Nvidia and the H200 landscape; Broadcom's strategic positioning

0:10Sarah Awad from Tech Contrarians, always great to have you on talking tech and updating us on what you think of that part of the market which is so much a part of the market. these days. Welcome back. Thanks for having me again. I'm really glad to be back. There's lots to talk about. It's definitely an exciting time with the news on the H200s coming out. Yes, it's a journey and we're not sure of the destination. So talk us through what you're thinking about these days. What are a few things that are top of mind for you and what are you most focused on? So there's a couple of different moving factors that we're looking at the moment.

0:44I'll get started with the H200 news because that's really the elephant in the room. And I think it's exactly what you're saying. No one really knows how to think about this because we've had so many on and off. Will NVIDIA ship to China? Will they not ship to China throughout this year? That it's kind of like the market doesn't know whether to trust what we're getting in terms of news from the U.S. because China also has a say in this now. Right. They're not just going to accept any chip, any AI chip that the U.S. is going to approve. And we saw that with Beijing warning domestic Chinese players against buying the H20 when we saw the reverse ban in July.

1:19So I think the market's really kind of confused about how to understand this news. Is it going to move towards NVIDIA having China sales or are we still far away from that because we still have a murky image from Beijing? Now, what we did get from Trump is that he talked to Xi about this and the reaction was positive. At the same time, we have a report from the Financial Times that's noting that Beijing regulators are looking into ways to permit limited access of NVIDIA chips. And so there's a lot of moving parts. So I think let's talk about what we know for sure. And then let's talk about what we're speculating about, just to differ between the two, because too often kind of intertwined in the markets today.

1:58To begin with, so we have the H200s are back on the table. This time around, different from the H20s, the U.S. is going to take a bigger cut, right? with the H20s, it was 15%. And this time around with the H200s, the US is looking to take a 25 % cut. To an extent, it's almost as if we're saying like we're balancing national security threats that could be, you know, posed by H200 sales to China for profit. You know, some have critiqued this as, you know, trading national security interests for an advantage in trade. But I think first and foremost, what this points to is what, you know, we've said time and again, which is that Trump is a businessman before anything else, right?

2:34And this is an attractive opportunity to re-enter China. And it also shows that, you know, Jensen's lobbying at Washington has really paid off because he has mentioned, even up until last week, he was mentioning that this is really a race against China. And earlier this year, he had commentary about how, you know, China is nanoseconds away from the U.S., if not already where the U.S. is. So it's really a murky picture that Jensen has kind of frame China in, in the sense that if the U.S. doesn't move now to guarantee NVIDIA's presence in China, there's a chance that we could be losing this market altogether with Huawei and other players kind of penetrating the space.

3:10And so we've seen that really pay off. At the same time, there's still a lot of, so those are the facts, right? But at the same time, there's a lot of uncertainty and there's unknown. So the two big questions I would say on everyone's mind now is how much supply is there already for the H200s that can exist in 2026? How much supply can NVIDIA guarantee for China for that market? I think it's less a question. There's a lot of people have been talking about a question of, is the demand even there in China? And I'm going to get to that in a minute, but I don't think there's even a question about whether the demand is there in China.

3:43You know, especially for the H200 product, that's going to be the best performing product within the entire Chinese market. So there is demand for that product, especially from the likes of Alibaba, Tencent and other domestic Chinese players. But the question is how much supply can NVIDIA really guarantee, especially with the shortage that we're looking at now when it comes to memory, when it comes to a lot of different components of AI servers and AI GPUs. And the second question is if China is going to say yes or no, right? Has the train already left the station? And so I want to address both of these.

4:16I think that overall, I'll tell you as a bit of a teaser from now, we're still very bullish on NVIDIA. we actually think that this news on NVIDIA potentially re-entering China is going to be the next big boost for the stock once the market has solid ground to stand on in terms of seeing this actually turn into action. So the ball is really now in China's court. So let's look at the metrics in terms of what would really allow China to accept H200 used by domestic players. If you look at performance, the H200 performs much better than any domestic AI chip in the market today. So the H200 wins big time against Huawei's SN910C product, which is the best performing product in China.

4:55In terms of total processing performance, the H200 outperforms substantially. In terms of terabyte per second, the H200 outperforms substantially, even in terms of the node that is built on, 4 nanometer for the H200 and 7 nanometer for the SN910C. So in terms of all the benchmarks, performance H200 versus domestic chips, H200 is the better option. And then even if you look at H200 versus the H20, you know, the estimates are for H200 having six times the power of H20. So all around, it looks like an attractive option for China to accept this product into the domestic market. Now, the problem for China is that at the end of the day, if they accept this product, they're facing this risk of, you know, relieving the pressure or the heat that they placed on their domestic chip market to really fill the gap that NVIDIA has left.

5:46So that's why we're seeing this, you know, one foot in, one foot out of the door kind of situation where Beijing may say, hey, we're going to have limited access where we can, for example, allow Chinese domestic players to submit a request for approval, explain why the domestic solutions were able to satisfy their needs. And then we would accept or reject their request for H200s. So that's kind of what we're seeing play out. But I think it's just a matter of time before we see China accept and welcome H200s or other NVIDIA products, depending on how things play out with the US. Although we don't think we're going to see approval for a more advanced product than the H200s, because then we're at Blackwell and that's where everyone is today.

6:27And, you know, the US can't afford to have China be on the same pace as the rest of the world when it comes to AI and technology and investment, because they want to keep that kind of lead secured. So our expectation is that Beijing is going to soften because at the end of the day, whether Beijing likes it or not, the performance gap between China and the rest of the world will continue to widen. And what do I mean by that? So there's been a lot of talk. We've gotten a lot of headlines. I'm sure you've seen these two. Since China got out of, sorry, since Nvidia got out of China, we've seen a lot of headlines about how well Beijing can do alone, right?

7:01It's almost as if we're talking about China trying to show off muscle as a negotiation tactic, right? So they held off on purchasing the H20s when the ban was reversed. And now we're seeing them even kind of hold off on giving clear commentary about this. So and a clear, you know, excited approval about it. And so I think this is a negotiation tactic. And it's worked well for Beijing, they held off on the H20s, and they indeed did get superior products. But I think that is, at best what it is, I don't think this is going to play out as longer than a negotiation tactic. I don't think we're going to see Beijing really close the door on NVIDIA re-entering the market because U.S.

7:39still has a chokehold on China's access to tech, right? Whether through, you know, expert restrictions that have to do with ASML when it comes to lithography and EUV machines or, you know, what advanced nick wafers from TSMC. So there's a lot of restrictions and China's really operating from the place of a chokehold, from a place of, you know, tide webs. So it's a matter of time before we have, you know, next year, Vera Rubin and MI450 from AMD move to smaller nodes, three nanometer nodes. And, you know, as time progresses, smaller and smaller nodes. And China is going to be left behind because they're capped at seven nanometers and below, they can't really access that.

8:22They're capped right there. So we're going to see China kind of cave because at the end of the day, the priority is catching up in this AI race, as well as creating that long-term self-sufficiency. But long-term self-sufficiency is going to happen anyway. So I don't think it's a huge issue to discuss today. I think today the problem is that Beijing used a strategy, they negotiated, they got what they wanted. And I think now we're likely to see them basically accept NVIDIA back into the market. And I think that's going to be great news for NVIDIA. The numbers that NVIDIA is printing now, while already beating numbers, don't account for those Chinese sales.

8:55So once we have that market to tap into. Again, I think NVIDIA is going to be a different ballpark. And I think a bit of the kind of cautiousness that we've seen the market treat NVIDIA stock with is going to ease. And what else would you say about, we have a few names reporting this week. We have Adobe, we have Oracle, we have Broadcom. Anything you would say as part of what you're describing or talking about or separate themes that you're looking at this week specifically? So there's a couple. So Broadcom is a big point of focus for us. And we've been bullish on Broadcom for a while. In fact, at one point earlier this year, and we still hold this notion, we believe that Broadcom will be a more rewarding investment in the longer term than NVIDIA could be today.

9:44We still maintain our bullish stance on NVIDIA, but the reason I say this about Broadcom is because to an extent, Broadcom is the NVIDIA of ASIC, right? NVIDIA is the NVIDIA of AI GPUs and it really spearheaded all of that. But Broadcom is that alternative for ASIC. And the way we look at the AI market as it matures and it's segmentized is that you're going to have ASIC for low to mid level inferencing. And then you're still going to need AI GPUs for high level inferencing as well as training. And so Broadcom, at the end of the day, is well positioned because it has massive scale. It has more scale than its direct rival within ASIC, which is Marvell, who reported last week, actually, and had interesting things to share on that earnings call, which I'll get to in a minute.

10:32And for Broadcom, essentially, because they have this scale and because they have exposure not only to the ASIC side of the business, where they can benefit from AI, but they also have exposure to the networking side of the business, where they can benefit from a growing AI GPU market, they kind of have the best of both worlds. And so that's why Broadcom is extremely well positioned. And then aside from that, on the ASIC front, Broadcom is the industry go-to. So they've actually enjoyed most of the tier one hyperscalers choosing them. Google's recent release of Gemini 3, the fact that it's been trained on TPUs and it's outperforming on multimodal understanding, speed, and long context reasoning against ChadGBT, is showing that ASIC is picking up in terms of being a point of attention for investors more and more.

11:16So I think heading into the report on Thursday after the bell, everyone's going to be watching to see if Broadcom will give, you know, attractive outlook for what's to come, similar to what we saw Marvell do last week. So last week, Marvell basically reported earnings. Their results weren't extremely strong. It was pretty much inline performance. Their data center segment, which is what includes their optics and their ASIC business, was only up 2 % sequentially, which you know how AI growth looks like nowadays. And you can't call the AI growth. 2 % sequentially is not the kind of percentage that you would attribute to a company that's considering itself an AI name, right?

11:53So if NVIDIA reported that kind of number, it would be horrendous. It would be disastrous. So for Morvel, they don't have that positioning in terms of fundamentals. And they reported earnings. The stock dropped 7%. And then they came on the earnings call and they gave an outlook for fiscal year 2027 and for fiscal year 2028. And then the stock shot up 13%. And so if we see any kind of outlook from Broadcom on Thursday, I think the stock is going to react very positively to that. At the same time, there's talks about Microsoft potentially getting into bed with Broadcom as another hyperscaler customer.

12:31If we see any progress on that front, the stock, I think, will also react positively. And then even if we see neither in terms of fundamentals, Broadcom has been performing extremely well. Do they guided last quarter for this October quarter above of consensus? they're going for 17.4 and consensus then was 17 billion. Since then consensus has come up and we're expecting on Thursday we're going to see Broadcom beat consensus and guide higher. So that's kind of our trajectory for Broadcom. We think that the tailwinds are going to increase especially towards the second half of 2026 because then you know the memory shortage is likely to ease a bit and then you can kind of have higher ramp volumes when it comes to their their growing hyperscaler customer base.

13:12So that's our logic on Broadcom. And then when we take that information and circulate it into what you and I were discussing in terms of, you know, ASIC versus AI GPUs and the, you know, the question on everyone's mind after the release of Gemini 3 and Meta approaching Google for talks to use its AI chips is, does this mean that NVIDIA could lose market share, right? From its current market share estimated to be, you know, somewhere between 70 to 90%. And the reality is that, you know, sooner or later, NVIDIA will lose market share. You know, anyone watching tech long enough, you know that no one maintains a stellar monopoly forever.

13:50And so it's less so a question of if and more a question of when and how will this play out for NVIDIA. In our opinion, it's too early to be having this conversation of NVIDIA losing share because ASIC is just now getting started, right? There's still so much of the, you know, the granular detail to ASIC that the industry is still figuring out. So I think this is more of a second half of 2026 discussion, which isn't too far out, but it's still not today. And between now and then, we see tailwinds for NVIDIA that the conversation of NVIDIA losing share could overshadow. And those tailwinds, specifically speaking, are its reentry into the Chinese market.

14:27So I think there's enough room basically for AI GPUs and ASIC to both exist as AI matures. And as of today, don't see a reason to freak out in terms of having a stake in NVIDIA and worry about its share getting shrinked in the first half of next year. Anything else to say about names reporting or themes that you're going to be looking at or also into 2026, what you're thinking about for these names and or others? Yeah, actually, looping in the part about the current shortage in the industry, it's no secret we're in a memory shortage. What some are calling the hyper cycle, the super cycle, excuse me, of memory, the memory super cycle.

15:11And it's an AI led shortage, right? Because AI is memory bound. AI needs as much memory and as much storage as you can get it. And we're seeing that reflected in Micron and Seagate and Western Digital. and a lot of these memory and storage peer groups. But things have escalated so much to the point, specifically for DRAM and HBM in specific, the shredders has escalated so much that it's putting a huge amount of pressure on price. So we're seeing prices go up. Actually, we're seeing insane surge in DRAM spot prices. So the DDR4 was up 158 % since September. The DDR5 was up 307%. So the shortage is, you know, it's pushing prices up and it's also causing memory suppliers to prioritize their capacity.

15:58And that's why we saw Micron earlier this month announced its intention to exit its consumer business in February 2026, if I'm not mistaken. So we're seeing this memory surge, this memory shortage really impact what we usually know about memory cyclicality. and it's special to point out not just for the sake of players like micron who we are bullish on and whose exit from the consumer business we don't think is a huge red flag because consumer business is not what's driving micron's outperformance you know at the end of the day what's driving micron's outperformance in 2025 and expects it to be in 2026 is hbm and so they're just allocating more capacity to this higher asp product that that they have hyper demand for but the benefit of looking at the shortage is thinking about who can who has the advantage when such a dynamic plays out and that's when we turn and look at Apple right so Supermicro and Dell have warned about this memory shortage Chinese smartphone manufacturers have warned about the summary shortage and I specify about Chinese smartphone manufacturers because those are the direct you know competitors have really been squeezing Apple over the past couple of years And so we're seeing Apple actually have an advantage.

17:13And that's something that we haven't been able to say about Apple for a while now. So we're seeing Apple have this advantage in the sense that it's not as pressured by shortage issues as its Chinese competitors. And the reason behind that is pretty simple. It's because at the end of the day, Apple is one of the OG players in the field. And it basically makes sure to have medium-term contracts to lock in price. and that's the first side of the coin and the second side of the coin is it makes sure to over forecast so it can gain a cost position so these two things protect it from basically being you know you know dumbfounded by a surge in memory prices and that basically entails that apple has product positioning which means that a lot of the reports coming out because of the shortage is that different chinese smartphone manufacturers which have been stealing share from Apple are actually looking at either having to see margin pressure or having to actually increase prices on their smartphones.

18:14We're seeing that from Samsung. We're seeing that from Xiaomi. So if we see all of these guys be forced to cave and increase pricing on their smartphones, then Apple has the advantage of not needing to increase prices and simultaneously only needing to slightly decrease prices to see its products fly off the shelves. Because at the end of the day, if you have an iPhone that's trading, sorry, that's selling at a discount versus any other phone from the Android ecosystem, you know, the historical trends have showed us that the iPhone wins favor. And so that's why we're more bullish on Apple, because we think that it kind of is shielded from this issue that the rest of the peer group has to face.

18:55What else are you thinking about tech-wise? What else do you feel like is valuable for investors to have in mind? I think in terms of heading into 2026, especially given what was a tough November for markets, I think in retrospect, our view was that what happened in November was a correction rather than a bubble pop. And I think as we saw recovery towards the end of the month and how we're seeing things play out so far in December really points to that really being the case at the end of the day. We saw the AI bubble deflate slightly and now we're seeing things pick back up. And we're seeing momentum around AI names, which is something that got scary for a minute there that the market was missing.

19:35And so I think that right now it's again time to stick to the fundamentals. So stick to the guys that can basically have the scale and the positioning to be resilient to the bigger conversation playing out, whether about the AI bubble or circular financing or even the offshoots of this entire AI boom, such as the memory shortage. So those are the guys that I think that fly under the radar. And I think the names that are been hated the most, like we saw similar to with Google earlier this year, the names hated the most are the ones that are likely to come back the strongest. And now on our radar is really, you know, Amazon with a Trinium 3 being updated for commercial use.

20:14That looks like an attractive area for potential growth in our opinion. So we're watching Amazon very closely. Anything that you see like market headlines, tech headlines that you would want to dispel for investors or for those listening? So I'm going to open up right here something that we have in our investing group, Tech Contrarians, which is our track everything tech worksheet. And we update this every Friday and during the week, and we kind of keep track of all the headlines and have two columns with a fact and one with debunking the context of it. And opening this up right now, the first thing that shines out is really the fact that Google is basically considering plants to monetize Gemini via introducing ads next year.

20:57And that's something that we've seen OpenAI kind of talk about a lot, but not really execute on. And I think that this just shows how much this, you know, Gemini versus ChatGPT race is heating up in terms of monetization and in terms of performance. We're seeing Google really gain the lead on both. So I think this is going to put OpenAI under more pressure, comparable to what we've already seen since Gemini 3 is released. But I think this is a bigger conversation that speaks to the fact that the AI dynamics that we have become familiar with can very quickly change. Earlier this year, Google was a loser of AI and OpenAI was, you know, anyone who made a deal with OpenAI stock would soar.

21:37Now people are much more cautious about open AI and you're not as safe being in bed with open AI today than you were earlier this year. So I think that there always needs to be room for these dynamics to change and for investors to catch on to that before it's already become the reality. Appreciate this conversation. As always, once again, you write under Tech Stock Pros on Seeking Alpha and run the investing group Tech Contrarians. Any final words? I hope you have a great new year and to everyone listening as well. And of course, we're always interested in having anyone in our investing group who's interested in kickstarting their investing journey, or if they've been doing it for a while and interested in deepening their tech expertise.

22:19Just a reminder, anything you hear on this podcast should not be considered investment advice. This is for entertainment purposes only, and you should seek advice from a licensed professional before investing. If you enjoyed the episode, leave a rating or review on your favorite podcasting app, and we'll see you soon with a new episode.

22:38We'll be right back.