China’s EV industry delivered a packed news cycle on May 12-13, led by Li Auto’s formal unveiling of its self-developed Mach M100 AI chip for the new Li L9 Livis. At the same time, XPeng detailed a new battery safety chip for its flagship GX SUV, GAC and Huawei-backed premium brand Qijing secured over RMB 1 billion in fresh capital, and geopolitical pressure around Chinese cars intensified ahead of Donald Trump’s China visit. Together, these developments show how China’s EV race is no longer just about batteries and price—it is increasingly about full-stack AI, supply-chain control, and global market access.
Li Auto Unveils the Mach M100, Chasing Full-Stack AI Leadership
Li Auto founder and CEO Li Xiang has officially announced the company’s in-house Mach M100 chip, describing it as the "world’s most powerful computing" solution for production vehicles. The chip will debut in the new Li L9 Livis, marking a major milestone in the company’s effort to build a vertically integrated AI stack spanning chips, operating systems, large models, and vehicle hardware.
According to Li Auto, the Mach M100 is:
- Built on a 5nm automotive-grade process
- The industry’s first dataflow-architecture edge inference chip
- Rated at 1,280 TOPS per chip
- Deployed in a dual-chip configuration in the new L9, for 2,560 TOPS total computing power
- Designed to reduce end-to-end latency by 40%
- Capable of cutting sensor-to-action delay to 200-300 milliseconds
Li Auto says the chip uses a dynamic dataflow architecture rather than the conventional von Neumann or instruction-driven GPGPU approach. In practical terms, that means the compiler pre-plans data movement, reducing reliance on broad cache structures and improving efficiency for AI inference workloads.
That matters because modern intelligent vehicles increasingly run:
- End-to-end assisted driving models
- Multimodal perception stacks
- In-cabin AI interaction systems
- Large language model-based interfaces
Li Auto also tied the chip closely to its own Mind VLA-o1 large model and 3D ViT vision model. The company claims the new Li L9 Livis can achieve faster reactions than a human driver in emergency scenarios, while its upgraded Mach VLA 2.1 system delivers a 10x increase in multimodal computing volume.
Why the architecture matters
Most automakers still rely heavily on Nvidia, Qualcomm, Mobileye, or Horizon Robotics for core compute. Li Auto’s move is significant because it is not simply designing a custom accelerator for cost reasons; it is trying to optimize the entire software-hardware pipeline around AI inference in the physical world.
Li Xiang has also pushed back against criticism that in-house chips are expensive vanity projects. His argument is that suppliers cannot solve every bottleneck in latency, efficiency, and model deployment speed, especially as vehicles evolve into AI terminals.
Li L9 Livis: A New Benchmark for Intelligent Vehicle Compute?
The new Li L9 Livis does more than just add silicon. It packages the chip into a wider intelligent driving and chassis-control architecture.
Key upgrades include:
- 3D ViT perception model combining lidar and camera inputs
- 50% longer visual detection range, according to Li Auto
- Steer-by-wire and brake-by-wire capability
- Active suspension with millisecond-level full-chain response
- Support for automated driving, autonomous parking, and smart interaction
Li Auto claims the dual-chip setup delivers effective usable compute as much as 5-6 times that of Nvidia Thor-U in its targeted workloads, though such comparisons depend heavily on software optimization and benchmark methodology.
Li L9 Livis / Mach M100 at a glance
| Item | Li Auto Mach M100 / L9 Livis |
|---|---|
| Process node | 5nm automotive-grade |
| Architecture | Dynamic dataflow |
| Single-chip compute | 1,280 TOPS |
| Vehicle total compute | 2,560 TOPS |
| End-to-end latency improvement | 40% |
| Sensor-to-action latency | 200-300 ms |
| Perception stack | 3D ViT + lidar + vision fusion |
| Claimed visual range improvement | 50% |
| Chassis tech | Active suspension, steer-by-wire, brake-by-wire |
This is a notable step in the broader Chinese EV trend toward full-stack vertical integration, where automakers seek to own the key layers of intelligence rather than simply assemble third-party components.
XPeng Targets Battery Safety with New EIS Chip
While Li Auto focused on AI compute, XPeng highlighted a different piece of the next-generation EV stack: battery intelligence and safety.
XPeng said its flagship GX full-size SUV will be the first in the industry to carry an EIS battery safety chip. The company says the chip can:
- Monitor battery conditions in real time
- Provide 2x earlier warning of cell defects
- Improve low-temperature charging speed by 27%
- Increase SOC accuracy by 50%
The GX, which opened pre-sales on April 15 from RMB 399,800, is positioned as XPeng’s first full-size luxury six-seat SUV and the debut model for its SEPA 3.0 physical AI architecture.
XPeng is framing the GX as more than a family SUV. It links the model to:
- Robotaxi-grade intelligent driving
- Flying car redundancy concepts
- Embodied intelligence chips and architecture
- AI-centric luxury positioning
That messaging underlines an important shift in China’s EV market: innovation is expanding beyond drivetrain efficiency and cabin screens into underlying chip-level safety systems.
Li Auto vs XPeng: Two different chip strategies
| Brand | Chip Focus | Main Purpose | Key Claims |
|---|---|---|---|
| Li Auto | Mach M100 AI chip | Assisted driving, multimodal AI, vehicle control | 1,280 TOPS per chip, 2,560 TOPS dual-chip system, 40% lower latency |
| XPeng | EIS battery safety chip | Battery monitoring and charging/safety management | 2x earlier fault warning, 27% faster low-temp charging, 50% better SOC accuracy |
The contrast is telling. Li Auto is prioritizing compute architecture for autonomy, while XPeng is using silicon to strengthen battery reliability and user trust. Both approaches reflect where Chinese EV makers now see strategic differentiation.
GAC-Huawei’s Qijing Adds Funding Ahead of GT7 Launch
Another major domestic development came from Qijing Auto, the premium intelligent NEV brand jointly built by GAC Group and Huawei. The company has completed a capital increase of more than RMB 1 billion, with investors including:
- CATL
- A Bosch-affiliated investment platform
- State-owned and government-backed capital
Previously known as Huawang Automotive Technology (Guangzhou), the company has now unified its corporate and brand identity under Qijing.
Its first model, the Qijing GT7, is scheduled for a launch event in June. The vehicle is described as an intelligent shooting brake with the following dimensions:
- Length: 5,050 mm
- Width: 1,980 mm
- Height: 1,470 mm
- Wheelbase: 3,000 mm
Notably, the GT7 is expected to feature:
- 896-line lidar, billed as the highest specification in global mass production
- Huawei’s Qiankun Chitu platform
The investor lineup is especially important. CATL brings battery credibility, Bosch adds supply-chain and systems value, and the state-backed funding suggests local governments still see intelligent EV manufacturing as a strategic growth engine.
China’s Supply Chain Edge Keeps Deepening
Beyond vehicle launches and chip announcements, the latest reports also reinforced a broader structural trend: China’s automotive supply chain is becoming harder for global rivals to match.
One report cited Japanese automakers increasing the share of Chinese components in response to brutal price competition. Chinese suppliers were said to hold:
- 30%-40% cost advantages over Japanese peers
- Roughly 10 months from order to mass production versus more than 18 months for some Japanese suppliers
- Little to no meaningful quality gap, according to executives quoted in the report
Battery demand data also remained strong. According to the China Automotive Battery Innovation Alliance, domestic power battery installations reached 62.4 GWh in April, up:
- 10.4% month on month
- 15.2% year on year
Meanwhile, sulfur prices—a useful upstream indicator for parts of the battery materials chain—have surged. China’s benchmark sulfur price reached RMB 7,300 per ton on May 6, with year-to-date gains around 80%, driven by Middle East tensions, lower inventories, and growing demand linked to the expansion of the EV and LFP battery market.
These pressures show that while China still enjoys scale and manufacturing speed, upstream raw-material volatility remains a real issue for the EV ecosystem.
Trump’s China Visit Highlights Growing US Anxiety Over Chinese EVs
The geopolitical backdrop is just as important as the product news. Ahead of Donald Trump’s May 13-15 China visit, Reuters reported that more than 20 US business leaders—including Tesla CEO Elon Musk and Apple CEO Tim Cook—were expected to be involved in the broader trip environment, with trade and technology cooperation central to discussions.
At the same time, US political and industry voices have been intensifying calls to block Chinese vehicles from the American market.
According to the source material:
- 74 Democratic and 52 Republican House members reportedly signed a letter urging Trump not to allow Chinese automakers into the US market
- US lawmakers are pushing legislation targeting connected vehicle data security risks
- Some proposals would effectively ban vehicles with advanced connected features if their design originates in China
- Wider US industry groups, including automakers, suppliers, steel interests, and unions, are broadly aligned in favor of tighter restrictions
This is happening because Chinese brands continue to gain ground overseas.
Reported market share gains for Chinese brands
| Market | Chinese brand share / presence |
|---|---|
| Europe overall | 6% in 2025 |
| Norway | 14% |
| Italy | 9% |
| UK | 11% |
| Spain | 9% |
| Mexico | 34 Chinese brands on sale, about 15% market share |
| Canada | Around 49,000 Chinese EV imports allowed annually |
The affordability angle is central. The report notes that the average new car price in the US has climbed above $51,000, while Chinese EVs in other markets often undercut Western rivals significantly. In Mexico, for example, a Geely EX2 reportedly starts at about $22,700, versus roughly $38,630 for the Tesla Model 3 in the US market.
That helps explain why the American debate has shifted from tariffs alone to concerns around software, data, and industrial competitiveness.
Why This Matters
These developments point to three defining themes in the Chinese EV market.
1. The competition is moving down the stack
The biggest battleground is no longer just vehicle styling or battery range. It is now about who controls:
- AI chips
- Operating systems
- Large models
- Sensor fusion
- Battery diagnostics
- Software-defined chassis systems
Li Auto’s Mach M100 and XPeng’s EIS chip are examples of Chinese automakers trying to own the most strategic parts of the technology stack.
2. China’s advantage is becoming systemic
Chinese EV makers benefit not only from battery leadership, but also from:
- Faster supplier iteration cycles
- Lower component costs
- High domestic deployment scale
- Tight integration between software, hardware, and manufacturing
That combination is difficult for legacy automakers to replicate quickly.
3. Global expansion will face political friction
As Chinese EV brands gain share in Europe, Mexico, and other export markets, resistance in the US is hardening. Even if Chinese automakers remain locked out of America for now, the debate itself confirms that the industry’s competitive center of gravity has shifted.
Forward Look: China EVs Enter the AI Industrial Phase
The most important takeaway from this week’s news is that China’s EV sector is entering a new phase: the AI industrial phase. Carmakers are no longer only competing on electrification; they are trying to become integrated AI hardware and software companies.
Li Auto’s Mach M100 launch is the clearest example. If the chip delivers in production as promised, it could strengthen the case for in-house automotive compute in China and raise pressure on rivals still dependent on external silicon roadmaps. XPeng’s battery chip push suggests that safety and energy management will become another front in the chip race, while Qijing’s fresh funding underscores how capital continues to back premium intelligent EV projects.
The next questions are straightforward:
- Can Chinese automakers scale proprietary chips without sacrificing reliability?
- Will in-house silicon create a durable cost and performance advantage?
- How much will geopolitics reshape export opportunities for Chinese EV brands?
Those answers will define not just the next product cycle, but the next era of the global auto industry.
