Arm and Shanghai Auto Show Signal EV AI Shift

China’s automotive and mobility tech sectors took another step toward an AI-defined future on September 8, as Arm used its Arm Everywhere China conference in Shanghai to outline a computing roadmap spanning cloud, edge and physical AI, while robotics firm Songyan Dynamics launched its new embodied intelligence brand Scalabot. On the same day, organizers of the 2027 Shanghai Auto Show confirmed that next year’s event will run from April 23 to May 2 at the National Exhibition and Convention Center in Shanghai, with an expected 400,000 square meters of display space. Taken together, the announcements show how the Chinese EV industry is expanding beyond vehicles alone and into a broader intelligent mobility stack that includes chips, software, robotics and physical AI.

Arm’s Message: AI Computing Is Moving Into Cars and the Physical World

At its 2026 Arm Everywhere China annual conference, Arm framed the industry’s transition as a move from the generative AI era to the “agentic” era, where computing stretches across:

• Cloud for model creation and large-scale inference
• Edge for personalized, always-on intelligence
• Physical systems such as robots, vehicles and autonomous machines

Arm said chips based on its technology have now shipped in excess of 350 billion units globally, underlining the company’s scale as a foundational platform supplier. More importantly for the EV and intelligent vehicle industries, Arm is positioning itself not just as a smartphone CPU designer, but as a common computing layer for software-defined cars, robotics and AI infrastructure.

This matters because Chinese automakers are increasingly becoming computing companies. As cockpits, ADAS, autonomous driving stacks, in-car voice assistants and vehicle-cloud systems grow more complex, the value chain is shifting from pure hardware integration toward platform control.

Edge AI: Why Arm’s Mobile Roadmap Matters for Smart EV Cockpits

One of the most relevant announcements for automakers and suppliers was Arm CSS for Mobile 2, a new-generation mobile compute subsystem designed for AI-native workloads.

Key components include:

• Arm C2-Ultra CPU with second-generation SME2 (Scalable Matrix Extension 2)
• Arm Mali G2-Ultra NX GPU
• Dedicated neural accelerators for what Arm calls “Neural Graphics”

Arm’s pitch is straightforward: future AI experiences will not run on a single processor type. Instead, workloads will be distributed across the CPU, GPU and dedicated accelerators depending on latency, efficiency and task requirements.

For the EV sector, that architecture is increasingly relevant in areas such as:

• In-car voice assistants
• Driver monitoring systems
• On-device multimodal AI
• Navigation and contextual recommendation engines
• Gaming and entertainment in smart cockpits

Arm also highlighted ecosystem adoption from partners including Alipay, OPPO and vivo for SME2, and gaming partners such as Tencent Games, NetEase, Unity China and Unreal Engine for neural graphics-related work. While these are not automotive companies, the software and silicon optimization work often migrates into vehicle cabins as cars inherit more consumer electronics functions.

Physical AI: Robots and Vehicles Are Starting to Share the Same Stack

Arm’s most strategically important message for the broader mobility market may be its push into physical AI. The company said it is extending its Arm Total Design ecosystem into this space with support from more than 80 companies, including AWS, Hugging Face, Liquid AI and QNX.

The first key initiative is a Robotics Capability Framework, intended to create a common language for describing and evaluating the capabilities of autonomous machines.

That may sound abstract, but for the automotive industry it reflects a very real convergence:

• Vehicles are becoming robotic platforms on wheels
• Advanced driver assistance and autonomy depend on sensor fusion, world modeling and action planning
• Supply chains for chips, real-time operating systems and middleware are increasingly shared across cars and robots

In other words, the line between an intelligent EV, an autonomous shuttle and a service robot is becoming thinner at the compute and software layers.

Songyan Dynamics’ Scalabot Push Shows Embodied AI Is Going Mainstream

That convergence was reinforced by Songyan Dynamics on the same day. The company officially launched Scalabot, a new embodied intelligence technology brand, and introduced HERON-World Model, the first world model under its HERON architecture.

The launch marks Songyan’s move from robot hardware into a fuller “body + embodied brain” strategy. Rather than simply attaching algorithms to existing robot hardware, the company says it is pursuing deep co-development, where sensing, compute, degrees of freedom and data interfaces are considered from the hardware definition stage.

What stands out in Scalabot’s approach

• World-model focus: The company wants robots to understand environments, predict future states and act autonomously.
• Cross-platform design: Model capabilities are not locked to a single robot form factor.
• Unified action representation: This should allow capability transfer across robots with different configurations and degrees of freedom.
• Data flywheel: Real-world interaction data feeds model improvement, which then enables more complex tasks.

The target application roadmap starts with home services and expands toward commercial services.

This is notable for automotive watchers because many Chinese EV players are already investing in humanoids, logistics robots and factory automation. Technologies developed for robotics—especially perception, planning, simulation and world models—are likely to feed back into intelligent driving and smart manufacturing.

Songyan also noted that it has spent the last three years building out robot body R&D and manufacturing capabilities. In October last year, it launched what it described as the industry’s first high-performance consumer humanoid robot, Xiaobumi, with a price reduced to the 10,000-yuan level, a sign that embodied AI is beginning to chase scale, not just demos.

Cloud AI: Arm Wants a Bigger Slice of AI Infrastructure

Arm’s cloud message was equally ambitious. Citing IDC, the company said the market for Arm-based rack-scale servers has already surpassed x86 in accelerated computing, making Arm a mainstream platform in this segment.

Arm outlined two major infrastructure building blocks:

• Neoverse CSS N4 for highly configurable, throughput-oriented scale-out workloads
• Arm AGI CPU as a high-performance deployable computing platform

Arm said OpenAI, Meta and Oracle are developing solutions based on the Arm AGI CPU, while Volcano Engine, ByteDance’s cloud arm, has already brought the first agent sandbox products based on the platform to market.

For EV companies, this cloud-side shift matters because intelligent vehicles no longer depend only on onboard compute. They also need:

• Training infrastructure for ADAS and autonomous driving models
• Simulation platforms
• Fleet learning systems
• Cloud-connected digital cockpit services
• AI development environments for in-car agents

As a result, the same architecture family can increasingly span the vehicle, the edge device and the data center.

Comparison Table: What the Announcements Mean for Mobility

2027 Shanghai Auto Show Will Reflect the Industry’s New Center of Gravity

The organizers of the 2027 Shanghai Auto Show confirmed that the event will be held from April 23 to May 2, 2027, at the National Exhibition and Convention Center (Shanghai). The show is expected to span 400,000 square meters, keeping it among the world’s largest auto exhibitions.

Its theme, “Build the Ecosystem, Lead the Future with Intelligence”, is revealing. The show will not only focus on automakers and suppliers, but also on:

• Embodied intelligence
• Physical AI
• High-end chips
• Smart hardware extensions beyond cars
• Collaborative industry ecosystems

This is a telling shift. In China, car shows are no longer only product launch venues for sedans, SUVs and MPVs. They are turning into platforms where automakers, chip companies, robotics firms, software developers and AI infrastructure providers compete for influence over the future of mobility.

The event’s plan to introduce a dedicated embodied intelligence exhibition area is especially significant. It suggests that robotics and automotive technologies are no longer adjacent categories—they are becoming part of the same strategic conversation.

Why This Matters

For anyone tracking Chinese EVs, these three announcements point to a deeper structural change.

1. The car is becoming one node in a larger AI system

An EV is now part of a computing continuum that stretches from cloud training to edge inference to real-world action. That aligns perfectly with Arm’s cloud-edge-physical AI narrative.

2. Automotive competition is shifting from hardware specs to platform capability

Battery range, charging speed and motor output still matter, but future differentiation will increasingly depend on:

• AI model efficiency
• Chip-software optimization
• Real-time operating systems
• Data loops and continuous learning
• Cross-device ecosystem integration

3. Robotics and EVs are starting to share technology pathways

World models, sensor fusion, planning and embodied reasoning all have applications across humanoid robots, autonomous systems and intelligent vehicles.

4. China is building an ecosystem advantage

The combination of domestic automakers, battery makers, chip designers, cloud providers and AI developers gives China unusual vertical depth. Whether the brand is BYD, NIO, XPeng, Zeekr or a newer mobility-tech player, the competitive battlefield increasingly includes software stacks and AI compute foundations, not just vehicles.

Global Implications

For global automakers and suppliers, the takeaway is clear: China’s EV market is evolving into a broader intelligent mobility market.

That has several implications:

• Chip architectures like Arm’s are gaining strategic importance across cars, robots and data centers.
• Embodied AI is moving from research narrative to commercialization roadmap.
• Auto shows in China may increasingly preview not just future vehicles, but also future machine intelligence platforms.
• Western incumbents that still separate automotive, robotics and AI infrastructure into distinct silos may struggle against Chinese ecosystem players that are integrating them faster.

What Comes Next

The next major milestone will be how these ideas show up in real products. Watch for three developments over the coming quarters:

• Automotive-grade adoption of more advanced Arm-based AI compute platforms in cockpits and autonomous driving domains
• Embodied AI pilots moving from laboratories into homes, retail and industrial services
• Shanghai Auto Show 2027 becoming a showcase not just for new EVs, but for the full stack of intelligent mobility—from chips and operating systems to robots and physical AI demos

In that sense, September 8 was less about three isolated announcements and more about one unified trend: China’s electric vehicle sector is rapidly merging with the AI computing and robotics industries. The companies that master that merger will shape the next phase of the global EV market.