Bloomberg Audio Studios, podcasts, radio news. Welcome to the Daybreak Asia podcast. I'm Doug Prisner. Trading in global equities was muted on Monday given a number of holidays. We have lunar New Year festivities underway in the Asia, Pacific and in the States. Markets were closed on Monday to observe Presidents Day. Now in the week ahead, disruption from AI
is likely to continue as a theme. Last week, nearly every part of the financial sector was hit, and strategist at JP Morgan Chase are urging caution on stocks at risk of AI driven cannibalization, and other Wall Street firms are creating tools to capitalize on this divergence. Goldman Sachs, as an example, has launched a new basket of software stocks that goes long firms that will benefit from AI adoption and short companies whose workflows could be REPLA placed
for a closer look. I'm joined by Stephanie Leungshi is the CIO at Stashaway. Stephanie joins from our studios in Hong Kong. Thank you for being here. It seems clear that AI has ushered in disruption across a range of industries. How do you understand this moment in the evolution of artificial intelligence.
I think two things.
A number one is that Anthropic released its latest model together with open AI, and they are much more powerful models than previously. And what these new models enable kind of I guess us to do is actually to build even more powerful agents. So if you think about the latest kind of I guess development, which is something called open call, it's basically an agent that you can deploy on your home PC. And actually, I don't know if
you've tried it. I've actually started deploying my own agents in my PC as well using open plot and it's quite easy, to be frank, and it can help you to do a lot of things right. It can help you to automate your daily routine, help you to automate
any workflow that you specify. So I think the fear is that once these agents are kind of so easy to deploy, then you don't need to buy any software, right you can just ask an agent to build your software and it'll come back with what you require in a very very short time and do it on a very cost effective basis. So there's a question about, okay, whether or not all these software companies will actually face
extential kind of crises. And I think we do have to separate these companies into two groups.
The first group is.
The ones that have mode, which I mean from my perspective, the modes are defined as for example, companies with a distribution strong distribution like Microsoft or SAPE. These are very very entrenched with the enterprises, and of course the software companies that have their own kind of understanding of the data layer.
For example, if you think about the securities.
Companies, right, they have a very deep understanding of how to kind of think about security and how to survey the data to to to to do their job. And these are not that easy to replace. You can't just tell an AI agent to build me a company with distribution or a company with with a deep understanding of cybersecurity. The second group of company are the ones that have not a lot of modes, uh, And I mean those are kind of I think companies that for example, serve
single purposes. For example, if let's say a company has a SAA business that just focus on providing a scheduling tool, I think that is a mode that is easy much easier for agent to crack. I mean, you can just spin up agents and and and tell it to build that. So I do think that there are descriptions to the sector.
And then also I think on a broader basis, if you think about simple kind of old school supply and demand, there's going to be a lot of supply of software that's coming right, and and the question is that is there a so much demand to have solve this supply?
And I think given that the supply.
Is also coming up at very very low cost, I mean that is a true disruption for the whole software industry. But I do think that the market has been selling down kind of quite in discriminately. And also these are are a bit overblown.
So one of the things that I want to focus on the disruption from AI that has the potential to impact nearly every part of the financial services sector, whether you're talking wealth managers, insurance brokers, even some of the big banks, boutique advisors. Now I know your firm, stash Way, is a digital investment platform. Do you think that this is necessarily going to engender a lot more in the way of competition?
Yeah, And I think, I mean, believe me, this is like the daily conversation we have in the c suite as well, right, And I think there are two things that I think that we need to think about in terms of whether or not these would disrupt the financial industry.
The first thing is, of course.
The regulations, because I mean all of us are are tightly regulated. For example, I mean we have we have licenses with regulators in all the markets that we operate, and I mean these are very very strict kind of I guess regulatory kind of restrictions that we have that we need to operate within.
The question is I mean if even if somebody.
Builds an AI agent, right, I mean you need you still need to comply with other regulations. You need to file a license with the regulator. So I mean there is a level of expertise that is required to do all these And I think from a regulator's perspective, I mean, regulators are typically slower to react to these kind of innovations, right, They don't have a framework to approve, for example.
An AI agent. Can an AI agent actually give advice or not?
These are sort of still unanswered question because I mean the AI agent, of course is not licensed. So I think that's the first kind of big question mark as to whether or not AI can or AI agents kind of independently can disrupt the industry. Secondly, if you think all these kind of financial companies, I mean we're not standing still, right, We're not sitting around and and not doing anything.
Uh.
In fact, I think a lot of the bigger financial industries companies are innovating and and sort of investing a lot of money into the into building out the own A capabilities.
Uh.
If you look at the for example, in JP Morgan or Goldman or or I mean even in the stashway ourselves, we're investing a lot in building out our AA capabilities, and I mean those are I think the bottleneck still remains the sort of I guess, the the resources that you have in order to build these out, because I mean using tokens or or using these alblems, it's not cheap, right,
it doesn't come at free cost. And therefore I do think that companies with platform, with resources and with regulatory kind of approvals will still remain the bigger players.
So we've seen how the adoption of artificial intelligence has impacted the hardware market, particularly in the Asia Pacific region, so much so that we're talking now about a shortage of memory chips, particularly the high bandwidth memory that's necessary to work with these various AI platforms. Are you seeing opportunity here? How do you understand the shortage that we're seeing in memory?
Yeah, I think the I mean, of course, a memory industry has gone through a lot of consolidation in the past, and now we have basically kind of three big companies supplying most of.
The memory chips. I mean, memory historically has been a cyclical industry, and I mean.
That is because basically it goes with the industry cycle, right when when I mean, when the economy is hot, there's a lot more demand for memory and these kind of chips. When the economy is not so hot, then basically there is oversupply. And typically, I think if you look at the AI development today, as we go from kind of model training to inference, there's a lot more demand for memory because because of the thingand kind of context windows.
So when we do inference, when we ask.
GPT a question and we carry on these conversations very very, very very long time, and therefore there's a lot more that needs to be stored in memory for the model to be able to have a prolonged conversation with the user and This is the main difference between inference and training. So in the past few years, when the industry or when token usage was actually mostly focused in training, the
demand on memory was not that big. Right right now, actually, we're just starting to see the increase and ramp up and demand for memory, and I think as the AI kind of agent usage proliferates starting this year, we're actually
going to see even more demand. And the problem is that because the memory supply is actually focused on these three companies, it's not that easy for them or it's not that fast for them to ramp up capacity, so I think at least for the next few months or the next year, So the visibility of a memory shortage is actually still remains quite large. And if look at some of these companies, they're still trading at pretty decent valuations because the market historically has viewed them as cyclical
rather than structural. So I do think that there is more to go. Of course, in an air term, a lot of these companies are quite overbought, but I think these are companies that you add to if there is a correction.
So when you consider how the technology has been evolving over time, and I'm wondering whether or not, there is a risk here that what we're talking about today is being cutting edge becomes obsolete six months from now, and that if you're investing in that, if you're making a capital expenditure in that type of technology, that you could be forced to remain competitive. You're going to be forced to have to upgrade.
Yes, But also I think because the demand is so vague the UH I guess, the risk of UH investing in assets that they appreciate very very quickly is today lower. So if you look at, for example, the Nvidia chips, the older chips actually don't follow that much in value. Indeed, if you look at kind of the recent pricing, they've actually been going higher. And that's because the way it kind of I guess the whole AI stack works, the
most cutting edge hardware is used for training. But today we have an addititional demand which is inference, right, which is basically I mean us talking to GBT and also using agents. These tasks actually don't require the cutting edge chips that are required as of training. So we have started another wave of of of of I guess AI application and these applications actually can make use of the older UH kind of uh, I guess older and less
advanced chips. I think the analogy maybe we can draw the iPhone, right, I mean, iPhone basically has a new version every year, but the old iPhones actually still retain the value because you can actually resell these to other kind of I guess, other users that do not demand such a cutting h iPhone. So perhaps I mean this is kind of how I would categorize or characterize the the the AI hardware, because the demand is so big
that needs to be satisfied. So and also the demand is a spectrum of high end, lower end, medium end. So I do think that the risk of kind of investing in technologies that depreciates or fates is it's lower.
Stephanie, before I let you go, we are obviously in the midst of the lunar New Year holidays, celebrating the beginning of the Year of the Horse. Can I ask you to offer some wisdom?
This year is the year of the fire horse, which tends to be I guess fast running and and I guess quite vivid. So I think in a sense, I mean it's sort of I guess shells with our that this year we're going to see a lot of upside downs and volatility and times. I mean, think about kind of what we've gone through already in the last month and a half. I think that's sort of a taste that was to come and problems. I mean, that's what the fire Horse is going to bring us.
Okay, Stephanie, Happy New Year too. Thank you so very much. Stephanie Ljung is the CIO of Stashuay joining us here on the Daybreak Asia podcast. Thanks for listening to today's episode of the Bloomberg Daybreak Asia Edition podcast. Each weekday, we look at the story shaping markets, finance, and geopolitics in the Asia Pacific. You can find us on Apple, Spotify, the Bloomberg Podcast YouTube channel, or anywhere else you listen. Join us again tomorrow for insight on the market moves
from Hong Kong to Singapore and Australia. I'm Doug Prisoner and this is Bloomberg