🤖 Robotics
Unitree Is Worth $50 Billion. It Shipped 5,500 Robots. That's $9.1 Million Per Unit.
Unitree's Shanghai debut valued the world's first profitable humanoid robot maker at $50 billion after a 486% first-day surge, pricing every robot it shipped in 2025 at $9.1 million. We reverse-engineered the unit economics, ran the DCF both ways, and calculated exactly how many robots the company needs to sell to make today's price rational.
On Tuesday morning in Shanghai, Unitree opened at 1,100 yuan per share, a 629% premium to its 150.8 yuan offering price, before closing the morning session at 883.87 yuan and valuing the company at roughly $50 billion. In 2025, Unitree shipped 5,500 humanoid robots. Divide those two numbers and each robot the company has ever delivered to a customer carries an implied market value of $9.1 million, roughly 370 times the $24,700 average price at which those robots actually sold.
That ratio captures something important about where robotics investing stands in August 2026. Unitree is not a story stock with nothing behind it. It earned RMB 600.1 million ($89 million) in non-GAAP profit on RMB 1.708 billion ($252 million) in revenue last year, with gross margins above 60 percent and the largest humanoid robot market share on Earth at 32.4 percent. It is profitable, growing, and shipping real machines that walk, run, and perform martial arts well enough to headline China's Lunar New Year gala. Among humanoid robot companies globally, no one else can say all of those things simultaneously.
And yet the market on day one priced the company at 562 times its 2025 earnings and 198 times its revenue, ratios that say less about Unitree's factories than about what investors believe those factories will become.
Three Numbers That Define the Gap
| Metric | At IPO Price | At Morning Close | At Opening Peak |
|---|---|---|---|
| Market Cap | $9.0 billion | $50 billion | $67 billion |
| Market Cap Per Robot Shipped | $1.64 million | $9.09 million | $12.18 million |
| Premium Over Average Selling Price | 66x | 367x | 493x |
| P/E (2025 Non-GAAP) | 101x | 562x | 753x |
| P/S (2025) | 36x | 198x | 266x |
Reuters Breakingviews estimated the implied 2026 P/E at 860 times, after accounting for the profit decline Unitree itself flagged in its prospectus: first-half 2026 non-GAAP net profit is expected to fall 6 to 22 percent year-over-year, with first-quarter profit already down 47.69 percent. Profits are shrinking while the valuation is expanding, two arrows pointed in opposite directions.
What $50 Billion Demands
We ran a simplified discounted cash flow in both directions to determine what Unitree's current market price is actually betting on.
If Unitree is a manufacturing company (terminal P/E of 25 times, 10 percent discount rate), a $50 billion market cap today requires approximately $2 billion in annual net profit within a decade. At Unitree's current 35 percent net margin, that implies roughly $5.7 billion in annual revenue. At the current average selling price of $24,700 per humanoid robot, Unitree would need to ship 231,000 robots per year, 42 times its 2025 volume.
If the average selling price continues falling toward $10,000 as Unitree pushes its G1 model at $13,500 and likely lower, the required shipment volume rises to 570,000 units per year.
If Unitree is a platform company (terminal P/S of 10 times), the same $50 billion requires roughly $5 billion in annual revenue by the mid-2030s, whether from hardware, recurring software subscriptions, or some combination. At $10,000 per unit plus $1,200 per year in cloud services, Unitree would need an installed base measured in the hundreds of thousands to approach that figure.
Neither scenario is impossible. But both require Unitree to scale manufacturing by at least two orders of magnitude from where it stands today, in a market where the largest buyer (China's government) has set a 2026 target of deploying just 10,000 humanoid robots nationwide.
Wright's Law Is Working
One fact working in Unitree's favor is how aggressively it has driven costs down. According to the prospectus, the average selling price of a Unitree humanoid robot fell from RMB 590,000 ($87,500) to RMB 166,400 ($24,700) over roughly two years while gross margins held above 60 percent. That means production costs fell roughly in proportion to prices.
Applying Wright's Law to Unitree's disclosed data, with cumulative production approximately doubling from early volumes to 5,500 units, yields an implied learning rate of roughly 55 percent per doubling. That rate is extremely steep, comparable to the early years of crystalline silicon solar panels (which achieved 75 to 80 percent learning rates over decades) and far faster than automotive manufacturing has ever demonstrated.
If Unitree sustains even half that learning rate, a $10,000 humanoid robot by 2028 and a $5,000 model by 2030 are plausible. Whether the market for $5,000 robots is 100,000 units or 10 million. Between those two figures lies the gap between a $10 billion company and a $500 billion one.
Figure AI Makes Unitree Look Cheap
The strangest feature of Unitree's valuation is that by the logic of its closest Western competitor, it might actually be underpriced. Figure AI, the Sunnyvale-based humanoid startup backed by Jeff Bezos and Nvidia, reportedly holds a private valuation of up to $39 billion despite having no meaningful revenue and no history of profitability.
At $50 billion, Unitree is valued at just 1.3 times Figure AI's last private mark, but it ships thousands of robots, posts $89 million in profit, maintains 60 percent gross margins, and holds the largest global market share. UBTech, the only other publicly traded humanoid maker (listed in Hong Kong), has been consistently unprofitable.
In a single morning of trading, the Shanghai market moved Unitree from "profitable manufacturer" territory (36 times revenue at IPO pricing) into "narrative platform" territory (198 times revenue at close). Whether it stays there depends entirely on which analogy the market decides applies: Fanuc or Apple.
A Geopolitical Ceiling Nobody Wants to Model
Two months before its IPO, the U.S. Federal Communications Commission banned imports of future models of foreign-made humanoid and quadruped robots, citing national security concerns. In June, the Pentagon added Unitree to its list of Chinese military companies, a designation that falls short of a sanction but bars the U.S. military from using its technology.
Unitree's existing models still have U.S. approvals, but future products face an effective American market ban. The company's prospectus shows overseas revenue accounted for more than 40 percent of sales during each reporting period, and while not all of that is American, losing access to the world's largest single economy constrains the addressable market at exactly the moment the valuation requires maximum addressable market growth.
Reuters reported that the core innovation behind Unitree's most successful robot dogs was based on U.S. Army-funded research published openly at MIT and Carnegie Mellon to stimulate progress in the field. American defense labs funded the basic science, Chinese engineers scaled the manufacturing, and now American regulators are banning the finished product. That circularity is striking.
What We Did Not Prove
Our DCF reverse-engineering uses simplified assumptions. The terminal P/E and P/S multiples we chose (25x and 10x) represent mature-phase manufacturing and platform companies respectively, but Unitree may command higher terminal multiples if it builds a genuine software ecosystem on top of its hardware, which no humanoid company has yet demonstrated at scale. The Wright's Law calculation relies on prospectus data that aggregates humanoid and quadruped robot costs without separating them cleanly, and the learning rate could flatten as volumes increase and easy gains are exhausted. The 562x P/E at morning close may not hold by market close or end of week, and early STAR Market trading in China is notoriously volatile: CXMT surged 466% at debut on July 27 before stabilizing at lower levels. A single morning's price discovery is a noisy signal.
Why $50 Billion Might Be Right
The best argument for $50 billion is that humanoid robots are where smartphones were in 2007, a pre-inflection market that looks absurd in unit terms and obvious in retrospect. Apple shipped 1.4 million iPhones in 2007. At its peak, it shipped 232 million per year. If Unitree is the Apple of humanoid robots, its current unit volumes are meaningless, and the learning curve data suggests costs will fall fast enough to make consumer-scale pricing inevitable within five years. DeepSeek, Tencent, and Alibaba, three of China's most sophisticated technology investors, all backed the IPO. Beijing has made humanoid robotics a strategic national priority with explicit deployment targets. And Unitree is the only company on Earth demonstrating that you can build humanoid robots profitably at scale.
But iPhones worked on day one. Unitree's robots run, dance, and perform martial arts, but the prospectus acknowledges that "few of its robots are being used in commercial settings" and "many of its products are sold to research institutions and universities." Smartphones had billions of potential users from the moment they functioned. Nobody knows yet how many businesses, households, or institutions actually need a humanoid robot, at any price.
What to Watch and What to Do
If you are an investor in robotics equities, the number to watch is not Unitree's stock price but its average selling price and gross margin in each subsequent quarterly report. As long as ASP declines while margins hold, the Wright's Law trajectory is intact and scale economics favor the bull case. If margins compress while prices fall, the manufacturing premium unwinds fast. If you are evaluating the humanoid robot sector globally, Unitree's public pricing now sets the valuation anchor for every private and pre-IPO competitor. Figure AI's $39 billion private valuation, which seemed extravagant last month, is now 22 percent below the public-market price of its only profitable peer. At least a half dozen Chinese competitors, including Deep Robotics, Leju Robotics, and AgiBot, are preparing IPOs that will reference Unitree's multiples. If you are building with or deploying robots, Unitree's cost curve says a sub-$10,000 general-purpose humanoid is likely within three years, and a sub-$5,000 unit within five. The buy-versus-wait calculation for industrial deployment just shifted.