Have you rethought or shifted your view at all in terms of the depreciation schedule for semiconductors, the chips that are being used in these big data centers? And I think that that is fundamentally almost the whole game in terms of the math. Either you think that these chips can
be used for six, seven, eight years, in which case the math makes sense, or you think these chips have to get upgraded, or there's going to be such an upgrade cycle that they're not going to be used the same way three or four years out, in which case the math gets a lot more complicated.
I think as is so often true in bubbles, it's really a religious argument. To some extent, it's hard to prove concretely one side of the other is absolutely right, in my opinion.
I think that there's a little bit of truth to both. At the same time, we have NVIDIA upgrading their chips at such a high rate every 18 months or so, and really creating chips that you almost can't avoid upgrading to if you are a hyperscaler because of the efficiencies that those new chips
bring. At the same time, the older chips, the chips that are now 18 months old, they still are very functional for different workloads. They're still functional for inference. They're still functional for other workloads that we've seen. These hyperscalers shift work to even back to
A100s, NVIDIA's chips that are five plus years old now. I think that in a weird way, both things are true. Because of that, when I think about the cycle, where are we at right now, I think that leads me to believe that we still do have more time here to go, because we haven't
gone down the path where we're starting to depreciate these things at an unrealistic clip, where we're still not monetizing them in reality.
走上這條路，開始以不切實際的速度折舊這些東西，我們仍然沒有在現實中實現它們的貨幣化。
01:46
As the Trump administration decided whether or not to allow NVIDIA to sell the H200 chips into China, that's been a discussion and something that increasingly people are speculating about.
Well, I've seen that speculation. That kind of decision sits right on the desk of Donald Trump.
我已經看到這種猜測。這種決定權正在唐納德·川普的桌上。
02:06
Right? He's got Jensen from NVIDIA who really wants to sell those chips, and he's got a good reasons for it. There's an enormous number of other people who think that that's something that should be deeply considered. And the benefit that we have is we have Donald Trump in the Oval Office. He is going to weigh those decisions. He understands President Xi the best. He will
decide whether we go forward with that or not. That's on his desk with lots of different advisors.
決定我們是否繼續推進這件事。這件事正在他的桌上，有很多不同的顧問。
02:31
The president loves to hear lots of different voices to make those kind of decisions, and he'll decide whether we sell those chips or not. And then we will go execute it. However, he decides to go forward. There's a bigger question here, Secretary Lutnik, about what national security is. Is it more of a national security risk to give China some of these high powered chips,
or is it more of a national security risk to not have US tech in China? Has your view on that changed?
或者不讓美國科技進入中國是否更具國家安全風險？你對此的看法是否有所改變？
02:57
Well, that's the question exactly, as well put, and in front of the president, which is, do you want to sell China some chips and keep them using our tech and our tech stack? Or do you say to them, look, we're not going to sell you our best chips.
We're just going to hold off on that. And we're going to compete in the AI race ourselves.
我們只是暫時不會這樣做。我們將自己參與人工智慧競賽。
03:19
So that is the question. It's in front of the president. He's going to decide—it's a really, really interesting question. He's got all the information. He's got lots and lots of experts talking to him. And he's going to decide which way he wants to go forward.
It's on this idea of depreciation, depreciating assets within these companies and the costs with the issue that Michael Burry and you've raised this, too—this idea that the life cycle of tech is becoming shorter. So these assets are depreciating faster, costs rack up quicker.
And it's not something that the companies have accounted for in their own figures.
而這不是公司在自己的數據中考慮到的。
03:51
Correct. How do you solve that?
正確。你如何解決這個問題？
03:52
Well, the companies themselves really don't know. I mean, this is one of the big known unknowns. And that is, how long are these chips going to be valuable? How long are they going to work? Some of the invidious chips are pretty hot and they kind of self-destruct after a while.
But look, I think the important thing to recall is that a couple of years ago, when you looked at the number of data centers in the United States, there were 4,000 of them.
不過，我認為重要的是要回想，幾年前，當你審視美國的數據中心數量時，它們有4,000個。
04:21
In other words, this is not like a new surge in data centers. And the reason that we're building all these data centers is because we need the capacity. I'm using AI all day as a research assistant. And everybody I talk to is looking to see how they can use AI. But there's lots of
other technologies that are being used now to increase productivity because we have to, because there's a shortage of labor for all sorts of different reasons.
其他技術現在正被用於提高生產力，因為我們不得不如此，因為由於各種不同的原因，勞動力短缺。
04:51
So that's one side of it—how useful is this stuff? Can you get the return on it? The other side are—do funding markets remain open for these companies to continue to tap to do the incredible amount of spending that they're doing? You have seen some nerves there, be it in CDS spreads of
所以這是一方面——這些東西有多大用處？你能從中獲得回報嗎？另一方面是——融資市場是否
05:06
Oracle, be it in the ability of private credit players to keep funding it if you have wobbles in their publicly traded stocks. Is that a concern for you that credit markets don't absorb some of the spending as well as they have been? It's a concern, but at the same time, I'm glad to see that the markets are disciplining the whole process. I mean, again, the fact that people are so
nervous about an AI bubble, the fact that some of these risk spreads for the bonds being issued by those who are trying to raise money for AI, the fact that the market is onto the notion that let's not overdo it—that's a good thing. That's a discipline on the fact that we might be prone
to excess, but maybe the markets will diminish that. Does that mean that the melt up risk, a week ago, I lowered it to 15% because I think the fact that the hype is sort of,
let's put it this way, that the air is coming out of the bubble rather than the bubble bursting is, in my mind, a very positive development. Do you think that Stacey, the dust has cleared a little bit from the big release last week? Everybody was waiting to hear what Nvidia had to say.
Turns out they had much better numbers than people had anticipated, much better guidance than people had anticipated, and yet the stock sold off, at least initially.
結果顯示，他們的業績和指引都遠超預期，但股票卻至少在初期出現了拋售。
06:26
Yeah, it initially didn't, right? Initially it went off, and then it went up, and then it gave back for the trading session, the first full trading session after the release of earnings.
是的，最初並沒有，對吧？最初它下跌，然後上漲，然後在第一個完整的交易日後回吐了收益。
06:37
You definitely saw a huge chunk of the stock taken out. Sure, sure, but it wasn't just Nvidia, right? Everything sold off. And in fact, Nvidia actually outperformed the semi-sector that that day. So it wasn't just—I'm not convinced that the sell-off we saw that day was necessarily
like Nvidia-driven. Although I do think you're right, people are starting to get nervous about AI. They've been nervous, right, about bubbles and everything, but that did sort of come to the forefront, partially just because the numbers are getting so big so quickly. There were some other events last week as well. We had the release of Gemini 3 from Google, which looks like a great,
great, great model. And some of the other semi-names, you mentioned AMD, for example, which was down a bunch. Some of the other names that are, say, more dependent on open AI sold off more, I think, on the fact that Google's Gemini looks really good. So I think there were a lot of factors going on last week. I don't think it was just Nvidia. Although I wouldn't have expected
Nvidia's results to drive a little bit better performance. Me too. I know you raised your own price target for Nvidia after you saw the company come out with its shares. You raised your price target, and you raised your earnings estimates pretty substantially. I guess if I'm trying to
figure out where things stand at this point, how much of this is it easy for you to do as an analyst? You say, these are my expectations for earnings. This is how I compute this out, and here I get to this. If you've got a lot that's riding on sentiment right now, how do you factor that in? Well, I mean, here's our earnings, and that's the easy part.
I mean, actually getting it right can be tough. And look, it's sentiment and everything is always part of this. I feel OK about Nvidia right now. It's not like everybody has to jack up the multiple
to get to the targets to where they are. And I've said this many times, on the station, the stock today is much, much cheaper than it was before any of this started a few years ago.
來達到他們的目標。我已經多次在節目中說過，現在的股票比幾年前開始時便宜得多。
08:36
The stock is up a ton, but the earnings have more than kept pace with it.
股票大幅上漲，但盈利增長更快。
08:40
I mean, it's trading in, oh, I don't know where it is, mid-20s price to forward earnings right now.
我的意思是，它現在的交易價格在，哦，我不知道具體是多少，目前的預期市盈率在20多倍左右。
08:45
It's not even aggressive if you think the numbers are anywhere close to right. So the earnings, and this is the issue, right? When you see something like that, what it means is that the market is anticipating that those numbers are not sustainable, that it is a peak. And so, I mean, that is the debate now. How long can this go on? And you start to look at it, though.
I've been saying this since this started. At some point, presumably they'll be an air pocket or something. Nothing goes up into the right forever. But it's clearly not now. And I don't know what it is, but it's not now. It's not this year. It surely doesn't look like a net. It's next year.
And I mean, it may not be 2027 either. Things look pretty robust at the moment.
我的意思是，可能2027年也不是。目前的情況看起來相當強勁。
09:24
I guess when you have these really fast developments in technology, and this seems faster than just about anything we've ever seen, you're right. The leadership positions can shuffle and jostle, like you mentioned, Gemini 3 coming out and being maybe something that
could steal a lot of thunder from OpenAI. But then you also have this threat of the ASICs, the custom AI chips that all of the big purchasers of NVIDIA chips, AMD chips are also looking to make their own. How do you get your arms around that?
Sure. And by the way, even with that, it sort of flows back to Google. The only company that's really deployed ASICs in sizable at scale volumes is Google. And to be here, they've been doing this for well over 10 years. This is why everybody else is much earlier in the process. It takes time.
Even Amazon, they've deployed, they have something called training, and they've deployed a little bit, but it's not anywhere near on the scale that Google has. That folds into the Gemini point earlier. They trained Gemini on TPUs, although they've been training all of their models on TPUs.
TPUs are their chip. They've been training all their models on TPUs for a long time.
TPUs是他們的芯片。他們很久以前就開始在TPUs上訓練所有的模型。
10:30
I get the competition worry and look, this is semiconductors. There's always competition.
我理解競爭的擔憂，看，這是半導體。總是有競爭。
10:35
There's always someone looking to eat your lunch, and it's imperative on you to make sure that doesn't happen. At the same time, though, I don't really worry too much yet. I think it's the wrong question. The question right now is not who's winning or losing, in my opinion. It goes back to the first one. Is the opportunity in front of us still big or is it not? Is it sustainable or
is it not? Because presumably, we're not in a saturated mature market. We better not be.
或者不是？因為據推測，我們不在一個飽和的成熟市場。我們最好不要。
11:00
And if the opportunity is still big, they're both fine. If it's not big, they're all screwed.
如果機會仍然很大，他們都很好。如果機會不大，他們都完蛋了。
11:06
We're not at the point where I think you necessarily have to worry who's winning or losing it. Then the pie is still hopefully getting bigger.
我們還沒有到必須擔心誰在贏或輸的地步。然後這個餅仍然希望變得更大。
11:13
OK, Stacy, we've got to run, but your favorite—first, second, and third choices and stuff.
好的，Stacy，我們得走了，但你最喜歡的——第一、第二和第三選擇等等。
11:20
Our general call this year has been owned the high quality AI names and ignore most of the rest. It's worth working pretty well. We like NVIDIA and Broadcom. I think you should still own both. The GPU versus ASIC narrative tends to go back and forth. You never know which one it's going to be. I think they're both winners. We like them both.
Matt, let's return to the AI optimism trade. Following a report that US officials are holding early talks to let NVIDIA sell its H200 chips to China, I spoke with Commerce Secretary Howard Lutnik earlier this morning. Do you want to sell China some chips and keep them using our tech
and our tech stack? Or do you say to them, look, we're not going to sell you our best chips.
和我們的技術棧？或者你對他們說，看，我們不會向你出售我們最好的芯片。
12:01
We're just going to hold off on that and we're going to compete in the AI race ourselves.
我們只是暫時不出售，我們將自己參與AI競賽。
12:05
So that is the question. It's in front of the president. He's going to the side—it's a really, really interesting question. He's got all the information. He's got lots and lots of experts talking to him. And he's going to decide which way he wants to go forward.
For more, we're joined by Bloomberg tech co-host Caroline Hyde. Caroline, a lot of moving pieces here. And we've heard from the Trump administration before that they would allow more chips being sold into China. There was that report earlier in the year that they'd take some revenue in trade for it. Howard Lutnik himself has had to kind of not issue a Miyacopa, but has had to
soften his tone when he said before, we want China addicted to our best technology. So are they going to allow these sales to go through? Well, we're wondering whether it ends up being a moot point because there's also Congress trying to work out whether they just stop any sort of Nvidia chips eventually going into China. But thus far, it seems to be this hawks versus non hawks
that have got the president's ear. Now, remember Howard Lutnik, but a month or so ago, we're saying, we don't want them having our best, our second best or our third best. Remember, even Trump himself was saying, we're only going to ship basically obsolete chips to China. And of course, China then decided, no, thank you. We don't want your age 20s. We don't want you getting 15% of
that cut. I might remind you that that hasn't even gone through and been passed into law yet. So age 20s can't be shipped without a license thus far from Nvidia AMD's equivalent. But it does feel as though there's this tension going on. Will we allow Grace Blackwell, Grace Hopper, in fact,
and this is what something we'd heard Trump talk about before. Would we even let a more dumb down version of the Blackwell architecture get into China's hands? At the moment, it's still just the previous iteration. And this is something that's just going to go on as this narrative continues to buck heads over in the administration. Of course, Nvidia is desperate to get back into
China. At the moment, it's able to win over and beat analysts' expectations on its earnings with 0% of sales going to China. But they are really keen on getting back into that particular market. All right. I hope you're all doing great today. Today, the Commerce Secretary confirmed on Bloomberg that the administration is considering allowing sales of Nvidia's age 200 in China.
As I've been saying, it is to the United States strategic benefit that countries around the world build on the American tech stack 50% of the world's AI developers live in China and where the developers go. Those are the platforms that will succeed. Now following the export control on Nvidia's age 20 earlier this year, China is now acutely aware of this fact. And so it will be
interesting to see if China would allow the age 200 in China since they appear to be very determined to expedite the development of their own domestic ecosystem to avoid being reliant on American technology. It's important to note that Chinese tech companies absolutely want Nvidia chips and they would gladly use the age 200 if they were allowed to. Many of them would also use
the age 20 if they were allowed to. At the current moment, the Chinese government is pushing back on the sale of Nvidia chips. But there is definitely demand from Chinese tech companies. Any opening for Nvidia to sell in China would be a positive surprise for the stock as both Nvidia and analysts have removed all China revenue from their forecasts and models. In other news, Michael
Burry has launched his paid sub stack. I have no intention of signing up for the sub stack. And so I'm not going to say much about it. But I've seen some people online claiming that I'm Burry sub stack. He's taking shots at Nvidia and equating it to the Cisco of the AI bubble. I have no idea if that's what he's actually saying or not, because I'm not signed up for his sub stack. But that's what people online are saying. At the height of the dot com bubble Cisco was trading at a triple
digit forward PE. Today, Nvidia is trading at a forward PE in the 20s. That is not egregious at all. That's basically a market multiple. And if Nvidia's earnings turn out to be even better than what analysts are currently estimating, which usually is the case, then Nvidia is likely even cheaper than that. Additionally, we need to acknowledge that while there are similarities
between this AI revolution and the dot com bubble, there are also key fundamental differences. I'll briefly mention one major difference between the two eras. Gavin Baker made this point recently at an event hosted by Andreessen Horowitz at the height of the dot com bubble 97% of the fiber that had been laid was dark. Today, there are no dark GPUs. Every GPU is spoken for as the world is
severely compute constrained in the face of very strong AI demand. This is not a bubble. It's insatiable demand. Think about what was fundamentally happening during the dot com bubble. There was a massive overbuild of supply and a shortage of demand. Now compare that to what we're seeing today.
We have a severe shortage of supply in the face of extremely strong AI demand. It's the complete opposite. Now it will certain parts of the AI ecosystem eventually enter into bubble territory.
Sure, some probably will, but that does not mean that Nvidia is a bubble and it does not mean that AI as a whole is a bubble. Really quick. I want to remind you that Nvidia will be presenting the UBS global technology and AI conference on December 2nd at 635 AM Pacific 935 AM Eastern.
This event will likely have an impact on how the stock trades in the short term given that the event is geared toward the financial community. There's been a lot of consternation online and some of the financial press about accounts receivable and day sales outstanding on Nvidia's recent earnings report. Nvidia is growing rapidly at a massive scale and these metrics are
in my opinion, nothing to be concerned about. When Nvidia reported earnings, we learned that day sales outstanding or DSO for short was 53 days. This is not unusual and it's nothing to be concerned about. If you look at DSO for some of the networking companies that are central to the data center build out, you'll see that many of them have even longer DSO than Nvidia,
some 60 or 70 plus days. And so Nvidia having DSO of 53 days is nothing unusual and it's nothing to be concerned about. And the same goes for the rise in Nvidia's inventory. We know that Blackwell Ultra production is ramping swiftly and Nvidia is preparing to launch Rubin in 2026 as
well. We know that demand is substantially greater than supply and that will likely continue for at least a few years. And like I said, Nvidia is growing rapidly at massive scale. And so growing inventory levels are not something to be concerned about at the current moment. Now looking at Nvidia stock, I don't know what the stock will do in a shorter, but from a long-term perspective,
I think Nvidia still has plenty of runway ahead of it. I don't think Nvidia is overvalued here and I continue to think that most analysts are underestimating Nvidia's long-term growth.
可能會有回調嗎？當然。我們應該為此做好準備，以防萬一。
17:43
Could there be a pullback at some point? Sure. And we should be prepared for those just in case.
But regardless of what the stock does in the short term, the long-term fundamental growth of the company remains intact. And that is what we should be focused on from the perspective of long-term investors. Of course, I expect Nvidia's data center business to continue growing strong as the world is compute constrained in the face of very strong AI demand. I also think that most
analysts are not adequately factoring in growth from some of Nvidia's other business segments into their long-term estimates, especially physical AI. There's a reason why I keep talking about Nvidia's physical AI business and Nvidia CFO nailed it on the earnings call when she said that physical AI addresses quote, a multi-trillion dollar opportunity and the next leg of growth
for Nvidia. I am completely serious when I say that I think Nvidia's physical AI business may eventually be larger than their current data center business. This is not getting enough attention, but I think it will get a lot of attention in 2026 and 2027. Nvidia sells the hardware for the data centers where the models are trained. They offer omniverse where the models
are taught and tested and Nvidia also sells the hardware that allows on-device real-time inference through Nvidia AGX, allowing robots to have intelligent interactions with the real world even when they are not connected to a data center. Notice that Nvidia is taking a holistic platform approach to physical AI and they're embedding themselves as the underlying foundation supporting
all of it. Over two million developers are already building on the Nvidia robotic stack and this is not getting enough attention. I continue to think that most analysts long-term Nvidia estimates are too low given what's ahead as for production ramps. Blackwell Ultra is ramping very quickly right now leading to revenue growth acceleration. Ruben is on track to launch in 2026. Then we're
expecting Ruben CPX at the end of 2026. Later on we're expecting the launch of Ruben Ultra in 2027 and Feynman after that in 2028. We have a clear data center product roadmap stretching into 2028 and Jensen believes there will be three to four trillion dollars of global AI factory build out
between now and 2030. That means Jensen is expecting growing AI demand and an expanding total addressable market underpinning all of this. I don't think we are anywhere near any type of bubble bursting type of event. With all of this in mind I seriously think that Nvidia still has plenty of runway ahead of it and I think this company will be worth substantially more in future
請記住在這個市場中保持冷靜。請記住保持長期的視角，不要做出任何倉促或非理性的決定。
19:45
years than it is today. At least that's my view of the situation. Remember to stay calm in this market. Remember to maintain a long-term perspective and do not make any hasty or irrational decisions.
With all of that being said I hope you all have a great rest of the day and I'm curious to hear your thoughts about Nvidia in the comments below. Please leave a like on this video so more people will see it and while you're down there please consider subscribing. It's free and you can always change your mind. Thanks for watching and hopefully I'll see you in the next video.
