect upon it come up with several versions pick the best one and then presents it to you so the amount of computation that we have to do even at inference time now is a hundred times more than what we used to do when Chad GPT first came out and so all of a sudden the combination of all these ideas largely related to reinforcement learning and synthetic data generation and and uh uh reasoning all of this is just causing compute demand to go diey high now tell me about price performance because this is one of those things that I think a lot of investors are trying to figure out they see the hyperscalers coming out with their own chips they say hey boy uh there are a lot of customers who don't have a lot of money to spend they want to get the most bang for their buck so they'll view trainum inferentia Microsoft chips Google's chips as potential competitors for NVIDIA but at the same time you talked on the earnings call about performance per watt right and so so it it seems to me at GDC and I know our Jim Kramer is going to be there with you next month that's where you introduce new products and and new powerful use cases if your performance of what you're coming out with next is that much higher than what's otherwise available on the market does that mean would you argue that your price performance actually ends up better than what's cheaper the reason why we sell so much is because our price performance is the best and it's absolutely the case um that performance per watt is incredibly important and the reason for that is because a data center is only so large that data center could be 250 megawatts or it could be a gigawatt but within that data center however however large it is you want the amount of revenues you can generate to be as high as possible so you want to generate you want to do two things you want to generate very high quality token tokens AI is ultimately expressed in tokens and that's what you monetize million dollars per million tokens you want to generate very high quality tokens because you could get better pricing on that better ASP on the other hand you want to get as many tokens out of that data center as you can and in order to do that your performance has to be excellent and your performance per watt has to be excellent and so the simple way to thinking about that is if your performance per unit energy is the high the revenues you can help a company generate is the absolute highest and there's if you look at look at the way we are driving our road map between Hopper and Blackwell our token generation rate for these reasoning AI models can be as high as 25 times that's a that's the same thing as saying that factory can generate 25 times more revenues using Blackwell than you could using using Hopper before that which is the reason why demand for black well is so great and then of course we on a road map that's once a year and every single year we're increasing our performance per dollar performance per energy performance W per watt so that everybody's data centers become more energy efficient on the one hand generate more revenues on the other hand so let me ask you about your China business um you talked on the call about how uh the percentage of Revenue is half what it was before export controls uh does the emergence of deep sea equ which some have cast as a workaround some of those restrictions does it tell us anything about the effectiveness of export controls um it's hard to tell whether export control is is effective the the thing that I can tell you is this um uh our our percentage revenues in China before export controls is twice as high as it is now there's there's a fair amount of competition uh in China export control and otherwise Huawei other companies are doing very are quite rigorous and very very competitive and so I think that that ultimately uh software finds a way maybe that's the easiest way of thinking about it you know software is always whether you're developing software for a supercomputer or software for a personal computer or software for a phone or software for a game console you ultimately make that software work on whatever system that you're targeting and you create great software and so that's kind of the beauty of software Engineers they're incredibly Innovative and clever in this way um our our architecture uh is about about the flexibility of software and which is kind of nice but ultimately here in United States if you look at where we are now compared to what is uh controlled uh gb200 is probably something along the lines of 60 times the token generation rate of what is being uh shipped in China that's currently export controlled okay and so the separation of performance is quite High um uh and uh uh you know ultimately we we uh what we know what we know we experience is a great deal of competition there real quick I think last quarter you said demand for Blackwell was insane this quarter you said it's extraordinary those about the same as one better than the other I would say that that um my feelings about Blackwell is better today than it was at last quarter and the reason for that is because we of course ramped up into production uh we exceeded our Target um and uh the teams did an amazing job as you recall we had a hiccup in our design flaw in Blackwell that we found uh early on last quarter or quarter before that and we recovered tremendously well and I'm very proud of the team for that and so for those reasons I feel pretty great from an execution perspective from a demand perspective you know deep seek was fantastic it was fantastic because it op sourced the reasoning model that's absolutely world class just about every AI developer in the world today has either Incorporated R1 using it's called distillation distilled from R1 um or using techniques that have been open sourced out of R1 so that their models could a lot more capable across the world AI has become better as a result of the last several months and so I'm excited about that and the demand for computation for inference time for test time scaling which is one of the reasons why Grace Blackwell MV link 72 is so exciting that feature is now more prominent than ever more demanded than ever and I think the the uh the the demand side of it is more exciting Too jamson Wong thank you CEO of Nvidia hope to see you soon thank you John great to see you all right I hope you're all doing well today but simply my thoughts on this earnings report are very positive let's go through some important points first Nvidia of course beat substantially on the top and bottom lines additionally nvidia's next quarter Revenue guidance came in well above consensus at $43 billion that is very very good now one thing that some Market participants are not happy about was that Invidia expects q1 margins to be lower at around 71% this is nothing to be concerned about Nvidia leadership told told us this would happen as Blackwell ramps and they transition to a new architecture and they address this directly on the earnings call margins will be slightly lower because Nvidia is quote Expediting manufacturing to ramp Blackwell faster once Blackwell has ramped then Nvidia expects to see increased margins in the mid-70s around the end of this year that is very reassuring news on the topic of China and Export controls Nvidia does not seem to be concerned about this situation and they actually expect China sales to make up the same percentage of Revenue in q1 as they did in Q4 one of my concerns going into this earnings report was the possibility of Nvidia giving softer guidance due to the looming threat of tariffs but thankfully that appears to not have been the case instead Nvidia appears to not really be concerned about the current geopolitical and international trade situation that is good news for investors and should help calm some fears not only in Nvidia stock but also throughout the broader Market we also got some good news about Blackwell and Blackwell Ultra on the earnings call Jensen acknowledged that Nvidia has fully recovered from the delays they faced in the rollout of Blackwell and on that note Blackwell Revenue was $1 billion for the quarter that is much better than what in video guided for on their last two earnings calls and about Blackwell Ultra Jensen said that the transition from Blackwell to Blackwell Ultra later this year will be much easier and much smoother than the transition from Hopper to Blackwell that is very good news and speaking of new products from Nvidia CFO Colette Cress told us that Jensen's keynote speech at GTC on March 18th will be a quote news packed event Jensen later told us that we will learn about Blackwell Ultra Reuben and even the next product generation after Ruben at GTC as a reminder GTC is March 17th through the 21st and that keynote from Jensen that I just mentioned is on March 18th mark your calendars for that event overall I am very pleased with this earnings report as well as invidious forward-looking guidance on both revenue and the production ramp of Blackwell I'm also even more excited for GTC now that we know we are indeed getting a newss packed event with new product announcements from Nvidia that is something to look forward to in the immediate term I have no idea how Market participants will react to these earnings but from a long-term perspective Nvidia is absolutely crushing it and I still think a long-term thesis is firmly intact I am still bullish on Nvidia from a long-term perspective with all of that being said I hope you all have a great rest of today and I'm curious to hear your thoughts about Nvidia in the comments below
Comments 0