Stuff Meets Yenchi Lee, MediaTek VP & GM for Smartphone Chipset Business

MediaTek's Global Strategy, AI, and Gaming Ecosystem

It’s easy to think of the silicon wars as a one-horse race, but while you weren't looking, MediaTek has been quietly powering a hefty chunk of the planet's pockets. Now, the Taiwanese titan is done playing the quiet underdog. We sat down with Yenchi Lee, the big cheese (VP & GM) of MediaTek’s Smartphone Chipset Business, to find out how they plan to shake off the 'Asia-only' label and properly crack the premium market in Europe and India.

From chipset capabilities that allegedly leave laptops eating dust to optimising your next Genshin Impact session with dedicated gaming collaborations, MediaTek is making a serious play for the throne. Whether you're an AI aficionado or just want your phone to be smarter than you are, here’s why your next handset might just have a Dimensity heart beating inside it.

What role does Asia and Europe play in MediaTek’s global strategy?

I think our main business right now is still in China, because our customers are mainly in China but it doesn't mean that they (Chinese OEMs like Oppo and Vivo) only do business in Asia, right? Their phone sells globally. They also have a great presence in Europe, India and many other places. So, I think the strategy for MediaTek is that we do not want to be seen as a China or Asia-only company. We are actually adapting our plan for the global scale. And that's why I think we are participating in a lot of events in Europe, like MWC (Mobile World Congress) and this event (Oppo Find X9 Series Global launch in Barcelona), and we're also hosting events in India, and around Asia. I think Europe is an important market for us because we can sell into more premium segments as well. But right now, a lot of devices are still from mid to lower end. But we are trying to get into the premium segments in Europe. I think you will start to see those things coming along with our strong partnerships with Oppo and Vivo, and others.

Is MediaTek doing something for AR and VR headsets, like Meta Ray-Bans smart glasses and looking into that product category as well?

I think we are certainly looking into that. Right now, we do not have a specific product for those particular devices, but we’re definitely monitoring it along with the requirements and the necessary technologies for it. And, we do have the capabilities and technologies to go after that market so eventually, we’ll try to find the right customers and the right devices that we go after.

How much impact will the NPU have on the smartphone market in the next decade?

If you look at the computing capabilities of our flagship smartphone chips, like our the Dimensity 9500, you could really compare the computing capabilities to an AI PC. We actually have a much higher computing power than AI PCs today. The AI PC today is probably about 40 to 50 TOPS, but our Dimensity 9500 is actually 100 TOPS. So if you look at the computing capability, I think we are not less than AI PCs. In fact, we actually do more than what an AI PC can do. It's just that when AI will actually take off on the smartphone it will come from the ecosystem, user scenarios and the user experience side of things. I think Google is making a bet on that. So we are also following those spaces and working with Google.

Since everyone has a smartphone, there will be unique user scenarios that are tailored to your own preferences to understand your needs. For an AI on a PC or the NPU on the PC, it’s more focused on productivity without really having the need for understanding the person's needs. It doesn’t have the always-sensing capabilities of a smartphone. This is also a very unique feature and no other device has, maybe the smartwatch, but the smartphone actually is in a very unique position.

A few years ago, the video edits I did on a computer are now happening on a smartphone like background removal and noise reduction. Where's the AI performance of smartphones headed in the next coming

I think the computing capabilities for the NPU is probably not as important as the power efficiency of an NPU. While the AI models are becoming increasingly intelligent. I would say given the same size of the model (file size), you can pack in maybe two or three or even five times more knowledge into it now, right? You don't really have to run big language models to get that intelligence. You can actually run the same 3 billion, 4 billion size of the model, and get maybe five times better intelligence than before. So really, the key is not about increasing the computing capability of larger models; it's more about how you run it more efficiently. That's very key. And for smaller form factors like smartphones, having it run like 20-30 billion models probably doesn't make sense because it's going to burn out your power.

I think running the models more efficiently is the key and for that to happen, sometimes you need always-sensing capabilities that really run some very small, tiny models and do some AI training on your phone so that your data is stored on your phone; it doesn't really go to the cloud.

We are looking into how we can make the smartphone chipsets' NPU capabilities available to third-party applications. We want the AI to run on the device but there are also situations where you still need to go to the cloud, however we are looking into how to run these models on the device so that they can create new user scenarios and use cases. For example, when you don't have the connections on a flight, you cannot do too much if the application needs to use the cloud to do certain things. Can we do that on the airplane or when you don't have good connectivity?

Do you think the tablet as a market is something that you would tap into with the same premium chipset?

I think we are actually doing that already. So if you look at the Samsung Tab S11. They’re actually using our 9300 Plus and 9400 Plus. So I think we are already putting our Dimensity 9000 series into tablets, and not just the 9000 series but 8000 and 7000 series are in the tablet as well. So we are actually tapping into those adjacent markets with our smartphone chipset, and there are certain customers that want to do industrial applications using the smartphone chipset. So we may as well just enable a lot of IoT and AIoT devices based on the smartphone chipset. For that, I think we will leave it to our customers to have more creativity for those different types of applications.

There was a rumor of MediaTek and Nvidia doing something. It hasn't shown up yet in terms of a chipset for a handheld gaming computer, or a gaming chipset, is that a thing?

Actually, this is my first time hearing that! But we do have a lot of collaborations for MediaTek in automotives (with Nvidia) and computing. So, I think we do have tight collaborations with Nvidia so you'll see in the future, I think.

Are you talking to game developers to optimize the MediaTek chipset for smartphone games?

I think we do have a team working with gaming studios on the chipset so they know what our chips can do. And I also talk to them on how they can optimize the games for the MediaTek's chipset. We work with the big names such as Mihoyo that made Genshin Impact, and also Tencent Studio, and many others in China. So I think we have very close collaborations with those gaming studios. They understand the capabilities of our chipset and we regularly tell them what we put into the chipset, and we tell them the best practice and the tools to work with them to optimize for our chips.

Do you know we have our MediaTek Dimensity Developer Conference (MDDC) in China? It's been held two times. So next year will be the third time. This is the biggest developer conference where we invite all these third-party developers to the conference and tell them about our capabilities that they can leverage. We also show them the examples, sample codes, best practices, and everything so that they understand. It's pretty much like WWDC.