SemiAnalysis: Unitree Set to Dominate Global Robotics Industry, G1 Gross Margin Estimated at 67%
0xBroomberg
SemiAnalysis published a deep-dive report estimating Unitree's G1 robot has a bill-of-materials cost of just $8,976 and a gross margin of 67%, arguing the company is replicating BYD's and DJI's hardware playbook to dominate global robotics.
The G1 is cheap — how is the margin still 67%?
The G1's pre-tax price has dropped from over $50,000 to $27,300 in 12–18 months. Some units have traded below $20,000.
Yet the estimated BoM is just $8,976, yielding a 67% gross margin. This means → low price does not equal buying market share at a loss — the manufacturing cost itself is low.
The key is in-house actuators — the robot's joints. Actuators account for 50%–70% of a humanoid robot's BoM, and Unitree's self-made cost runs 30%–40% of Western equivalents.
In plain terms = competitors buy joints off the shelf; Unitree builds its own, slashing the single biggest cost item by more than half.
BYD had batteries, DJI had flight controllers — what does Unitree have?
SemiAnalysis frames Unitree inside a proven Chinese hardware playbook: control the most expensive, hardest core component → scale to drive cost down → each generation opens a new market.
BYD's core component was the battery, once 30%–40% of an EV's BoM. DJI's was the flight controller — the 2013 Phantom 1 sold for $679, roughly half the cost of a self-assembled drone.
Unitree's core component is the actuator. It validated the path with quadruped robots: the 2018 Laikago cost $45,000; today's Go2 starts at roughly $1,600–$2,800 — a 94%–96% price drop in six years.
This reflects something larger: the quadruped line was not just a product — it built the actuator tech, control systems, supplier base, and production processes that humanoids now inherit.
Why can China's supply chain iterate on a weekly cycle?
China produced 31.3 million vehicles in 2024, 40.9% of them new-energy. The drone supply chain has roughly 3,000 component suppliers. About 200 humanoid-robot companies share and feed back into this ecosystem.
Brushless DC motors, drivers, encoders, batteries — parts built for cars and drones transfer directly to robots.
The real edge is speed: suppliers are a few hours away by train, samples arrive the same day or the next, iteration cycles run weekly, not quarterly. Component prices can be 20%–40% below Western equivalents.
This means → Unitree's moat is not just "we're cheaper" — the entire supply chain is accelerating its cost reduction.
Early G1 units were not great — how much has improved?
The H1 and early G1 batches were "not very capable": arms extended could briefly hold 2 kg; bent-arm load of 2–3 kg lasted about 2–3 minutes. Motors frequently overheated.
Unitree has since lowered motor current draw, smoothed magnetic fields, improved copper fill, and upgraded thermal management. The October 2025 G1 batch added active cooling near the pelvis.
Current state: bent-arm carry of 5 kg for 10–15 minutes; extended-arm hold of 5 kg for about 1 minute. This means → payload roughly doubled, sustained duration roughly 5× longer versus early units.
What can the G1 actually do today — and how far is it from replacing workers?
The current role is "light duty": 2–4 kg tote handling, empty-bin transfers, and other narrow tasks — not a general-purpose worker.
SemiAnalysis benchmarks against Agility Robotics: Agility demonstrated 66 totes per hour at a conference; GXO deployments handle 2–4 kg totes.
As of 2025, an estimated maximum of roughly 250 Unitree humanoid robots are in productive pilots or deployments, with one enterprise running about 30 units. Unitree is reportedly on track to ship its 10,000th humanoid within weeks.
In plain terms = the G1 is not walking onto factory floors to replace people tomorrow — it is planting a flag in the simplest handling tasks first, then expanding from there.
After the IPO, which three metrics actually matter?
SemiAnalysis says Unitree's IPO "marks the official start of the robotics era."
But the report also notes the market should look past near-term shipment headlines and watch three things happening at once: whether G1 real-world deployments expand, whether BoM keeps falling, and whether AI capability can reduce the share of teleoperation.
This means → only if deployment volume, cost, and autonomy all move forward together does "dominate global robotics" become more than an aggressive call.
We are witnessing the birth of another Chinese hardware giant.
SemiAnalysis
AI industry-chain research firm
(June 9, 2026, deep-dive report: Unitree Will Dominate Global Robotics)
Content is for reference only, not financial advice.