- 1 week ago
Category
🗞
NewsTranscript
00:00How are you going to unlock that value in AI?
00:03Yes, thanks for having me today.
00:05I feel very excited to join Baidu since last July.
00:09After joining, I realized so many pressures and so many treasures, I'd say,
00:14sitting on the Baidu's balance sheet.
00:16And many assets are waiting to be unlocking the value.
00:19I think the backdrop is, if you remember, early part of 25,
00:23Deep Seat come out, and right now I think all the focus are coming from
00:27the training of the foundation model to applications.
00:30If you think about today, the AI softwares and the robotaxis and cloud and the GPU
00:37are all linking to applications.
00:39And I think Baidu are sitting in a very good crossroad
00:42that can unlock in our historical technology assets
00:46to more like the monetization on the business models and all those things.
00:50I think one thing first I want to mention is about robotaxis.
00:52I think the big advantage for us, I think we do enjoy
00:56the full stack of capabilities from data foundation model
01:00and also the algorithm as well.
01:02And we are doing the robotaxis today.
01:03If you remember the numbers, we deliver almost a quarter
01:06a quarter million rides per week, which is actually very similar
01:10on the level with Waymo globally.
01:12And that is actually a very good user case
01:14as we use our AI technology.
01:16So where are the main priorities?
01:17I mean, we'll talk about chips.
01:19Don't worry.
01:20We're going to talk about Kunlunxin and the chip
01:22and also Ernie Bot and the AI side.
01:24But since you went there first, let's talk about robotaxis.
01:28Apollo Go is essentially a service.
01:31You've gotten the license in Dubai.
01:33I understand you're going to start doing testing in London as well.
01:36Yes.
01:37Hong Kong is underway.
01:39But what's the most important aspect to get it right when you're competing
01:44on a global stage with Tesla and its CyberCab as well as Waymo from Google and others?
01:52They're deep-pocketed.
01:53Yes.
01:54The data is very important for robotaxis.
01:56So it is not a new business for us.
01:58We have been working on this business for almost 12, 13 years.
02:01We have a huge number of mileage of all experience of the driverless experience.
02:06I think that's actually one major advantage of us.
02:09The second part in terms of the model capabilities, it is not only about building rules in a model
02:14or in a car.
02:15It's really about building the entire AI capabilities from using the foundation model
02:20to linking the algorithm to the sensors and the lidars to making that very important connections.
02:26And the third part is operational excellencies.
02:28Because if you think about the waiting times and the distance between the car drop you off
02:34with your house and, for example, the user cases and the user experience on the mobile apps,
02:40these are all very important investments actually we are making on that.
02:43And the last point is very important.
02:45If you think about Google's Waymo, Google also has a Google map.
02:48And also Baidu has a Baidu map, which is actually the leading map products in China as well.
02:53So we have a lot of information regarding the point of interest and how the traffic looks like.
02:57And that will improve a lot of user cases.
02:59And, for example, people using Google and Baidu map, they can also call a car.
03:04And the most important thing is, I think, given the past few years,
03:07the EV car supply chain in China is improving quite a lot.
03:10So we can produce a car as a very cost advantage model in China
03:14and using that cost advantage to actually working with the global markets.
03:18So today we are seeing great benefits when we have a partnership with Google, with Lyft, with Uber and with Lyft globally,
03:26especially in Europe countries and the Middle Eastern countries.
03:29We enjoy a lot of headroom in terms of potential uplift of the margin profile
03:33because we have the cost advantage, but also we enjoy the markets globally.
03:37Right. So if you combine all your AI endeavors, including the robo taxis,
03:43but including as well the chips division, which is likely to soon go IPO in Hong Kong.
03:49In addition to your large language model, the various multimodal Ernie,
03:53I think it's Ernie 5.0, Ernie Bot right now.
03:56How much combined has Baidu invested in AI?
03:59When are you going to be profitable as a contributor?
04:03You know, it's increasingly contributing to revenue.
04:05But when can it all combined be contributed to profitability?
04:09Yes, sure. It's a very important question for CFO.
04:12So in the November earnings call last year, we actually disclosed one numbers.
04:16Since earnings came out back in 2023, we have invested over 100 billion RMB in terms of AI investment,
04:24which is a huge investment just on the one side.
04:27On the other hand, we do see a great return on that.
04:29For example, right now, our cloud revenue increasing about 128% on a worldwide basis in Q3 for the AI cloud.
04:37And also, we are seeing over kind of 200% growth on AI search transformation.
04:43So for example, last year, we only have 3% of the contents are generating for the AI for our traditional search.
04:49But right now, this number is increasing to about 18%.
04:52And we're seeing kind of three digits increasing on the traditional search to the AI new search.
04:57And also for the Robotaxi, as I mentioned, it's actually delivered over a quarter million drives every week.
05:03And we're seeing that number is on accelerated growth paces.
05:07So I think the investment, we are seeing the good trends of the monetization and the computing on the return.
05:13How do you see the difference between how AI is evolving in China versus, you know,
05:17what we've all been following very closely in the MAG-7 and what's happening, OpenAI and others in the United States.
05:24Essentially, there are concerns about over-allocation of capital and not getting the kinds of returns.
05:30So there could be potentially a bubble.
05:32China, there are bubble concerns as well.
05:35But there is a bifurcated AI space in China as well.
05:39The big boys like you and Tencent and ByteDance and Alibaba.
05:44And then there are new up-and-comers that are hitting the market in Hong Kong like Minimax and others.
05:49Those LLMs that are starting to make a dent.
05:52But you guys have the capital base.
05:55Is there a bubble forming as well in the investment in China because of the over-investment?
06:02I want to call it two tails of the city.
06:05Obviously, I think if you want to decompose the drivers, I think there are four things.
06:11The data, model, computing power, as well as applications.
06:16So I think if you analyze these four different things, U.S.-China does have a lot of different contexts.
06:21For example, right now in the U.S., there's a lot of investment on infrastructures.
06:25But today, if you look at electricity, cable network in China has already been built in the past few years.
06:31So if you really want to compare on the IOI or ROE on the infrastructure side, I think China does have a certain advantage on that.
06:38And also in the past ten years of the mobile internet age in China, there's a huge accumulation of data.
06:44The data become important for today's inferences because a few years ago, everyone was competing for the foundation model.
06:50So to your question, I think in my view, any company have a closed loop or full stack of the computing power, data, and ownership of application user cases will be sustainable on the AI.
07:03That's probably not about it.
07:04How much more needs to be invested in your chip division?
07:07Because the Kunlunxin chips essentially power the servers in the data centers, right?
07:13Which is a real, I guess, strong point of NVIDIA.
07:17And I think the mandate really from authorities is China must be self-sufficient there.
07:22So how much more needs to be invested in Kunlunxin and the chips that power those servers going forward?
07:29I think that's a very important question.
07:32But my view is if you want to rank the number one factor, I think it's the talent.
07:38Because we are a designing company, just like any of the chip companies in the U.S., we're designing the chips, but we're producing the chips with partners.
07:47So the investment into the key talents and the team and finding the ecosystem partner is an important investment from both strategy and R&D perspective.
07:57So that takes capital as well to hire the right people and to spend the money wisely in what some say is a bit of a frothy environment.
08:07So how much do you need to raise in this IPO in Hong Kong?
08:11Yeah, I think it's pretty early to say that.
08:13$2 billion is the number.
08:15Some say, I've also heard $1 billion.
08:17So let me put it this way.
08:18So we made a public announcement last week.
08:20So we are making good progress in terms of filing for the IPO.
08:24So hopefully, I'm confident in terms of the market.
08:27If the market can keep the way and enjoy and see the benefits of AI and computing power, I think our coolant chip will eventually unlock that value to the market at a fair level.
08:37But again, my point is because Baidu today have a full stack of ability, any investment getting to the coolant machine will be linking to the cloud.
08:46And our end customer eventually using our cloud to use the application and generating real results for them.
08:52I think it's a full loop of the monetization path, which is putting Baidu at a very unique competitive advantage with other peers.
08:59Because other peers are only on one side, but we are actually the full stack on them.
09:03Let me come at that question from a little bit of a different angle, because I can tell you can't really tell me the details of that IPO that is potentially upwards of $2 billion in Hong Kong.
09:12But the Hong Kong market is hot for AI right now, and it's not just the hardware players, which saw huge gains, whether it's B-Ren or Moore's Threads,
09:25but also now going to like Minimax, the large language models and this thing.
09:29So wouldn't it be prudent to list now while the market is still hot as soon as possible?
09:35Sure.
09:36That's a good question, which I can answer.
09:39Given the growth trajectory of our trip department and cloud department, I think the timing of listing will be very reasonably decided by the pacing of the business.
09:50Because the business is growing, I don't think there's an urgency to do it just because of the valuation.
09:55So I think let the business grow, but also my confidence level is as long as the business is growing and generating results for the clients,
10:03our valuation will reflect on that.
10:05And also, you know, to a point, I do also see many of the news coming out from Bloomberg,
10:09so I will encourage many of my friends just to look at Bloomberg reports.
10:12Many of the sales are recently answered or they gave a valuation target.
10:15Many of the sales arecentballs in Pensacola and Windows in so join them.
10:32So we have numbers that go into Bitcoin.
10:35We appreciateIGHT!
Be the first to comment