Why Unitree’s Slump and Xpeng’s $900 Million Bet Tell the Same Story
Unitree has lost around $30 billion in market value since its post-IPO peak. At almost the same moment, Xpeng agreed a financing package worth approximately $900 million for its robotics business. These may look like opposing signals. Together, they reveal something more useful: a promising technology and a sound investment are not the same thing.
5-minute read · 25 August 2026
At first glance, the robot economy appears to be sending two contradictory messages.
Shares in Unitree, one of China’s best-known humanoid and quadruped robot makers, fell roughly 45% from their debut peak after an extraordinary first-day surge. The company had briefly reached a market value of around $66 billion before approximately $30 billion was erased.
One day earlier, Xpeng announced a financing agreement for Dogotix, its robotics subsidiary. The transaction is expected to provide approximately $900 million and implies a post-transaction valuation of $6.3 billion under the assumptions set out in the company filing.
One robot company was being sharply repriced. Another was attracting some of the largest private financing yet seen in China’s embodied-AI sector.
The apparent contradiction disappears when we separate four things that markets often combine: technological progress, commercial adoption, industrial capability and valuation.
The technology did not fall 45%
Unitree’s robots did not suddenly become less capable. Its factories did not lose 45% of their capacity in three trading sessions. What changed was the price investors were willing to pay for the future they imagined.
The scale of the earlier move matters. Unitree priced its Shanghai IPO at 150.80 yuan per share, valuing the company at more than 60 billion yuan. Retail demand was more than 8,000 times the shares available to that group. On its first trading day, the stock closed 460% above its offer price and at one point gave the company a market value of approximately $66–67 billion.
The subsequent fall is therefore a warning about expectations, but not yet evidence that humanoid robotics has failed. Even after the decline, Unitree remained worth several times its original IPO valuation.
That distinction is important. A falling share price can expose an excessive valuation without invalidating the underlying technology.
When possibility is priced as certainty
Unitree is not an empty technology story. It is already profitable, which distinguishes it from many early-stage robotics competitors. It is also one of the world’s larger producers of humanoid and quadruped robots.
But the commercial evidence remains less mature than the market excitement suggested. Reuters reported that many Unitree sales have been one-off purchases, often by universities and research institutions, and that relatively few robots are operating in broad commercial environments. Adjusted net profit also fell 53% to 40 million yuan in the first quarter of 2026.
Before the listing, the IPO valuation already represented about 219 times 2025 earnings and 36 times sales. Reuters Breakingviews calculated that the peak reached on the first trading day implied around 857 times projected 2026 earnings.
None of those figures proves that Unitree cannot grow into a much larger business. They show how much successful growth investors had already placed into the price.
This is the first lesson from the two stories:
A technology can be transformative while the price of accessing that transformation is still too high.
What Xpeng’s $900 million announcement really contains
The Xpeng headline also needs to be read carefully.
According to the company’s Hong Kong Stock Exchange filing, Dogotix entered into a conditional share-purchase agreement covering approximately $900 million. Of this, external investors conditionally agreed to contribute $600 million. An Xpeng subsidiary agreed to invest $200 million, while companies owned by two Xpeng executives agreed to invest a further $100 million.
IDG Capital is leading the external investment, with Gaorong Ventures participating and Tencent and Alibaba joining as strategic investors. The filing gives Dogotix an implied pre-transaction valuation of $5 billion and an implied post-transaction valuation of $6.3 billion under its stated assumptions.
The distinction is not merely technical. The full $900 million is not new external capital, and completion remains subject to closing conditions. Xpeng itself warns that the transaction may or may not proceed.
Still, the agreement is a substantial signal. It indicates that sophisticated investors are prepared to fund the attempt to move humanoid robotics from demonstration towards production. It also gives Dogotix a separate valuation and broader access to capital while Xpeng retains control.
Funding is evidence of confidence and financial capacity. It is not evidence of customer demand, profitable production or a working market at scale.
Why carmakers may have a structural advantage
Xpeng’s most interesting asset may not be the IRON humanoid robot itself. It may be the industrial system surrounding it.
Modern electric-vehicle manufacturers have already built capabilities that Physical AI requires:
batteries and power management;
sensors, cameras and computing hardware;
AI models and autonomous decision systems;
hardware-software integration;
safety engineering and real-world testing;
complex supply chains and mass-production processes;
service networks and physical locations for initial deployment.
Xpeng plans to use its stores and industrial campuses as early environments for IRON and has said it aims to reach a monthly production rate of 1,000 units by the end of 2026. Commercial deliveries in China and international markets are scheduled to begin in 2027.
This creates a credible route from research to controlled deployment. A robot can first be tested inside the company’s own environment, where tasks, failures and improvements are easier to observe. The resulting data can then feed back into the hardware and software.
For a standalone robotics start-up, much of that system must be assembled from the beginning. For an established vehicle manufacturer, part of it already exists.
An advantage is not a victory
The carmaker thesis is compelling, but it should not be stretched beyond the evidence.
A humanoid robot is not simply an electric vehicle with arms and legs. Reliable manipulation, balance, safe interaction with people, task generalisation, maintenance and productive operation over long periods introduce different engineering and economic problems.
Deploying robots in a company’s own shops also does not prove that independent customers are willing to buy them. Internal deployment can generate valuable evidence, but it can also hide weak external demand if it is treated as commercial validation too early.
Xpeng’s production targets and 2027 sales timetable are plans, not observed outcomes. The financing may help the company reach them. It cannot tell us in advance whether the robots will perform useful work at an acceptable cost.
That is the second lesson:
Industrial capability can improve the probability of success, but it does not remove execution risk.
The evidence that matters next
The next phase of the robot economy will be measured less by demonstrations and more by operating evidence.
Five signals should matter more than viral videos or funding headlines:
Independent commercial orders
Are customers outside the manufacturer’s own organisation willing to pay?Productive utilisation
How many hours can a robot complete useful work rather than demonstrations or supervised trials?Unit economics
What does each productive hour cost after energy, maintenance, supervision and downtime?Repeat demand
Do early customers expand their fleets after using the first units?Production evidence
Can announced output targets be reached with consistent quality and sustainable capital requirements?
These measures will not produce the excitement of a robot running, dancing or walking with unusually human movement. They will reveal whether Physical AI is becoming an industry rather than remaining an impressive collection of machines.
The Hikari Nova perspective
Markets often compress a long and uncertain development path into a single price.
During periods of strong momentum, technological possibility can quickly be treated as commercial certainty. When expectations reverse, the falling price can then be mistaken for evidence that the entire long-term thesis was false. Both interpretations are too simple.
Unitree’s decline does not tell us that the robot economy is over. Xpeng’s financing does not tell us that mass-market humanoid robots have arrived. Together, they show that capital is beginning to test the difference between an exciting product, a scalable company and a defensible valuation.
The long-term opportunity may be real. The path will still contain mispriced companies, missed production targets, weak use cases and periods when market momentum moves far ahead of operating evidence.
For investors, the useful question is therefore not only whether humanoid robotics will become important.
It is also:
Which companies can convert technological progress into reliable customer value — and how much of that success is already reflected in the price?
Unitree and Xpeng are not opposite stories. They are two parts of the same reality test.
Editorial sources
Reuters: China robot maker Unitree’s post-listing slump sparks bubble fears
Reuters Breakingviews: Nobody knows how to price China IPOs anymore
Xpeng filing: Dogotix share-purchase agreement, 24 August 2026
Editorial disclaimer
This article is for information and analysis only. It does not constitute investment advice, a recommendation or an offer to buy or sell any financial instrument. Market prices, company plans and transaction terms may change after publication.