Nick: Hello, everybody. This is Nick Robinson from abrdn and you're listening to the Emerging Markets Equity Podcast, the show that explores the factors that underpin our thinking on emerging markets. We ask our expert guests the big questions from key individuals to evolving trends, all with the goal to identify and profit from opportunities in the region. So today we're going to discuss one of the biggest market themes at the moment, artificial intelligence, and specifically in the context of emerging markets. Since the beginning of the year with the launch of ChatGPT and then two incredible quarterly earnings result releases from the semiconductor company Nvidia that showed just how much investment is going into AI chips. The market has been gripped by AI and what it means for the world economy and companies. So today I wanted to focus on how this AI revolution could impact emerging markets specifically. So, I'm delighted to be joined once again by Pruksa Iamthongthong, who is our expert in Asian technology, based out of our Singapore office. Pruksa I think this is your third time on the podcast, so you’re a pretty regular guest. Welcome back. It's great to have you on again.

Pruksa: Yeah. Nick, thanks for having me back and it's exciting to talk about this AI topic, which has been a very hot topic this year.

Nick: Yeah, it's really exciting and I can feel like we've left this topic a few months to discuss, which gives us hopefully a bit more insight now that things have settled down a bit, post the kind of excitement around March and April. So, firstly, I wanted to talk about AI in general and how it's impacting emerging markets and how the impact might be a bit different from more developed markets. And example I was thinking about was how wireless telco had a different impact in EM than DM, you know, in the developed world prior to wireless, communication was pretty good. You know, most people already had a landline or an easy access to one, but this wasn't really the case in emerging markets, particularly in rural areas. So actually when mobile phone became ubiquitous, it had a much larger impact on productivity, as now most of a population could suddenly be connected without the time and expense of building landline connections. So, do you think we might see something similar today where the AI revolution could have a similarly different impact on EM versus DM?

Pruksa: Yeah, the short answer is yes, Nick, and I think it's important to put this into the context as well, where it has been something that’s been under research since the 1960s. But I think the question is why now and we feel that the inflection point in terms of the more commercial adoption really came over the last few years, and consumers have just become more familiar last year when ChatGPT became a bit more consumer facing as well. I'm sure you have gone onto to try to ask ChatGPT some questions. So, I think that's something that has gotten a bit more traction and putting AI in the forefront of consumer's mind. But I think what's important is that given the timing of today, EM, Emerging markets stands to reap the benefits of number one, years of prior R&D and the dollars have gone into it. So perhaps we have sort of like gone past the proof of concept phase and we are now entering more into what you call the implementation phase when the technology is out the lap and becoming a bit more applicable to the real world today. And this is really possible because we have a combination of a few things today. Number one, more data. This is the ability to generate, collect and make sense as well of all the data that is coming through. And number two, lower computing costs as semiconductor advances continues. So, yes, I think we are seeing the power of leapfrogging, you know EM would have less legacy infrastructure so you can actually design the optimal data centre architecture, for example, from the start without having to worry too much or having to reconfigure existing legacy infrastructure. So, this makes the spending a lot more efficient as well. And if you were to talk about the broader opportunities of AI and more specifically generative AI, we are very big believer that it has the potential to transform many industries and that ChatGPT that you and I know are just the tip of the iceberg. If we talk about, you know, its ranges from nearer-term opportunities in software applications like minute taking of the podcast we are doing today, and the ability to summarise this, that drives higher productivity. But there are other interesting areas like healthcare, which we can talk, go into a lot more details later as well. This is happening today in seven companies that we are invested in. So, I’d say very, very exciting and lots of potential to come.

Nick: Yeah, thanks. Thinking about emerging markets specifically there’s obviously been a lot of focus on AI and developed market companies like Microsoft, Google, Nvidia. Yeah, how does emerging market benefit, do you think, in terms of companies and the supply chain? And yeah, as I mentioned, now that we're several months into this investment cycle, where else would you see benefiting in emerging markets beyond the more obvious initial beneficiaries like Taiwan Semiconductor, which make all the Nvidia chips?

Pruksa: Yeah I think after the past few months we are seeing more beneficiaries actually. So, I would say overall the excitement has not died down, but in fact it has broadened out to include a wider range of industries and companies as well. But just starting to look at DM, the reason why DM companies are gaining lots of excitement, that's really because that’s where the large tech companies are from in terms of, you know, the Hyperscalers activities and they are at the forefront of the various language learning models, as well as investment into the end applications when you think about open AI’s ChatGPT and of course Nvidia that you mentioned with is very powerful GPU to power the models computation work basically. Now if you compare that to EM all this excitement will actually need to be built somewhere. So, EM continues to play its traditional role of being the factory of the world and the needs from the end perspective will need to be manufactured and supported in Asia, because this is where the bulk of the technology supply chain sits, including the most advanced areas of technology. So, plenty of opportunity here and I think I said a few times now that we are very excited about this opportunity. I think the key here to understand is that in order for generative AI to fulfil its full potential, models will need to be trained, computational capabilities would need to be built. So, this is like the first wave of the investment opportunity that we hear about you know when you talk about Nvidia GPU, which is a key component in the data centre AI server and Nvidia their manufacturing is done as TSMC. So, this is the first layer of opportunity and this explains autofocus in segment today. But what we think is equally interesting or even more interesting is the second, the opportunity that lies in the rest of, say, the data centre architecture and we see the need for data centre architecture upgrade and redesign. And this means is that the various components that goes into the workings of the data centre will need to be upgraded to handle more complex tasks as well and more complex calculation. I thought during one of the visit that we went that we did in Taiwan this year in May, one of our companies, which is a mid-cap company, check this analogy with me and I thought it is very timely to share with you today. The company makes network, they make data centre switches. And if you imagine that we have a lot more data traffic and this traffic are like cars on the highway and this is going to get more congested, basically the highway will need to be upgraded and they would need to the lanes would need to be extended. The traffic light, which is the switch in this case, will also need to be faster and smarter to control the cars that may be driving in many more directions. So, you can see this upgrading of the highway and its components could be a very long-term multi-year structural trend.

Nick: Yes, I mean, that's something that's been quite clear, I think, from the Nvidia earnings calls, just how the role of the data centre is now changing from data storage to much more compute. And that seems to be quite a big capex driver. If we yeah, if we think about the investment opportunities kind of moving away space from the direct supply chain of AI and thinking about more, you know the companies that should benefit from access to the various AI driven tools. Yeah, what kind of companies would you highlight there? And I suppose as a follow up, is there a risk, do you think, that a lot of the value that's created by artificial intelligence accrues to big tech in the US?

Pruksa: I think at the first stage you will probably see a lot of that value being accrued to the big tech companies in the US. So just as soon for the second question first, and I think that's really because of the huge investment that they have been making and they are trying to find ways to basically monetise the investment that they have been made. But I think what's quite clear for us going forward is that as the cost of AI adoption gets cheaper, as people actually understand more about how to apply this technology, I think you will see the potential of this flattening out across many industries. And I think right down to the smaller companies and the smaller guys, especially the cost of adoption gets cheaper and you may not need to, you know, no one, everyone doesn’t need to have their own model, but they can have like the application layer above the big models that need a lot of investment. So, I think that's something that we can look forward to in terms of how this would move on from say the big players into the smaller players as well. In the earlier question, I touch upon the benefits of health care as well as the semiconductor opportunities, but I think if you were to extend beyond that, the health care we focus on, the productivity that can be achieved through what you talk about drug discovery already, you talk about imaging diagnostics on cancer detection, for example, companies that are involved in the business of, say, content generation and creative, which is something that we initially thought that AI wouldn't be creative is actually benefiting from this as well. So, this can be companies that are involved in games development, the entertainment industry, advertising industry, for example. Anything that you have to basically use manpower to create content will be able to benefit from the use of generated AI to lower costs. Apart from this we also see it being beneficial for I.T outsourcing companies. Again, very similar concept because the use of AI should help to relief the labour intensive nature of software coding. So, you might have heard something about copilots in terms of coding and that allows the engineers to free more time to work on more value added work at again increasing productivity. So, I think in a nutshell, any industry that can benefit from productivity improvement that comes from AI adoption will stand to benefit from a cost perspective and allows you to work on more value added stuff. So, there is also a cost end goal to this whole AI adoption.

Nick: Great. Yeah. And an I suppose if we think about it as being a tool to enhance productivity and if AI is a potential future engine of productivity, you know, how significant is it, do you think, to Chinese growth that their access to these chips and equipment to make the chips has been limited?

Pruksa: Yeah, this is a timely question because the updated export control rule from the US has just been published, and I think the short answer is that in the short term we probably won't see much impact because the leaders within this in China of the large Internet companies that have been at the forefront of this, like the U.S. companies have stocked up quite a fair bit of inventories. So, they have a good level of inventories to run through and train their models, which is the current situation. In the medium term when we look at chip capability of the Chinese, there is a bit of a technological gap if they cannot get their hands on the right chip that are not permitted to be exported to China. And this is taking into consideration of a strict scenario where basically the approval for export would not be granted. So, what this means is that given that you have basically lower performance chips in the Chinese market, you probably will take more chips to train the model and it's going to be more expensive to do so. So, in simple terms, that means that it might take longer and more costly to achieve the same outcome in terms of training and the way that your algorithm work. Longer term though, we will see that China is very likely continue to improve its technology as well as its localisation drive so that the gap will narrow over time. And again there is also a different way of designing the data centre architecture, making it very suitable for specific tasks and you don't have to go through the route today. So, I think there should be some efficiency improvement as well to help to mitigate this limitation. And I think ultimately it might just also be the case that the models are trained currently, they are good enough we don't have to have a very, very sophisticated model and the future focus is really on the application end, and that means it requires less complexity and less usage of supercomputers. But I think over the long term we will need to see how this develops and it's something that we are monitoring very closely.

Nick: You know, one example of maybe how China is managing to get by despite the restrictions was this recent launch of the Huawei phone that uses seven-nanometre chips. Yeah, what's your take on this phone? And what it implies is it signs that restrictions aren't working as anticipated or perhaps that Chinese onshore technology is more advanced than a lot of people have anticipated.

Pruksa: Yeah, this is something that I think caught us by surprise and there are many versions of details out there that hard to verify given that Huawei is a private company. But I think this is how we will try to analyse the situation. So, if you look at the chip that is in question, it is quite likely that this is met through a process called multiple patterning, which is basically a more expensive process than using EUV tool and this tool is not available in China it has not been able to be exported to China. And what this means is that it is also quite likely that this comes with a trade-off in terms of lower production yield when you think about manufacturing costs, you have a higher cost and also from a chip specification perspective, it may not be advanced as, you know, the chip we find in an iPhone today, which runs at three nanometres to five nanometres as well. However, if you look at it from an overall phone product perspective, this chip may be good enough and perhaps the hardware shortfall can be mitigated by software and the higher cost may also be absorbed by Huawei given that it is a high is it's a private company, we actually don't know with regard to the margin transparency there as well. So, I think consumer so far has had good reviews and they are adopting this. So, I think the experience has been pretty good from a consumer perspective. But I think what's quite clear is that this is a showcase of China's localisation drive that I've touched about before that I think will remain the long-term goal of the Chinese government. And so, while the restrictions have tightened up and I think this might continue to be tightened up as we move forward, the direction of travel for China to its chip localisation will continue over the long term.

Nick: Thanks. Yes, I suppose I asked you earlier about the risk of benefits of AI accruing to the big U.S. tech companies, but perhaps an optimist would say that given decoupling of China and U.S. tech, that's going to be less of a risk to China going forward.

Pruksa: Yes, that's right. And I think from a self-sufficiency reason, I think we talk about how they will be able to handle the short to medium-term picture at the expense of costs and perhaps productivity. But longer term, I think we see investment opportunities in both the China supply chain as well as the EX-China supply chain as well, given the localisation drive that we have talked about.

Nick: Great, well on that optimistic note towards China, perhaps that's a good place to draw the podcast to a close. So, the only thing left for me to do is thank my guest Pruksa. Thank you.

Pruksa: Thanks Nick for having me.

Nick: And thanks to everyone who took the time today to listen in. If you enjoyed today then please download our other podcasts from our website or wherever you normally get your podcasts. Watch out for our next episode and tune it.

Individual

Institutional

Podcast

transcript

Next steps

Related Stories

Related articles

Small caps: Quality as a cornerstone

Equities: Diversification still the only free lunch

Infrastructure equities – firm foundations for a volatile world

Mission to Mumbai: on-the-ground investment insights

Small but mighty: why European equity investors should consider a meaningful small-cap allocation