Valerio Baselli: Hello and welcome to Morningstar. After having experienced a huge rally in 2024, artificial intelligence stocks have been falling over the last weeks amid market turmoil and the DeepSeek disruption.
Today I’m joined by Josh Sambrook-Smith, global equity investment manager at GAM Investments.
So, Josh, what is your take on the recent sell-off we have seen in the AI sector? Was it maybe a necessary correction? Is it now an investment opportunity?
AI Sector Correction: Opportunity or Warning?
Josh Sambrook-Smith: Yeah, I think there’s a couple of things going on. So firstly, it was the DeepSeek news that really catalyzed the first round of selling. And that was really because this new model came out and it indicated that you could do training of large language models at significantly lower cost than the market had expected. And so that means that, you know, you don’t need to spend as much money training your model, you probably don’t need as much hardware to do the training. And therefore, the sector started to sell off. And so, this was against the backdrop of very high valuations pretty much across the board for anything with AI exposure. So, this is kind of the bad news the market was looking for. The stock sold off as a result.
I think the second point was just that there has been a, well, information that before this news was being treated kind of indifferently by the market, is now being treated as a sell signal. So, Microsoft talking about switching of data centers and Microsoft CEO out there talking about overbuild, and news that some of the GPU designers and asset design companies had reduced their orders at TSMC that has contributed to the selling. I think the final point is just there is a lot of uncertainty in the market at the moment because of the tariff news coming out of the US. People really didn’t know what was going on. And markets hate uncertainty.
AI Innovations to Watch in 2025
Baselli: Right. And in light of this, what are the main innovations you expect from here on in the AI industry?
Sambrook-Smith: You know, I think the most interesting one at the moment is epigenetic AI, and that is where you have an AI that is intolerance that can act on its own. It can make decisions. It can basically take a lot of work off your plate and do it for you. And so that uses a variety of technologies from the, ChatGPT so large language models, it relies on machine learning, natural language processing, and all of which needs a lot of heavy hardware to do the job but then combines all those technologies and almost is your perfect virtual assistant. There are a lot of interesting potential applications for this stuff. It’s very hard at the moment. The most obvious ones are things like the system level, you know, buy flights for you, that will negotiate, bills for you and do your personal finances, but then it becomes hopefully a lot more advanced than that. And we can see these things creating videos, creating music. Doing all of this autonomously. And then eventually, maybe one day these technologies will be capable of doing really exciting things like discovering cures for diseases, developing new chemical compounds, designing completely novel materials. And I’m already beginning to see evidence that with Google and AlphaFold model, which is already out there discovering new drugs and new chemical compounds that are being put to use in the real world.
The US-China AI Battle: What It Means for the Global Tech Landscape
Baselli: That’s very interesting. We saw what happened when DeepSeek launched its much cheaper AI model, the R1, do you think artificial intelligence will be another battleground between the US and China? Do you think Chinese tech companies could really undermine American AI giants?
Sambrook-Smith: I think it already is a battle ground. And I think, everything that the Biden administration in the US was doing has indicated that, so, the export controls that are making it harder for Chinese companies to buy the super advanced GPU chips that Nvidia makes, as well as the export controls that are making it extremely difficult for Chinese companies to buy the manufacturing hardware. So, the EUV tools from ASML, to name one company, that make it very hard for, Chinese companies to make this stuff. So, without access to the chips under the access to the tools that you need to make the chips, you know, there’s clearly some, there’s clearly a conflict going on there.
I think if you look more widely at the model space itself, so, you know, DeepSeek and the ChatGPT in the world, there’s going to continue to be a lot of innovation going on in the model. Chinese companies are being forced to innovate in different ways because they don’t have access to the hardware. And that’s arguably what led to the creation of DeepSeek itself. So, I think the two countries are going to take different routes, but ultimately, they’re going to keep innovating. And, and I think that’s basically what will be happening for the next year.
3 AI Stocks to Invest in Now
Baselli: Finally, to be more concrete, can you name three companies, three AI stocks, that you particularly like today and briefly explain why?
Sambrook-Smith: Yeah. So, my three would be Broadcom, ticker is AVGO. They are a custom chip designer in the background to the thesis is really that there is an exploding demand for customized, accelerator chips. So, number one, buyers want to get away from the Nvidia tax. Nvidia is very costly for the best devices market, but they cost a lot. Number two, the big buyers like Google, Amazon, Microsoft, etc., they want hardware that is specifically customized for their own applications. They have their own list of demands. They need the ability to consume power as efficiently as possible to really give the best value services to their customers. So, this is the opportunity that, to design their own chips. Broadcom the largest and most well-resourced custom chip designer in the world. They’ve got a long pedigree of designing custom chips. They got a lot of experience on the leading edge. So, at seven nanometers, three, two, etcetera. And they’ve got a long-time list of the, you know, the largest, most sophisticated buyers out there. So, Google, ByteDance, Microsoft, possibly, maybe even Apple. But there’s a lot of very big companies that are going to Broadcom to get this stuff designed. So that’s one.
I think the second one would be ASM international ASM. So that’s a Dutch company. They are a specialist in something called atomic layer deposition. So that’s this little machine which will deposit a very thin film of metal or an oxide, onto a wafer. And so, it’s a little bit arcane and ridiculous, but the use cases for that technology are exploding. And the reason is because, you know, in order for AI to keep progressing, transistors, microchips need to get smaller and smaller. And the technology you need to do that, becomes more and more sophisticated. And atomic layer deposition is essentially one of the only technologies available for specific types of advanced semiconductor manufacturing.
And I think the third one would be Applied Materials AMAT. So, they’re an American company, though, the largest semiconductor, equipment manufacturer in the world. They supply all the big foundries from Intel’s TSMC and Samsung. But they have the one of the broadest and one of the most sophisticated toolsets out there. They do deposition, they do cleaning, they do all sorts of the most important manufacturing process domes. And what we know over the next few years is the demand for this stuff is going to increase as more advanced, foundry capacity is built. And so Applied Materials are one of the best positioned companies there. They also have very high return on capital. Great business. And it will continue generating value for many years to come.
Baselli: Thank you so much for your time, Josh. For Morningstar, I’m Valerio Baselli. Thanks for watching.
The author or authors do not own shares in any securities mentioned in this article. Find out about Morningstar's editorial policies.
