Qianxun AI has completed a new funding round of 1 billion yuan on April 7, 2026. This round was jointly led by Shunwei Capital and Yunfeng Capital, with participation from prominent investors including Dachen Caizhi, a leading RMB fund, Galaxy Yuanhui, Turing Fund, Xinding Capital, and Gengxin Capital.
This comes just 30 days after the company raised nearly 2 billion yuan in February. With a total of 3 billion yuan secured in two rounds within one month, Qianxun AI has become one of the most rapidly funded embodied intelligence companies in 2026.
Unlike other funding cases in the embodied intelligence field, the highlight of this round lies in the rare joint appearance of tech leaders Lei Jun and Jack Ma. Shunwei Capital was founded by Lei Jun, while Yunfeng Capital was co-founded by Jack Ma. This marks the first time these two heavyweight entrepreneurs have formed a co-investment in the embodied intelligence sector. Qianxun's latest funding round reflects a shift in investor preferences—no longer solely prioritizing technology, but emphasizing the integration of technology and commercial applications. This move signals that industrial capital is betting heavily on the critical transition of embodied intelligence from technical validation to commercial realization.
Global embodied intelligence approaches are converging significantly. In the same week as Qianxun's funding announcement, Generalist AI released its GEN-1 embodied foundation model on April 2, 2026. The model improved the average success rate of physical tasks from 64% to 99%, increased completion speed to three times that of the most advanced existing technology, and required only one hour of robot data per achievement. Many have compared GEN-1 to the GPT-3 moment, as it opens the door to AI in the physical world with master-level proficiency.
Two technical details are noteworthy: First, GEN-1 was trained not on robot data but on millions of activity data points generated by humans wearing low-cost wearable devices. Second, Generalist AI explicitly pursues a data-driven Scaling Law approach. This aligns closely with Qianxun AI's technical direction. The underlying logic for continued investor confidence is the growing consensus in the embodied intelligence field, where technological breakthroughs are advancing daily.
Qianxun AI has long focused on "diversity" as a core element of its Scaling Law strategy, having accumulated over 200,000 hours of multi-type real interaction data (covering internet videos, teleoperation, wearable device collection, and real-machine deployment). The company expects total data volume to exceed one million hours by 2026, closely mirroring Generalist's scaling path. In terms of data collection, Qianxun has reduced data acquisition costs to one-tenth of traditional methods through its self-developed fifth-generation wearable data collection devices, achieving 95% data accuracy.
These capabilities were demonstrated in the Spirit v1.5 model released in January 2026. In the RoboChallenge real-machine evaluation ranking, Spirit v1.5 achieved a 50.33% success rate, surpassing the U.S. leading model Pi0.5 and ranking first globally. A key feature of Spirit v1.5 is "zero-shot generalization," enabling it to perform complex tasks such as wiping objects, operating hinges, and handling flexible objects without additional training. This indicates that robots can autonomously complete operations in unfamiliar scenarios by understanding physical laws, rather than relying on pre-set programs.
In both Scaling Law methodology and low-cost, diversified data collection techniques, leading domestic embodied intelligence companies have reached a high degree of consensus with international peers like Generalist. In this field, data serves as the fuel for model evolution, while technical direction determines a company's trajectory.
Han Fengtao, co-founder of Qianxun AI, acknowledged past gaps but emphasized the significance of Spirit v1.5's top ranking: "We were catching up in the first three stages, but in the fourth stage, China has truly advanced side by side with overseas competitors." Additionally, domestic companies have made breakthroughs in hardware design. For instance, Qianxun opted for a three-fingered dexterous hand, which is more challenging than Generalist's two-fingered scissor design. Although the three-fingered hand requires finer degrees of freedom and force control, it offers higher operational limits and stronger generalization capabilities, providing richer data dimensions for embodied models.
Meanwhile, Qianxun AI's data team is expected to expand to 1,000 members by April 2026, bolstering the company's scaling strategy and representing a significant advantage for domestic embodied intelligence firms.
Commercial validation is accelerating. While several embodied intelligence companies secured funding of around 1 billion yuan in February, leading to a concentration of top players, Qianxun's additional 1-billion-yuan funding in April indicates that investors are placing greater emphasis on commercialization capabilities alongside technological advancement.
On March 19, Qianxun AI signed a strategic cooperation agreement with JD.com. Moz robots were fully integrated into JD MALL's smart retail场景, achieving stable real-machine operation in high-precision, non-standardized coffee-making scenarios. Each cup of coffee produced contributes real-world training data to the model. In industrial applications, Qianxun deployed the world's first humanoid embodied intelligence production line at CATL's Zhongzhou base. Moz robots have produced nearly 1,000 batteries on the PACK production line, achieving mass production.
This addresses the earlier question: Why are industrial investors willing to increase their bets within a month? Qianxun AI is pursuing a dual strategy—aiming for top-tier embodied models technologically while ensuring practical, grounded implementation. This approach creates a potential closed loop between research and commerce, likely making the company attractive to investors recently. From CATL's super factory to JD's smart retail, every real-machine operation contributes high-quality training data to the Spirit model. The dual drive of industrial and commercial applications allows Qianxun's technological capabilities to be cross-validated in two distinct scenarios, while also providing diverse, real-world data sources for its data flywheel.
Industrial capital is entering the embodied intelligence sector collectively. As an independent investment institution under Lei Jun, Shunwei Capital's investment moves often reflect independent judgments on leading players in a sector. Its investments neither directly compete with Xiaomi's ecosystem nor strategically align with its technological direction, a trend evident in areas like dexterous hands and embodied intelligence core technology. Under this investment logic, Qianxun AI has become Shunwei's most heavily invested company in the embodied intelligence field. In the February funding round, Shunwei participated as an existing investor; in the April round, it led the investment with significant capital, demonstrating rare consecutive backing within a month.
This highlights a key difference between industrial and financial capital: while financial capital can bet on sectors and wait for industry breakthroughs, industrial capital must validate capabilities. Whether it's Foxconn, BYD, CATL, or Xiaomi, all industrial investors seek promising targets but adopt a more pragmatic selection approach. This also explains the first-time joint investment by Lei Jun and Jack Ma in the embodied intelligence sector.
Han Fengtao predicts, "2026 for embodied intelligence is equivalent to 2023 for large language models in China. People will see that scaling is feasible and believe that the capabilities of embodied large models are about to rise rapidly."
Comments