원문 보기
what did you think about the last 24 hours hey Nicole thanks for having me back on Happy New Year first time in January so good to see you again so what happened yesterday uh obviously was a huge disruption for the markets um but I don't think it's really about deep seek this is something that you and I have been talking about for for several months now and I think the bigger story here is the expansion of AI workloads and that's training and inference from the cloud uh to the edge and and what we're seeing is essentially a Confluence of of two things one the models are becoming much more efficient which means it's a lot faster to train the models and it also requires a lot less computational power less energy less water all these resources that that we normally assigned to uh big heavy uh AI training workloads uh so that's that's because of these efficiency improvements we're seeing less and less uh need for this uh and then the other thing is the chips are also becoming much more efficient which is like the idea behind nvidia's Blackwell they're much better chips they're they're a little bit still expensive on The Upfront but uh on on the back end in terms of the operational cost you're getting a lot more Roi because these chips can perform a lot better so better software better training more efficient models and also more efficient chips and what deep seek was has done essentially is sort of U shown the possibility of of what can be done with this uh so I don't know that we'll even be speaking or talking about deep seek 6 months from now but we will be talking about this broader theme of the the efficiency the improving efficiency of models and ships essentially being a a a sort of a a much lower cost alternative to the way that we're thinking about AI training uh and even in uh infrastructure Investments now if there is any doubt a Chinese software application is not going to cause Nvidia to lose its Market leadership actually you have an opportunity to buy a lot of Nvidia and power grid companies due of Monday's deep PL in the US Stock Market caused by Deep seek this was undoubtedly not the case on Monday as investors expressed worry over allegations that deep seek a Chinese artificial intelligence service was both Superior to open Ai and significantly less expensive all investors should be reminded that the stock market is essentially a crazy crowd at this time the fact is that crowds don't think they just respond in actuality IQ decreases with crowd size therefore the stock market's Furious reaction during Monday's sell-off was especially foolish since investors were behaving strongly without thinking as strongly the stock market appears to be regaining its Clarity the Deep seek incident was referred to by President Donald Trump as a wake-up call for American businesses to reclaim their leadership in artificial intelligence because Silicon Valley will have the same degree of support from the Chinese Communist party as Chinese Enterprises but in a capitalist setting it is anticipated to Prevail in the AI Wars meanwhile according to a Wall Street Journal article deep seek asserts that it's our 1 and V3 models outperformed chat GPT or came very close despite export restrictions put in place under the Biden Administration to restrict the sale of nvidia's cuttingedge AI chips deep seek has succeeded the entire plan for the development of AI is currently under scrutiny because of this some investors are unsure if the energy intensive and Chip hungry us AI business will prevail in the AI race though details on the precise percent very greatly deep seek boasts that its AI search algorithms consume a lot less power than chat GPT and other AI search engines this implies that deep seeks regenerative AI is operating on a very different architecture than the one that Nvidia invented there might be a more straightforward answer we carefully consider which degrees of freedom to incorporate in our quantitative models in order to prevent the inclusion of uncorrelated data factors 1 2 and three provide significantly more information than factors 6 7 and 8 when attempting to identify a vital path to Discovery I believe deep seek is avoiding using power on uncorrelated data and restricting the degrees of freedom in order to streamline their AI searches though clever and evocative of the British deciphering German codes during World War II this tactic may be repeated and enhanced without requiring a complete overhaul of the AI architecture currently used by American businesses I was reading about the founder um The 40-Year-Old founder in China of deep seek and um throughout you know what he's been doing but it was cited that he had said at some point the models were trained on fewer than 10,000 Nvidia a100 GPU um and so the question is is you know we've seen push back from Elon Musk and others then I read today that they weren't using Nvidia trips or but you talk about training and you talk about inference is there any way to know whether or not Nvidia help to make deep seek because that could help Nvidia too right I mean that could mean more demand for NVIDIA and I know that we're going to get to how maybe we're not looking at this the right way but just that little fun fact I mean important fact do you know anything about that uh well no not yet and I don't want to speculate about which chips were used it's it's kind of not really relevant uh I mean it's it's it'll be interesting if it's Nvidia chips or not uh maybe it's a mix of Nvidia chips and other chips uh they're they're not the only player in town uh but I think what's interesting is it it is a proof point for the the increasing efficiency of the models themselves and the increased efficiency of of uh the chips themselves and Nvidia there uh ironically is I you know you're you're showing uh some of the the losses yesterday envidia is really well positioned to take advantage of these more efficient models and I think that if we reset the conversation back if we if we go back before yesterday's announcements uh and and sell off and we go back before the previous week so last week's announcements of the Stargates uh infrastructure investment project and go back to what uh Jensen was saying at CES about what Nvidia was working on and what the the product road map was and the revenue road map I think that's still uh that's still on target nothing has changed it's just we raised our expectations last week with Stargate in terms of Investments uh and then deep seek basically just pulled the rug out from under that saying no you don't need to invest that much um but it fundamentally what Nvidia was saying two weeks ago is still true you still have this expansion of of AI uh trading and inference you have this expansion of it in the cloud and at the edge and to end points and so we're we're entering into a more hybrid AI training and workload model uh that still benefits Nvidia uh it's just you know last week's numbers I think were a little bit inflated so alphabet meta platforms Microsoft open Ai and other AI search companies are now investigating deep seeks capabilities and those of their own AI search engines few developments are more inspiring to Silicon Valley than a wakeup call therefore I have little doubt that deep seek success will not last even if it has a short-term AI Advantage recall how cloud computing and the move to mobile Computing both posed enormous complicated obstacles but are now commonplace and controlled by American Tech firms those assisting in the expansion of the electricity infrastructure for new data centers were particularly severely struck on Monday following the deep- seek Revelation as a result companies like Eaton Powell Industries Argan Comfort Systems USS ER Core Group Power Solutions International Verve Holdings and vistra Corporation are suddenly begging to be bought it is extremely unlikely that deep seek will lower the demand for AI power furthermore it is absurd to think that deliberate utility corporations will scrap their plans to boost grid capacity due to a Chinese software application China may also be launching a propaganda effort to humiliate Silicon Valley and the Trump Administration as deep seek caused a market sell-off that momentarily established China's AI dominance versus silicon valleys recall that short selling which is common in the United States is prohibited in China due to potential Chinese propaganda I would advise investors not to sell Nvidia and the businesses assisting in the expansion of the power grid if a new AI economic War breaks out remarkably Nvidia did not take as much of a beating on Monday as the businesses assisting in the expansion of the electricity grid this is actually a positive indication that Nvidia was showing some relative strength and that the smart money was purchasing more of the company during its Monday decline on pullbacks astute investors buy up solid equities following the announcement of their next quarterly profits Nvidia and businesses assisting in the electric grid expansion should see a 20% to 30% increase in value hangler Investments thank you both for being here with us right now let's talk about Dan the epicenter of all of this and you've been a notable Bull on many of these companies many of the technologies that are at the center of the Meltdown today does this worry you I mean to my in my view it's a top three buying opportunity that I've seen in the last decade for for Tech and the reason I'm saying I spend the whole weekend talking to Tech Executives cios individuals that I know that are going down the sort of AI arms race or going through use cases in no way of really 25 companies does this change their sort of plans and a lot of these could be seven eight you know even upper eight figure type budgets I just continue to view this as it's a model and a lot of the details I think still need to be vet out in terms of 6 million do they actually have the h100s how they actually built it but this is something we're going to look back on and I think the street and the market got this way wrong in no way does it Dent our bothies that's what we've been telling clients really since the weekend these are the times you own these names Dan not all of the companies at the center of the selloff today are necessarily affected the same way chip stocks Nvidia Kelly mentioned it a half a trillion dollars in market cap lost that's more than the size of Exon Mobile that's roughly the size of United Health Group in terms of an entire company lost in market value some are affected more than others Mega Cap Technology not as much so across the board as say Nvidia per se when this is a buying opportunity how much of a buying opportunity and how specific do you have to be well first un envidia that continues to be I think the Golden Goose in terms of what Jensen and that company are ultimately there is one chip in the world there one GPU that's fuel in Ai and and it's Nvidia and when you think about these build outs when we talk about the Enterprise the two trillion of AI CB backs next three years that doesn't change from deepsea you know we're talking about one 1,000 of ultimately the buildout that you're going to need for a lot of these Enterprise use cases you look at names like paler you look at the hyperscalers I could argue with inferencing if you actually then costs are coming down it's actually bullish for even some of the software players names like sales for service now Oracle and others I just think this when when you look at Nvidia I think we sit here 6 9 12 months now look at this as a moment it was a golden buying opportunity and in no way changes our bull thesis for where this is all head the Revelation sent massive aftershocks through the tech equities on January 27th causing major Tech indices to plummet many specialists however hold a different view after deep seek released its R1 model which apparently outperforms top Western AI models like gp4 despite being produced with less resources boa brushed down worries about a decline in demand for AI semiconductors boa thinks this concern is exaggerated even though the model's Effectiveness might indicate a decreased need for expensive AI chips the R1 model probably uses foundational models such as meta llama which accounts for the majority of infrastructure expenditures the increasing need for computing resources is demonstrated by llama's parent company's intentions to raise 2025 capital expenditure by more than 56% to 6065 billion boa anticipates that larg scale AI inference derivative approaches and foundational models will fuel Rising computational demands additionally Dan Ives a wed Bush analyst described it as a chance to purchase US tech equities Atif Malik a city analyst kept his buy recommendation on Nvidia on January 27th he said a price objective of 175 the analyst talked about the issues brought up by Deep seeks llm R1 one which was said to have performed well for a fraction of the price of its American Rivals although Malik saw the promise of deep seeks accomplishment he questioned if it was possible to achieve its success without sophisticated gpus perhaps by use methods like distillation Malik thinks American businesses still have an advantage because they have access to cuttingedge processors even in the face of potential International competition additionally he cited recent announcements of AI Capital expenditures like Stargate as evidence of the ongoing need for state-of-the-art gpu investors they took advantage of the dip buying more than 550 million in shares that's the largest amount on record going back to 2014 according to Venda research here but the larger question too is off of yesterday's activity how much of that and the bounce that we're seeing a little bit here today how much of that will hold are we in the dead cat bounce uh territory or is this something that is more pronounced or prolonged now everybody would be rights to also look at a longer term chart yesterday's decline included but then also if you're looking at some of the moving averages Nvidia broke through a 200 day moving average the larger question that I have is are we going to see especially extended hours pre-market as we've gotten back above that 200 day moving average is this a sustained move back above that or is this the lead to what Ray doio is calling out in some of the early uh setup that is similar to what took place in the Doom bubble yeah I think if history is a guide it is a bit concerning right when you take a look at the past as some of the larger sell-offs that we had specifically with Nvidia and going back to the last eight times it actually fell more than 17% % the next day it only closed higher three of those times so far less than half so we could see potentially a bit of a reversal here as we do wait for the trading day to get underway and then of course throughout the day the one silver lining though I think for investors and very much has been the focus it's been in a lot of notes here this morning is just a concentration of yesterday's selloff yes we did see deep sea of red here when you take a look at many of those larger C names when you take a look at specifically at the NASDAQ with the NASDAQ closing off just about 3% obviously a bit worrisome here when you take a look at a like Nvidia given the size the magnitude how important it has been to the markets overall momentum here over the last several months falling and and falling so dramatically but when you take a look I say the S&P equal weight index that that that was essentially unchanged yesterday when you take a look at some of that price action even beyond that certainly a bit encouraging here majority of the S&P stocks actually ended the day higher you had the advanced decline index and that that measures stocks uh Rising minus those that are falling that showed a positive reading of 199 even as that Benchmark fell just about 1 and a half% so again that puts some of this selling into context and yes while there is reason to be concerned and you see why some of the larger uh well-known investors like aay Delio are maybe drawing larger comparisons to what has happened in the past I think it is a bit reassuring here for investors this morning that you can easily I think make the argument that the bull market is still very much intact yeah options Traders fervor for the chipmaker shares was unabated by Monday's big sell off in AI darling nvidia's shares as the stock rebounded on Tuesday Traders were ready to pour back into bullish Nvidia contracts concerns about a lowcost Chinese artificial intelligence model that challenges the supremacy of us competitors caused Nvidia shares to plummet around 177% or nearly 593 billion in market value on Monday this is a historic one-day loss for any corporation on Tuesday the stock was up almost 7% options Traders were already taking advantage of the selloff to position themselves for a recovery but investors who had become accustomed to the stock Rising steadily were startled according to Steve snik Chief strategist at interactive brokers it has undoubtedly been a speed bump for many but it hasn't stopped activity after falling to a more than two-month low of 1.36 to1 on Monday call options which are usually purchased to indicate a bullish opinion on a stock outnumbered put options 1.6 to1 on Tuesday which is almost in line with the one-year average we've witnessed amazing dip buying snik stated according to trade alert data by 1:30 p.m.
around 5.4 million Nvidia options contracts had changed hands which is almost double the typical rate in the days preceding the sell-off options Traders had become exceedingly optimistic about Nvidia and AI technology in general this optimism was bolstered by President Donald Trump's promise last week that the private sector would be investing heavily in the construction of additional AI data centers much of the volatility was caused by positional exess according to Brent kachuba creator of the options research firm spot gamma having said that kachuba claimed that Nvidia options trading was far more balanced and two-way
Comments 0