- 1 hour ago
Every robotics revolution needs its foundations. Before the robots ship, someone has to build the chips, the operating systems, and the development platforms that make them possible. D-Robotics has made that infrastructure layer its business, providing the hardware, software, and application stack that robotics developers worldwide build on. In this conversation, Co-founder Xuan Jiang and Hermitage Capital's Henry Zhang discuss what it takes to back and build the essential layer beneath the robotics boom, how capital flows from East to West in deep tech, and who will own the stack when the dust settles.
Category
🤖
TechTranscript
00:25Good morning, Paris.
00:26How are you guys doing?
00:28My name is Henry Zhang.
00:29I'm the president and managing partner of Hermitage Capital.
00:32We're a $1.5 billion AOM tech-focused fund.
00:36We invest in AI, robotics, deep tech across Asia and the U.S.
00:41And today it's a privilege for me to host this panel on robotics infrastructure.
00:47And I would like to begin with a question to the audience.
00:49So raise of hand, how many of you have a robot at home?
00:54Being robot vacuum or more long robots?
00:57Raise of hand.
01:00Okay.
01:02How many of you would like to potentially buy a human or robots?
01:06What is a human like in the next five years?
01:08Raise your hand.
01:11Great.
01:11So if you're interested in robots, today is a session.
01:14So I think before we begin, I would like to introduce our speaker, Shun Jiang, who is a co-founder
01:20and CFO of D-Robotics.
01:23D-Robotics is a leading robotics company out of Asia.
01:27They're number one in terms of valuation.
01:30They just raised around at a few hundred million, a billion dollar valuation, raised a few hundred million dollars.
01:36One of the hottest companies coming out of Asia on robotics.
01:41And I think today we're going to talk about the robotics infrastructure, including the software, the hardware, integrated approach.
01:49Why don't we begin with that?
01:51Shun Jiang.
01:51Definitely.
01:52Yeah.
01:52It's a pleasure to have you.
01:54I think, first of all, I would like to hear the story.
01:56You know, how did robotics came out and why do you have this name?
01:59What is the story behind it?
02:01Oh, as for the name, it's an interesting question.
02:05So D-Robotics is actually a spin-out vehicle from Horizon Robotics, which is a listing company at Hong Kong
02:14Stock.
02:16Horizon Robotics is focusing on automotive, but D-Robotics is actually focusing on the robotics itself.
02:23So back to 10 years when Horizon Robotics was actually founded, we want to be the platform for robotics, the
02:34computing platform for robotics.
02:35So starting from 2015, we start to do the R&D on the computing chips, computing algorithms, and all the
02:46ecosystem for robotics back in 10 years.
02:49So eight years ago, we spin out D-Robotics from Horizon Robotics to focusing more on the applications of robotics,
02:59a whole industry.
03:01So we see a lot of pinpoints of the current robotics industry, because if you can see, there's a lot
03:12of companies focusing on different kinds of robotics.
03:15For example, the vacuum machine, the long movers, the drones, the human robotics.
03:24Every company has their own set of chips, algorithms, training tools.
03:31So everybody's starting from scratch, starting from zero.
03:35They spend a lot of money, a lot of R&D efforts to do repeating works.
03:40So our dream is to accelerate the whole R&D speed for each of the companies so that they can
03:52accelerate the whole process of pushing out their new products so that they can iterate their products and to serve
04:01the needs of consumers.
04:02So in that sense, D-Robotics is actually focusing on the processors, on the computing platforms, on the operating system,
04:13the ecosystem tools, to help all the companies to accelerate the development of robotics itself.
04:22In that sense, we already have over 400 customers spreading from maybe the low-end vacuum machine to the middle
04:34-end, for example, the drones, the robot dogs, and to the high-end, the human robotics.
04:41So in that sense, we are actually the number one computing platform in China nowadays, and also we are shipping
04:49our solutions to the whole world.
04:53For example, the Europe, some of the vacuum machines, the long movers are using actually our chips.
04:58And also we are shipping the solutions to a lot of developer communities around the world, for example, in Singapore,
05:09in Japan, and also in Switzerland.
05:12In that sense, we are actually building the whole ecosystem in a world world for the developers and also the
05:21enterprise to build up their robotics in order to accelerate the whole development of robotics industry.
05:29Yeah, it's definitely a very crowded and interesting space, right?
05:33You have, you know, in the U.S., you have Figure, you have Optimus under Tesla, and you have 1X
05:39initially came out of Europe, now expanding into U.S.
05:42In China, you have Unitree, you have GalBot, and a few other robotics.
05:46Some of them are here at VivaTech.
05:48What do you think about, you know, your role, D-Robotics, on the edge computing side?
05:54How do you actually solve the bottleneck and further develop, you know, human robots and robotics in general?
06:00How do you solve that challenge?
06:01I think it's a good question.
06:03So we are dividing the computing for robotics into cloud computing and edge computing.
06:09Actually, D-Robotics is focusing more on the edge computing because we think that the robotics at your home, at
06:17the factories, at everywhere else in the physical world, they need to be low latency, low power consumption, and long
06:28lifetime.
06:30And also, you have to control the cost.
06:33That's why our product is mainly focusing on the edge computing.
06:37But not only the chips, we are actually providing the operating system and also the training tools and also some
06:47reference design of the whole robotics.
06:51But we are not making robotics.
06:53We are providing the tools to each of the robotics companies.
06:56In that sense, I think that's why we are actually the common ground, the common supplier for most of the
07:05robotics companies.
07:07For example, as you mentioned, the Unitree and EGI Bot and also the most popular and have the most shipment
07:16companies in China now, also in the world.
07:20I think the next question is on the developer community, right?
07:23I think similar to NVIDIA in the early days, I think D-Robotics over the years have very successfully built
07:29a very large global developer community base with over 100,000 community developers around the world.
07:37How did you achieve that?
07:39And what do you think is the importance of open source contributors, such as developers, to the entire sort of,
07:46you know, robotics community?
07:48It's very essential for open source algorithms for most of the developers.
07:53Actually, we have opened our ecosystem to three layers of developers all over the world.
08:00First one is the individual developers, which is the most rooted and also maybe lack of funding and very young
08:09age developers.
08:10But actually, they are the future of the world and actually future of the robotics.
08:15So actually, we are providing the toolkit for those individual developers and also the reference algorithms and also a community,
08:24which they can interact with each other to communicate on the efficiency of the algorithm.
08:31What we get back is actually the feedbacks of the toolkits and also our iteration of our processors.
08:40So this is the first layer we want to enlarge for the individual developers.
08:45And I think till today, we have over 200,000 developers worldwide.
08:53It's actually one of the biggest robotics community in the world.
08:58And second layer, we are actually assisting the startups for robotics industry.
09:05Maybe the startups only have like 10 or 20 developers in that company, but they have a brilliant idea.
09:14So we actually support the computing power, actually the chips and also the development tools for them to have their
09:26own ideas come true.
09:28So this one is a second layer.
09:30And in this layer, we are reaching out to almost 500 startups to help their idea come true.
09:39And the third layer, which is a leading robotics company in the world, for example, as you mentioned, those human
09:48art robotics and also the consumer robotics,
09:52which they have the very clear understanding of what the consumer needs and they have a lot of funding.
09:59They are actually the world leading technology light to bring all the robotics companies to follow.
10:11So we are actually working with them to get the revenue.
10:14For example, last year, we shipped 2 million of processors worldwide.
10:21And we are covering like 80% of the leading human art robotics companies in the world.
10:28So in that sense, we built our whole ecosystem from the individual developers to the startups and also the leading
10:36human art robotics companies.
10:39There are different applications, right, when it comes to robots.
10:43There is obviously manufacturing, logistics.
10:46There is also home use, embodied AI.
10:49Which application do you think has the biggest near-term demand for something like the robotics offers, the chips?
10:58You mean for the chips?
11:01Actually, in the sense of the shipments, actually we have the chips for the consumer products.
11:09For example, just as we mentioned, the long movers, the vacuum machine, the drones, and also the companion robots, and
11:20also some robotic dogs.
11:21These have the biggest shipment volume, but in the sense of the technology advancement, actually we ship a lot of
11:30processors and also computing platforms to the human art robotics.
11:38So for the consumer robotics, the ASP, the average price is low, but for the human art, it's actually more
11:47expensive.
11:48In that sense, I can't really tell which is more advanced, but it's actually our product strategy.
11:55So we are not only relying on the industry with huge volume, but also we rely on the advancement of
12:05technology in the future, which is the human art robotics.
12:08So we are getting our revenue and also technology iteration very quickly from them.
12:14And we actually get data, get feedbacks, get all the corporation engineering services, so that we make them a flying
12:23wheel.
12:24In that sense, we will achieve our commercial achievement in the future.
12:30Obviously, you know, D-Robotics being one of the hottest companies in Asia on robotics, within the robotics sector, the
12:38valuation has come a long way, right?
12:40I mean, Hermitage Capital, we were a very proud investor in Series 8 of D-Robotics just two and a
12:47half years ago at just a little over $200 million.
12:50Today, we're looking at, you know, potentially 10x of that or even higher valuation.
12:55In your day-to-day conversation with different investors, what do you think the investor really care?
13:01What are the metrics or performance that they value the most, right, out of D-Robotics versus other robotics companies?
13:11I think, obviously, the competition mode is what the investor is most concerned about because they don't want to invest
13:22in a company that everybody can do it.
13:25So, in that sense, I think D-Robotics is positioned itself in a very good position because, in a sense
13:33of the computing platform, I think, nowadays, there's not much players.
13:37I think, other than us, it's NVIDIA on the edge side.
13:42But I think for NVIDIA, they are more of the lab testing.
13:51Let me give you an example.
13:53If a company wants to do a demo or prototype, normally, they will use a GPU because it's general-purpose
13:59use and it's got power, huge computing power, let alone it has a huge power consumption, it's water-cooling.
14:11It doesn't matter because it's just a prototype.
14:14But when they want to go to the 1 from 10, they will normally choose another player like D-Robotics
14:24because we are actually providing low-cost, more efficiency on the power consumption and also the local service support, which
14:37can help them to iterate their products in a very fast pace.
14:40In that sense, I think we are different from NVIDIA, but we are helping on the mass production of most
14:50of the robotics company.
14:51That is the special standing point of this company, which makes us different and gives us a very competitive mode
15:01in the whole industry.
15:04And second is the continuous business model of the company, which the investor wants to see from a target.
15:15Some of the companies, they have a prototype, but their competitor is a lot.
15:24So in that sense, D-Robotics actually have a very continuous business model.
15:32We don't only serve the prototype or the human robotics, which don't have much shipments nowadays.
15:40We also help with the vacuum machines and also the drone companies so that they can have more shipments and
15:50help us with the negotiation on the supply chain.
15:54In that sense, I think we got a lot of revenue and we got fundings from the investors.
16:00We are running on a very healthy runway.
16:03That is another thing that the investors pay more attention.
16:09And maybe the third one is nowadays, I think the algorithms is not converged yet.
16:16So a lot of companies are hesitant to invest into the R&D of algorithms and also other applications.
16:27In that sense, we are providing actually the reference design of some algorithms so that they can just try and
16:36test.
16:37In that sense, we also get feedbacks from them.
16:40But our customers is very widely spread, so we can get the common feedbacks from the whole industry, make us
16:50the very leading players of the whole industry.
16:54And we are actually gathering the common ideas from most of the leading players.
17:00In that sense, we always stand in the front of the industry.
17:06So in that sense, I think investors like Havantage want to invest into the D-Robotics.
17:15Obviously, within robotics, the different layers of stacks, right?
17:18There's the cheap layers such as D-Robotics.
17:22There's more sort of hardware-focused layers such as Unitree and others from China and elsewhere.
17:29And obviously, there's, you know, model focus.
17:33Young Le Quint was at the session yesterday.
17:36I think he founded a company called Army Labs, which focuses more on the world labs.
17:40And on the algorithm side, there's, you know, two school of thoughts, VLA versus world model.
17:46But I think from our perspective, we think that D-Robotics capture a very unique stack in between, where basically,
17:53no matter how the body side of the hardware side evolved or how the algorithm side evolved, the chip is
18:00always there.
18:01And not many players can really play in that sector.
18:04So in that sense, we really like this particular stack within the whole robotics ecosystem.
18:11I think the next question I'll have is, what do you think is the biggest bottleneck or challenge for, you
18:19know, human robots to further advance?
18:21Because my analogy is, if you think about, compare, you know, human robots or general-purpose robots versus autonomous driving,
18:29we're probably like layer, sorry, you know, L1 or L2 stage on the way to become, you know, level 3
18:38and above.
18:38Do you agree with standard energy?
18:40And where do you think is the biggest bottleneck for this to be even more advanced to the level people
18:45really want to adopt?
18:49I think the first hurdle is actually the lack of the real-world data for training.
18:56Because nowadays, you see a lot of world models, even some startups focusing on the world models come on.
19:02But actually, there is limited real-world data for different kinds of applications.
19:09You know, the robotics industry is very fragmented.
19:13So you have to have a specific training platform to generate data.
19:23But you have to validate from the real-world data.
19:27So I think that one is, you know, some of the scenarios, you can't really get the data, for example,
19:33your home.
19:34So it's very difficult to validate your models nowadays, even though we have the world models to generate.
19:41But, you know, for the real world, it's difficult and there's gaps.
19:46I think the second is actually, I just mentioned, the algorithms is not converged yet.
19:54So in that sense, there's a lot of different directions that people will try.
20:02But I think it's not yet the JPT time.
20:06So I think still a lot of funding and talents come into the industry.
20:12But maybe in three or five years, the direction will be converged.
20:19But within the three and five years, people will be lost direction.
20:24And maybe the third is actually the commercialization of robotics.
20:31You know, it's hot topics for robotics, but in the real world, the applications, maybe in the factories, the manufacturing,
20:40also in home applications, is still very limited.
20:45It's only a prototype.
20:48But as you just asked the audience who want human robotics, I think only one-third want the human robotics.
20:57It may be just the current performance of a human robotics is not that good, not up to the standards.
21:06So it's still got a long way to have these algorithms and processors more efficient so that the applications will
21:14be up to the performance.
21:17Yeah, maybe let me ask some questions on the outlook, right?
21:20What's your prediction for the China and also the global robotics industry in the sense of, you know, how many
21:28robots will be there, you know, in terms of annual capacity, annual productions and adoptions for both China and globally
21:36for the next three to five years?
21:39I think the future time is definitely a trillion, several trillion markets, but I can't really tell the number.
21:48But the growth rate is really fast, especially in China market.
21:53Maybe the China consumers is more embracing the new technology and also they are tolerant for the maybe data privacy.
22:03It's actually the China market, but also China market has a very strong supply chain for the robotics.
22:10You know, in Shenzhen, the south part of China, you can have your own robots being formed within one month
22:16because they have very sophisticated supply chain.
22:19So I think the growth rate in China may be quite high in the recent five years, you know, just
22:26like Tom's driving.
22:27But I think in the next five years, because of the population change and also the lack of the labors
22:36in international markets,
22:39I think the global market will catch up the speed.
22:44And also they have much more values in the international market.
22:48In that sense, I think 10 times or even 20 times growth in the future is very foreseeable.
22:55And for coming back more specifically to the robot infrastructure for you versus your competitors in the market you just
23:03mentioned, right?
23:04It's going to be a trillion dollar market.
23:05How much of the chip market share do you think the robotics can capture potentially in the next three to
23:12five years?
23:12It's a very interesting question.
23:14I think in the sense of computing platform is always winner takes all.
23:19I think the second and third place is have limited space to life because all the algorithms and efforts you
23:28have to put on one computing platform.
23:31In that sense, we want to be the number one and takes all.
23:35I think we are running out of time.
23:37We have 30 seconds left.
23:38Maybe any final thoughts for the audience in terms of how they can potentially best capture the robot economy?
23:46Maybe a buzzword or, you know, one comment.
23:50I think nowadays the infrastructure that's commonly used by all the players is the most valuable investment target in the
23:59future.
24:02Thank you, Shuan, for your insights, ladies and gentlemen.
24:05That wraps up the session today.
Comments