Maurice:

Hello everybody, and welcome to another edition of C&F Talks. It's a great pleasure to have with me today, Esther Wong, who's the Founder and CEO of 3C AGI Partners in Hong Kong. And Esther's going to be speaking at the AI and Digital Innovation Summit on the 1st of July, which is being held as part of City Week at the Royal Garden Hotel. Esther, welcome.

Thanks for having me.

Very good to have you with us.

As AI capabilities mature rapidly, particularly with developments in agentic AI, what do you see as the critical success factors for realising AI's full potential across the various sectors where it's most important?

This is a very good question. I think AI development in the last few years has gone leaps and bounds. As a matter of fact, we call it the AI 2.0 era, where AI 1.0 is when you know that it sort of works, but not everybody uses it. AI 2.0 is when the birth of the birth of ChatGPT. Everybody, I'm sure you're using it, I'm using it, a lot of our friends are using it. So, I would call it the true democratised era of AI, and that literally just happened two years ago.

And precisely because it happened so quick and so soon, I do believe that a lot of the infrastructure that is needed to support this kind of growth is not really in place yet. I mean, the reason why NVIDIA share price went up 30 times in five years is really because it happens to be the device that can do AI training, the algorithm model training. However, 2025 is the year that for the first time in the history of humanity, more AI compute is going to be used for inferencing than for training.

What do I mean by that? So training is when the 20 or so big companies in the world, they train the models, right? But not everybody trains the models, only the AI super hyperscalers do. But for inferencing, we have 8 billion people on this planet. And imagine if all of us use this AI, that's inferencing. So, you can imagine the ultimate demand for inferencing is going to be literally a billion times bigger than training.

So, my perspective is that the AI infrastructure that we have currently is completely not ready for it, far from ready for it. So going back to your question, Maurice, the biggest opportunity that I see in AI right now is actually on restructuring or rethinking the AI inferencing or AI infrastructure so that it can be better prepared for the age of inferencing.

Yeah, and I think that's particularly true when you consider that alongside this, there's the electrification of so much of industry as people try to combat climate change and the energy requirements for AI on top of that. I mean, the stress on the grids around the world is going to be huge. And the time it takes to rebuild and strengthen and increase the capacity of those grids is quite a break on what you can do in these other areas.

But turning to developments in China, the rise of advanced models like DeepSeek has sparked global attention. What are the broader implications of China-based foundational AI models? How do they shape competition and collaboration across the AI ecosystem?

Well, this is a wonderful question because I think there are three reasons why DeepSeek is so impactful in my mind. First and foremost, this is the world, I will argue, is actually truly open source. So, some would say, oh, but Lama is also open source.

But then the way that DeepSeek open source, they really tell you all the minute details on how to replicate it. They actually do expect you to replicate it around the world, right. So basically, 24 hours after the first R1 model was open source, there are literally hundreds of variations of Hugging Face already.

And those variations are from all over the world, including UK, Europe, South Africa. So basically, it's the first model that people can truly replicate relatively easily because they give you all the ingredients and step-by-step details, unlike Lama. So that's number one.

Number two, it actually has profound implications on how to do top-notch model on relatively speaking, not-so-top-notch hardware, right? Because DeepSeek is not trained on the latest GPU, as we know. So basically, we often these days think about AI as a race, right?

Between, let's say, the two biggest nations, China and US. But what we don't realise is that there are also 200 countries in between. And not every country can actually afford the largest cluster of GPUs, the most amount of compute power. So, what about those countries? Are they being left out? The answer that DeepSeek provides is no.

Because as long as you have some kind of relatively rudimentary structure and chips, you can actually still run the most advanced model. So that's the second implication, right? The third implication is that it's coming from China.

I mean, to be honest, right, if DeepSeek is coming from, say, a Stanford lab, I mean, the shock value will not be as big. But the fact that a Chinese company, relatively not known outside of China, as a matter of fact, not so well known outside the AI circle, they came up with that and its completely open source. I think it's quite impactful.

And the implication is actually beyond China. Because as I mentioned, for the first time, people with not-so-great compute power, they can actually replicate a model. Can you imagine the amount of potential brainpower unleashed as a result of that, right?

So, it's not limited to China at all, number one. But number two, it definitely put the China AI advancement sort of on the map for a lot of investors and users alike. Because previous to that, people just would assume, and quite rightly so, that usually China would have more users.

So, application-wise, it's more advanced. However, when it comes to the underlying model-wise, the US tends to be more advanced. So, DeepSeek breaks that impression.

And I think that it actually gives a lot of local confidence, so to speak, by the local developers, and also the application developers, and also the users, and also the government, right? You can see that the government is actually also having a renewal effort to enhance overall to support the AI sector. So, I think that's actually a pretty positive cycle locally, and globally, for that matter.

Yeah, it's remarkable, isn't it?

As you say, China's AI development is progressing at a very rapid pace. But I guess it also faces unique regulatory and governance challenges. How do you assess China's strengths and vulnerabilities in the global AI race? And what ripple effects do you see internationally, particularly perhaps in Europe?

Well, I think when you call something a race, it means that there's a winner and there's a loser, right? So however, I feel like in the world of AI, the advancements happen every day, right? And it's on top of each other. And the development is also building on top of each other. So, I see why people like to build it as a race, but it's not finite. In fact, the progression, it's actually infinite.

I think the advancement in AI. So, with that in mind, I will answer your question. What is China good at and what are its limitations? I think China has a large population, right? 1.7 billion people. So, you can see that in China, because the population is quite big.

So, they have a second move advantage when it comes to mobile internet. So, you see that there's a lot of wonderful mobile apps that came from China and can spread to all over the world, right? For instance, you have the TikToks of the world and you have the Alibabas of the world, all come from China.

So same thing is happening in AI. Because with AI, you need four things. The number one ingredient is obviously the data. So, China has a lot of data. Number two, you need to have the hardware that China is not quite there yet but definitely catching up. Number three is algorithm.

Algorithm is the people, right? So, China obviously has the world's biggest number of STEM students. And you can see the result is we have guys like DeepSeek that came out, which is also quite impressive.

But number four is often overlooked, is the feedback loop. So, AI without application is nothing. So, with China, you actually have a relatively short feedback loop.

What do I mean by that? If a product is good or if it's bad, you need to get some feedback and then fine-tune it, right? Just like an Algol. So, in the case of China, because the population is so big and they tend to be young, and they want to try new things. So, you can fine-tune the quality of a product relatively quickly, right? So, the feedback loop is very short.

And that is a unique advantage in China compared to the rest of the world. It's not just that people are very smart, which I would like to think that the scientists are obviously very impressive. But one of the major advantages is because the population, the feedback loop is very short.

So that's why you see that the world's top 2C AI APPs, more than half of them actually have the origin in China. So, you try in China first and then you can actually go overseas. And that obviously benefits everybody, including Europe as well.

Yeah. Yeah, you're absolutely right. There are some unique advantages, I think, that China has.

But as the geographical landscape intensifies, especially between the US and China, should venture capital be more actively looking towards AI opportunities in Europe and Asia? And what factors should VC's weigh when evaluating regional innovation systems?

Hmm. This is what I sometimes follow with every risk, every crisis, there's always two sides, there's always opportunity. So unfortunately, we're living in a world that geopolitics dominate a lot of headlines, and it has implication in everyday lives, and it has implication in investment.

But Europe, in this particular case, it's actually, I would argue, is a beneficiary of that. Okay. Why do I say that? Because as recently, you can see, especially intensifying the last few weeks, a lot of the foreign talents studying in the US universities are studying PhDs. And the degree of how welcome they are is being questioned, so to speak. So, a lot of the opportunities is being left with, can they go back to their original country?

Sometimes they stay in Europe, sometimes they are in South America, sometimes they're in Asia. So that's an opportunity for these homegrown talents to stay at home, which is not necessarily a bad thing. However, I do feel a pity, because when you think about the Nobel Prize winner, 40% of them is actually foreign born, even though they have a US citizenship.

So however, when America does not assume the place to be the black hole for all the talents, the talents are still smart, they need to go somewhere. So, I think Europe can be a huge beneficiary of that. And I think what is stopping Europe in the past, from my personal experience and from talking to friends, there are two things, right?

Number one, the appetite, the pocket, the depth of the pocket for VC money is not as vast as those in US and I would argue not even in China, because China has a very deep pocket, but they have other sets of issues. But in Europe, it sounds like it's great opportunity if I were a VC to invest in Europe, because the same company used to go to US, now they would have more thoughts and maybe they think about staying homegrown instead. That's number one.

Number two is the regulation. My understanding is that in Europe, there's a lot of regulations that actually you need to get a lot of approval for some new technology to be tested. So, in that sense, I think Asia and Europe can have a very good synergy.

Because in China, as I just mentioned, China is a place where we crave experiment, and we strive in new business models. And so sometimes the stuff that works, the business model that works in China can be potentially export to globally. And of course that will include Europe as well. So that could be a very good marriage between these two regions, for sure.

Yeah, absolutely. Fascinating time. So much potential in so many areas.

I think we've run out of time, sadly, Esther. So, for our viewers, it'd be great if you could join us in person at the AI and Digital Innovation Summit. As I say...

I look forward to that.

Excellent. On the 1st of July, it's being held as part of City Week at the Royal Garden Hotel in London. Details are available at the website and registration forms and so on, www.cityweekuk.com. So please do go and have a look at the further details about the programme over the three days.

Just remains for me to say, Esther, thank you so much for joining us today. I very much enjoyed it.

Thanks for having me. Thanks, Maurice.