AsianFin -- China and the United States are battling for dominance in the next frontier of artificial intelligence: humanoid robots.

Inside a cavernous warehouse in Shanghai's Pudong district, dozens of humanoid robots designed by Chinese AI firm TMTPost App are engaged in repetitive tasks—folding T-shirts, opening doors, and assembling sandwiches. Each movement is recorded for up to 17 hours a day, generating massive datasets to fine-tune the robots' performance.

Meanwhile in the U.S., Tesla's Optimus robot is showing off new capabilities, from dancing and executing voice-activated commands to performing chores and factory tasks. Figure AI, another key American player, has announced plans to scale production of its next-gen Figure 3 humanoid robot to 100,000 units annually by 2029.

Driven by breakthroughs in generative AI over the past two years, including OpenAI's ChatGPT and China's SenseChat, humanoid robotics is experiencing a surge of momentum. Tech firms on both sides of the Pacific—including China's Unitree Robotics, Fourier Intelligence, Galaxy General, and the U.S.'s 1X Technologies and Apptronik—are racing to industrialize humanoid robots that could soon enter homes, factories, and public venues at scale.

By 2050, Citibank forecasts the global humanoid robot market could reach $7 trillion, with some 650 million robots deployed worldwide. China alone is expected to inject nearly $140 billion into robotics and related sectors over the next two decades—far outpacing the U.S. and Europe.

Despite the optimism, industry experts caution that humanoid robots remain in their infancy. Debate is intensifying over whether they are the best path toward Artificial General Intelligence (AGI).

At a recent conference hosted by Tsinghua University, Chinese Academy of Sciences academician Zhang Bo questioned the fixation on human-like designs. "Two hands are sufficient for factory work. Why give robots two legs, which increase costs and reduce reliability?" Zhang said. "Humanoids aren't always optimal. Four-legged or flying robots may be better suited for many tasks."

Elon Musk, however, remains bullish on Optimus, even as he acknowledged the threat of Chinese competition. "No one can match Tesla on humanoid robots," Musk said during a recent earnings call. "But positions two through ten could all be Chinese companies."

Investor enthusiasm is exploding across China's robotics ecosystem.

"This is a period of capital frenzy—everyone is rushing in," said Cao Zheng, Managing Director at SDIC Fund Management.

Local governments in Beijing, Shanghai, Shenzhen, and Chongqing are rolling out funding plans and robotics strategies. State-backed investment in the humanoid sector has already surpassed 70 billion yuan ($9.7 billion), targeting core components, intelligent systems, and real-world applications.

China is also home to nearly 900,000 robot-related companies, according to Tianyancha. In the first five months of 2025, over 100,000 new firms were registered—a 44% year-on-year surge.

More than 82% of the 300+ global robotics investment deals in 2025 have taken place in China, according to Xiniu Insights. Total financing in the sector has already topped 20 billion yuan, exceeding the full-year figure for 2024.

Galaxy Universal Robotics, for example, raised 500 million yuan in an Angel+ round, bringing its total funding to 1.3 billion yuan within just two years. Its backers include IDG Capital, Meituan, SenseTime, and government-linked AI funds.

Internationally, Texas-based Apptronik raised over $403 million in a single round, while Figure AI and 1X Technologies secured $675 million and over $100 million respectively.

Despite China's advantages in manufacturing and AI talent, scaling humanoid robots to real-world applications still faces formidable barriers—chief among them being access to high-quality data.

"Compared to autonomous vehicles, humanoid robots can't passively accumulate data," said Midea Group's Hou Che. "We must either simulate environments or rely on manual operation, both of which are costly and slow."

Open-source datasets are a start but remain too generalized. Tailored, task-specific data is needed for robots to function autonomously in homes or industrial settings.

Gu Jie, CEO of Fourier Intelligence, says full-stack integration is key. "In some areas, we co-develop with partners—for example, in force and vision sensors," Gu said. "For large models, we work with SenseTime and also track Nvidia's progress. Localization and fine-tuning are essential."

UBTECH, the world's first publicly listed humanoid robot company, is leaning into the industrial opportunity. In partnership with FAW-Volkswagen, its Walker S robots are already helping assemble vehicles. Clients include BYD and Foxconn, and each industrial unit sells for around $100,000.

The company plans to launch a $20,000 consumer-grade home robot later this year, targeting China's aging population and rising demand for companionship robots.

Tan Min, UBTECH's Vice President and CBO, says the sector is undergoing "brutal competition" that will weed out weak players. Despite losing over 1.1 billion yuan in 2024 and a 45% drop in its stock price over the past year, Tan remains optimistic: "This fierce competition helps the industry improve overall efficiency."

Midea Group has also entered the fray, with its humanoid robot already deployed inside a washing machine factory. Meanwhile, academia and industry leaders are calling for measured optimism.