00:00You've been around the world right now, really selling your devices.
00:04Tell us a little bit about why your chips are such a differentiator from, say, the likes of NVIDIA
00:10that has such a big demand from hyperscalers.
00:13Yeah, exactly, you know, I mean, let me explain one example, you know.
00:17For example, you know, Hyundai robots, they are, you know, are using our chips.
00:22Based on our chips, they can, you know, perceive, decide, and react to the physical world.
00:29But in the past, these capabilities, you know, normally, you know, require the GPU technology,
00:36which consume a lot of power and generate excessive heat, you know, making them very unsuitable
00:43for the on-device applications.
00:45So instead, I mean, robots can use our chip because it consumes a lot of 5 watts.
00:50But NVIDIA chips can consume, like, 40 watts or over 100 watts.
00:56This is a really, really high power consuming thing.
00:59So that's why, because our, you know, low power, you know, technology, the robots can, you know,
01:06automatically, you know, you know, process any kind of, you know, labors.
01:11So that's why, in my opinion, so that's why we use, you know, the current technology
01:15like Samsung Foundry 5nano or 2nano technology.
01:18So in my opinion, so right now, it is a time to use the NPU.
01:24This is a low power, you know, technology for AI applications.
01:29Yeah, so it's really about allowing AI functions to run directly on devices and robots as well.
01:34And in-device AI and embodied AI have been really key themes at CES for this year.
01:39But to that end as well, it's been perfect timing for you to launch this new chip, the DXM2.
01:43So tell us about that.
01:47Yeah, you know, one of the biggest challenge in AI today is the high cost and power constraint
01:55because of, you know, the current data center centric model.
01:59So to address this challenge, we designed a new chip.
02:03It's a kind of, you know, generative AI on-device chip.
02:07So it process up to 200 billion parameter generative AI locally right on the devices, not in the data centers.
02:16So by using this, we can, you know, process like the TGPT service without the broadband network, without the data center reliance.
02:25And you're partnering in Samsung and using their foundries to actually manufacture this.
02:29I'm really curious about your partnerships because you're also partnering with Baidu, but also your partner with Dell here in the U.S.
02:36So what are your expansion plans, especially in this very tricky geopolitical environment?
02:41Oh, yeah, there are some regulation in the geometry issue, but it regulates around above the 100-tops capacity of the AI chips.
02:51But our product is under 100, 100-tops.
02:55Like, you know, DXM1 provides 25-tops and DXM2 provides 80-tops.
03:01That's why.
03:02So we have freedom to sell our products all over the world.
03:07It is, though, coming up against a push, and you don't just see it from China.
03:11I mean, there's a push here in the U.S. as well to reshore, onshore, semiconductor capabilities.
03:17So where do you see the biggest strengths for you as a business moving forward?
03:21And are you concerned as well about that push for self-reliance and self-sufficiency, also restricting sales of your products?
03:29Yeah, I mean, you know, right now we cover all over the world, like in China, U.S., South Korea, Taiwan, Japan, and even Europe.
03:38So, you know, in China, you know, there is huge movement about on-device AI, you know, services.
03:45That's why I can see many players there.
03:48At the same time, U.S. is focusing on the, you know, AGI or ASI-level algorithm, you know, technology.
03:55So, also, we can see that many, you know, manufacturing and industry in Japan and South Korea, including, you know,
04:04or Panasonic, LG, Samsung.
04:07Do you think you can actually go against NVIDIA?
04:10No, no, in my opinion, you know, it's like Intel and AI technology.
04:16Their technology, you know, for the high performance.
04:19But we have covered low power in the small devices.
04:22Like, you know, based on GPU, they control around 100 watts.
04:26But we cannot use that chips for the on-device, you know, applications like, you know, robots or battery power.
04:32But it's not the trend, but we're going more towards on-device.
04:35Exactly.
04:36So, you know, normally experts, you know, forecast that, you know, the major part of the AI will be the physical AI.
04:44That means embodied, you know, generative AI capability.
04:48So, in my opinion, in the future, we can see a lot of, you know, handheld devices can use the generative AI.
04:53We know you've been looking at a listing.
04:56How is that going, that process?
04:59And would you consider, instead of perhaps pursuing an IPO, also looking to make yourself an acquisition target,
05:04just because we've seen the likes of AMD, NVIDIA, Qualcomm making several high-profile acquisitions in the last few months?
05:12Yeah.
05:13So, also, we got some acquisition, you know, requests, but we, one year ago, we rejected it.
05:18That's why I cannot share the information.
05:20Anyway, for DBEX IPO, we are taking very thoughtful approach while seeing the strong interest from the capital market.
05:30Two months ago, I was invited into the, you know, International Day at NYSE.
05:36And at the same time, you know, we discussed with the NASDAQ and the Singapore Stock Exchange as well.
05:43We're really grateful for that heavy interest.
05:46But, you know, our top priority is on the, you know, business strength or business foundation.
05:55So, that's why we will choose the IPO timing very carefully when we can have the strong business foundation.
06:03Could that be next year?
06:05Maybe two or three years later.
06:06Two or three.
06:07So, when you say you actually had interest from people that wanted to acquire you, what sort of businesses are we talking about regionally as well?
06:14Because they were, you know, normally semi-contact companies.
06:17You know, right now, on the device, you know, semi-contact, there are conventional players in this area.
06:22So, like NSP, Renesas, and TI, they are maybe interested in the, you know, on the device, you know, chip, you know, designs.
06:30That's why.
06:31That's why.
06:32But, for us, I would like to see the change in this world without any, you know, IPO and, I mean, the inclusion.
06:38What's sort of the, I guess, the valuation then that's also being given to you in the market?
06:44Around 1 billion.
06:46So, we are talking with investors in the global side or domestic side, so.
06:51And how much is that?
06:52Do you expect to see that growing this year just because you've had such an explosion in demand for this sort of technology that can power in device AI?
07:00Where do you expect to be at the end of the year?
07:02I mean, you know, we can receive, you know, I mean, we are not quite, you know, sensitive to the investment.
07:07Actually, we have, you know, some capital money.
07:10And at the same time, Samsung Foundry supports our chip designs and our development.
07:15That's why we are more, you know, consider about the partnership in the world so that we can have the variation of the products.
07:25What is your client portfolio like?
07:27Because we talked about Dell, for example.
07:29I know that you're also working with Hyundai Robotics Lab.
07:31What are some of the big names that you're supplying to right now and what does it look like?
07:37Let me share the example of our, you know, mass production.
07:41So, for example, we are collaborating with Baidu on the large-scale deployment of AI-enabled IT services.
07:49So also now we are expanding to the physical AI, you know, applications such as robots, drones, and physical, I mean, smart factories.
07:58So, in parallel, the mass production of commercial robots from Hyundai Motor Robotics Group is set to begin this year.
08:09At, you know, POSC DX, our chips are being deployed across multiple industrial automation applications.
08:17So, there are over 50 customers in the world.
08:21So, there are so many different stories.
08:23So now, that's the segurança companies that come to us.
08:24We want to share these stories.
08:25Thank you very much.
08:26Thank you very much.
08:27Thanks.
08:28Thank you very much.
08:29Thanks.
08:31Thanks.
08:32Thanks.
08:34Thanks.
08:35Thanks.
08:40Thanks.
08:41Thanks.
08:43Thanks.
08:44Thanks.
08:47Thanks.
Comments