Huawei Habo has made a major move in the field of physical artificial intelligence by investing in GigaAI, a young but fast-growing startup focused on world-model-based general intelligence. The investment marks a key milestone in Huawei’s strategy to expand deeper into embodied intelligence and next-generation autonomous systems.

Funding Overview

GigaAI recently completed its hundred-million-yuan Series A1 funding round in early November. The round was jointly led by Huawei Habo Investment and Huakong Fund, showing strong backing from major industry players.

This comes just two months after GigaVision, part of GigaAI’s ecosystem, secured several hundred million yuan in its Pre-A and Pre-A+ rounds in August. Together, the rapid funding rounds signal rising confidence in GigaAI’s technology and potential.

About GigaAI

Founded in 2023, GigaAI is positioned as China’s first startup dedicated to world model research in physical AI. The company focuses on developing world-model-driven general intelligence that can operate effectively in real-world environments.

Its mission is to build the next generation of intelligent systems that understand, predict, and interact with the physical world, enabling breakthroughs in autonomous driving, robotics, and embodied AI.

GigaAI’s Product Ecosystem and Technology

GigaAI operates through a full-stack ecosystem that combines both software and hardware:

• GigaWorld Platform: Built for driving intelligence and advanced embodied intelligence.
  
• GigaBrain Foundational Model: A core model enabling adaptive, world-model-based AI decisions.
  
• Maker General Embodied Ontology: A structured knowledge layer for physical agents and developers.
  

By integrating these components, GigaAI provides end-to-end solutions for complex physical environments. Its technology is already being applied to areas such as autonomous driving, robotics, and general embodied AI systems.

Huawei’s Strategic Vision Behind the Investment

The investment reflects Huawei’s broader shift from traditional VLA (Vision-Language-Action) models, which rely heavily on language models, to a WA (World Action) strategy.

WA emphasizes using vision and physical signals directly to control vehicles and intelligent agents. This approach, supported by Huawei Intelligent Automotive Solutions BU CEO Jin Yuzhi, is now a central part of Huawei’s long-term AI roadmap.

By backing GigaAI, Huawei aims to accelerate developments in world models, physical AI, and embodied intelligence, strengthening its position in the next wave of autonomous and intelligent mobility technologies.

Impact on Everyday Consumers and Global Competition

For the general public, this development could translate into practical improvements such as safer autonomous vehicles, smarter home and service robots, and more reliable AI assistants that can understand and react to real-world situations, not just digital inputs.

Analysts believe this shift toward physical AI may trigger a technological transformation similar to the rise of smartphones, bringing intelligent machines into daily life in more meaningful ways.

On a global scale, Huawei’s move increases competitive pressure on the United States, particularly in fields like robotics, autonomous driving, and world-model research. As China strengthens its capabilities in physical AI, it could intensify the ongoing technology rivalry and prompt the US to accelerate its own advancements in next-generation autonomous systems.

Industry Significance and Future 

Industry experts view this investment as more than financial support. It represents a strategic partnership that could speed up the development of world-model-driven AI and expand Huawei’s capabilities in real-world automation.

GigaAI’s strong full-stack capabilities and rapid funding growth position it as a rising leader in physical AI. The collaboration is expected to drive new advances in autonomous driving, robotics, and large-scale embodied AI deployment, marking a significant step toward the future of intelligent systems in China.

The post Huawei’s Surprise Investment Could Redefine the Future of Robots and Self-Driving Cars appeared first on Gizmochina.

XPENG has taken a major step into the future of mobility and artificial intelligence with the unveiling of its latest technologies at the 2025 XPENG AI Day, themed “Emergence.” Held at XPENG Science Park in Guangzhou, the event showcased the company’s biggest achievements yet in Physical AI, including the launch of XPENG VLA 2.0, Robotaxi, Next-Gen IRON humanoid robot, and flying car systems under ARIDGE.

The event marked XPENG’s transformation from a “future mobility explorer” to a “global embodied intelligence company.”

A New Era of Physical AI

Chairman and CEO He Xiaopeng opened the event by explaining XPENG’s vision for the coming decade. He said that just as AI revolutionized the digital world and electricity replaced oil in the physical world, the next revolution will come from the combination of both, Physical AI.

XPENG is now the only company in China to have built a full-stack self-developed Physical AI system, covering chips, large AI models, and intelligent hardware. This system will serve as the technological backbone for its AI cars, Robotaxis, humanoid robots, and flying vehicles.

He Xiaopeng announced XPENG’s upgraded identity as “a mobility explorer in the Physical AI world and a global embodied intelligence company.”

XPENG VLA 2.0: Redefining Large Models for the Real World

One of the biggest highlights of the event was the unveiling of the XPENG VLA 2.0 large model, described as the “operating system of the Physical AI world.”

Unlike traditional models that depend on language translation, VLA 2.0 follows a “Vision–Implicit Token–Action” structure. This allows the model to generate direct actions from visual signals, enabling true end-to-end intelligence.

The model can perform self-learning and understand the real world, making it adaptable across different platforms, from cars to robots to flying vehicles. With 72 billion parameters, trained on 100 million driving clips (equal to 65,000 years of human driving), it’s one of the largest and most advanced models ever created.

XPENG’s powerful 30,000-GPU cloud cluster and optimized software allow VLA 2.0 to run efficiently, even on vehicles. The model brings features like “Narrow Road NGP”, which improves performance in complex urban environments by 13 times, and gesture recognition for real-world interaction.

VLA 2.0 will begin user testing in December 2025 and roll out to XPENG Ultra models in early 2026. XPENG also announced that the model will be open-sourced globally, with Volkswagen as the first launch partner.

Robotaxi: China’s First Full-Stack Self-Developed Autonomous Taxi

XPENG also revealed its Robotaxi, China’s first fully self-developed, mass-producible autonomous taxi. The Robotaxi is powered by four Turing AI chips with a combined computing power of 3000 TOPS, the highest in the world.

It relies entirely on pure vision, removing the need for lidar or high-definition maps. Designed for driverless operation, it features a dual-redundancy system for safety and integrates XPENG’s VLA 2.0 and VLM models for real-time decision-making.

The Robotaxi includes the industry’s first sun visor external display, allowing it to “communicate” with pedestrians through visual signals.

XPENG also introduced a new intelligent trim called “Robo”, launching in 2026. It shares the same technology as the Robotaxi but is built for personal use. The company plans to open the Robotaxi SDK for global collaboration, with Amap (Alibaba’s mapping platform) as its first global ecosystem partner.

Next-Gen IRON: The Most Human-Like Robot Yet

Another major launch was XPENG’s Next-Gen IRON humanoid robot, featuring 82 degrees of freedom, bionic muscles, a flexible skin, and a humanoid spine. It even supports custom body designs and performs graceful, catwalk-style walking.

Powered by three Turing AI chips with 3000 TOPS, it combines three large models, VLT, VLA, and VLM,  to achieve high-order intelligences like conversation, movement, and interaction.

XPENG emphasized “extreme anthropomorphism” as its design philosophy, aiming to make robots more adaptable to real-world scenarios. The Next-Gen IRON also features all-solid-state batteries, offering lightweight and safe performance.

A new “Fourth Law of Robotics” was introduced: “Privacy data does not leave the robot.”

XPENG plans mass production by the end of 2026, starting with applications in industries such as retail, tourism, and factory inspection. Baosteel Stock will partner with XPENG to deploy the robots in industrial settings.

ARIDGE Flying Cars: Expanding into Low-Altitude Mobility

XPENG’s flying vehicle division, ARIDGE, showcased two parallel systems for low-altitude travel:

1. A868 Full Tiltrotor Hybrid Flying Car: a 6-seater business model capable of 360 km/h speeds and 500 km range.
   
2. Land Aircraft Carrier: a personal flying vehicle designed for short-range travel, already nearing mass production with 7,000 global orders.
   

The Land Aircraft Carrier includes an intelligent cockpit and a single-stick four-axis control system, making it easy for beginners to fly. Production has already begun at XPENG’s new flying car factory, capable of producing one aircraft every 30 minutes.

In 2026, ARIDGE will collaborate with the Dunhuang Government to launch China’s first low-altitude self-driving tourism route.

XPENG Science Park: Where Innovation Meets Collaboration

The event was held at XPENG Science Park, home to over 10,000 employees working across AI, robotics, cars, and aviation. The company says its multidisciplinary teams are driving daily innovations that fuel the emergence of Physical AI.

Looking ahead, XPENG will begin mass production and rollout of its AI-driven mobility products from 2026, paving the way for a new era where cars, robots, and flying vehicles coexist under one intelligent ecosystem.

XPENG’s message is clear: The future of mobility is not just smart, it’s embodied. With its breakthroughs in Physical AI, the company is positioning itself at the forefront of a global transformation that blends intelligence with motion, bringing tomorrow’s innovations into today’s world.

The post China’s XPENG Shocks the World with Full-Stack Physical AI: Meet the IRON Robot and Flying Car appeared first on Gizmochina.

AgiBot, a robotics company focused on embodied intelligence, has achieved a major milestone by successfully deploying its Real-World Reinforcement Learning (RW-RL) system on a pilot production line with Longcheer Technology. This marks the first-ever real-world application of reinforcement learning in industrial robotics, bridging years of AI research with practical use in manufacturing.

Transforming the Future of Manufacturing

For decades, precision manufacturing has relied on rigid automation systems that require complex fixtures, extensive tuning, and high setup costs. These systems are difficult to reconfigure when new products are introduced or production layouts change. Even modern robots equipped with vision and force-control systems often suffer from high sensitivity to parameter changes, resulting in long deployment times and expensive maintenance.

AgiBot’s RW-RL solution aims to solve these long-standing challenges by enabling robots to learn directly from real-world interactions. Instead of depending on pre-programmed instructions, the robots can adapt to changing conditions on the factory floor and refine their performance in real time.

How AgiBot’s RW-RL System Works

The Real-World Reinforcement Learning system allows robots to acquire new skills within minutes instead of weeks. Once deployed, they can maintain industrial-grade stability and perform tasks continuously without performance degradation. The system requires minimal hardware adjustments when switching between tasks or products, significantly reducing downtime and costs.

AgiBot’s technology also enables robots to autonomously adjust for part variations and manufacturing tolerances, ensuring consistent precision. This adaptability makes the system ideal for flexible, multi-product production lines, paving the way for smarter and more versatile factories.

From Research to Real-World Deployment

The RW-RL breakthrough is built on years of academic research into the stability and efficiency of reinforcement learning. Led by Dr. Jianlan Luo, Chief Scientist at AgiBot, the team successfully transformed complex AI algorithms into a deployable industrial system. The pilot deployment with Longcheer Technology validated the system’s performance under near-production conditions, confirming it as the first verified industrial application of reinforcement learning in manufacturing robotics.

Following the successful pilot, AgiBot and Longcheer plan to expand the use of RW-RL to consumer electronics and automotive component production. Their goal is to create modular, easily deployable robotic solutions that can integrate seamlessly into existing factory systems.

With this achievement, AgiBot has not only demonstrated a major advancement in robotic intelligence but also taken a crucial step toward the future of self-learning, adaptable manufacturing systems, bringing the promise of embodied AI closer to industrial reality.

The post AgiBot Makes History: First Robot to Learn Directly on the Factory Floor appeared first on Gizmochina.

Embodied intelligence refers to the ability of robots to combine physical actions with advanced reasoning, perception, and decision-making. Unlike traditional machines that follow fixed commands, embodied robots can interpret complex instructions, adapt to new situations, and perform tasks with human-like flexibility. This appears set to become a reality in the not-too-distant future, as an increasing number of players and nations are entering this competitive arena. Currently, the race for embodied intelligence is intensifying.

The Embodied Intelligence Race

Tesla is placing huge weight on its humanoid robot project, Optimus. The company expects it to become a major part of its future, representing up to 80% of Tesla’s overall value. Elon Musk has set an ambitious target: producing one million humanoid robots per year by 2030. Optimus is designed primarily for factory use but is also being developed for consumer applications, ranging from household assistance to personal support.

Boston Dynamics, long known for its groundbreaking robotic designs, has turned its Atlas model into a fully electric humanoid platform. Atlas is capable of performing complex multi-task demonstrations, such as handling tools or navigating obstacle courses. With Toyota’s large Behavior Model supporting its development, Atlas is being positioned as an advanced industrial solution. However, like Tesla’s Optimus, Atlas still faces the challenge of moving beyond staged demonstrations toward everyday usability.

NVIDIA is entering the robotics race with a software-first approach. Its Isaac platform and GR00T foundation models provide robotics developers with tools for simulation, reasoning, and adaptability. NVIDIA describes GR00T as “the next wave of AI,” aiming to make robots more capable of decision-making in real-world settings. This positions NVIDIA as a key enabler of robotics rather than a hardware maker, offering the “brains” for embodied intelligence.

The Challenge: Demos vs. Dependability

Despite impressive demonstrations of robots running, jumping, or manipulating tools, most humanoid robots struggle with real-life tasks. Simple activities like folding laundry, serving food, or caregiving remain much harder than advertised. The gap between marketing demos and practical dependability is still wide, slowing adoption in homes and industries.

China’s Disruptive Entry: X Square Robot and Wall-OSS

A new player from China, X Square Robot, is taking a bold step with its open-source model, Wall-OSS. This is China’s first foundational model for embodied intelligence and is designed to make robots adaptable in unpredictable real-world environments. Unlike proprietary approaches from Tesla or Boston Dynamics, Wall-OSS will be freely available on platforms like GitHub and Hugging Face.

How Wall-OSS Works

Wall-OSS uses a Shared Attention Mechanism to focus only on relevant cues, improving reaction time and reducing errors. It’s also paired with a Task-Routed Feed-Forward Network (FFN). In most legacy systems, all sensory streams, such as camera vision, spoken instructions, or motor commands, are funneled through a single processing layer, forcing the model to juggle unrelated inputs in one space. The result was bottlenecks, slow adaptation, and a tendency to misprioritize commands.

Wall-OSS addresses this differently. Shared attention allows the model to selectively focus on the most relevant cues in a scene, while the task-routed FFN processes different types of input, vision, language, and motor actions through specialized pathways. For example, visual data is processed along one optimised pathway for object recognition and spatial mapping; linguistic commands are parsed and processed through a separate pathway. Concurrently, motion processing operates independently, accounting for physical constraints and real-time feedback.

This enables robots to understand and act in context, such as following the instruction “Pick up the apple on the table and place it in the bowl” instead of treating “see apple” and “pick” as separate tasks. Routing multimodal inputs effectively mirrors human cognition, where vision/sign, hearing, and motor planning seamlessly integrate. For robots, this translates to faster response times, fewer errors, and better performance in unfamiliar environments.

Another key feature is Chain-of-Thought (CoT) reasoning integrated into its architecture, where the robot plans out multi-step actions before execution. For example, a command like “clean the table” is often treated as a one-off trigger by robots in traditional models. They may forget to wipe it down after picking up tableware. While CoT reasoning gives the ability to simulate multiple steps before moving. It generates an internal plan instead of reacting in isolation. Responding to the command ‘clean the table’ involves recognizing clutter, sorting items, removing dishes, and wiping the surface, all done logically rather than through trial and error.

This, of course, extends beyond simple housekeeping. In industrial or service settings, multi-step planning allows robots to adapt without explicit programming. A warehouse bot could deduce how to stack differently-sized packages without knocking them over, while a healthcare assistant could prepare instruments by following procedural sequences.

Training for Real-World Adaptability

Unlike older systems trained on narrow datasets, Wall-OSS has been trained on billions of Vision-Language-Action samples. These come from real-world robotic logs, generative videos, and synthetic environments with different lighting, textures, and clutter. This broad training gives model resilience beyond controlled labs, and also means that WALL-OSS-powered robots are less likely to falter when faced with unusual household layouts, different object shapes, or sudden changes in context.

The Quanta X2 Robot

To showcase Wall-OSS, X Square Robot has developed the Quanta X2. This robot uses a wheeled base, a 7-degree-of-freedom arm, and a dexterous hand with lifelike gestures. It can handle up to 62 degrees of freedom for natural movements and features rotating clamps for 360° cleaning. The Quanta X2 is designed for use in service industries, households, and industrial settings.

Open-Source vs. Proprietary Futures

While other robotics companies maintain proprietary stacks, X Square Robot is set to release Wall-OSS openly on GitHub and Hugging Face. And if adoption spreads, Wall-OSS may position open frameworks as a credible counterpoint to closed systems.

With roughly US $100 million in funding, X Square Robot is betting that open-source collaboration can solve what proprietary efforts have not: dependable performance in unpredictable environments. For startups, Wall-OSS provides a ready-made brain; for hardware makers, it offers a generalizable intelligence layer. The race is no longer about choreographed performances but about bridging the gap between shallow demos and real-world needs. 

The post Embodied Intelligence Race: How China’s Open-Source Robotics Is Challenging Global Leaders appeared first on Gizmochina.