Issue #037 — Weekend Notes

Unitree Robotics filed for its IPO this week, and honestly, I can’t decide if it’s a victory lap or a warning shot for the rest of the humanoid industry.

I’ve been staring at the ticker and the spec sheets all weekend, trying to reconcile the hype cycle we’re in with the cold, hard reality of the public markets. It feels like just yesterday we were debating whether bipedal robots were even viable outside of a YouTube promo, and now we’re talking about valuations and lock-up periods.

I ran into a veteran hardware founder last week—we’ll call him "Dave" to save him from HR—who put it bluntly: "Ken, the hardware is hard, but the unit economics are murderous."

He’s right, and that’s exactly why Unitree’s move is so fascinating. It forces us to ask the uncomfortable question: Is the humanoid market actually ready for prime time, or are we just great at burning cash?

Let’s dig into the noise.

The IPO Rabbit Hole

The news dropped via The Robot Report (per their coverage of the filing), and the immediate reaction in my group chats was a mix of cheerleading and sheer panic. On one hand, you have a Chinese player actually pulling the trigger on a public listing. On the other hand, you have to look at the context.

According to the catalog we track over at Maze/FinBrain, we are currently monitoring roughly 600+ robot models across about 350 companies globally. That is a lot of metal and code chasing a very small amount of actual revenue. When a player like Unitree steps up to the plate, it’s not just a fundraising event; it’s a stress test for the entire "humanoid as a service" thesis.

I’ve seen some wild numbers thrown around this week regarding valuation and market cap, mostly unsourced speculation on X (formerly Twitter). I’m not going to repeat them here because, frankly, without the final prospectus in hand, they are just guesses. But the *signal* is clear: The Chinese supply chain is moving faster than the Western regulatory and investment frameworks to commoditize this tech.

If Unitree succeeds, it validates the "China speed" narrative—we can iterate hardware and dump it on the market faster than you can say "generative AI." If it stumbles? Well, it’s going to freeze the LP market for every other humanoid startup in the West that’s burning $2 million a month. Our estimate is that the top-tier humanoid players are burning cash at a rate that would make a SaaS founder blush, and an IPO flop is the last thing they need right now.

The "Skin" of the Game

While everyone was doom-scrolling the IPO news, a quieter story slipped out that I think actually tells us more about the near-term future of robotics. Unichem acquired Loomia this week (also reported by The Robot Report).

I love this story because it’s so unsexy it’s brilliant.

Loomia makes tactile sensors. Basically, robot skin. Unichem and R&Y are automotive suppliers. Why does an auto supplier care about robot skin? Because the biggest bottleneck for humanoids right now isn't the brain (LLMs are handling that); it’s the body.

We’re obsessed with the "humanoid" form factor—the walking, the waving, the parkour. But for these things to actually work in a factory or a warehouse, they need to feel. They need to know if they’re crushing an egg or holding a steel beam. If you pair the Unichem acquisition with the news that Schaeffler plans to mass produce gearboxes for humanoids in 2027, you start to see a pattern.

The old guard—automotive and industrial giants—is waking up. They aren't trying to build the *entire* robot (like Boston Dynamics or Tesla). They are trying to build the guts.

Schaeffler is talking about mass manufacturing gearboxes using forming technology by 2027. That’s three years out. Unichem is pushing into skin. This suggests that the industrial supply chain views 2027 as "Go Time."

It’s a subtle distinction, but I think it’s the most important trend of the year. The "brains" of the robot are being commoditized by open-source models and NVIDIA. The "bodies" are being commoditized by companies like Schaeffler and Unitree. So, where is the moat for the startups?

If I’m a VC right now, I’m not looking at the guy building the general-purpose biped. I’m looking at the guys building the sensors and the gearboxes that go inside them. Or, I’m looking at the guys applying the tech to boring, specific problems.

The Generalist vs. The Specialist

Speaking of specific problems, let’s talk about Theker. TechCrunch reported that they just raised $85 million to build a factory robot that "doesn't specialize in anything."

I had to laugh when I read that. The tagline is brilliant marketing, but terrifying engineering. A generalist factory robot? That sounds like a machine that can do ten things poorly instead of one thing perfectly.

But I get the appeal. The dream of the universal robot—the one that you can reconfigure on the fly to screw in a bolt on Monday and weld a chassis on Tuesday—is the holy grail. It’s the only way the unit economics make sense. If you build a specialized robot, you’re stuck in a niche market. If you build a generalist, you have a TAM the size of the global manufacturing sector.

However, contrast that with Pudu Robotics, which launched its MP2000 autonomous forklift this week. The MP2000 is an "AI-native" beast designed to haul 2,000 kg loads (per The Robot Report). It’s not a humanoid. It doesn’t have a face. It doesn’t do backflips. It just moves heavy stuff around a warehouse efficiently.

This is the bifurcation I’m seeing in the data.

On one side, you have the "Platform Players" (Unitree, Tesla, Figure) trying to build the iPhone of robots—a general-purpose chassis that runs apps. On the other side, you have the "Utility Players" (Pudu, Theker-ish) solving specific logistics pain points.

The MP2000 isn't going to win any design awards for looking like a human, but I bet its deployment economics look a lot healthier than a humanoid that needs $50,000 worth of actuators just to stand up. It raises an open question I’ve been wrestling with: Is the humanoid form factor a solution looking for a problem?

In a warehouse, do we really need a robot that walks on two legs to carry 2,000 kg? Or is that just our own biological narcissism projecting itself onto machinery?

The Geopolitical Elephant

I can’t wrap up this week’s notes without mentioning the elephant in the room, or rather, the ban hammer.

TechCrunch dropped a bombshell reporting that the US government is banning new foreign-made humanoids, robot dogs, and solar inverters, citing risks to national security. The report notes this largely affects imports from China, which dominates the market for these specific goods.

This changes everything.

If Unitree’s IPO signaled the commoditization of hardware, this ban signals the bifurcation of the market. We are effectively creating two robotic ecosystems: the Western sphere and the Chinese sphere.

For a Western analyst, this is scary. China has been aggressive in driving down costs. If US companies can’t import cheaper Chinese components—or fully assembled robots for testing and R&D—their burn rates are going to skyrocket. We talk a lot about "sovereign AI," but get ready for "sovereign Actuators."

The ban on robot dogs specifically is fascinating. It feels like something out of a sci-fi novel, but it underscores a real fear: autonomy + mobility = threat.

From an investment standpoint, this makes due diligence infinitely harder. You aren't just looking at a startup’s tech stack anymore; you’re looking at their supply chain exposure and their regulatory risk profile. A startup relying on Chinese strain wave gearboxes (even if they are superior and cheaper) might suddenly find themselves unable to scale in the US market.

The Weekend Takeaway

So, where does this leave us?

We have a Chinese IPO that will test market appetite. We have automotive giants building the organs (skin and gears) for 2027. We have $85M going into generalist robots and massive forklifts launching. And we have a government slamming the door on cross-border trade.

It feels chaotic, but I think it’s actually the market growing up. The "Wild West" era of robotics—where anything with a joint got a term sheet—is ending. We are entering the "Infrastructure Era."

The winners won’t necessarily be the ones with the coolest demo video. They’ll be the ones who navigate the supply chain mess (Schaeffler/Unitree), the ones who find the boring but profitable use cases (Pudu), and the ones who survive the regulatory crossfire.

I’m going to spend next week digging deeper into the "skin" and sensor market. If the hardware is becoming a commodity, the edge might just be in tactile sensing. It’s not as flashy as a humanoid doing a cartwheel, but it might be what actually gets these things out of the lab and into the workforce.

Stay tuned.

— Ken