Economist Fu Peng: Humanoid Robot Benchmark Company Yushu's R&D Lower than Pig Farming Company Muyuan
Economist Fu Peng pointed out in a financial dialogue that the humanoid robot benchmark company Yushu Technology's R&D expenses in 2025 will be only 145 million RMB, lower than the 1.648 billion RMB R&D investment of pig farming company Muyuan in the same year, which is more than eleven times that of Yushu. He judges that the humanoid robot industry is still in its very early stages and states that Yushu's pre-IPO valuation has far exceeded market consensus, with its stock price fluctuations essentially representing a venture investment, unsuitable for ordinary investors seeking stable returns.
The prospectus shows that Yushu's R&D expenses for 2023 are 49.9518 million RMB, 70.017 million RMB for 2024, and 145 million RMB for 2025, totaling approximately 265 million RMB over three years. The R&D expense ratio will decrease from 31.39% in 2023 to 8.53% in 2025 as revenue expands. During the same period, peers like UBTECH and Cloud Deep have R&D expense ratios mostly above 20%, with UBTECH's mid-2025 R&D expenses accounting for 35.1% of revenue. Yushu's revenue in 2025 is expected to be around 1.699 billion to 1.708 billion RMB, with R&D breakdown approximately 85.04 million RMB for personnel salaries, 28.97 million RMB for direct materials, 15.15 million RMB for cloud computing and services, and 7.78 million RMB for depreciation and amortization.
As of September 2025, the company has about 480 employees, with approximately 175 to 184 in R&D. Early R&D focuses on the body structure and motion control, referred to as the "cerebellum"; starting in 2024, they will increase investment in the embodied large model, referred to as the "brain". However, the prospectus admits that during the reporting period, they did not collect real data on a large scale or conduct factory deployment training, and their self-developed general embodied model has not been scaled up. As of the end of January 2026, Yushu holds 262 domestic and international patents, of which only 20 are domestic invention patents; UBTECH has accumulated about 2800 to 3000 authorized patents. Founder Wang Xingxing has stated that the cost of pure AI R&D is too high, and the body is the main business, collaborating with Deep Thinking and NVIDIA on the brain.
The issue price is 150.80 RMB, with a price-to-earnings ratio of 219.23 times, higher than the industry average of 38.56 times. On the first day of trading, it opened at 1100 RMB, a 629% increase from the issue price, with a total market value reaching approximately 444.9 billion RMB at one point. Subsequently, in about six trading days, the market value evaporated by nearly half, with the stock price falling back to around 591.53 RMB, with a low point of 571 RMB. Of the planned fundraising of 4.202 billion RMB, 2.022 billion RMB is allocated to intelligent robot models, 1.11 billion RMB to the body, and 624 million RMB to the manufacturing base. In the first quarter of 2026, revenue increased by 68.49% year-on-year, while net profit excluding non-recurring items decreased by 52.55% year-on-year to 40.2536 million RMB, with R&D expenses increasing by 38.328 million RMB year-on-year.
In 2025, humanoid robot shipments exceeded 5500 units, while quadrupeds approached 20,000 units, with humanoid business revenue surpassing quadrupeds for the first time, and humanoid gross margin at approximately 62.9%. The company became profitable in 2024, with a net profit excluding non-recurring items of about 600 million RMB in 2025. International sales historically account for more than half of revenue. Management states that there is an extreme shortage of personnel in all positions, and the expense side has entered an expansion phase.
In terms of market mechanism, this is a valuation digestion rather than an order collapse. Buyers are theme funds from the Sci-Tech Innovation Board and first-day momentum traders, while sellers are those who took profits at high positions and speculative funds. Capital quickly shifted from the pricing power at the issue price of 150.80 RMB to the risk pricing near 600 RMB in the secondary market. Beneficiaries are pre-IPO shareholders and first-day sellers; those under pressure are ordinary investors who bought humanoid robots as stable growth stocks and the company’s expense sheet that must use fundraising to fill the "brain" gap. The incremental information compared to Muyuan is that pig farming companies have scaled R&D into breeding and disease investment, while robot companies still rely on less than 200 million RMB to support the narrative of global shipments.
Source: Public Information
ABAB AI Insight
Yushu is following a cost-reduction path in consumer electronics, not a money-burning path in laboratories. Wang Xingxing replaces hydraulic systems with external rotor brushless motors, making quadrupeds into hardware that can be shipped at scale, and uses the Spring Festival Gala dance to bring humanoid robots into the public eye. The prospectus clearly states: first the body and cerebellum, then the brain. This explains why 145 million RMB in R&D can support the first shipments and why the expense ratio drops from 31% to 8.5%—the revenue is growing faster than the R&D budget, not that the research is completed.
The capital path is to first prove profitability with hardware profits, then invest half of the fundraising into models. Of the 4.2 billion RMB plan, 2 billion RMB goes to the "brain", which acknowledges that the secondary market is buying unscaled software capabilities. Deep Thinking and NVIDIA are brought in for computing power and model interfaces, indicating that Yushu does not intend to build a foundational model factory alone. Muyuan's 1.648 billion RMB R&D corresponds to breeding, feed conversion, and disease, which is a mature industry writing R&D into production costs; Yushu's 145 million RMB corresponds to a motion control team of less than 200 people. The incomparable aspect is the industry stage: one is optimizing pigs, the other is proving that robots can walk steadily.
The comparable object is not Tesla's Optimus capital narrative, but DJI in its early days: a very small team, extremely high hardware iteration, absolute R&D expenses looking poor, and shipments outpacing competitors. UBTECH follows a patent accumulation and higher expense ratio, resembling traditional robot manufacturers. The current stage is "the body has been commercialized, the model has not been productized". The drop from 1100 RMB to about 600 RMB reflects a correction from science fiction premiums back to "a hardware company that can ship plus unverified brain options".
Structural judgment belongs to capital concentration before technological substitution is completed. The market first priced the scarce A-share humanoid robot stocks at around two thousand times the price-to-earnings ratio, then slapped back with absolute R&D values in six trading days. The mechanism is: shipments and gross margins prove the hardware closed loop, while patents and expense ratios expose software gaps; thematic funds only recognize the first stock, not the R&D bills. Once the 2 billion RMB model expenditure enters the profit statement, pricing power will shift from narrative to whether expenses can be converted into billable industrial tasks.
ABAB News · Cognitive Law
First shipments do not mean research is completed
When absolute R&D values cannot compete with narratives, stock prices will calculate for you
Profitable hardware companies, after going public, are buying brains that have not yet grown