XPeng’s humanoid robot arm grabs $900m to scale physical AI
XPeng has just closed the largest single-round private capital raise in China’s physical AI industry, netting over $900 million at a $6.3 billion valuation for its IRON humanoid robot unit. The financing, led by IDG Capital with Tencent and Alibaba joining as strategic backers, hands XPeng fresh firepower to push its in-house robotics stack from lab to factory floor.
Inside the full-stack push
XPeng says the money will back long-term investment in what it calls a full-stack physical AI development pipeline—software and hardware R&D, physical-AI model training, high-quality data generation, and end-to-end mass production lines modeled on its EV playbook. Executive and staff incentive plans are also part of the package, underscoring XPeng’s bet on talent retention as competition for robotics engineers heats up.
Hardware with 76 degrees of freedom
IRON’s hardware is built from the ground up: a fully enclosed flexible lattice frame for safety and aesthetics, 76 degrees of freedom across the body (21 per hand), and three in-house Turing AI chips delivering up to 2,250 TOPS for on-device inference. XPeng argues this local compute stack enables low-latency execution of its physical AI foundation model, letting the robot tackle complex tasks without cloud dependency.
What it means for investors—and the industry
Even as XPeng’s core EV business grapples with domestic and Tesla-led competition, investors have priced the robotics spin-out at $6.3 billion, a valuation that now trades separately from the parent’s falling share price. The round signals growing confidence that humanoid robots can be a standalone, high-margin opportunity rather than just an R&D curiosity.
Why it matters
The scale of this raise underscores how quickly physical AI is moving from prototype to investable venture. For peers and startups, it sets a new benchmark for capital intensity and execution risk. For end users, the bet on in-house chips and automotive-grade production suggests robots may reach factories—and homes—faster than many analysts expected, but only if hardware and software can be tightly integrated at volume.
Source: AI News. AI-assisted editorial synthesis — TechnoExpress.