Josephine Teo on CNBC: Singapore's bold AI push

On CNBC's Converge Live, Teo uses Changi Airport as her frame: Singapore does not own aircraft technology, yet handles nearly 40 million passengers a year because it found roles where it could add value — Singapore Airlines, maintenance and overhaul, aerospace manufacturing. AI works the same way. The most sophisticated large language models may not be developed here, but Singapore can still partner with them and build value propositions that work for commercial users. Resilience does not mean owning the supply chain. It means having choice — diverse sources, diverse technologies, diverse system designs.

DeepSeek changed the cost calculus. Two or three years ago, enterprises told the government AI was too expensive; with cheaper open-source models, enthusiasm has surged across finance, healthcare, transport, and logistics. On infrastructure, Teo points to Stargate's $500 billion and France's roughly €100 billion investments and says, on a hundred-year horizon, this is unlikely to be overcapacity. US firms alone invested roughly $50 billion in digital infrastructure and utilities in Singapore last year, and Singapore's data-center density per GDP is already among the highest in Asia. The green DC roadmap pursues two strategies: use less energy (including rethinking cooling standards for tropical operations) and use green energy, with research extending even to greener algorithms.

On safety, the global conversation has shifted from Bletchley's existential risks to Paris's near-term ones — discrimination, cybersecurity, AI as an attack tool. Singapore's approach is risk-based, sectoral, and targeted. First check whether existing law covers a risk; pass new law only when it doesn't — as Singapore did in 2024 for AI-generated content during elections. This produces a patchwork compared with the EU AI Act, but Teo prioritizes speed and specificity for now. On skills, she highlights one program: Singaporeans 40 and older can return to school full-time or part-time for up to two years, with subsidized fees and a S$3,000 monthly training allowance — roughly S$72,000 over two years to reskill and retool.

Caption language: en · Fetched: 2026-05-02

When you have a technology like AI that is potentially so transformative, how can you take full advantage of it? So does Singapore need to spend $500 billion on AI infrastructure? I can tell you that a lot of companies are spending that money. Singapore has sought to establish itself as a central technology hub in Asia and the wider world through regulatory clarity in many areas to fostering a startup culture. But with technology rapidly advancing with a focus on artificial intelligence, can Singapore compete with the US, China, and countries in Europe which are all vying to become leaders? Thankfully, I've got Josephine Tio here to talk about this. Josephine is Singapore's Minister for Digital Development and Information.

Thanks for joining me today on this special episode of Beyond the Valley Recording in front of a live audience here at Converge Live in Singapore. Happy to be here. Arjun Josephine, can we just talk about uh AI of course uh and from a perspective of yourself from a government minister? What's the biggest opportunities right now in AI for Singapore and for the rest of the world? Well, the fundamental duty of any government is to its citizens. And in our case, it's a question of how we create more opportunities for our people and for our businesses. So that is very much top of our list of priorities. So when you have a technology like AI that is potentially so transformative, the key question is how can you take full advantage of it?

And the starting point has to be what are your people doing in terms of jobs, in terms of careers, what kinds of businesses are already here and can AI help them to individually as well as the enterprise level achieve new peaks of excellence. That's really how we're thinking about it. So the idea that um AI serves the public good for Singapore and the world starts from this point. You want people to have skills so that they can contribute even more to their own businesses as well as to their employers.

uh you want businesses uh that uh certainly have to compete very intensely in a global market to find that Singapore gives them an edge and that's really um where all of our efforts are directed towards building up the skills building up the enterprise capabilities so that both at the individual the workforce as well as the economy level we can be more competitive together. We'll we'll dig into skills and safety as part of this discussion. But let's start with Singapore first, Minister, because we talk about AI in the world. A lot of the conversation always centers around the US, around China and the capabilities and companies those two nations have. A lot of smaller nations are left out of that conversation. Even big regions like Europe are left out of that conversation.

So when we think about Singapore, uh how would you assess Singapore's position here in some of these key technologies? You know, it's a really interesting question and one that we grappled with round about 2022 2023 when we were trying to have a rethink of our national AI strategy. The analogy I'd like to draw is if you think about our Changi Airhub where you are right now and you think about the fact that um we serve close to 40 million passengers a year now out of a population of not more than well not even six million just around that. Um how how is it possible? How is it possible that uh you don't even own aircraft technology? Um there are commercial aircraft uh you know manufacturers Boeing and Airbus. They are not Singapore companies but it doesn't prevent us from conceiving of an air hub.

Uh you would want to have your own champions and I think over the years we were very fortunate that there were some really excellent people outstanding efforts to build up a worldclass airline like Singapore Airlines. We've got very good maintenance, repair and overhaul companies based here. We've also got um aircraft um you know manufacturing aerospace industry is very significant. None of these were built up on the basis of Singapore owning aircraft technology. They were built up on the basis of identifying areas where the value ad could be contributed out of Singapore that made sense commercially to these companies. I think if you adopted that sort of a mindset um the same could apply to AI too.

the the most sophisticated the highest end large language models may not be developed here but it doesn't prevent us from working in partnership with these kinds of inventions these kinds of tools and create a value proposition that is relevant in our context and also useful to commercial entities so many people are looking at AI and how foundational it is and resiliency is a key word that's bandied about when we talk about some of these technology ies like AI, like semiconductors, the need to have a a trustful, a secure supply chain and and there are many countries in the world right now saying that means we need to try to either reshore or own a lot of that supply chain. So from Singapore's point of view, you you've got a different perspective here. It's you're happy to work with a lot of these global partners.

I think you're spot on in saying that at the heart of it, the desire is not always to have things that are created and uh maintained by yourself. At the heart of it, your main objective is to be able to achieve resilience. In other words, you want to have choice. You want to have options in case something happens that is not favorable. And if you want to achieve resilience, there are different ways of doing so. Um achieving diversity in terms of how uh your supplies are obtained. Achieving diversity in terms of the technologies that are being deployed. Achieving diversity in terms of uh even the design of various systems that you put in place. Those are all contributing towards resilience. And that's how we think about it.

Um it's not always the case that you uh have the luxury of owning the entire supply chain or even the most critical components. You've got to be able to look at it and say what is sensible for you yourself to be able to produce, manufacture or store at least uh but what is sensible for you to um diversify in terms of getting the supplies. So that's the way we think about it. It's hard to have a conversation this year about AI without mentioning deepseek because of the ripples it had through through global markets but also through the technology sector as well. But what it did show I think was the importance of open source and the development of open source technology which may give uh smaller nations as well as smaller companies the ability to create real applications.

Because one of the conversations I've been having over the past few months was that actually a lot of people in tech think you know the models are going to be commoditized and the real value really is in the applications there and almost what what DeepS and other open source models are giving is a chance for companies and countries to create valuable applications. So how do you how do you view kind of what's happening open source right now and what it means for the development of of say some of these AI applications within Singapore? One of the concerns we had um you know two to three years ago was to what extent the AI hype would actually translate into real benefits in the business context.

And as we interacted with uh a whole range of industries as well as companies of different sizes operating in different sectors facing different challenges. uh one very common theme was AI is still super expensive to use. So if we did the costbenefit analysis, uh yes, we're very interested in the potential that AI can bring, but the amount of money that we have to put into developing um the use of these AI tools because it's not just always so easy buy off the shelf and you can implement it. uh you may have to make some adjustments within your own organizations. The companies told us that it was still very costly for them to do so.

I what's consistently the feedback uh in today's context with the introduction of deepseek is that now the cost considerations have become quite different and from that perspective in terms of being able to promote AI adoption and boost the development of AI ecosystems in many more different contexts I think is very exciting and we see this happening um not just you know in in um uh within the government in in terms of how we are hoping to use AI tools to improve public service delivery. We see this amongst um the various industries that are operating here. Whether it is in finance, it is in healthcare, it's in transport, logistics. The enthusiasm I think just went up because the cost considerations have become more favorable. And that brings me on to infrastructure. I think you know $500 billion Stargate project in the US.

Recently we heard France with a a roughly hundred billion euro investment in infrastructure as well. Are these countries spending too much then or or it depends on what time frame you ask. Yeah. I think if you know you fast forward a 100 years and you say what's happening today I I think it's very hard to say that it will be too much. Okay.

uh but these things you know they happen in cycles and uh certain points in time you may have gone ahead of the demand and um find yourself with over capacity this happens in many other industries and sectors too so I wouldn't be overly surprised if it happens but the long-term trajectory does suggest that the needs are going to grow and if you look at the Singapore context uh in terms of our data center capacity relative to our GDP relative to our population is already one of the most densed in the whole of Asia. And uh even with that sort of buildup, we're still looking at creating more headroom because if you want to be able to host the most sophisticated users of AI, latency does matter. Some of the AI workloads cannot be so easily uh configured out uh of um you know the the geographical location.

So even if you developed a very robust and extensive network of data centers in our part of the world, uh where the activity is most concentrated, you will still need to have the compute capacity. So does Singapore need to spend $500 billion on AI infrastructure? I I I can tell you that a lot of companies are spending that money. I think I was uh with um the former uh US Secretary of Commerce um Miss Gina Ramondo when she was here. I'm thinking is maybe September of October last year and we were trying to put together what the investments were like just in digital infrastructure and um utilities investments by US companies not including other companies. I think it came to about 50 billion US dollars in Singapore alone. That's that's the kind of scale they are looking at.

But I think actually if you look at the numbers elsewhere in the region, they are even bigger because the gap is wider and so there is a buildup that is necessary to try and close the gap. Yeah. And there's other things to think about, right? In terms of infrastructure with AI, it's energy, water, it's cooling, water. How do you address some of those challenges? We face up to these challenges. So we say that for data centers we have a green DC roadmap. Um there are essentially two strategies. One you use less energy. Two you use green energy and both are necessary. So in terms of how we use less energy um one very practical thing is that what are the standards for operating uh data centers in a tropical setting? Do you need it to be cool to such a great extent or the standards can accommodate uh tropical weather?

So that's one of the things that we're doing but we're not stopping there. We think that even at the algorithmic level there must be ways of writing them in a more green manner. So we are also putting in some research resources to try and make this happen. You were recently at uh the AI action summit in in France in Paris. uh I was there too. We didn't cross paths, but a big f focus of that of course was was around safety and regulation as AI continues uh to develop as well. What what would you say yourselves and other global leaders concerns at the moment around the development of AI? Because I've heard warnings around AI getting out of control, out of side of human control, uh AI being incredible biases written in to some of the algorithms and the models as well.

What are some of the the concerns that you share with with the global community? Well, it's really interesting because before um we had the AI action summit in in Paris um there was the AI safety summit in Korea in in Soul and before that the origination was in Bletchley. This was the AI safety summit and the conversation has actually shifted quite a bit because in Bletchley we were really talking about extreme risk. Yeah, you know what happens when um the AI is able to operate entirely autonomously and there's no longer uh humans in the loop. Existential risks were also on the table. The conversation shifted so significantly. I think in Paris it was centered a lot more around AI adoption. And with AI adoption, it doesn't mean that there are no risks. Near-term risks include, you know, risks of discrimination.

What if you have generative AI models that are not only inaccurate, they reproduce biases that we already observe in society? Uh there are other kinds of risk. Um for example, the cyber security risk of AI. The AI models itself can be susceptible to attacks. Um that's one type of risk. The other type of risk associated with security is AI being used to develop even more sophisticated tactics that can infiltrate systems apart from you know social engineering and implementing scams. So these kinds of near-term risk I think also deserve a lot of attention. It doesn't mean at all that we forget about the longer term risk. Some would argue that they are not so long-term. I think we keep a very close watch but at the same time our citizens want us to give them the reassurance that you are also dealing with near-term risk.

So the approach we take in Singapore is that we look at the kinds of legislation we already have and we ask ourselves to what extent they already uh address the kinds of risk that you will see with the proliferation in the use of AI. For example, child sexual uh abuse materials. um if you use AI it's very damaging to the families involved and um even without AI we already have laws against it. It's a matter of updating what we already have in law to say that if you use AI the law covers it as well but there will be occasions where the law does not adequately address the risk.

So in the context of elections for example if AI generated images or AI generated uh audio are being used to confuse the public and our democratic institutions uh come under great assault uh because of that then you perhaps need to put in place new laws and that's exactly what we did last year. We put in place a new law to address the specific risk of uh AI generated content being used during election periods. So, so is that the approach in Singapore to regulation then? Because we've seen in in the European Union the EU AI act and that's looking at a risk based approach and use use of AI. In China, they've gone also for quite specific laws targeting the use of AI in context such as deep fakes for example. Will Singapore be looking into a broad sweeping AI piece of regulation or more targeted smaller pieces of uh legislation?

At the moment, I think the approach we would prefer to take is to see how the risks materialize and the environment we have in Singapore supports law making. If you look at um the way in which parliament uh debates bills um I think credit to all the members it's on very solid basis mostly evidence-based as as as best as you can get. uh so which means that if you put forward a reasonable uh proposition to members of parliament you have the prospect of actually getting the bills passed and that's what we've continued to do. We assess the risk and very quickly try and come up with the legislative response to that go to parliament explain to our legislators why this is so important.

It sometimes results in what may appear to be a patchwork of you know different laws and regulations but the tradeoff is that they are much more targeted and they are much more specific. I would say that in time to come there may be a need for us to reorganize some of um the bills and and the acts that have already been implemented. But at this stage we are prioritizing speed of response and being targeted in terms of the legislative measures that uh you know address the problems that arise with uh digital development and um and of course uh AI implementation too. you you mentioned that the sort of three AI AI summits that have happened and how the conversation slightly moved on but but the whole point of them was to try to bring together countries to discuss frameworks and uh legislation in a global manner.

M um there was a a letter that was signed by many of the attendees except for the US apart from the UK as well and and it felt to me that uh as this conversation towards global frameworks continues there are so many divisions yet. So how do you go about how one for you how important is a global framework around AI safety and secondly how do you go about getting that when there feels to be a lot of competition and tension right now between nations? Well, you know, the way with these meetings, I I wouldn't be overly perturbed, you know, if um a declaration didn't have all the signatures of everyone who attended. The reason is quite straightforward. In your organizations, if there is a commitment you have to make publicly, then there will need to be an internal process to see whether you are able to live up to those commitments.

And there will be countries who look at it and say that there's too little time. we need to talk to our capital, we need to talk to all the different agencies. So from that perspective, I completely understand why there are occasions where a declaration, you know, did not get signed. It it all partially depends on, you know, when this draft was surfaced. So I wouldn't be too concerned about that. Having said that, I think at some point global frameworks are going to be critical. If you talk to businesses operating you know in multiple jurisdictions and in the digital domain the parocity means that you are in in many m many markets and to have to deal with a different set of rules each market that you go to. The lack of interoperability is a great impediment to business expansion.

It is also difficult to ensure that the citizens are of each of these jurisdictions are protected to a comparable degree comparable degree and I think over time these frameworks are going to be increasingly important. The difficulty is that the more you want people to sign on to it, um the tendency is that it becomes more bland. It becomes, you know, where the the it's it's it's not perhaps at the right level of ambition because it has to satisfy everyone's, you know, sense of what's right and what's wrong. So this is I think an ongoing conversation. I don't expect that uh it will come to any conclusion anytime soon. However, I am very hopeful and I'm optimistic that there are like-minded countries that are very keen to continue to move forward together, particularly in the area of AI safety.

Just because we want to advance AI adoption and the buildup of AI ecosystems in all of our countries doesn't mean that we throw caution to the wind doesn't mean that we no longer care about AI safety. If you want this uh to be truly transformative, we have to give our citizens the assurance that it's being implemented with great sense of responsibility and also safety. Minister, it's hard to talk about technology without talking about the rapidly changing global political landscape, regulatory landscape as well. I was at a Mobile World Congress very recently on stage with the CEO of Mistral AI, one of Europe's leaders in large language models, Arthur Arthur.

and and he vocalized something that's kind of been bubbling under the surface for a long time in Europe and he said we think that it's time for European companies to to look to reduce their reliance on American technology which is a a discussion that's been being had in the background but the first time I'd heard it publicly vocalized uh by a European tech company. How and it comes at a time of course which where it looks like the US is going US first uh when it comes to technology and many other areas as well. How do you navigate that landscape particularly as it relates to what you were talking about at the start of the conversation around resiliency? You know, this idea of over reliance is not new at all.

Um, even as an economy that was developing through the 70s, the 80s and the '9s, we looked at Singapore and say that, you know, are we relying too much on a particular sector and how do we diversify our economy? uh I believe the first recession that we have P's independence was in ' 8586 and it was very prominent in our minds that look you know if you're only depending on a few engines then when the chips are down you you you may not you know have enough um to keep going. So the idea that we would diversify your economy, you would diversify the kind of activities that people were engaged in and within each of those activities um industries, you would want to have a diversity of players in those um domains. I think that is quite you know embedded in our way of thinking. So that's not new in any way.

Having said that, I think you also have to recognize that um some countries are very advanced in uh the areas um um of excellence that they have. AI is one of them. Quantum will be another one. We will certainly try to build up our own capabilities. But it is not at all realistic for Singapore to say that you know you want complete independence and you know you want to be self-sufficient. We decided a long time ago that that was not an idea that was going to work for us. And we continue to try and engage with all of our partners. Doesn't matter where they come from um at the enterprise level and and in many different ways. We look primarily for comp for a combination of performance and also security as well as resilience and that's that remains the fundamentals that will drive decision making.

So, I just wanted to talk about the future and in this age of AI, how you're thinking about fostering the next generation of tech leaders and things like reskilling as well, which are becoming uh increasingly important. I think I saw a story in the last couple of days that schools in Beijing will introduce AI courses in in primary and secondary school as you know, countries around the world think about education, think about reskilling as well. um what what are you doing at this point uh when you're thinking about the younger generation here and how to build them up into a you know a nation of of tech forward and forwardinking people. We think not just of the younger generation, we think of every generation because they are going to be part of our workforce for uh a lot longer.

So um we've paid attention to what we call pre-employment training. So you know universities, polytenics, uh institutes of technical education those are very important to us but continuing education training is probably going to be even more important. The c the ability of a young mind to absorb what's of current relevance is very strong. So we don't need to do a lot you know to encourage young people to be savvy in the digital sense. We don't need to do a lot to persuade them that uh it's kind of useful to learn coding. Um and even if you were not able to do the most sophisticated kind of coding simple stuff you know block coding is not that difficult and they pick it up. So I I don't have a great big worry there. I also think that uh our educational system uh has been tried and tested over many years.

We've got teachers who are very invested in helping their students prepare for what we call 21st century competencies. What we did interest uh interestingly decide was that um instead of instead of us trying to impress upon these educators what's relevant in this new era, it's easier for us to invite them to a setting like this and listen to people talk about how technology is changing the world. The educators themselves bring it back and redesign the curriculum accordingly. So we have a program that we call the smart nation fellows to do just that. In other words, don't attempt to as a non you know expert in curriculum design change curriculum. You need the curriculum designers to uh be embied by it, embrace it and change it for themselves.

But I wanted to shift a little bit of attention on to continuing education and training people after they leave their workforce. If we think that what they left school with is going to last them that their entire working lives, I think that's not realistic. I think in their uh you know 40, 50 years active in the workforce, they might well have to pivot three, four, five, six times, maybe even more. Who's to tell? But the idea is to give them the support and an infrastructure that supports that makes it possible for them to make this pivot.

One program, for example, that I'm particularly happy about is that once you reach uh the age of 40 in Singapore and you decide that you have to go all the way back to school and acquire qualification, formal qualifications all over again, we now have a scheme that can support you going back to school for two years, up to two years. You can do it on a full-time basis. You can do it on a part-time basis. What the country, what the state decides to do is that your financial commitments, we will help you fulfill them. meaning that not only when you go back to school, there are, you know, subsidies for course fees, there's actually a training allowance that goes to you. Um, it's in Singapore dollars, it's $3,000 a month. So, if you take two years for this program, that's $72,000 for you to reskill, retool, and be ready for the future.

That's how we're looking at it. Uh, that's a great point to to leave it on. Thank you for all your insight, Minister. Thanks for joining me on the podcast today. Thank you. That's it for this episode. If you want to contact us, email us at beyondthe valley@cnbc. com. Please follow and subscribe to the show and you can leave us a review. We'll be back soon for another episode of Beyond the Valley. Goodbye. [Music]