Everyone is reading the Unitree IPO as a competitive shot at Tesla. Revenue up 335%, 5,500 humanoid units shipped, 62.9% gross margins on the humanoid line. The surface-level framing writes itself. Unitree has proven unit economics, Tesla hasn't shipped a single Optimus externally, so the pressure is on Elon to respond. But the more I dug into the prospectus details, the less clean the story got.

Here's what's actually happening with that margin. Unitree's primary customers are universities and research labs. These buyers are purchasing experimental platforms, not factory labor. The base G1 at $12,750 ships locked down (no SDK, no dev access). The version researchers actually buy runs $30-60K.

And the total addressable market for it is.. maybe 10,000 units globally? Unitree could saturate that in two years at current pace. Meanwhile, 50-70% of non-research humanoid revenue comes from reception desks and tour guide duties. Not exactly the trillion-dollar labor replacement thesis investors are pricing in.

The Tesla comparison compounds the error. Optimus V3 hasn't even had its formal unveiling yet. Production starts this summer at best, in volumes Musk himself called very low. But Tesla isn't trying to sell research platforms to labs. It's trying to crack general-purpose industrial deployment, a problem nobody has solved. Comparing Unitree's shipped units to Tesla's zero is a category mismatch, not a competitive gap. The real question (who cracks industrial margins at scale) remains completely open. I wouldn't size into either side of this yet.

Unitree Files for $610M IPO on China's STAR Board, Reports Strong Humanoid Robot Commercialization
Hello China Tech
robotics
Unitree's 335% revenue growth and 60% margins validate humanoid robot unit economics. IPO signals investor appetite for robotics commercialization at scale.

Arm Launches AGI CPU, Entering Data Center Chip Market for AI Infrastructure
The Register
compute
Arm's data center entry fragments CPU supply. Server OEM design cycles now include Arm option alongside x86, reshaping processor roadmaps.

Microsoft Forms Third Major AI-Nuclear Partnership in 60 Days
Tom's Hardware
energy
Three hyperscalers now pursuing nuclear power. Power, not compute, is the binding constraint for AI scaling; nuclear partnerships become table stakes.

Frontier model economics shifting toward closed-source moats

66
+14 pts

Agent payment infrastructure collapsing outside enterprise rails

60
+15 pts

Today’s briefing tracked 5 developments across the AI infrastructure stack.

You saw the summary. Open the terminal for the full picture.

Open Tessara →