Robot Valuations Take Off First, as Capital Starts Rethinking the Math

Two star companies in the robotics industry have delivered major news back to back.

On October 27, Bloomberg reported that Physical Intelligence, a US robotics AI startup, is in talks to raise at least $1 billion at a target valuation of roughly $20 billion. The company, founded just over two years ago, has no revenue yet.

A day later, Unitree Technology, already listed on the STAR Market, released its third-quarter financial report. Revenue for the first three quarters reached 1.817 billion yuan, up 55.71% year on year, while third-quarter net profit fell 17.37% year on year.

Compared with many robot startups that have yet to establish stable sales, Unitree has already achieved preliminary commercialization results. In 2025, the company sold more than 5,000 humanoid robots and turned a profit. Yet set against the capital market, that performance still looks thin: as of October 29, Unitree's total market capitalization reached 267.9 billion yuan, about 158 times its revenue for the previous year.

The two companies represent two technology routes—general-purpose AI models and robot hardware manufacturing—but they face the same reality: their existing commercialization results remain very far from the valuations the capital market has assigned them.

Unitree sells thousands of robots, yet still cannot support a hundreds-of-billions valuation

On August 19 this year, Unitree Technology officially listed on the STAR Market, opening at 1,100 yuan on its first day, up nearly 630% from its IPO price of 150.80 yuan, with total market capitalization briefly exceeding 440 billion yuan.

The stock then fell steadily from its highs, dropping intraday to a low of 404.03 yuan on October 15, more than 60% below its first-day opening price. Just a week later, the trend reversed again: the stock hit the daily limit up on October 23 and rose another 7.69% on October 26. As of October 29, the share price had recovered to 662.39 yuan, with a market capitalization of about 267.9 billion yuan.

In just over two months, Unitree's market capitalization swung by hundreds of billions of yuan, while the company's fundamentals showed no change during that period sufficient to explain such violent volatility.

The third-quarter report released on October 28 shows that Unitree's revenue for the first three quarters was 1.817 billion yuan, with net profit attributable to shareholders of 387 million yuan. Third-quarter revenue was 665 million yuan, up 69.93% year on year, while net profit fell 17.37% year on year to 113 million yuan.

It is worth noting that the 270.54% year-on-year increase in net profit attributable to shareholders for the first three quarters was also affected by a one-off share-based payment expense in the same period last year. Unitree had recognized a one-off share-based payment expense of about 349 million yuan in 2025 arising from a low-priced capital increase to its employee shareholding platform, lowering that year's profit base. Excluding non-recurring gains and losses, the company's net profit attributable to shareholders for the first three quarters of this year actually fell 18.78% year on year, and net cash flow from operating activities also declined 54.69%.

One important factor behind the profit pressure is increased R&D spending. In the third quarter, Unitree's R&D investment reached 127 million yuan, up 249.58% year on year, with funds mainly directed to embodied intelligence models, computing power and new product development.

In terms of operating scale, Unitree is already at the forefront of the global humanoid robot industry. In 2025, the company generated revenue of about 1.699 billion yuan, with humanoid robots contributing 868 million yuan and quadruped robots contributing 698 million yuan. It shipped 5,511 humanoid robots for the full year and actually sold 5,215.

Yet compared with a valuation in the hundreds of billions, these revenues remain insignificant. Based on the market capitalization of about 267.9 billion yuan on October 29, even if first-three-quarter 2026 revenue is simply annualized to 2.423 billion yuan, Unitree's price-to-sales ratio would still be about 111 times, and its price-to-earnings ratio calculated on annualized net profit attributable to shareholders for the same period would be about 520 times.

Unitree itself has warned about valuation risk. In an announcement on abnormal stock price movement issued on October 27, the company disclosed that as of October 26 its trailing price-to-earnings ratio had reached 465.22 times, far above the average of 38.54 times for the general equipment manufacturing industry. Unitree also cautioned that the humanoid robot industry is still in the early stages of commercialization, and that if the pace of large-scale adoption falls short of expectations, there is a risk of valuation reverting in the future.

Zero revenue, a $20 billion valuation: Silicon Valley capital bets on robots' "brains"

Unitree at least already has real revenue and profit, whereas Physical Intelligence's valuation is built almost entirely on imagination about the future.

Founded in 2024, the company's core team comes from institutions including Google DeepMind, Stanford University and UC Berkeley, and its goal is to develop general-purpose AI models that can adapt to different robot hardware.

In November 2024, Physical Intelligence completed a funding round of about $400 million at a valuation of $2 billion. In November 2025, its valuation rose to $5.6 billion, and earlier this year it was raised again to $11.5 billion. Now a new round of financing talks is seeking to push its valuation to $20 billion. If the deal is completed smoothly, Physical Intelligence's valuation will have grown nearly tenfold in just two years.

More surprising still, Physical Intelligence currently has no revenue. The company's popularity stems mainly from the potential commercial value of its general-purpose robot foundation models.

Most robots today still require task-specific programs and trained control systems. Physical Intelligence hopes to use vision-language-action (VLA) models to let robots complete operations based on visual information and human instructions, and to transfer learned skills to different tasks and devices. Put simply, it is trying to provide a shareable "brain" for robots of different brands.

In recent years, Physical Intelligence has launched a series of models including π0 and π0.5, demonstrating manipulation capabilities such as folding clothes and tidying objects. On October 13 this year, Physical Intelligence also opened a beta version of the Pi API, allowing external developers to upload data, fine-tune models and run them on their own robots.

Physical Intelligence has also begun exploring commercialization, partnering with companies such as Weave Robotics and Ultra to conduct real-world tests in scenarios including laundromats and e-commerce order packing.

However, completing a technical demonstration, achieving actual deployment and generating sustainable revenue are three entirely different stages. Physical Intelligence has not yet disclosed a mature commercial pricing system, and the model's cross-hardware adaptability, long-term operational stability, inference costs and customers' willingness to pay all still require further verification.

Yet at a $20 billion valuation, even if the market were willing to grant a 20-times price-to-sales ratio in the future, Physical Intelligence would need to generate $1 billion in annual revenue. For a company with no revenue at all, that means it must move from technology R&D to product sales, and then extend its model services to enough robots and application scenarios.

Venture capital naturally tends to seek out the next-generation technology platform early. Silicon Valley's sustained bets on AI foundation models have also given companies like Physical Intelligence ample funding support.

The question is whether the robotics industry will eventually form a large-scale third-party model market, and how much revenue Physical Intelligence can capture from it—at present, there are no clear answers.

Behind hundred-billion-level valuations, the capital climate for embodied intelligence has changed

The core argument supporting high valuations for robot companies is a potentially enormous market. Beyond optimistic expectations for the industry's prospects, the capital market's own trading mechanisms have also amplified robot companies' valuations to some extent.

Unitree's shareholding structure deserves particular attention. According to its listing announcement, the company's total share capital after issuance is about 404 million shares, but only 30.0877 million shares were freely tradable in the initial period after listing, accounting for 7.44% of total share capital.

As of October 29, although Unitree's total market capitalization reached 267.9 billion yuan, the market value corresponding to unrestricted tradable shares was only about 19.9 billion yuan. The limited free float amplifies the sensitivity of the share price to changes in capital supply and demand, and also subjected Unitree to violent market-cap swings after listing.

Physical Intelligence operates in a different market environment. As an unlisted company, its valuation is mainly set through negotiation among a small number of investment institutions; even with no revenue, as long as investors are willing to pay for future technological potential, its valuation can keep climbing across successive financing rounds.

Over the past two years, this logic has driven robot company valuations steadily higher. But entering the second half of 2026, the capital market's attitude toward embodied intelligence has begun to show some subtle changes.

Looking at the global market, money is still pouring in. According to statistics published by Crunchbase on October 27, global funding for robotics startups reached $9.9 billion in the third quarter of 2026, more than double the same period last year. Cumulative funding for the first three quarters of this year rose about 185% year on year, with China and the United States remaining the two largest funding markets.

But the pace of valuation and funding growth has also bred market doubts about valuations. On September 21, Reuters reported that Chinese regulators have applied stricter scrutiny to the IPO applications of some humanoid robot companies. Beyond excessive valuations, regulatory officials are particularly focused on whether revenue that some companies obtain through locally government-supported data collection centers and joint-venture projects represents sustained commercial demand from independent customers.

One investor interviewed said that valuations of some embodied intelligence projects in the primary market have already been cut by 30% to 50%. Another source even estimated that, excluding revenue brought in by data collection centers, some robot companies' valuations could fall by 60% to 70%. These are estimates from industry insiders and cannot yet be independently verified with sufficient public financial data.

This valuation dispute is not unique to the Chinese market. US humanoid robot company Agility Robotics previously announced plans to go public via SPAC at a transaction valuation of about $2.5 billion. However, documents it filed with the US SEC show that the company's 2025 revenue was only about $1.78 million, with an operating loss of about $140 million. Even with robots already deployed in real warehouse environments, the company's revenue is similarly far from supporting a valuation in the billions of dollars.

Money keeps flowing into the robotics industry, but more and more investors are beginning to ask more carefully about these companies' business models and potential: can they support today's high valuations in the future? Valuations that were previously pushed up round after round are now being tested by financial data and the public markets.

Leading companies such as Unitree and Physical Intelligence remain at the center of this capital boom, while some other robot startups have already begun to encounter valuation cuts and stricter financing scrutiny.

A recent investigative report by Hard Kr, a hard-tech outlet under 36Kr, revealed many details of this shift: some leading institutions that previously invested actively in embodied intelligence have begun to tighten their purse strings, and some financial advisors (FAs) are no longer willing to take on robot financing projects. A financing that might once have closed in a month now takes three to six months.

In addition, according to an October 29 report by Securities Times, in the past two months early shareholders of some embodied intelligence companies have begun transferring old shares at a discount, quoting only 65% to 80% of the previous financing round's valuation—including some leading companies.

Technical competition in embodied intelligence has only just begun, yet the capital market has already launched a contest over future profits.

Now that contest is producing new disagreement: some are still willing to pay higher prices for a distant future, while others are beginning to re-examine how much of today's valuation can actually be delivered.

Join the official Coincamps community:

X: https://x.com/coincamps

Telegram: https://t.me/coin_camps