Ollie Smith: Now to China where an updated AI model has upended assumptions about the cost of the AI revolution and sent the stock prices of some of the biggest companies with skin in that game through the floor. Here to answer your questions is Morningstar’s Kenneth Lamont from our manager research team. Ken, thanks so much for your time. Just as a starter for 10, what’s happened?
Kenneth Lamont: Well, the disruptors have been disrupted. Some of the largest AI players that we all know and love, those key members of the “Mag Seven″, have really, really suffered and it’s all due to one very small Chinese startup, which is quite incredible if you think about it. Specifically, what’s happened? Well, this small Chinese startup has essentially released an AI model, which is largely disruptive in two specific ways. One, they’ve managed to find a way of building these models at a much, much lower cost. So the training cost is much, much lower than the large AI players that you’re familiar with. So this is disruptive in its own right. This means this will put price pressure on the largest players. They will have to up their game, improve their models, and almost certainly reduce their costs to compete with this, which is disruptive in its own right. That’s great news for us as users of AI, but not so great for the bottom line of the big companies that are dominant in this space.
But the second reason, and I would argue even more disruptive, is that the traditional investment rationale, let’s say, behind AI, is that there’s an assumption that the best models are essentially a combination of the deepest and broadest datasets, mixed with brute force power. So it’s really been a scale game. And so this is why you’ve seen this dominance of, again, the names that we mentioned, your Microsofts, your Googles, et cetera, because they really have the scale. And this is the assumption that they would just therefore make better models. What this development, this Chinese startup has really shown is that actually the algorithms, you’re very, very smart in how you build the algorithms themselves, you’re much, much less reliant on the brute force power element. So this has huge consequences. So the biggest stock in the world, your Nvidia NVDA, is entirely, well, it’s in that position because it provides the best hardware for computational purposes for AI. And if computation becomes less important, then clearly this changes the investment rationale for Nvidia and others, ASML ASML and others in the space.
Smith: Talk to me about concentration risk because you mentioned the Magnificent Seven. I mean, that’s a story of phenomenal equity headline generation, isn’t it, baked into just seven companies. What does this say about concentration risk for investors?
Lamont: Well, this has been a topic of conversation. It’s something my colleagues in the Morningstar research team have published frequently about we’ve really seen historically high levels of concentration, actually globally, but particularly in US equity markets. And this could well really be seen as a correction. Interestingly, many of my colleagues, the prevailing view was that these AI or these companies that because of the scale component, the barriers to entry, et cetera, were almost natural monopolies. And actually, many thought the only way they’d actually lose their market position would be through government intervention.
Smith: Yeah, and the market got there first.
Lamont: Yeah, the market seems to have at least shown they can be disrupted.
Smith: Is it investable? Is DeepSeek an investable phenomenon? If someone’s watching this thinking, ‘well, why don’t I just invest in the disruptor of the disruptors?‘ Can they?
Lamont: So this is less of an opportunity. So the answer is no. It’s a small Chinese startup. We can’t just walk in and invest ourselves. And yet as investors, we are all clearly exposed to the disruptive risk of this company. So it’s more, we should be looking at potentially protective measures, maybe perhaps allocating a bit more to equal weight or sort of tilting away from the from the traditional highly concentrated indexes.
Smith: Sure. Kenneth, thank you so much for your time as ever. For more on this unfolding story in the world of AI, check out Morningstar.co.uk or any of our international websites. Until next time, my thanks to Ken. I’ve been Ollie Smith from Morningstar.
The author or authors do not own shares in any securities mentioned in this article. Find out about Morningstar's editorial policies.
