J Studios
Listen here or on the go via Apple Podcasts and Spotify
Sara Awad from Tech Stock Pros and the investing group, Tech Contrarians, explains the DeepSeek story and contextualizes Nvidia concerns (1:15). Downgrading AMD (8:00). Palantir’s run (10:00). Recent earnings from semi caps – 3 stocks to like (18:30). Fundamentals vs. hype, supply and demand (22:30).
Transcript
Rena Sherbill: Today we have Sara Awad from Tech Contrarians, which runs the investing group, Tech Stock Pros, or perhaps it’s the other way around.
It’s Tech Stock Pros and you run the investing group, Tech Contrarians. Clear that up for us, Sara. First of all, welcome to Seeking Alpha. Welcome to the show. It’s great to have you on Investing Experts.
Sara Awad: Thanks so much. It’s great to be here. I can say for myself and the whole team, we’ve been fans of the podcast for a while, so it’s a pleasure to be on it this time around. And yeah, it was actually the other way around.
So we have Tech Stock Pros, which is our public site on Seeking Alpha. And then behind the paywall, we run our investing group, which is Tech Contrarians, where we give our contrarian take on tech.
Basically we focus on semiconductors and software hardware. We kind of do it all in terms of tech. So we have a coverage of over 50 companies across semis and software hardware.
RS: Tech is very much in the conversation when it comes to what’s going on in the markets, what people are paying attention to.
You talked about being a contrarian in the tech space. So talk to us how you’re approaching the tech space right now, smack dab in the middle of February 2025?
SA: So basically there’s a lot happening. There’s a lot of chatter, there’s a lot of rumors.
And I think for the past at least 3 years, there’s been a lot of hype around AI. And so January, everything that happened with DeepSeek and how things are playing out now even with the Trump administration and more information coming out on DeepSeek day after day, there’s a lot happening. There’s a lot to unpack.
I would say the contrarian side that we take is, what we try to do is we try to filter through a lot of the market noise coming out because there’s so many headlines that kind of confuse the retail investor. So we’re trying to filter through the hype and kind of get down to the fundamentals to really give investors, firsthand, an education on what they’re investing in.
To armor them with that, going through and navigating the tech world as so much unfolds. And at the same time to let them be ahead of the curve in terms of when that hype is going to die out and when those fundamentals are going to show up.
RS: So how are you separating hype from fact? How would you articulate, how would you contextualize what’s happening with DeepSeek and the AI space?
SA: For DeepSeek, I think part of the hype that came out around it is that, DeepSeek isn’t new. And that’s something that we try to emphasize whenever we get the chance to.
It’s been around since let’s say like late 2023. But that doesn’t mean that it’s been here for a while. We’re talking about something that’s around 20 months old. And there’s a lot of misinformation going around it and a lot of misinformation that cause that massive sell-off, I think now two Mondays ago.
And a big part of that misinformation is that everyone woke up on Monday morning, and they were like, wow, there’s this new AI model that’s somehow doing the work that competes really closely and even beats OpenAI on specific benchmarks at such a lower cost. The cost in the report was something like under $6 million.
Well, you know, we have all these hyperscalers, Meta (META), Microsoft (MSFT), Amazon (AMZN), and Google (GOOG) (GOOGL), spending hundreds of billions. Like last year alone, 2024, their CapEx spend was 61% higher year-over-year. So they were looking at a total spend for the year of something like 230 billion.
So it’s crazy that this new AI Chinese startup came out and they were like, we did this with under $6 million and it didn’t compromise on the performance.
To begin with DeepSeek, the way it operates and it’s important to put some context here because it helps us understand how they came up with this R1 model.
We saw this reaction from the market in late January, but DeepSeek actually dropped their V3 a day after Christmas, so on the 26th of December, but wasn’t getting much attention from the market there. And then around a week before that red Monday, they started getting some attention from US media outlets. And then over the weekend, everyone downloaded the app, it dethroned ChatGPT on the Apple store as the number one downloaded app.
And Monday pre-market everything was red. And then we saw NVIDIA (NASDAQ:NVDA) close 17% lower and all these AI names sell off throughout the day because there’s all this panic about maybe we don’t need as much computing power as we thought we needed to make AI happen.
So there’s positives and negatives for that. And that means different things for different parts of tech. And at the same time, it means different things for different timeframes.
I think the first thing that we break down from our contrarian take is that for the near term, for the very near term, I must emphasize, it doesn’t really have massive implications.
So we think NVIDIA is safe for the current quarter. And the reason is because NVIDIA is ramping their Blackwell series, and everyone is going to be trying to get part of those. So everyone’s going to be putting in orders because there’s a shortage.
And if there’s one thing we know about the market and supply demand chain, whenever there’s a shortage, especially on something as competitive as NVIDIA’s AI GPUs, everyone wants a piece of the cake because you can’t win a lottery if you don’t buy the lottery ticket to begin with.
So you have to put down that investment, even if we have this DeepSeek that came out. And that’s potentially showing us that we can do the same high performance even at a lower cost.
I’d say that’s the first thing that we would disclaim. That doesn’t mean that NVIDIA is safe and going to be on this upward trend of hypergrowth that we’ve seen over the past two years for the rest of the year, but we think for the near term, this has less implications than maybe the market realizes, at least for now.
Now, down the line, we think that this means something really big. DeepSeek is really big for AI infrastructure. And I want to take this moment to differentiate between AI infrastructure and AI endpoint.
AI infrastructure is all the spend that’s going into buying these GPUs from NVIDIA, everything that NVIDIA is really pocketing in terms of that. So we think that computing power base needs to be lowered.
These numbers that we’re looking at in terms of like the CapEx spend, what Meta is expected to be, you know, 53% higher for 2025, Microsoft’s expecting 45% higher for 2025, Amazon’s 20% higher for 2025, Google’s 43% higher for 2025. We think that we’re going to see potentially these numbers stretched out.
Not only for 2025 and looking at a different CapEx spend for next year, we potentially could see these numbers stretched out because that AI infrastructure spend could be lowered. And so that could cause a lot of hiccups, or we really do believe will cause hiccups for NVIDIA come around July quarter when that double ordering shows up.
And that double ordering is all the ordering that people are doing now with the Blackwell ramp and the shortage. Everyone will put in double orders because they want to guarantee that they have a slice of the pie.
RS: Any other names that you’d care to mention when you’re talking about who it’s going to be, who it’s going to benefit, who it’s not going to benefit?
I’m curious also, anything to comment on Palantir’s (PLTR) surge in the past week or so, in terms of that name in the AI space?
SA: I’ll get to that first part of the question. In terms of who it’s going to benefit. So for near term, we’re not too worried about NVIDIA, but of course for the midterm and the long term, we think that’s a different conversation to be had.
We do think that NVIDIA is a name to revisit once that AI infrastructure base is lowered. And I’ll get into the nitty gritty of how this works in terms of training and inferencing and kind of like the fundamental aspects of that.
But another name worth mentioning that we’re pretty negative on and have been negative on for a while, and being contrarian you face a lot of backlash, so we face some backlash on it, is (AMD).
So we actually downgraded AMD earlier this year, around January 8th, because for a while we felt that AMD really didn’t have that fundamental edge that NVIDIA does have. They lagged behind in terms of the design win cycle.
And even with management putting up all these different AI GPU expectations and outlooks, they were really going to just fall short. And the market was hyping up AMD as the next NVIDIA for a while now.
This continued all throughout last year, even when AMD missed outlook on consensus for like 5 consecutive quarters, the stock price kept moving against those fundamentals. So AMD is a name that we think is not going to be a winner for this year.
In fact, they recently reported earnings. And what they said is that basically their outlook for the first quarter of 2025 is that they’re expecting sales to drop 7% sequentially even though they’re expecting AI growth in terms of their AI GPU sales, they’re not expecting their AI GPU business to do well in the first half of the year compared to the second half of last year, they’re actually expecting it to be flat.
So we stay negative on AMD, although a lot of backlash coming forth is that AMD is so cheap, they’ve priced in so much of the downside already. We don’t think that they’ve priced it enough yet.
We think that the market has still hyped up AMD’s position inside the AI accelerator market beyond what is actually going to be the reality. And we think once their ZT acquisition closes this summer, if I’m not mistaken, things will just get worse for AMD from that front.
So that’s another name that we think won’t be doing too well, but we’ve been bearish on for a while. And we think that a DeepSeek news just makes it that much worse for the company and for the stock.
But we’re not saying AMD is dead, we think AMD will come back. But we just think that the stock price now is not at all at attractive levels and expectations in terms of their position in AI are still way too high.
Palantir is another name that’s interesting. You know, Palantir’s been really on a run. And a big part of it is, thanks to their data analytics and the kind of the foundation they had when this AI boom happened a couple years ago.
So if I look at Palantir, yeah, they’re up like 55% year-to-date, which is impressive because the valuation is saying something that we should see a bubble pop soon. So we actually were Hold rated on Palantir for a while because we felt that some things have to give, this stock is too overvalued, it’s priced for perfection, and we could see that bubble pop soon.
And then on our public site, where we mostly cover quicker trades, we upgraded it back because we think just investor confidence in itself will push the stock price higher from the chatter and from the sentiment post earnings and from management’s commentary there.
Something worth noting is that we think that their commercial revenues could see, and this is the thesis that we’re still building our research on, but we think their commercial revenues could see some pressure from DeepSeek considering that DeepSeek has such a lower barrier to entry and that you don’t need to spend now even the 6 million that DeepSeek spent to create your own LLM, or your own AI model. You can just take it because it’s open source and then build on that.
So we think that they could see some pressure on their commercial revenues which they’ve been working very hard to kind of grow against their government revenues.
We do think that the government revenues are looking positive. The government revenue, I think, came in at around 40% of their total sales for this quarter. So with the pro-AI, anti-China kind of sentiment, we might see more government spends flowing into Palantir in the near term.
But we do see this higher risk profile because of that threat to the commercial side of the business and because of, you know, the stock is priced for perfection. So any slip ups here are not affordable.
And if I’m not mistaken, you guys had a podcast about shorting Palantir and opportunity there recently, right?
RS: Yeah, we sure did. Julian Lin was on talking about that. I was going to ask you in terms of like valuation, because that was one of his main contentions with Palantir, that there was a lot to like fundamentally, but the valuation wasn’t there.
Is that a big metric that you are looking at Tech Contrarians?
SA: So yeah, we look at valuation for sure. I would say that we focus a lot more on the fundamentals because we think that it’s all a balance between balancing the valuation with the fundamentals at play because we’ve seen NVIDIA continue to outperform although it was trading at premium multiples.
We saw the same with Reddit (RDDT) last quarter and we’re going to have Reddit’s report today after the bell. So I think that there’s enough investor confidence, you could see valuation be less of a factor. But I do think that this bubble of Palantir will need to reset and need to kind of deflate a bit soon. Then there’s a question of when that will happen.
Now with everything happening in the tariffs, something that management mentioned on the call is that they actually saw some more traction from different data analytics being run around the tariffs and the potential impact that it could have.
And we had gotten new tariffs on steel and aluminum just this week. There’s more being foreshadowed. So we could see more momentum there kind of offsetting whatever downside there might be to the commercial side of the business.
So I think this is, it’s a waiting game. We’re definitely going to revise our rating ahead of next quarter, but we’re still looking into the valuation there. But I think it’s something that’s definitely bound to reset soon.
RS: Just because you brought it up, anything to say ahead of Reddit’s earnings? I know this will come out after the earnings, but interested to hear if you have any takes?
SA: Yeah, I don’t want to give too many spoilers away. But yeah, we’re bullish on Reddit. It’s done really well for us. It’s up around 170% since we said Buy.
I think something that could play into Reddit is the political landscape at the moment, which could bring a lot more traction to Reddit itself. I would say something that we were hoping for or eyeing that is less likely to happen now is their AI data licensing deals with Google.
We were eyeing to see if that would expand to other hyperscaler names as well. But now with DeepSeek, that’s less likely to be the case.
And bringing the conversation back to DeepSeek, something interesting about DeepSeek is that, aside from that lower price tag that everyone focused on, there’s a couple of different things that DeepSeek’s model does differently that actually enabled that lower price tag that I’d love to get into, because I think it really sets the outlook for how this is going to impact both the semis and the software hardware space.
Something that DeepSeek did differently from, let’s say, the American counterparts working on AI is that they didn’t do the supervised tuning. So what that means is that basically you manually label data and then you do reinforced learning.
So for example, I show AI a picture of a duck, tell it it’s a picture of a duck enough times and then do that reinforced learning. What they didn’t, they were like, screw that. What we’re going to do is we’re not going to label anything, we’re going to skip right into the reinforced learning. And so that lowered the compute power necessary and that enabled that lower price tag as well.
But also looking at that price tag, it’s not you, we can’t take it at face value. So within any AI model, you have inferencing and then you have training. So if we look at NVIDIA, we could say it’s 50% training, it’s 50% inferencing.
And so what AI training is, it’s basically AI training data is a set of information or inputs that are used to teach AI make these accurate predictions or give us these accurate models and decisions. While everything on the other end, which is inferencing, is everything that we do as end user.
So you log into OpenAI, you ask it a prompt, and then everything the algorithm does from there is inferencing. So when you’re working and you’re using it, that’s inferencing.
On the inferencing side, DeepSeek was able to lower the compute power necessary for inferencing. So that’s important because it kind of helps us target where our concerns should be in terms of looking at NVIDIA price targets or looking at where that lower compute power is going to be necessary.
When it comes to training, we know less about exactly how DeepSeek went about the training. And that’s why the context of DeepSeek’s foundation is important.
DeepSeek actually, there’s a lot of rumors that flew around to begin with like, oh, DeepSeek did this without NVIDIA with navigating a US export ban in terms of advanced semis and all of that is true. But you also got to take the context into account.
The first piece of context there is that they used an old open source model from OpenAI and that open source model had a lot of money funded into it. So they didn’t necessarily recognize with that price tag. And they didn’t necessarily have to, right, have to because at the end of the day, that was open source, so they got it for free. So they didn’t really pay money on it.
The second thing to disclaim there is that aside from that open source model that they got DeepSeek is basically operated as an independent AI research lab under the umbrella of the hedge fund that the CEO of DeepSeek founded and so basically he took the research and development spend from that hedge fund and he used it to build a foundation to fund DeepSeek.
And part of what that came with is NVIDIA GPUs that were used for that fund because it was a quant fund. So all of that really comes together to build a better picture about how much compute power maybe went into this R1 model that isn’t necessarily accounted for in that $6 million price tag.
But that doesn’t necessarily mean what a lot of people jump the gun to say, which is that, okay, now anyone with $6 million can go and create this new open source AI model. It’s more complicated than that because you don’t even need to go ahead and build a new AI model because you have DeepSeek, which is open source, we can just take and build upon and make your own use of.
So that’s really how it changes the game. So you don’t need to reinvent the wheel now that DeepSeek is open source. It actually just puts a lot more pressure on companies like OpenAI and companies like Microsoft, because it places essentially pricing pressure, right? Because now Copilot, which is, let’s say, a standard at $20 per month, can’t keep that price tag on because now they have this competition that’s doing this for free and that’s open source.
And that’s why I mentioned that risk for Palantir’s commercial side of the business, because again, you have an open source model that you can train to kind of fill in the blanks for you.
RS: Anything else that you took away from recent tech earnings?
SA: Within Tech Contrarians, we come out with earning updates to give some flavor there across the board.
Something that we do have on our radar are the semi cap names. We think that those guys don’t get enough attention, especially from the retail investor. So essentially, semi cap, these are the first guys in the supply chain.
There’s three names that we like a lot, (ASML), Applied Materials (AMAT), and Lam Research (LRCX). And ASML stands out in particular.
So they had their earnings really early this quarter. They gave some flavor on the rest of the year. And basically, Wall Street has been pretty negative on the semi cap. And so, ASML reported earnings and they maintained their full year outlook in terms of revenue and that got them to outperform and actually erase their DeepSeek losses just a couple days prior.
And so, we think ASML could do very well. It’s one of the stocks that we’re bullish on for the long term. And the reason is because they have this monopoly over EUV tools.
The market they operate in is the lithography market. And so anyone who’s going to make chips, and the most advanced chips needs to buy these EUV tools from ASML. And ASML is the only one that makes it. So the second runner ups in terms of lithography are Nikon and Canon, and they’re not even close. They don’t have the EUV tools, they have the DUV tools, which are the less advanced version.
Part of our bullish thesis on ASML is that regardless of what happens with DeepSeek, you’re going to need these EUV tools because tech will keep moving forward. We will continue as an industry to uphold Moore’s law.
So smartphones will move to smaller nanometer nodes, AI will move to smaller nanometer nodes. And so companies like TSMC (TSM), which are a major customer of ASML, will continue to buy these EUV tools. And we’ll see higher industry adoption from that.
And that trickles into Applied Materials and into Lam Research, which don’t work in lithography, but work in etching deposition, which also go into that semiconductor supply chain for companies like TSMC to be able to fabricate the chips. So these are a couple of names that we’re bullish on. And TSMC is also on our list. We think TSMC really gets a bad rep because of the whole US-China-Taiwan kind of love triangle.
RS: That’s a nice way to put it.
SA: Yeah, I mean it’s the opposite of a love triangle, really. But it’s gotten a bad rep because there’s this constant narrative being pushed that China is going to invade Taiwan, so TSMC is going to be in trouble. But that hasn’t happened and the stock’s been doing really well and their earnings have been really well.
So if you compare actually TSMC’s earnings to AMD, the hype AMD got was much more deserved by TSMC, who has like Apple (AAPL) as a top customer and NVIDIA as a top customer, and who should continue even if they lose this AI tailwind in terms of DeepSeek in the midterm and what’s happening there, in terms of a lower AI infrastructure spend, they should have an offset to it, which is a lower threshold for AI endpoint.
And that was the trouble most of last year. Why haven’t we seen these AI smartphones or these AI PCs kind of get mass adoption? It’s because they always came at this higher price tag.
So now that that price tag is lowered, we’re looking at potentially seeing faster pace adoption for AI PC potentially, for, AI smartphone. So we think TSMC and these semi-cap guys are good long-term fix, to add on pullback and to keep in an investor’s portfolio.
RS: What would you say is important to highlight for investors wanting to know what it means to be a contrarian looking at the tech space specifically?
We’ve had Courage & Conviction Investing, he’s an analyst on Seeking Alpha. We’ve had him on a few times talking about the nuance, the artistry involved in being a contrarian and also needing to have a long-term view a lot of the time, and being able to not be knocked off your course.
What would you say about being a contrarian in the tech space? What would you share with investors?
SA: I would say that being a contrarian in the tech space, it’s really a challenge. It’s an uphill battle because it’s very often you’ll see the market just moves based on hype.
I think Tesla (TSLA) is the first example that comes to mind for that and a really accurate one. I mean, even with missing earnings, even with a lot of unfulfilled promises from Elon Musk’s side, the stock kept going up. I mean, it is down over the past, I think, 5 days, a good amount.
But in terms of fundamentals, yeah, the stock price performance doesn’t really reflect that. So I think in terms of being a contrarian, it’s finding that balance between knowing the fundamentals and understanding the industry and more so understanding the supply and demand dynamics.
So really thinking about things a lot more logically than Wall Street makes it out to be a lot of the time. So it’s knowing those fundamentals and really trusting that 1 plus 1 does equal 2 and that the odds won’t suddenly happen where 1 plus 1 equals 4.
Something that is really helpful is really circulating information and not looking at stocks as if they exist in a vacuum. So that really applies to all aspects of tech. So for example, even circulating semis into software for analysis and for research is something that we do very often.
When you have a good idea of the supply chain in terms of what’s happening in the end markets for the semis and what that outlook looks there, you have a much better grasp at being able to basically predict what’s going to happen to the software hardware side of things.
For example, we can circle back to ASML. When you see these EUV orders coming into ASML, then you understand that yeah, TSMC is going to be producing more chips because they were able to buy these EUV tools that are extremely expensive, have long lead times.
And then you can look at a company like Apple and try to predict better there what’s happening in terms of demand. And then the same thing around. So it works from top bottom and bottom up. So when you see mass adoptions, for example, in terms of Microsoft Windows 10’s end of life approaching.
So AI DeepSeek aside, just in terms of data points and plugging them into each other, that could mean a refresh cycle in terms of PCs. And that could be a benefit for a couple other names in that supply chain.
Looking at the supply chain and following the fundamentals there, like a story, I would say is essential without getting wrapped up in all the hype around Wall Street, because very often Wall Street will follow the hype and pump up stocks, price them to perfection, and then completely step out.
I’d say especially for the longer-term investors, that’s the most essential thing, to really be a Tech Contrarian. You’ve got to look at the fundamentals and you got to stick to that. I think the hardest thing about that is you’ll find that a lot of people just don’t use that logic, right?
A lot of the time the market doesn’t follow that logic or doesn’t follow the fundamentals and that’s something that will always pay off. Even if it takes a bit longer for everyone to catch up, sooner or later it will happen.
RS: I was interested in your point that you mentioned the different investing timelines and different investor profiles in terms of, let’s say, NVIDIA specifically, like what it looks like in the coming quarters, what it looks like beyond.
How would you expand on that in terms of maybe the different profiles briefly and how they should be looking at some of these top names that you’ve been discussing?
SA: I think this goes back especially for NVIDIA, it goes back to this idea of inferencing and training.
So if we think of NVIDIA as 50% inferencing and 50% training, and what we know is that DeepSeek has enabled inferencing at a much lower cost and at a lower computing power, then we can think about NVIDIA stock as needing to discount at least part of that 50% that belongs to inferencing to be able to get in at the right price.
Now there’s a lot of concern about, okay, will NVIDIA be dead now that, for example, we can do this with a lot less computing power than we had expected. And the truth there is that from an industry standpoint, there will always be a need for training.
Training will always be part of this equation. So even if the argument there is that basically even there’s a lot of people using AI now, computing power won’t go away because you’ll always need to train, even if we train AI for the problems that we have today, there’ll be new problems to train tomorrow and then there’ll be new problems to train the day after.
You’ll always need that computing power. So NVIDIA is never really out of the equation in terms of that. The demand will be there. It’s just about getting in at the right price when that AI infrastructure base is lowered.
So I would say that with that in mind, we were eyeing NVIDIA seeing some very high risk in the July quarter. And we think that once the stock is discounted there to below $100 per share at least, that’s when there’s a right time to jump back into the stock because that training will kick back in and the need for NVIDIA will still be there.
And that extends to Broadcom (AVGO) and Marvell (MRVL) as well in terms of their in-house ASIC business. So basically their in-house ASIC business is actually a positive, DeepSeek is a positive for their in-house ASIC business in the long run, because now that we have that lower threshold, a lot more people are going to be using AI. And in-house ASIC allows that at a lower price and at a customized price.
So for example, what Facebook is doing is different from what YouTube is doing is different from what Microsoft Copilot is doing. And each of these guys are going to be able to basically build in-house ASICs that’s custom to exactly what they want this AI to do.
So we’re also bullish on Marvell and we’re also bullish on Broadcom for the long run. Especially where Broadcom is concerned we think their software business is also going to carry part of that outperformance, especially if they focus on bigger enterprise customers and in terms of their business models.
Micron (MU) is a name that we follow closely and we think if this AI endpoint thesis plays out in terms of more adoption for AI because of the lower threshold there, we could see a lot of outperformance and a lot of space for Micron to basically participate in that.
And the same thing can be said in a different aspect for Meta. So Meta is one of the names that actually didn’t go red on the Monday, where you know, on that red Monday where the DeepSeek news came out.
And the reason is because now Meta is able to use AI for their own, let’s say, ad targeting, or even for this new commerce vertical that they’re growing more efficiently and at a lower cost. So we’re bullish on Meta. That’s definitely a name that we’re looking at. And we think that this can extend to a lot of places.
But what we’d say is that we’d advise investors to just be wary about the new AI promises that are coming in and that there’s going to be a lot of money floating around from NVIDIA to, for example, Palantir and then if Palantir’s bubble pops to something else.
There’s always going to be an AI name that’s going to be, you know, getting a lot of attention, getting a lot of traction, and that bubble will always pop. We just say that being ahead of the curve and kind of following the fundamentals there is the most essential.
RS: Tech Stock Pros, that’s your free analysis on Seeking Alpha. Tech Contrarians is the investing group. There’s a one-month trial offer for full access. So really nothing to lose.
SA: I also wanted to let the listeners know that this coming weekend we’re going to be doing a flash sale on Tech Contrarians because we don’t want a price tag to be a hindrance to people getting good information and breakdowns for the tech and learning about the tech that they’re investing in.
So we’ll have a 50% flash sale for the weekend starting Saturday morning and ending Monday morning for those interested in getting the contrarian side of things.
Editor’s Note: This article discusses one or more securities that do not trade on a major U.S. exchange. Please be aware of the risks associated with these stocks.
