DBS CEO: Market Repricing Risks on AI Financing

Joining us now is a DBS Group CEO, Tan Su Shan. Really great to have you with us and we appreciate your time. Well, let's start off on the theme of the event. How much AI financing exposure does the bank have?

So we are originate a lot of these AI deals in Asia. We are often the lead manager for a lot of the syndicated loans, etc. Uh but we originate and we syndicate them out. So we don't actually end up holding a lot of it, but um year to date we've done about 25 billion or so.

origination, but we've held a fraction of that.

Uh are they across sort of energy, chips, infrastructure, data centers and if you sort of break it down even further, which are the ones that you're holding on to and I guess are you looking to increase the number of those deals?

Well, a lot of it's going to be in the semiconductor space. So that's a lot of a demand for for memory chips. There's a lot of demand for the upstream supply chain uh and all the way down to data centers. So the whole stack, right, from from the hardware and the semiconductors, the connectors, the cooling systems, uh down to the data centers and and even the traders. So there's there is, you know, the whole the whole semiconductor down to to data stack hardware. Mostly hardware.

Is there a sense as as banks, not just DBS, of course, but really across the region are increasing that exposure to AI financing? Are there concerns about risks and to that end, do you think there should be some sort of official or unofficial cap being discussed?

Well, I think, you know, we've always been very structured around our lending around the AI and data center space. We've always been looking at all the different risks. You know, we look at it being triple net, for example. um and who's the who's the takeout? Is there a Is there a back stop? Is is a big semiconductor company putting on a back stop, for example, on their chips? So, it it really is heavily structured right now.

But the market has spoken. I think the market has spoken. Rates have gone up over the last two year, after the over the last two weeks. Um and the bond prices of some of these, you know, um uh hyper scalers have also reacted. So the market has spoken and it's not a bad thing because now people are asking for higher returns and people are also looking at structures and they're making sure that whatever money is now being hooverd into the whole AI CAPEx ecosystem will lead to some kind of long-term ROI that, you know, that can then be repaid back to to to to debt or or bond or equity holders even.

What's really interesting at the moment is as you say, you know, you used to have these hyper scalars that would fund their AI through cash flow through equity issuance, right? But now it's coming from bond issuance. The bond influence is the gangbusters at a time when interest rates are rising. Does this confluence of factors worry you at all?

Well, I think, you know, the market will always innovate, Heidi. I think, you know, certainly, we're seeing in the US data center deals now as they mature, being packaged into ABSs, you know, securitized and sold off in chunks. Uh as you said, the hyper scalars have started to tap the bond market as well. They've tapped the equity market with secondaries. Some of them are the LLMs are tapping the IPO market with big IPOs. Um and we are now seeing the bond market being tapped as well. There will be some credit rating reversions potentially down for some of the weaker players. Um but the market is repricing risk today. And I think it's healthy because the market wants to ensure that these financial, these big financial commitments are actually financially sustainable. And and that's why I think the market's asking the right questions, the market's demanding a risk-rated returns, um and it will adjust.

I wanted to ask about the broader business too, because in the backdrop of the pressure that we've seen on bond yields, there has been a view that there's going to be further pressure on the banks, right? Is that something that you see as a likely scenario?

Well, that's interesting because technically as as as interest rates go up, you know, banks do better as their interest uh margin income, their marginal, you know, income for for interest rates go up as well. Obviously, there is a mark-to-market on the bond portfolio and and and um and and that will also affect many banks. So, it is a confluence of both. You know, rising rates does help bank's income, interest income, net interest income, and rising rates does hurt your mark-to-markets on some of your bond holdings. So, it's also how banks reprice their loans versus how they reprice their deposits that determine your net interest income margin. And this is where if banks know how to dynamically uh be nimble about managing their balance sheet, you know, if they're able to lower their elasticity of their cost of funding on the deposit side, and yet have that sort of right up as interest rates go up on the asset side, then naturally they will make more money on interest rates.

But of course, the trading side will will, you know, depending on your position, your trading side will will see ups and downs as as interest rates go up and down.

Yeah. I wanted to ask about how the wealth side of the business is going too. Are there sort of further gains to be expected and and where would you see those opportunities? I know Taiwan has been a highlight, one geography that you see a lot of opportunity.

Yeah. Well, wealth management has long-term structural tailwinds. um A because Asia our our our backyard is is creating wealth. Um B, there is an inter- intergenerational wealth transfer, and C, the capital markets here are really, you know, becoming more liquid, becoming deeper, more mature. So, and there are deals to be done, whether in the private uh you know, markets or the or the public markets. So, wealth has structural long-term tailwinds. In the short term, though, yes, you know, as markets go up and down, you know, the the flows will be affected. But we're seeing that new money growth being consistently strong. We're seeing that, you know, structuring and not needs for structuring and long-term wealth management is also continuing to be strong. So I think, you know, long-term trends are still still intact. Short term, there will be some volatility depending on how the markets move.

In terms of those uh high potential markets like Taiwan, are you looking at different ways of growing those markets? M&A in view?

Well, you know, we've had a, you know, we we bought Citi's business a few years ago and that's been a great uh great acquisition for us in Taiwan. Taiwan has very good structural growth trends. You've obviously a big beneficiary of the AI hardware, semiconductor, the NV- NVIDIA ecosystem. Taiwan's been a big beneficiary of that. Uh second, Taiwan's stock market is now $5 trillion, number five in the world. And thirdly, the onshore wealth and and domestic market is also uh picked up. It's very, very strong growth growth um tailwinds there as well. So Taiwan is is doing well and um hopefully will continue to do so.

The other aspect of AI that's get garnering a lot of attention is really some of these cybersecurity concerns, the hacking concerns, right? The ability of institutions to be able to control some of these models. What work is DBS doing in terms of these safety guard rails? And is that really one of the main things that concerns you in terms of balancing the opportunity and risk from new technology?

So, every new drop that has some patching, you've got to patch. So, patch, patch, patch, right? So really it's how do you shorten your patching cycle and what are the mitigants that that you have in place? Do you have mitigants? Is it internet facing? Is it customer facing? If it's all of the above, if you don't have mitigants, then we have to really speed up the patching cycle. Otherwise, the normal patching cycle, you have to go through the whole SDLC, software development life cycle, SIT, UAT, PAT, for example. And that does take time. But today, as you know, this this flow of of patches come through with each new model drop, the team is is really focused on ensuring that, you know, you we'll patch as soon as we know. Uh and really cyber security is not just your cyber team's job. It's everyone's job, right? from the front of the house, middle of the house, back of the house. Everyone has to be cyber aware and and on top of these of of all these new developments.

So, living, breathing this every day.

And to that point, the bank is rolling out AI technology across the internal workings as well, right? Are you able to tell us how that translates in terms of productivity, net jobs gained, net jobs lost perhaps as well?

Okay. Well, the way we look at AI, we've got three three sort of different pillars. The first pillar is your traditional AI, sort of machine learning, AB testing, and that continues to yield us good results and we've always been tracking this through the AB test and the delta between the A and B, and that's about one one over a billion dollars right now. Then of course, you have the generative and agentic AI respects. You've got sort of big horizontal, everyone can use AI kind of productivity saves. Heidi, that's a little bit harder to crystallize in terms of quantitative returns because everyone saves time. Everyone's more productive. Then it's really how people use that time saved or that productivity saves to do new things. The vertical use cases, however, are easier to quantify because you can say, oh, okay, by rolling out these agents, I've saved how many FTEs, through the FTEs saved, you know, the the employment saved, through the head count saved, how am I deploying the capacity that I have for growth? And that we can measure through through through through through through human productivity. And the last is we're also rolling out enterprise use cases and those are what I call the 12 big use cases. You can't do too much of those because those are a massive effort. And we've got 12 big journeys. We call them operate um operational model transformations. That means really relooking how we operate, really relooking our workflows and work processes, and coming up with big outcomes for those. And those, we're starting to measure those. It's and it's early days right now, but we're optimistic. We are beginning to see, for example, fee growth. We're beginning to see new account opening at a faster speed. We're beginning to see faster turnaround time. So it's time and it's fee and it's volume and we're trying to crystallize those right now.

In terms of your question on job losses and gains, well, you know, I've always said to my colleagues, we're not here to protect any jobs. We're we're not here to protect jobs, we're here to protect humans. We're here to protect our workers and our colleagues. And really, it's all about our workers and colleagues having the right attitude, being able to reskill and upskill themselves and being able to take on newer jobs, higher order jobs. So your mundane everyday kind of, you know, searching, researching, pulling information together, uh looking at legal docs for example. All that, if an agent can do it, let the agent do it, and then you can spend more time to help us grow, keep us safe, right? Help us grow new clients, talk to customers, bring them on board, give them advice, etc. So, I'm optimistic that as long as, you know, our staff and and really humans are incredibly nimble. The human brain is one of the most nimble things in the world, right? And I think as humans, um and and our certainly the DBS workforce has shown to be quite agile and quite nimble, we will move up this AI order and do higher order things.