원문 보기
why is deep seek not a concern for NVIDIA look um competition in the AI Revolution is going to happen and and we've always viewed commodation of llms of models as ultimately the path now in terms of deep seek in terms of 6 million to produce a model this good with no Invidia h100s I mean it's the same chance that I'm playing in the NBA for the Lakers uh you know in terms of the the chances that that's actually right so I think part of this issue is the street navigating what's the facts versus what aren't the facts in no way does this disrupt the AI Revolution thesis two trillion of AI capb backs is going to happen over the next 3 years regardless of deep seek Dan IES downplays the impact of deep seek on Nvidia calling claims that it built a competitive AI model without Nvidia gpus highly improbable and exaggerated he asserts that AI spending estimated at $2 trillion over the next 3 years remains on track with Enterprises continuing significant investments in Nvidia powered infrastructure recent Market sell-offs tied to deep seek he believes our overreactions and present a buying opportunity Ives emphasizes that while open-source models like deep seek showcase Innovation they don't match the scale or Enterprise level capabilities Nvidia provide demand demand for NVIDIA gpus far outpaces Supply and major players like Microsoft and meta remain committed to AI growth he sees no disruption in the AI spending trajectory due to deep sea lastly IES underscores the US's leadership in AI Hardware particularly physical AI driven by Nvidia Tesla and Google and protected by export controls he remains bullish on Nvidia and the broader Tech sector viewing the Deep seek debate as a minor bump in a long-term growth story that c Ben it's Nvidia Microsoft Amazon Palante tier and others no me have for but we were basically saying with the uh Lakers analogies it's a super low probability that it's actually true it's it's Fiction it's Fiction it's a Netflix show got it got it that deep seek spent 6 million and didn't use Nvidia gpus to produce I think this was one of the probably top three most unwarranted sell-offs I've seen in Tech since Co 2020 deep seek in no way changes the path of AI spend whether it's 80 billion from Microsoft 65 billion from meta and I just had a small llm model that consumers are interacting with is a lot different than literally 10,000x the capacity and Al algorithms needed to operate in an Enterprise so first we believe they had 30 to 50,000 Invidia h100 gpus to create so which is to level set in other words I don't believe what based on all the experts I've talked to as well as many that they were able to produce what they're saying with no Hardware no GP so I think to levels I think that's an important to to start with and then I would also add model open- Source models that's what meta has done the price of models and inference is going to continue to go down but that's a positive because you only have 4% of Enterprises in the US that are going down the AI path that means more computing power for use cases cheaper that ultimately more Enterprise sizes are going to go down and right now the demand to supply for nvidia's chips are 15 to1 we've talked to 25 Enterprises over the last 48 hours that are going through seven and eight figure AI deployments zero of them have even flinched because of deep seek and say like deep seek is the teu of AI just like teu was going to kill Amazon that's exactly how we view deep seed and these opportunities 25 years cover in Tech the Bears sound smart when they come out of their caves when they're in hibernation mode the Bulls make money um let's just say if you go to Time Square and someone selling heres sneakers for $30 that's how I would equate deep seek in AI it's a great model no one doubts that but to say that it was made $6 million with no invid gpus is a fictional Netflix show why do you think so many people took it so seriously though like especially the the engineers um in Silicon Valley they all seem to have taken pretty seriously no it's taken seriously in terms of the model that it's built the reaction from investors is way mispricing what it's going to do to spend it in other words this is not some Black Swan moment that's going to change the AI spending trajectory it's a great model Chinese built that shows open source is going to continue to close the gap with you know close loop like open AI being bullish on AI the llm models are just one small piece it's all about the data the use cases the algorithms and that's that's our view in other words like this is a panicked selloff that's hugely misguided even though it's it's a scary headon right so my follow next question was actually going to be on the potential implications to the HPM suppliers but my guess is that your view is that nothing changes my my view is I I'm not saying that you can have regards to deepsea we are in the early days of a $2 trillion AI tap X and that's not changing because the opensource model coming out of a lab in China is going to take down the US tech world that's really at the end of that's sort of our um our view after talking with dozens and dozens of Executives AI Engineers companies going down the path it's not just me being like you know my own view yeah yeah what about in the near term over the next couple of quarters um do you see companies at least saying you know the traditional buyers of these uh Nvidia um gpus let us reassess at least even a little bit no because then you go to the back of the line there's no lack of Confidence from anywhere in the ecosystem I think this is going to be viewed as more of a golden buying opportunity and every year the last few years even going back to co we get call it five six times a year pockets of that you could own these names and and and it's no different that's our Playbook because we believe we're less than halfway through a tech ball market with the Trump Administration that's going to clearly limit more and more export controls around chips especially as the bite Dan's Tik Tock situation is playing out as we speak and I just continue to view that us will continue to have a massive lead over China because of the bigger players in the AI Revolution what about other types of Hardware like cumate robots autonomous I think when it comes to physical AI they will clearly be behind because physical AI it's not even a debate that would be in videoed and the export controls will hugely limit their ability on physical AI deep Seek You could argue all day long in terms of like what did they use is is 6 million is 100 million but no doubt a successful model they've built physical AI whole other ball game that I believe there are years behind us when it comes to physical AI because of the nature of Tesla Nvidia and of course what we have with you know with wio and and Google let's shift gears to Apple yeah and I think Apple it's a good time to because what I noticed is um like Apple stocks been up during this whole period right and the question would be like why would Apple be up when the rest of tech is is really down it's our view Apple wins either way cuz it's not about llms they will be a platform that 25% of consumers around the world will access AI through whether it's us China Korea Middle East deep seeks claim of building a top tier AI model for $6 million without Nvidia gpus is widely doubted with analysts like Dan Ives dismissing it as improbable and overhype he asserts that this narrative caused an overreaction in the tech Market but does not alter the $2 trillion AI spending trajectory or nvidia's dominance open- Source models while advancing cannot replace nvidia's Enterprise grade Solutions and the demand for gpus remains strong Ives views this as a buying opportunity emphasizing the US's sustained AI leadership particularly in physical AI bolstered by export controls can open-source AI models realistically compete with nvidia's Enterprise dominance in the near future what role do you think export controls play in maintaining us leadership in AI technology
댓글 0