00:00Matrix Robotics is a company specializing in physical AI and embodied intelligence.
00:05The company unveiled its third generation humanoid robot, Matrix 3.
00:09It took place earlier this year.
00:11Min Min Law, a China correspondent, joins us now from the conference in Shanghai
00:14with the folks behind the machines. Min Min.
00:21Hey David, yes, you can see a model of the robot behind me.
00:24But joining me now is the CEO and founder, Alan Zhang of Matrix Robotics.
00:28Alan, great to have you here.
00:30You just launched your third generation robot.
00:32Tell us about what it can do, what are its capabilities.
00:35As you can see, we bring multiple third generation of Matrix 3 come to the Dyswag.
00:41And that one is offering the AI Brista.
00:44And it can provide the coffee from the empty cup and also fill with the coffee.
00:48And it can use the fingertip to control the touchscreen
00:51and finally using his right hand to deliver the coffee to the client.
00:55And also we have the other generation robot
00:59and the help to, you know, take the tour around the wag
01:02and also do the interaction.
01:04And also another, you know, far, far away from there
01:07it can provide a scissor rug and paper at the game with the guests.
01:10Your robot is now priced at upwards of 500,000 RMB.
01:15Tell me about your client uptake rate.
01:17Are you seeing a lot of orders from coffee shops and the like?
01:20Actually, we are currently working with a very big coffee shop chain here.
01:24Not only in China, but also worldwide.
01:26And also we're receiving some, you know, gaining interest from the Middle East,
01:29especially from the luxury resort.
01:31And they want to order our major three and providing the reception service.
01:35So at that level of pricing, I mean, I imagine a Chinese company,
01:39they may need to be able, they can hire maybe two staff.
01:43Is your robot offering that level of productivity
01:45or are they using your robot as more of a novelty factor right now?
01:50Actually, instead of providing the productivity,
01:53we think the current stage of the humanoid
01:55will provide more emotional elevation
01:58and especially promoting the traffic to the store.
02:01It is more, you know, working like NPC and very interesting.
02:05So, for example, as a regular workshop,
02:07every day they can serve around 100 to 150, you know, cups per day.
02:11But if they acquire the major three,
02:13they can, you know, increase 30 to 50 percentage of the revenue per day.
02:18Okay, that's quite paradoxical
02:19that robots are providing the emotional value right now.
02:21Yeah, and also driving the traffic, right.
02:23Your background is in Tesla.
02:25You helped create a Tesla Optimus robot.
02:27And some people are saying that your company is a copycat.
02:30What's your response to that?
02:32Actually, yes, some of our team is from the Tesla Optimus team,
02:37for sure, that's the truth.
02:38And also, currently, we think we,
02:40comparing with the Tesla Optimus,
02:42we are more looks like the human.
02:44As you can see that,
02:46Gen3 is, you know, standing at 1.7 meter tall
02:49and also weighted like 75 kilograms.
02:52And also, we are using a tendon drive with the Dexha's hand,
02:55which is five more degrees of freedom
02:57than the Tesla currently offering on the market.
02:59And we're the only one who can produce
03:01and deliver worldwidely.
03:03So, you know, so we're more looks like the human.
03:06You've previously talked about wanting to bring the cost down
03:09or the pricing down to 30 to 50,000 RMB.
03:12That's more than a tenfold decrease.
03:14How achievable is that?
03:16And what's your timeline to getting there?
03:17We think our timeline is about,
03:19we're talking about 24 to 36 months from now.
03:23Yeah, so that's the target.
03:25And because we truly can foresee in the future
03:29that once the house and all the public services
03:32surrounding with the different type of humanoids
03:35or the physical robots,
03:36so the cost, you know, drop down will be quite steeply.
03:41Especially we are leveraging the supply chain
03:43from the China market.
03:44That's actually very, very ambitious.
03:46What is the source of this cost reduction?
03:49Is it economies of scale?
03:50Or where is it going to come from?
03:52We think there's two very important core concepts.
03:55The first is how we keep iterating
03:58and re-engineering the humanoid itself
04:00and try to cutting some of the materials from the body
04:03and also, you know, reduce slightly a little bit
04:05of the degree of freedom
04:06and try to make the whole system more integrated
04:10and more easy to produce.
04:11And the second part is we really need to scale up
04:14the autonomous of the intelligence,
04:17especially we are currently working on the foundation model.
04:20So once the ability of the foundation model
04:22can do some type of autonomous work,
04:24that will also help us to get a lot of interest
04:27and the demand from the market.
04:29So once the volume goes up,
04:30the cost will go down automatically.
04:32Obviously, your robots have to be trained on a lot of data.
04:35It requires, as you said,
04:36the foundational compute as well.
04:37So are you facing any constraints
04:39when it comes to GPU availability
04:41or are the export controls affecting you?
04:45Yeah, this is a very good question.
04:47Actually, we are still okay in the data engine
04:51because, you know, in China,
04:52we can leverage a lot of the low-cost labor
04:54and help us to get the low-cost data set
04:57as we need to pre-train the neural network.
04:59But, you know, to be honest,
05:02a little bit, you know,
05:04restrict on the we are acquiring the GPU from overseas.
05:08But today, we also see a lot of the Chinese local,
05:11you know, chipset manufacturers,
05:12for example, like Huawei,
05:14they provide some super parts
05:16and help us to do some pre-train work.
05:18But we still have a long way to go.
05:20Yeah.
05:20Is there a shortage now
05:22if you want to buy from the Chinese suppliers,
05:23how long do you have to wait for it?
05:25Yes, it is quite a shortage, you know,
05:28because the large language model,
05:31they just, you know,
05:33absorb too much GPU in China,
05:35so which do not leave too much
05:37for the physical intelligence.
05:38So which means sometimes we'll wait for
05:40maybe extra two or three weeks
05:42as we can get ready to pre-train the network.
05:45Is this a major constraint
05:47on the pace of your innovation?
05:49Because time is money here,
05:51given how competitive it is.
05:52There are a lot of constraints,
05:53not only the pre-train shifted.
05:55What are the biggest constraints?
05:56The biggest constraint is
05:57how quickly and how low-cost
05:59we can absorb the, you know,
06:01physical data that we're collecting the data
06:03from when the human doing some manipulation stuff
06:06or working on the working station.
06:08So that's the main constraint.
06:09So speaking of cost,
06:10you're still making losses per robot
06:12as of now, right?
06:13How do you plan to...
06:15Actually, we can reach the break-even per robot.
06:17So we're selling at a cost.
06:19You're selling at a cost now?
06:21Yeah.
06:21Okay, and when can you reach profitability
06:23on a company-wide level?
06:25Profitability, we can see, you know,
06:27aggressively maybe next one month, for sure.
06:30Because we're keep iterating the matrix thing.
06:33Maybe the next generation,
06:34let's say matrix fall.
06:36It's super affordable
06:37and also we can have a very good profit margin.
06:39So talk to me about your growth target.
06:41I know you are expanding your capacity
06:43to I think 100,000 robots by next year,
06:46which is again very aggressive.
06:48Are you looking at that sort of demand?
06:52I think that's the current period of the milestone.
06:55We think this year we are talking about 1,000 units
06:57and the next year we are going to deliver
06:59about 10,000 to maybe 20,000 units.
07:03And the year after next year,
07:04which is 20,28,
07:05we are talking about over 50,000 units per year.
07:09So just like the beginning of the electric vehicle,
07:12so the volume will keep growing gradually.
07:14And how much,
07:16talk to me about any fundraising plans that you might have.
07:20Currently we are looking about,
07:22you know, 200 million US dollar fund
07:24as the per A currently.
07:28And given the overall robotics environment now,
07:31I mean here at WAIC,
07:32there's so many things on display.
07:35How do you assess the competitive environment here?
07:37And is the price competition, do you think,
07:39is it too unhealthy right now?
07:41I think the competition
07:43is just not about the price.
07:46You can see every exhibitor here,
07:48they're providing the different robots
07:50and the specific and focus
07:51on the different working scenarios.
07:53Some for the factory and for us,
07:54working like the type of service
07:56and the restaurant and the coffee shop.
07:58And they are not apple to apple.
08:00So the competition today is still very early.
08:03So we can see maybe two or three or four years from now,
08:07then it will be narrowed down
08:09to several very big giant players in the market.
Comments