Funding

XPENG Robotics Raises Over $900 Million in Record Embodied AI Round

XPENG Robotics has raised over $900 million in its first external funding round, valuing the humanoid robot unit at over $6.3 billion. The round, led by IDG Capital with participation from Gaorong Ventures and strategic backing from Tencent and Alibaba, marks the largest single-round private financing in China's embodied AI industry. Funds will support R&D, mass production of the IRON humanoid robot, and global expansion.

Neura News

Neura News

Neura Market Editorial

August 24, 20266 min read
XPENG Robotics Raises Over $900 Million in Record Embodied AI Round

XPENG Robotics has secured more than $900 million in its first external funding round, a deal that values the humanoid robot unit at over $6.3 billion and marks the largest single-round private financing in China's embodied AI industry.

The announcement, made on August 24, 2026, brings together a lead investment from IDG Capital, participation from Gaorong Ventures, and strategic backing from Tencent and Alibaba. The round is more than nine times the size of the unit's 2022 Series A, which raised over $100 million.

A Record Round for Embodied AI

XPENG says both the size and valuation set new records for the sector. The capital will fund software and hardware research and development, training and iteration of Physical AI models, high-quality data generation, construction of end-to-end mass production facilities, and global commercial expansion.

The round extends IDG Capital's relationship with XPENG Robotics, which began with that 2022 Series A. At the time, the unit was known for a quadruped robotic pony aimed at households, and the round was the largest single investment in China's bionic robotics sector in two years.

Now the focus has shifted to IRON, XPENG's next-generation humanoid robot. The robot is expected to enter mass production by the end of 2026, with initial deployment at XPENG's own stores and campuses before a commercial launch and deliveries in China and overseas markets in 2027.

He Xiaopeng, chairman and CEO of XPENG, tied the raise to years of internal development. "Our ambition is for IRON to become a trusted partner for people and a meaningful part of everyday work and life," he said, citing 12 years of full-stack in-house R&D as the foundation for the robotics push.

The IRON Humanoid Robot

IRON is not a small machine. The humanoid has 76 degrees of freedom across its body and 21 in each hand, giving it a range of motion that mirrors human movement. It runs on three Turing AI chips delivering up to 2,250 TOPS of on-device compute, and XPENG says it executes tasks autonomously without remote operation.

The design uses a fully enclosed flexible lattice structure, and the robot's form factor allows it to draw on behavioral data generated through everyday human activity. XPENG's AI Day presentation in November 2025 detailed the underlying stack: a humanoid spine with bionic muscles, fully covered flexible skin, all-solid-state batteries, and a combination of VLT, VLA, and VLM models.

Chips, controllers, motion modules, and dexterous hands are all developed in-house. The company applies automotive-grade quality standards and existing EV manufacturing infrastructure to robot production, a carryover that Gaorong Ventures highlighted in its comments on the round.

Strategic Logic and Investor Support

The strategic logic is the carryover from XPENG's EV business to robotics. XPENG's physical world foundation model, Turing AI chips, and AI infrastructure are cited as the basis for the raise. The company's EV business provides chips, controllers, motion modules, and dexterous hands that transfer directly to the robot line.

IDG Capital described the embodied AI industry as moving from technical breakthroughs toward scalable manufacturing and commercial deployment. Tencent and Alibaba's cloud and distribution assets complement a physical AI push, which is why both joined as strategic investors.

The #1 Newsletter in AI

Stay ahead of the AI curve

The most important updates, news, and content — delivered weekly.

No spam. Unsubscribe anytime.

The structure of the deal is notable. This is external capital coming into a controlled subsidiary of a dual-listed public company, not a startup round. XPENG will retain controlling ownership of the robotics business when the deal closes, and the unit will remain consolidated in the group's financial statements.

The round establishes a market valuation for the business and strengthens long-term incentives for senior executives and key talent. Share purchase agreements have been signed but still require closing.

The Data Flywheel Argument

The data argument is as important as the hardware one. XPENG owns both the robot fleet and the factory, which means each deployed robot feeds a production-data-model flywheel that improves the foundation model. This flywheel framing is the standard pitch across physical AI.

The company says it can use behavioral data from everyday human activity because IRON's form factor mirrors human movement. That data feeds the models that make the robot more capable, which in turn makes it more useful in real settings.

Veeda AI, which raised a $90 million seed round earlier in the month, is a comparable example of the data flywheel pitch in physical AI. The company's world-model training data approach reflects the same logic that XPENG is applying at a much larger scale.

Market Comparables and Context

The humanoid market has other large players. DeepSeek's recent investment in Unitree's Shanghai IPO valued Unitree at $9 billion, a figure that shows how far the sector has come. Gravis Robotics raised a $200 million Series A, an order of magnitude smaller than XPENG's round.

The round is the second act of a much smaller first one. The 2022 Series A of more than $100 million was led by IDG Capital, with XPeng and other long-term investors participating. That round funded a quadruped pony, not a humanoid, and the sector was far less developed.

Now the sector is drawing strategic investors from outside robotics. Tencent and Alibaba's participation signals that cloud and distribution assets are seen as complementary to physical AI. The round's size and valuation set new records for the embodied AI sector in China, according to XPENG.

The announcement lays out a dated sequence for mass production and commercial launch. Mass production by the end of 2026, initial deployment at XPENG's stores and campuses, and then official launch with deliveries in China and overseas markets in 2027.

The deal structure means XPENG keeps consolidation while the unit gets an outside valuation mark and strategic investors. That arrangement gives the robotics business financial independence without losing the parent company's manufacturing and technology base.

The round is a bet on the physical AI sector's trajectory, and the numbers are large. Over $900 million raised, a $6.3 billion valuation, and a robot with 76 degrees of freedom and 2,250 TOPS of compute. The sector is moving from research to production, and XPENG is positioning itself at the front of that shift.

Related on Neura Market

More from Neura News

Industry

Travelers builds its own LLM to cut AI costs and sharpen insurance answers

Travelers Insurance has developed its own proprietary large language model, TravelersLLM, to reduce AI costs and improve performance on insurance-specific queries. The model, unveiled in June 2026, is cheaper to run than frontier models and is used alongside them, with queries routed based on task complexity. This move reflects a broader industry trend of managing AI expenses by using multiple models and routing tasks based on cost and quality.

Aug 24·4 min read
Developer

Roblox's "Prompt to Prod" Aims for Fully Autonomous Software Development

Roblox is advancing toward fully autonomous software development with its 'Prompt to Prod' initiative, led by Senior Director Andrew Swerdlow. The company uses AI agents for code review and experiment authoring, achieving a 68-70% acceptance rate on AI suggestions. By mining institutional knowledge from 700,000 pull requests and implementing layered security, Roblox aims to reduce human touchpoints and accelerate development while managing risks.

Aug 24·59 min read
Developer

Canonical funds Bristol PhD to test AI translation of C codebases into Rust

Canonical is co-funding a three-year PhD at the University of Bristol to explore using large language models to translate large C codebases into safe, maintainable Rust. The project will test the approach on components like snap-confine and AppArmor, with the goal of improving software reliability. The Register expresses skepticism about the feasibility, particularly the challenge of decomposing code.

Aug 24·6 min read