Japan EV Shift: Autonomy Push Meets AI Reality

Japan is accelerating its autonomous driving agenda just as real-world testing continues to expose how far the technology still has to go. According to Nikkei, Japan’s government will publish a dedicated communications infrastructure strategy on July 14 to support Level 4 autonomous vehicles, with a national target of 10,000 self-driving buses, taxis, and trucks on the road by fiscal 2030. At the same time, fresh Chinese testing data shows advanced driver assistance systems still struggle with edge cases such as parking-lot navigation, garage waypoints, route deviations, and unexpected disengagements—underscoring that autonomy is not just a software race, but a systems-and-infrastructure challenge.

Japan’s new plan: build the network before scaling autonomy

Japan’s Ministry of Internal Affairs and Communications is preparing a national strategy focused on communications infrastructure for autonomous driving. The plan is aimed at supporting Level 4 automated driving, where a vehicle can operate without driver input in limited conditions.

The policy matters because Japan is not treating autonomy as a standalone vehicle feature. Instead, it is building a broader operating framework that includes:

• Remote vehicle monitoring systems
• Cooperation between automakers and telecom operators
• National technical standards for transmission speed and latency
• Standardization of intelligent transport systems at intersections and merging areas
• Demonstration funding in the FY2027 budget request

Under Japan’s Road Traffic Act, vehicles operating in Level 4 mode must be remotely monitored in terms of both vehicle status and surrounding traffic conditions. That makes connectivity a regulatory necessity, not just a performance upgrade.

Japan’s headline targets

Japan first set part of this direction in 2023, when it targeted autonomous mobility services in more than 100 locations by 2027. A cabinet-approved development plan in January 2026 then sharpened the long-term ambition: 10,000 autonomous commercial vehicles in operation by fiscal 2030.

Why infrastructure is becoming the real battleground

The Japanese strategy reflects a growing reality in the global autonomous driving industry: the limiting factor is no longer only perception, compute, or mapping. It is increasingly the ability to deliver stable, low-latency, standardized communications across roads, intersections, depots, and mixed-traffic environments.

This is especially relevant for:

• Robotaxis and autonomous shuttles operating in geo-fenced zones
• Autonomous buses and trucks requiring remote oversight
• Smart intersections where inconsistent standards can cause integration problems
• Telecom-backed vehicle monitoring needed for compliance and safety assurance

Japan is also trying to solve a common but underappreciated issue: fragmented supplier standards. Current intelligent transportation systems used at intersections and merge points vary by vendor, so Tokyo plans to work with industry groups to unify standards during the current fiscal year.

That is a practical move. Without standardization, scaling autonomous mobility beyond pilot zones becomes expensive, slow, and operationally brittle.

China’s testing data shows the software still has gaps

While Japan focuses on building the infrastructure layer, recent Chinese autonomous driving evaluations highlight the limitations of current systems in real-world scenarios.

Data compiled from 51 autonomous driving tests in June—including editor tests in Beijing, public tests nationwide, and the Tianjin smart-driving competition—paint a more cautious picture of capability. The competition added tougher scenarios such as:

• Parking-space-to-parking-space navigation
• Garage waypoint routing
• Campus-style unguided roaming

The result: average scores fell sharply, by nearly 20 points versus prior conditions. The top score was only 86 points, second and third place were below 80 points, and the event even recorded a historic low of 28 points.

The core takeaway is clear: many so-called advanced smart driving systems perform well in controlled or familiar NOA environments, but still struggle when navigation becomes less structured and more human-like.

June smart-driving leaderboard highlights

The June results showed that:

• Only Momenta improved month-on-month
• Most suppliers saw lower average scores after tougher test conditions
• Leaderboard leaders changed in sub-categories such as scenario handling, safety, and efficiency
• NIO entered the main ranking after meeting data thresholds

In other words, even in China—the world’s most competitive smart EV market—autonomous driving remains uneven across edge cases.

The human comparison is harsh, but useful

One striking interpretation from the Chinese testing commentary compared current autonomous systems to child development. The argument: today’s systems are only approaching the mobility ability of a one-year-old child—able to move, but not yet reliably capable of handling context, exploration, and fluid decision-making the way a three-year-old can.

That framing may sound dramatic, but it captures a real engineering problem. Modern ADAS and urban NOA stacks can often manage:

• Lane following n- Basic route execution
• Some merges and turns
• Structured highway or urban navigation

But they still fail too often in:

• Unmapped or weakly structured environments
• Parking facilities and internal roads
• Complex route changes and detours
• Traffic light interpretation edge cases
• Unexpected NOA degradation or disengagement

For regulators and automakers alike, this means the path to commercialization is not linear. Strong demos do not automatically translate into scalable driverless operations.

What Japan can learn from China’s road testing

Japan’s infrastructure-first approach may actually prove more durable than a pure vehicle-first strategy. China’s testing data suggests autonomy improves not only with better software, but with a stronger ecosystem around it.

Japan appears to be preparing for that by emphasizing:

• Roadside communications support
• Unified standards across suppliers
• Remote supervision requirements
• Cross-industry collaboration between carmakers and telecom firms

This matters because Level 4 services, especially buses and robotaxis, do not operate in a vacuum. They need dependable networks, fallback procedures, regulatory clarity, and predictable interaction with public infrastructure.

Waymo in the U.S. and Baidu in China have already commercialized Level 4 robotaxi services in limited environments. Japan is clearly trying to build the conditions for a similar rollout, but with a strong public-sector hand in network architecture and standards-setting.

A parallel AI story: embodied intelligence is getting cheaper

There is another interesting thread in the broader mobility technology landscape: the falling cost of embodied AI hardware.

At WAIC 2026, startup Xingji Guangnian is set to showcase dexterous robotic hands and miniature integrated joint modules that dramatically lower the cost of advanced manipulation systems. While this is not an automotive product story in the narrow sense, it is relevant to the future EV and autonomous ecosystem because robotics, embodied AI, and autonomous mobility increasingly share core technology stacks in:

• Sensors and control systems
• Real-time response architectures
• Edge computing
• Modular actuators
• Data collection for physical-world AI models

Some of the announced figures are notable:

The significance is not that EV makers will suddenly sell robotic hands. It is that China’s broader AI hardware ecosystem is moving toward lower-cost, modular, industrial-grade components—a trend that can spill over into automotive supply chains, factory automation, and even future in-cabin or service robotics.

Why This Matters

Japan’s new autonomous driving communications strategy is important because it addresses a piece of the puzzle many markets underweight: infrastructure readiness. The latest Chinese testing results, meanwhile, are a reminder that autonomy remains fragile in messy, real-world environments despite rapid progress in smart EV software.

Together, these developments point to a more realistic view of the autonomous future:

• The winners will combine vehicle intelligence, telecom integration, and regulatory compliance
• Level 4 deployment will likely scale first in commercial fleets and fixed-route services
• Real-world validation remains more important than marketing claims
• China continues to lead on competitive software iteration, while Japan may carve out an advantage in standards-led deployment architecture

Global implications for EV and autonomous markets

For the global EV market, the lesson is that autonomous driving is becoming a national industrial capability, not just a premium feature. Countries that want meaningful deployment will need to align:

• Automotive OEMs
• Telecom operators
• City transport planners
• Digital infrastructure providers
• Safety regulators

This is particularly relevant for Chinese EV brands such as NIO, XPeng, Zeekr, and Huawei-backed smart car ecosystems, all of which are pushing advanced driver assistance and urban navigation features. As these companies expand internationally, they will face increasingly different infrastructure, compliance, and connectivity environments.

In that context, Japan’s plan may become a useful template: define the network rules early, standardize the interfaces, and make remote monitoring part of the operating model.

What comes next

The next milestones to watch are straightforward.

First, Japan’s July 14 strategy release should clarify how aggressively the government plans to set technical standards for autonomous driving communications. Second, the FY2027 budget process will show whether Tokyo is putting enough capital behind demonstration programs. Third, ongoing Chinese road tests will remain one of the clearest barometers of how close smart driving systems are to handling truly nationwide, full-scenario autonomy.

For now, the message from both markets is the same: autonomous mobility is progressing, but commercialization will depend on far more than clever software. The future belongs to the companies and countries that can connect the car, the cloud, the road, and the rulebook into one reliable system.