Humanoid AI Startup Delta Intelligence Raises $14M

Beijing's Delta Intelligence just closed a roughly $14 million round to teach humanoid robots how to walk, think, and manipulate the world in three dimensions.

Delta Intelligence, incubated by the Beijing Institute for General Artificial Intelligence, has completed three funding rounds totaling over RMB 100 million, with participation from GL Ventures and strategic robotics OEMs. Founded in January 2026, the company is building humanoid foundation models (基础模型) paired with a proprietary 3D world model engine to support full-body loco-manipulation—think robots that can navigate factory floors and operate tools with human-like dexterity. Its founder, Ma Xiaojian, trained at Tsinghua and UCLA and previously worked with Google Robotics and NVIDIA Research, leads a team that includes alumni from Peking University, UCLA, Google, DeepMind, and NVIDIA.

The startup has not just built technology in a vacuum. It co-initiated the Tongzhi Brain Alliance with Leju Robotics and has already delivered industrial applications with FAW Hongqi, STEP, China Southern Power Grid, and China Mobile. Its technology has been showcased at prominent venues like the Zhongguancun Forum and the World Humanoid Robot Games, underscoring a dual aim: push core AI capabilities while proving the industrial value of humanoid platforms.

For observers of China’s robotics ecosystem, Delta Intelligence’s funding trajectory signals a deeper push to pair domestic AI foundation work with real-world hardware deployments. The emphasis on humanoid models—rather than narrow task-oriented systems—speaks to a longer-term bet that general-purpose, adaptable intelligence will shrink the gap between software breakthroughs and factory-floor execution. In practice, that means a convergence of AI labs, automotive and industrial OEMs, utilities, and telecom incumbents around a shared platform for locomotion, manipulation, and perception.

Two strands of Chinese context help explain this momentum. First, policy and funding streams have increasingly favored efforts to build self-reliant AI and robotics capabilities in-country, reducing reliance on foreign models and tooling while accelerating domestic hardware-software integration. Delta Intelligence’s incu bation within a Beijing research institute, combined with strategic investors and OEM partners, fits a pattern of state-linked, market-driven collaboration that China policymakers have been quietly encouraging. Second, the visible corporate anchors—FAW Hongqi in automotive, China Mobile in telecom, and China Southern Power Grid in energy—point to a future where humanoid and service robots support core industrial tasks across sectors that are central to China’s modernization push.

Ownership clarity remains limited. The company’s press materials describe private funding and an academic incubator affiliation, with GL Ventures described as an investor and strategic OEMs as participants. In practice, that mix typically yields a hybrid structure: private-led governance with state or state-affiliated backing, potentially easing access to pilots, data, and procurement channels.

For global manufacturers evaluating supplier ecosystems, Delta Intelligence highlights a plausible pathway for China-based robotics vendors to scale from lab-grade demos to site-wide deployments. The key value proposition—foundation-model-driven control of humanoid bodies in real environments—has clear implications for factories seeking flexible automation and for robot-component suppliers aiming to ride a domestic wave of demand.

What to watch next: pilot results with industrial partners, the scale of compute and data investments behind the 3D world model engine, and how the Tongzhi Brain Alliance evolves into more industrial collaborations. If the trajectory holds, Delta Intelligence could become a bellwether for how China translates cutting-edge AI research into deployable humanoid robotics across manufacturing and services.

Sources

Delta Intelligence Raises USD 14 Million+ to Build Humanoid Foundation Models