Robot Hands and the Business Case for Humanoids

0:00 Trying to copy the human hand is a mistake akin to building flying machines with flapping wings. One of the most talked-about robot companies in the world went public in August at a valuation of roughly $9 billion. On its first day of trading, that valuation briefly climbed to roughly $66 billion. Since then, the stock has come back down.

0:20 9% is how many of Unitree's humanoids actually went into industrial applications. Overall, 74% of their robots went to research nothing really industrial in nature. The hand is the problem. Each Optimus hand and forearm has more than a hundred screws and small parts. Some equipment at those stations reportedly can't line up the parts precisely enough.

0:44 Boston Dynamics made their own announcement. They got rid of the pinky. Every finger needs to earn its place on that hand. If the focus is on the hands, do you really need legs? Humanoids can walk across something, they can flip, they can jump, they can throw themselves into molten steel. But at the end of the day, manipulation is the product.

1:06 Welcome to the Business for Automation, the show that cuts through the hype and tells you what's exactly happening in the world of automation, robotics, and physical AI and what it means for your business. I'm Aaron Prather, Director of Market Intelligence here at A3, the Association for Advancing Automation.

1:20 And on our show, we go behind the headlines and actually see what's really happening. We cut through the noise and find the signal. Let's get into it. Robot safety isn't paperwork you fill out at the end of an automation project. It's part of design, integration, and operating the system from the beginning.

1:37 For more than 35 years, the A3 International Robot Safety Conference has brought together the people developing robot safety practices with the people applying them in the real world. This year's conference is November 3rd through the 5th. And for full disclosure, IRSC is an A3 event, and A3 is who I work for.

1:56 See you in Detroit. This week on the business of automation, one of the most talked-about robot companies in the world went public in August at a valuation of roughly $9 billion. On its first day of trading, that valuation briefly climbed to roughly $66 billion. Since then, the stock has come back down.

2:16 But here in the United States, the biggest humanoid program in America wants to evaluate building 20,000 robots a week. It's reportedly producing several hundred right now, but one of the biggest bottlenecks isn't the brain or the legs, it's the hands. And this week, a three-person team out of Columbia University raised $4.5 million for a robot hand with three fingers, with Toyota Ventures among the investors leading the round.

2:41 Three fingers on purpose. On this show, we talk a lot about humanoids, but I'm starting to wonder, are we focused on the wrong part of the humanoid? Because it seems like we're more interested in our hands, because this is what makes us human the ability to use our hands to manipulate things in our world.

2:57 Today, one question in three parts. First, humanoids are starting to be priced by the market. Second, we have a company that is already telling us that we're having issues with the hands, and that's holding back production. And then finally, actually looking at those companies and other teams that are developing the hands.

3:16 So let's get into it. For our first story, we're going to talk about stocks. And full disclosure, I am not a financial analyst, I am not a stock trader. I'm just a guy that looks at numbers, and that's what we're doing here. We're going to look at exact numbers for this company. Unitree Robotics listed on the Shanghai Stock Exchange on August 19th.

3:35 At its IPO, Unitree was valued at roughly $9 billion. During that first day of trading, though, the stock surged, briefly pushing the company's valuation to roughly $66 billion. Now, to be fair to Unitree, when we say the stock came down, we're talking about how it came down from its IPO high price of $66 billion.

3:56 At this point in time, in October 2026, things could all dramatically change. So we're just saying the stock went up and then it dramatically came back down. And that's what we're going to discuss now. This is a fast-growing business based on the numbers they submitted for the IPO. Revenue grew from roughly $58 million in 2024 to about $252 million in 2025.

4:18 Humanoids represented roughly 52% of that revenue, or about $129 million. Unitries shipped more than 5,500 humanoids, and H1 2026 revenue was projected at roughly 157 to 168 million, up 36 to 45% year over year. But there's another number I think matters more. Now here's that number I want to talk about because it's 9%.

4:45 9% is how many of Unit's humanoids actually went into industrial applications. Overall, 74% of their robots want your researchers. Now, why does this matter? It's because we know industrial applications is where you'll get a repeat sale. It's not highly likely that you're going to get a repeat sale from academia.

5:05 But even that 9% is sort of suspect because of out of that 9%, 74% of it was the robot was doing factory tours or inspections. Nothing really industrial in nature. And that sort of matters with this. Now from the market intelligence test, three things. First of all, the stock price is the stock price.

5:27 Stocks go up, stocks go down. We could be at a very different place in a couple months, even a year. Number two, this sort of opens up a lot of questions, especially in China. Regulatory agencies are already starting to ask questions of which companies should go to an IPO marketplace to begin with. And if their numbers aren't really that strong and they're not really showing growth, then maybe they shouldn't go to the marketplace.

5:52 Third, even though Unitree is still a multi-billion dollar company, it's all based on a promise of what the humanoid could bring. And now we're actually starting to have questions of is the humanoid really going to bring that much value? Before we get to that, let's talk about how one company actually just retired their old humanoids.

6:20 Yes, that is real. Figure is actually allowing these robots to jump into molten steel in a plant in Finland. And Arnold actually stuck around to do something with it. Figure two has been decommissioned. So let's talk now about another company, Tesla. On September 25th, the information reported problems with the hands and suppliers in Tesla's Optimus program.

6:45 Let me be clear about the source. This comes from one report. Tesla has not confirmed the hardware details I'm about to share. According to the report, each Optimus hand and forearm has more than a hundred screws and small parts. Now, before we get into that problem, let's actually look at the humanoid as it is right now, because we've done an excellent job of creating very strong and powerful legs.

7:12 Yes, we are able to actually boot a small little robot across the room. But does anyone have the hands to catch them? According to the report, it is people that are assembling the Optimus components. Some equipment at those stations reportedly can't line up the parts precisely enough. That means some units have to be reworked.

7:32 The report also says the tolerances are tighter than they typically are for automotive parts. Touch sensors have also caused problems. Tesla is reporting developing a replacement sensing glove for a future model. Production reportedly increased from a few dozen robots a week during quarter two to several hundred by August.

7:52 The testable target is 1,000 per week by the end of the year. Elon Musk has talked about eventually reaching 20,000 humanoids per week, but he describes that figure as aspirational. And his observation about the engineering challenge is difficult to argue with. The human hand is an incredible thing. The hand is a problem.

8:14 So many things go into our own hand. Now we're trying to create it in a robot. And that is a very challenging thing because all these parts is what's going to make them work. And it's really sort of ironic that it takes a person to actually make a hand for a robot that then that's gonna train on how to do the job.

8:36 And is that making the hand as well? So what is our takeaway here? Well, first of all, this report is only coming from one outlet. Tesla has not confirmed it. However, it's not hard to take this case up and see how hard it is to make the human hand. So the biggest thing we're gonna have to watch here is where is this going?

8:58 Because you can build a ton of humanoids, but what happens if their hands aren't really worth anything? So let's talk about some of these companies and other organizations that are actually trying to solve the hand issue. Because this week we actually had two major announcements. First, Tangent Robotics raised $4.5 million from Toyota Ventures.

9:18 If that's not a signal telling us how important the hand is, I think it just did. Tangent is a Columbia engineering spin-out based in New York, and its hand is called the TR02. Look at what the hand is doing. This is not mainly about moving boxes, it is about connecting parts, seating gaskets, threading, snap fitting, meshing gears, and wiring.

9:41 These are assembly jobs, and they are hard for standard grippers and suction cups. They require contact, control, and precise movement. Tangent calls its approach a fine motor skill layer. It sits between the AI model that understands the task and the controls that move the hand. The goal is to cut training time from months to days to hours.

10:04 But that is still a goal. It is not yet a proven production result. And then one of their co-founders said something that refrains this entire conversation. Trying to copy the human hand is a mistake akin to building flying machines with flapping wings. What we can do is build robot hands that are capable of dexterity, even with non-human-like components and articulation.

10:29 Let's sit on that for a second, because that's a very interesting point, because we have built airplanes that don't look anything like birds. So when we look at the human hand, do we really need all five fingers? Or could we actually get away with fewer digits to get the job done? One day after that announcement from Tangent, Boston Dynamics made their own announcement.

10:52 They announced a new hand that has 13 degrees of freedom versus their old hand that had seven. Now here's the interesting thing. They got rid of the pinky. Boston Dynamics designed the hand to pinch small objects, form strong three-point grasps, reposition objects without putting them down and operating tools, drills, torque drivers, grinders, nail guns, even welding torches.

11:18 And the hand includes dense tactile pressure sensors across the fingertips and palm so Atlas can detect contact and adjust its grip. But Boston Dynamics also had to answer a very basic question. Do we need our pinky? And Boston Dynamics took that question very seriously, to the point that their product and technology lead actually asked the team to test that.

11:42 And when they did, they came back. No pinky. This wasn't just about looks. Boston Dynamics says a fifth finger would add three more degrees of freedom. That means more actuators, more size, and more power. It also means more cost and more ways to fail. The team didn't think the extra complexity was worth it.

12:04 The goal wasn't to make Atlas less human, it was to use only the anatomy needed for the job. So the design focused on what matters dexterity, strength, durability, cost, repairability, sensing, simulation, and mass production. Now Boston Dynamics is clearly stating that they are building this hand for production capabilities.

12:24 So that means every finger needs to earn its place on that hand. And if it is not bringing value to the final product, it's not needed. It also should be noted that Boston Dynamics is on a clock here because they have to deploy Atlas into Hyundai's factories and Georgia. So they are already trying to find these hands to be ready to go, ready to work day one.

12:48 So a lot of work is going into this one. Now, before we go anywhere else, let's actually sort of recap here because we now know that Tesla is struggling with making their hands work. We have Tangent with backing from Toyota Ventures doing three finger hands. And we have Boston Dynamics doing four finger hands.

13:06 So the question isn't really is five fingers better than three? It's actually matching how many fingers do you need to get the task done. Now, a little bit of information from the market intelligence desk. As I've said, this is Toyota Ventures that is funding Tangent, not Toyota directly. So Tangent is still getting their funding to get going, and there is a hope that it'll eventually come with to a product that Toyota Maine will want.

13:32 But at this point, remember, this is Toyota Ventures making this investment. We still have to actually see products flow and purchase orders actually being made. Now here's my verdict. I don't really care if it's three fingers or four fingers or five fingers. The idea here is make something that the client needs and can use.

13:51 And if that's only three fingers, use that. The other issue is how strong are these going to be? Because if they're going into industrial settings, that means that they actually got to be able to function and survive that kind of environment. And that could be a huge challenge. We know from other folks that are making these types of hands that that is one of the main issues.

14:12 They don't survive long in industrial settings. And then finally, the other question is what's the humanoid really needed for? If the focus is on the hands, do you really need legs to deliver the hands to the operation? Or could that be any type of robot? So we definitely know we need hands, but now the question is, do we need the humanoid?

14:35 So here's my final thoughts. It is very clear that humanoids are here and they're going to be operating among us. So the question really isn't maybe humanoid or hand, it's more of are you getting the manipulation that you need to do the task? Because we know that humanoids can walk across something, they can flip, they can jump, they can throw themselves into molten steel.

14:55 But at the end of the day, the real solution that we're looking for is a manipulation solution that can actually do the task we're putting it on. And if it's attached to a humanoid, great. If it's attached to another type of robot, even better. But it is coming down. Manipulation is the product. That's our show for this week, but the show is just the beginning.

15:13 Go to our Substack and you can actually see even more details that we just talked about. With each one of these stories listed out, all the numbers, all the details, and even additional commentary. The Substack is free, and as you subscribe, it'll be dropped into your email basket every Friday. I'm Aaron Prater, Director of Market Intelligence here at A3, the Association for Advancing Automation.

15:35 Let's get this weekend kicked off.