While humanoid robots can now walk, jog, and run half-marathons, they still struggle with one of the most basic human tasks: reliably picking up a glass of water. This gap between movement and manipulation has become the defining challenge of the robotics industry, and it is attracting serious venture capital attention. Beijing-based PokeBot closed a $100 million Pre-A funding round just 120 days after incorporation, positioning robot manipulation as the field's primary battleground.

Why Is Robot Manipulation So Much Harder Than Walking?

The past decade of robotics progress solved what seemed like the hardest problems. Navigation, the ability for a robot to locate itself and plan a path, is largely solved in controlled environments. Locomotion, the ability to walk, climb stairs, or maintain balance on uneven terrain, has been so thoroughly addressed that it has become almost routine in the industry.

But manipulation requires something fundamentally different. Where locomotion demands precise joint control and balance, manipulation demands simultaneous understanding of an object's physical properties, its spatial relationship to the environment, the correct contact point for grasping, the expected physical consequences of applying force, and the higher-level task goal the action is meant to achieve. An error in locomotion causes a fall that can be recovered from. An error in manipulation, the wrong grip on a piece of tofu or the wrong angle on a knife, causes task failure with no recovery path.

Independent analysts and robotics researchers have consistently identified dexterous manipulation as the field's primary remaining bottleneck. A PatSnap Eureka 2026 report found that while robotic actuators now outperform human muscles in speed, endurance, and power density, "dexterous manipulation and tactile perception remain below human-level capability." Samsung's newly formed RX robotics division, announced in late July 2026, explicitly cited manipulation as the sub-problem receiving dedicated research investment, noting that manipulation drives 31 percent of humanoid costs and is widely cited as the single greatest barrier to commercial deployment.

What Makes PokeBot's Approach Different?

PokeBot was founded specifically to attack the manipulation barrier, rather than treating it as a downstream capability that locomotion-focused platforms will eventually acquire. The company's founder, Xu Huazhe, brings an unusual research background. He is a tenure-track assistant professor and doctoral supervisor at Tsinghua University's Institute for Interdisciplinary Information Sciences, the institute founded by Turing Award laureate Andrew Yao, and he created TEA Lab, China's first research laboratory dedicated specifically to robot manipulation.

Xu earned his PhD at the Berkeley AI Research Lab under Professor Trevor Darrell, one of the most cited computer vision researchers globally, and completed a postdoctoral fellowship at Stanford's Vision and Learning Lab under Professor Jiajun Wu. Over four years at Tsinghua, he published more than 100 papers at top-tier venues including Science Robotics, NeurIPS, ICLR, and CoRL. His paper "RoboCook: Long-Horizon Elasto-Plastic Object Manipulation with Diverse Tools," co-authored with collaborators from Berkeley and Stanford, won CoRL 2023 Best System Paper, demonstrating a robotic system that could manipulate deformable materials across long sequential tasks using multiple physical tools.

The funding round was co-led by Shunwei Capital and IDG Capital, with follow-on participation from Zhongding Capital, Jiukun Ventures, Junshan Capital, SEE Fund, Liepin Investment, and Yuanno Capital. All existing shareholders, including Yunqi Capital, Xiaomi Strategic Investment, Honghui Fund, Inno Angel Fund, and Dongfang Jiafu, chose to participate further.

How Does PokeBot Demonstrate Real-World Manipulation?

Three months after PokeBot was incorporated, the company released a nine-minute cooking demonstration video showing a robot autonomously preparing a complete dish of mapo tofu, from raw ingredients to a finished, plated result, without any human intervention. In isolation, cooking demos are not new, but what distinguishes the PokeBot video is the simultaneous presence of five manipulation challenges that individually push the limits of current robotics systems and, in combination, constitute a realistic test of general-purpose household manipulation.

These challenges include:

• Long-horizon planning: Nine minutes of cooking involves dozens of sequential steps in which each action's success is a precondition for the next, requiring management of error accumulation over time.
• Deformable-object handling: Tofu is soft, fragile, and deforms under contact force, requiring the robot to cut it without crushing it.
• Tool use and object interaction: The robot must manipulate multiple utensils and ingredients in realistic kitchen conditions.
• Tactile feedback integration: The system must respond to physical feedback from objects and surfaces in real time.
• Task generalization: The robot must adapt its approach based on the specific properties of each ingredient and cooking step.

What Are the Broader Implications for the Robotics Industry?

The robotics industry is experiencing a wave of significant funding and public market activity. Unitree Robotics, the world leader in robots shipped monthly with 25,000 robots shipped cumulatively, has announced plans to go public in August 2026. UBTech is already public, and Deep Robotics has filed to go public in China. Agibot and Galbot are also heading in that direction by the end of 2026.

However, the industry faces regulatory headwinds. The Federal Communications Commission (FCC) issued a ban on Chinese-made humanoids and quadrupeds being imported into the United States, citing national security concerns. The rule requires 65 percent of the bill of materials for a humanoid to be domestically sourced. This creates a significant challenge for U.S. robot makers, as approximately 60 percent of actuators, which represent 45 to 60 percent of a humanoid's cost structure, currently come from China. Even high-performance actuators from China are typically 20 to 40 percent cheaper than U.S. alternatives.

Additionally, all actuators require rare earth elements, which come almost entirely from China. This regulatory environment creates a potential vulnerability for U.S. companies if China retaliates by restricting rare earth exports, which would make it impossible to build robots domestically.

How to Evaluate Robot Companies in This Emerging Market

• Clear use cases: Look for companies with compelling, scalable applications rather than general-purpose robots that lack specific market focus.
• Competitive advantages: Assess whether the company has proprietary technology, research credentials, or manufacturing capabilities that create defensible moats.
• Funding trajectory: Track the speed and quality of funding rounds as an indicator of investor confidence and market validation.
• Regulatory compliance: Evaluate how companies navigate emerging rules like the FCC's domestic sourcing requirements and potential supply chain vulnerabilities.
• Research foundation: Consider whether the company is built on peer-reviewed research and academic credentials, which often correlate with technical credibility.

The robotics industry is at an inflection point. While locomotion has been largely solved, manipulation remains the frontier. Companies like PokeBot that focus directly on this unsolved problem, backed by credible research and serious capital, may define the next generation of practical robotics. The convergence of technical progress, significant funding, and regulatory pressure suggests that the next few years will determine which companies and approaches dominate the industry.