Agibot

Agibot, legally Shanghai Zhiyuan New Creation Technology Co. Ltd. (上海智元新创科技, Shanghai Zhiyuan Xinchuang Keji) and trading as 智元机器人 (Zhiyuan Jiqiren), is a Chinese humanoid-robotics company headquartered in Shanghai's Lin-gang Special Area, founded in February 2023 by Peng Zhihui (彭志辉), the engineer known on Bilibili as "Zhihui Jun" (稚晖君) who had previously been recruited into Huawei's "Genius Youth" (天才少年) program. Agibot develops the Yuanzheng (远征, "Expedition") A1 and A2 humanoid robots, the Lingxi (灵犀, "Spiritual") X-series and D-series platforms, and the Genie G-series, and as of 2026 is among the most-valued private humanoid-robotics companies in China alongside Unitree and UBTECH.

At a glance

• Founded: February 2023 in Shanghai by Peng Zhihui and a small founding team of former Huawei and academic engineers.
• Status: Private. Reported valuation in the $10 billion to $15 billion range as of late 2025, based on industry coverage of completed and in-progress funding rounds.
• Funding: Multiple private rounds from Hillhouse Capital, HongShan (formerly Sequoia China), Tencent, BYD, Lenovo Capital, and additional Chinese venture firms. Cumulative disclosed funding not itemised; individual rounds reported in the hundreds of millions of dollars.
• CEO: Peng Zhihui, founder, also serving as chief technology officer.
• Open weights: Partial. Agibot has released selected open-source components including the AgiBot World dataset (one of the largest publicly released humanoid-manipulation datasets as of 2025) and selected hardware reference designs. Core foundation models and proprietary AI stack remain closed.
• Flagship products: Yuanzheng A1 (October 2023, the company's first humanoid demonstration); Yuanzheng A2 (2024 production humanoid); Lingxi X1, X2 (smaller-form-factor humanoid and wheeled-biped platforms); Lingxi D-series (industrial dual-arm manipulator); Genie G1 (2024 to 2025 announcement, mass-production humanoid).

Origins

Peng Zhihui first achieved national visibility in China through his Bilibili technology video channel "Zhihui Jun," which by 2023 had accumulated several million followers with content covering robotics builds, hardware-engineering deep-dives, and personal-project demonstrations. His most-viewed videos included a self-driving bicycle, a robotic arm capable of threading a needle, and a homemade gantry plotter. The Bilibili audience translated directly into recruiting pipeline and consumer brand recognition for Agibot at founding.

Peng was recruited into Huawei's "Genius Youth" program (天才少年招聘计划) in 2020 directly out of his postgraduate engineering studies, on the maximum compensation tier offered under that program. The Genius Youth scheme was a Ren Zhengfei initiative to recruit the most-exceptional engineering graduates in China directly into Huawei's research and engineering organisations, and Peng's recruitment was publicly profiled in Chinese technology media as one of the program's most-prominent cases. He left Huawei in late 2022 or early 2023 to found Agibot, taking a small team of Huawei colleagues with him.

The first Agibot product, the Yuanzheng A1 humanoid, was unveiled in August 2023 at a Shanghai launch event approximately six months after the company's founding. The development timeline was unusually compressed for a humanoid-robotics company; comparable Western companies typically take 18 to 36 months from founding to a first public demonstration. The A1's launch was widely covered in Chinese technology media and was the first humanoid-robotics company launch to draw the level of consumer-and-investor attention that has since become standard for the Chinese humanoid cluster.

The A2 production humanoid followed in 2024, with reported initial shipments in the high hundreds of units. The Lingxi X1 and X2 (smaller form factors and wheeled-biped variants) and the Genie G1 mass-production humanoid expanded the product line through 2024 and 2025. By 2025 Agibot had positioned itself as the foundation-model-and-data-focused Chinese humanoid competitor, in contrast to Unitree's cost-and-volume positioning and UBTECH's enterprise-customer positioning.

Mission and strategy

Agibot's strategic positioning is on AI-foundation-model integration and on data-collection-and-training infrastructure for embodied intelligence. The company has emphasised the AI side of humanoid robotics in its public communications more heavily than the mechanical-engineering side, in contrast to Unitree's positioning. Peng Zhihui's public communications have consistently framed Agibot's competitive position around three pillars: hardware platforms, robotics foundation models, and data infrastructure for training those models at scale.

The AgiBot World dataset, released open-source in late 2024 to early 2025, is the most-visible expression of the data-infrastructure pillar. The dataset comprises millions of teleoperation episodes covering manipulation tasks, collected from a fleet of Yuanzheng A2 platforms operated by trained human teleoperators in dedicated data-collection facilities. The release positioned Agibot as the Chinese counterpart to Physical Intelligence and Skild AI on the data-and-foundation-model side of robotics, in contrast to the hardware-platform positioning of Unitree, UBTECH, and the others.

The product strategy spans research-grade, industrial-grade, and consumer-adjacent SKUs. The Yuanzheng A2 platform targets industrial and enterprise customers. The Lingxi X1 and X2 target research and developer markets. The Lingxi D-series targets industrial-manipulation customers (dual-arm stationary manipulator rather than full humanoid). The Genie G1 is the mass-production-targeted SKU intended to drive volume for the foundation-model training pipeline.

Models and products

• Yuanzheng A1 (2023). The company's first humanoid platform. Bipedal, electric-actuator-driven, two-arm manipulation, approximately 1.75 metres tall and 53 kilograms in weight. Development-prototype tier rather than production volume.
• Yuanzheng A2 (2024). Production humanoid platform. Reported initial shipments in the high hundreds of units. Targeted at enterprise and industrial customers.
• Lingxi X1 (2024). Smaller-form-factor humanoid in the 1.3-metre height range, targeted at research and developer markets.
• Lingxi X2 (2024 to 2025). Wheeled biped variant; lower-cost developer-and-research SKU. Open-source hardware reference designs released for portions of the platform.
• Lingxi D-series (2024). Dual-arm stationary manipulator (not a humanoid). Industrial manipulation customer focus.
• Genie G1 (2024 to 2025). Mass-production humanoid. Target volume and target price point have been less publicly itemised than competitor SKUs.
• AgiBot World dataset (open-source, late 2024 to early 2025). Multi-million-episode teleoperation dataset for robotics-foundation-model training.

Benchmarks and standing

Like the broader humanoid category, Agibot's standing is measured through production volume, deployment breadth, and proprietary capability rather than through standardised benchmarks. The company's distinguishing position in the Chinese cluster is on the foundation-model-and-data side rather than on the hardware side, and the AgiBot World dataset release has been the most-visible public artifact establishing that position.

Production volumes have not been officially itemised. Industry coverage in 2025 placed cumulative Yuanzheng A2 shipments in the low thousands of units, primarily to industrial-pilot deployments with enterprise customers in China. The company's manufacturing capacity expanded substantially through 2024 and 2025, with manufacturing facilities in Shanghai's Lin-gang Special Area expanding multiple times to support higher volumes.

The robotics-foundation-model capability has been demonstrated through internal model releases and through partnerships with academic robotics-research groups, but specific capability claims and external benchmarking have been less detailed than the language-model category produces. The most-visible signal of the underlying capability is the company's hiring and partnership pipeline, which has drawn substantial talent from Chinese universities (Tsinghua, Shanghai Jiao Tong, Zhejiang University) and from former employees of DeepSeek, Alibaba Qwen, and Huawei research organisations.

Leadership

• Peng Zhihui (Founder, Chief Executive Officer, Chief Technology Officer). Engineering background with a postgraduate degree from the University of Science and Technology of China (USTC). Recruited into Huawei's Genius Youth program in 2020 at the maximum compensation tier. Maintained a high-profile Bilibili technology channel as "Zhihui Jun" with several million followers prior to and during the founding of Agibot. Continues to lead Agibot through the current expansion phase.
• Senior engineering team. Approximately 700 to 1,000 employees as of late 2025, with senior leadership drawn from former Huawei, DeepSeek, Alibaba, and the major Chinese university robotics-research groups. The company has been notably aggressive on recruitment compensation, with reported senior-engineer packages at or above the levels offered by the major Chinese frontier-model labs.

Funding and backers

Agibot has raised multiple private rounds since founding in early 2023. Reported investors include:

• Hillhouse Capital. Multi-round participant; one of the largest single backers.
• HongShan (formerly Sequoia China). Multi-round participant.
• Tencent. Strategic investor.
• BYD. Strategic investor; potential customer for automotive-factory humanoid deployments.
• Lenovo Capital. Multi-round participant.
• Bilibili. Strategic minor investor (the Bilibili connection mirroring Peng Zhihui's content-creator background).
• Additional unspecified Chinese venture firms participated in subsequent rounds.

Specific round sizes and post-money valuations have not been publicly itemised in detail. Industry coverage in late 2025 placed the valuation in the $10 billion to $15 billion range, with multiple unconfirmed reports of in-progress rounds at higher post-money values. The company has not filed for IPO as of the current date, though industry speculation has consistently included Agibot among the candidates for a 2026 to 2027 Chinese-exchange listing.

Industry position

Agibot is the leading Chinese humanoid-robotics company on the foundation-model-and-data axis, paired in the domestic-leader position with Unitree (cost-and-volume axis) and UBTECH (enterprise-customer axis). The three-way structure is the most-frequently-referenced framing in Chinese technology media coverage of the humanoid cluster, with the other entrants (Fourier, EngineAI, XPeng, Xiaomi) positioned around the edges of those three primary competitors.

Domestically the company competes with the full Chinese humanoid cluster. The competitive question against Unitree is whether Agibot's AI and data infrastructure produces enough capability advantage to justify the higher likely production cost. The competitive question against UBTECH is whether Agibot's enterprise pipeline can match UBTECH's longer-established customer relationships with automotive manufacturers. The competitive question against the smaller competitors is around speed of execution and capital efficiency.

Internationally Agibot competes with the US cohort (Figure AI, Tesla AI, Apptronik, Agility Robotics, 1X) on the AI-foundation-model axis. The competitive dynamic mirrors the broader US-vs-China dynamic in foundation-model development: Agibot's positioning depends on whether Chinese AI-and-data infrastructure produces capability that competes with US-cohort offerings, and on whether the Chinese cost structure produces sustainable pricing advantage. The AgiBot World dataset release in late 2024 to early 2025 was a notable signal that Agibot intends to compete with Physical Intelligence and Skild AI on the data-asset axis, where US-cohort companies have historically held an advantage.

Competitive landscape

• Chinese cluster: Unitree, UBTECH, Fourier Intelligence, EngineAI, XPeng Robotics, Xiaomi Robotics. Direct domestic competitors with different vertical positionings.
• US cohort: Figure AI, Tesla AI, Apptronik, Agility Robotics, Boston Dynamics, 1X. International competitors with stronger Western enterprise relationships.
• Robotics-foundation-model labs: Physical Intelligence, Skild AI. Direct competitors on the AI-and-data-infrastructure axis, though those labs do not produce their own humanoid hardware.
• Other geographies: Sanctuary AI (Canada), Neura Robotics (Germany), Rainbow Robotics (South Korea), Kawada Robotics (Japan), Engineered Arts (UK).

Outlook

Open questions and watchable signals over the next 6 to 18 months:

• Foundation-model capability disclosure. Agibot has emphasised the AI-foundation-model side of humanoid robotics in its public positioning, but specific capability claims and external benchmarking have been limited. Detailed external benchmarks, dataset disclosures beyond AgiBot World, and any third-party capability comparisons against the US robotics-foundation-model labs will materially shape the company's standing.
• Production volume trajectory. The high-hundreds-of-units initial-shipment volumes from 2024 need to scale into the thousands or tens of thousands annually to justify the company's reported valuation. Manufacturing capacity expansion in Lin-gang has been publicly visible; resulting output volumes have been less so.
• Enterprise customer disclosures. Agibot has been less visible than UBTECH on specific automotive-factory and industrial-pilot customer relationships. Named-customer announcements in 2026 would shift the competitive position against UBTECH.
• IPO timing and venue. Multiple unconfirmed industry reports have placed Agibot on track for a Chinese-exchange listing in 2026 or 2027. Filing-of-record timing and venue (Shanghai vs Hong Kong vs Hong Kong Stock Exchange) will be informative for the company's strategic direction.
• Peng Zhihui's role evolution. Founder-CEO and founder-CTO dual roles are sustainable at smaller scale but increasingly difficult above 1,000 employees and across multiple product lines. The company's senior-leadership build-out around Peng (named COO, named president, board structure) will be informative for the long-term governance trajectory.

Sources

• Agibot official website.
• Agibot World dataset on GitHub.
• Peng Zhihui's Bilibili channel, the public-content lineage that preceded the company.
• Companion profiles: Unitree Robotics, UBTECH, XPeng Robotics, Xiaomi Robotics, Fourier Intelligence, EngineAI for the broader Chinese humanoid cluster context, and Figure AI, Tesla AI, Apptronik, Agility Robotics, Boston Dynamics, 1X for the US cohort context.