Bosch and China EV Tech Reshape Global Mobility

China’s EV industry delivered two very different but equally important signals this week: a major breakthrough in the global auto tech supply chain, and a stark reminder that safety regulation is catching up with fast-moving design trends. Bosch confirmed its fourth-generation multifunction camera, MPC4, has secured large-scale production nominations across more than 40 vehicle lines and 200+ projects worldwide, with mass production set for Q3 2026. At the same time, China’s market regulator published what Reuters and Bloomberg described as the country’s largest single-wave automotive recall, covering 4.276 million vehicles over poorly identifiable emergency mechanical door releases.

Together, these developments show where the Chinese EV market is headed next: deeper global integration in high-value technologies such as ADAS chips and perception systems, paired with tougher standards around real-world safety, usability, and regulatory compliance.

Bosch’s MPC4 Gives Horizon a Global ADAS Breakthrough

The most strategically significant news comes from Bosch’s new MPC4 front-view camera platform. On paper, it is an upgrade in hardware and software. In practice, it marks a notable shift in the automotive electronics supply chain: Bosch has selected Horizon Robotics’ Journey 6B as the computing platform for the new system.

That matters because earlier generations relied on overseas chip suppliers:

• Bosch’s second-generation front-view integrated unit used Xilinx from the US
• The third generation switched to Renesas from Japan
• The fourth generation moves to Horizon, a Chinese automotive chip company

This is not just another vehicle nomination for Horizon. MPC4 is a Bosch-standardized global product that will feed into more than 200 projects across:

• China
• Europe
• Australia and New Zealand
• Latin America
• The Middle East
• Broader Asia-Pacific markets

Mass production is scheduled for Q3 2026, and Bosch says the overall program has reached the 10-million-unit scale.

What the Bosch MPC4 Upgrade Actually Changes

The front-view multifunction camera is typically mounted behind the windshield and provides perception input for core ADAS functions such as:

• Automatic emergency braking (AEB)
• Lane keeping assist
• Traffic sign recognition
• Pedestrian and vehicle detection

With MPC4, Bosch is not merely refreshing the camera module. It is overhauling the full perception stack:

• 8-megapixel camera resolution
• 120-degree horizontal field of view
• Up to 300 meters detection range
• Bosch-developed perception algorithms now incorporating Transformer architectures

That last point is especially important. As ADAS perception moves from traditional CNN-heavy pipelines toward Transformer-based models, compute requirements rise and software-hardware co-optimization becomes more demanding. In other words, Bosch’s decision reflects not just chip sourcing, but confidence in Horizon’s toolchain, software adaptation, production readiness, and engineering support.

Why Horizon’s Win Matters More Than a Typical Design Award

China’s auto chip companies have long supplied domestic automakers directly. The larger shift now is that they are entering the standardized product stacks of global Tier 1 suppliers.

That changes the distribution model completely.

Multinational automakers such as Toyota and Volkswagen often buy integrated systems from suppliers like Bosch or Denso rather than selecting every underlying component themselves. Once a chip enters a Tier 1 standard product, it can scale across multiple brands and markets without the chipmaker having to win each OEM separately.

Horizon’s current momentum

According to the source report:

• Journey 6B first entered mass production in May 2026
• It debuted on the GAC Toyota Bozhi 3X
• By August 2026, it had reached:
  • 25+ automakers
  • 50+ vehicle models
  • More than 20 million units in lifecycle nominations

Bosch is also already working with Horizon beyond MPC4. A mid-level assisted driving solution based on Journey 6M has also entered mass production.

This suggests Horizon is not winning on price alone. It is demonstrating the full set of capabilities global Tier 1s require:

• Functional safety readiness
• Software compatibility
• Validation efficiency
• Volume manufacturing support
• Long-cycle delivery capability

Bosch MPC4 at a Glance

The Cost Pressure Behind China’s ADAS Supply Chain Rise

One reason Chinese suppliers have become more competitive is that they were forged in one of the world’s toughest automotive markets.

The D1EV report notes that a front-view camera module that sold for roughly RMB 1,000-1,500 a few years ago has now fallen to around RMB 600-900, even as performance keeps improving.

That dynamic is familiar across China’s EV supply chain:

• Vehicle development cycles are shorter
• New features rapidly trickle down into lower-priced cars
• Intense EV price wars push cost pressure onto suppliers
• Component makers must raise performance while reducing system cost

This is why Horizon’s Bosch nomination is so significant. It suggests Chinese smart-driving suppliers are no longer just cost-competitive domestically; they are becoming globally credible in mature, price-sensitive, safety-critical automotive systems.

China’s 4.276 Million-Vehicle Recall Puts Door Handle Safety Under the Spotlight

If Bosch and Horizon represent China’s tech ascent, the week’s recall news shows the limits of design-led experimentation.

On August 21, 2026, China’s State Administration for Market Regulation posted a batch of recall filings covering nine automakers:

• Tesla
• Xiaomi
• Leapmotor
• XPeng
• Zeekr
• Chery New Energy
• Dongfeng
• BAIC BluePark Magna
• FAW

The common issue was not batteries, motors, or autonomous driving software. It was emergency escape usability: in-car mechanical emergency door releases were difficult to identify, which could delay evacuation if a severe crash causes low-voltage power failure.

Recall scale and why the number matters

The actual total linked to the door-handle issue is 4.276 million vehicles.

That is significant for two reasons:

1. Reuters and Bloomberg described it as the largest single-wave recall in Chinese automotive history
2. It exceeds the scale of China’s entire 2025 NEV recall volume of 2.652 million units, which was spread across 105 recalls

The report also clarified confusion around inflated online figures such as 7.11 million or 6.934 million vehicles. Those numbers double-counted separate recalls, including:

• Tesla’s 2.7406 million-vehicle recall related to assisted driving
• A 92,900-vehicle Geely recall involving LiDAR on Galaxy M9 and four Lynk & Co models

Of the 4.276 million vehicles in the door-release recall, Tesla alone accounts for 2.976 million units.

What Automakers Are Actually Fixing

The current remedies are limited. They focus mainly on improving passenger self-rescue, not fully solving external rescue access.

The measures include:

• Adding a small label to identify the emergency mechanical release inside the cabin
• In some cases, replacing trim covers with clearer markings
• OTA software updates to improve window control logic so windows can lower before the low-voltage system fully fails after a severe collision

That means these are largely mitigation steps rather than full mechanical redesigns.

Why external rescue remains a problem

To make doors easier to open from outside after a power loss, automakers would need to change door handle structures and restore mechanical redundancy. That is much more expensive.

According to the source, retrofitting already-sold vehicles with such hardware changes could cost thousands to tens of thousands of yuan per vehicle, making large-scale rework commercially unrealistic.

China’s New Door Handle Standard Could Change EV Design

This recall wave also needs to be viewed alongside a major regulatory change.

On January 28, 2026, China approved GB 48001—2026, Safety Technical Requirements for Automotive Door Handles. The report describes it as the world’s first mandatory national standard to effectively ban fully hidden pop-out handles and fully electronic door handles without proper mechanical redundancy.

The new standard requires every door to have:

• A mechanical-release exterior handle
• A mechanical-release interior handle

Compliance timeline

This transition period matters because China had already accumulated more than 10 million vehicles with hidden door handles by the end of 2025.

In other words, the market embraced aerodynamic styling and flush surfaces much faster than safety regulation could respond.

Automakers Are Already Reversing Course

Even before the standard takes full effect, several automakers are moving back toward more practical designs.

Examples cited in the source include:

• A new-generation Xiaomi SU7 variant switching from fully hidden handles to a semi-hidden design with mechanical releases
• Independent external mechanical door release provision
• Backup power for door locks
• Emergency mechanical releases on all four doors
• The updated Li Auto L9 also adding a mechanical emergency structure

More broadly, the industry is seeing a return of:

• Exposed door handles
• Semi-hidden door handles
• Mechanical redundancy in both interior and exterior access systems

This is a notable reversal for a market that once treated hidden handles as a near-default EV design signature.

AI Radar Is the Next Frontier Beyond Cameras and LiDAR

A third theme emerging from China’s EV ecosystem is sensor evolution. In an interview with SuperRadar, Chengtai Technology co-founder and CTO Zhou Ke argued that the industry is entering the era of AI radar.

His framing is useful because it highlights a broader shift happening across automotive perception stacks. Traditional radar uses hand-engineered processing chains and fixed firmware logic. AI radar, by contrast, is defined by deployable, trainable models.

As Zhou put it, the transition is from:

• Traditional radar: fixed pipelines such as ADC sampling, FFT, CFAR, DOA, clustering, tracking, classification
• AI radar: models learning directly from raw electromagnetic signal data, aiming for end-to-end semantic understanding

He describes this as moving from software-defined radar toward model-defined radar.

AI radar vs 4D imaging radar

Zhou makes a clear distinction:

• 4D imaging radar improves hardware capability and helps radar “see more clearly”
• AI radar changes the development paradigm so radar can “understand better” and evolve faster

Rather than competing, the two should reinforce each other:

• High-resolution radar data provides better training input
• AI models extract more value from richer radar outputs

Why AI Radar Is Gaining Momentum Now

The interview identifies one demand driver and three technology inflection points.

Demand driver

Physical AI is expanding rapidly across:

• Autonomous driving
• Robotics
• Embodied intelligence

These systems need robust perception in conditions where cameras struggle, including:

• Rain
• Fog
• Smoke
• Glare
• Transparent obstacles

Three technical inflection points

1. Data

   • Millimeter-wave radar resolution has improved sharply
   • Point clouds can now reach thousands to tens of thousands of points per frame
   • Single-frame radar data volumes can approach LiDAR-class levels

2. Compute

   • Radar sensors themselves have much more onboard memory and processing capability
   • Vehicle E/E architectures are shifting from distributed ECUs to centralized computing with hundreds to thousands of TOPS available at the vehicle level

3. Algorithms and tools

   • Open toolchains such as PyTorch and ONNX are lowering development barriers
   • Foundation model ideas from computer vision and multimodal AI are increasingly transferable to radar domains

The strategic takeaway is that China’s intelligent vehicle market is not only iterating batteries and body design. It is also pushing into the next generation of machine perception, where camera-centric ADAS, radar, AI chips, and centralized computing architectures will increasingly converge.

Why This Matters Globally

These three stories—Bosch-Horizon, the massive recall, and the AI radar push—are connected more closely than they first appear.

They show that China’s EV sector is moving from scale to systems maturity in three parallel ways:

• Global supply chain integration: Chinese chipmakers are entering Bosch and other Tier 1 platforms, not just exporting alongside Chinese vehicle brands
• Regulatory normalization: flashy EV design elements such as hidden door handles are being forced to prove real-world safety value
• Perception stack competition: the next battleground is no longer just batteries or infotainment, but how cars sense and interpret the physical world

For global automakers, suppliers, and investors, this matters because the competitive map is shifting. China is no longer simply a giant EV manufacturing base or battery leader. It is becoming a source of:

• Automotive semiconductors
• Perception algorithms
• ADAS system integration
• Sensor innovation
• New regulatory templates for EV safety design

What Comes Next

Expect two trends to accelerate into 2027 and beyond.

First, more Chinese smart-driving suppliers are likely to enter global Tier 1 stacks, especially in cost-sensitive but high-volume domains such as camera perception, domain controllers, and radar. Horizon’s Bosch nomination could become a template for others.

Second, safety-driven redesign will continue to reshape EV packaging and user interfaces. Hidden handles may not disappear overnight, but the era of fully electronic, mechanically redundant-free door access is clearly ending in China.

The broader lesson is simple: the Chinese EV market is no longer competing on one dimension. It is competing on cost, software, perception, compliance, and global scalability all at once—and that is what makes it the most consequential automotive market in the world right now.