Preface

TrueView

For nearly a decade, the narrative of technological breakthroughs has dominated discussions about autonomous driving. However, the real turning point has little to do with code and parameters—it is a triple shockwave of capital, industry, and division of labor.

We have witnessed the industry's shift from the laboratory to the mass adoption in millions of new vehicles. Yet, we must also confront the fundamental laws of the business world: mass production is merely an entry ticket, not a guaranteed path to profitability, and a decline in hardware costs will not automatically translate into profits.

As the tide of euphoria recedes, the struggle for ecological niches and the test of cash flow become the true themes of the second half of the race.

This article will focus on analyzing

1. How will the three types of players reshape the industrial power landscape?

2. Why has a significant drop in costs pushed the industry further away from overall profitability?

3. Where is the survival space and core competitive moat for independent Tier 1 suppliers?

4. What is the true path to commercialization for Robotaxi?

Content/Jin Huan

Editor/Yong E

Proofreader/Mang Fu

The series of 'major events' in the autonomous driving industry are accelerating the arrival of a critical moment—not a technological singularity, but a violent fluctuations (fierce shake-up) in capital and industrial structure.

On July 8, Momenta went public on the Hong Kong Stock Exchange as the 'first physical AI stock.' Before the cheers from the bell-ringing had faded, two signals from different directions jolted the entire industry back to reality.

NVIDIA's autonomous driving division underwent a major restructuring, with former Xiaopeng intelligent driving lead Wu Xinzhou consolidating core authority and pushing software and hardware full-stack integration to the extreme. Meanwhile, ByteDance quietly assembled a world model team, entering this once heavily fortified industry through an unnoticed side door with its AI infrastructure stance.

Together, these three events paint a true picture of the second half of the autonomous driving race. When mass production is no longer a get-out-of-jail-free card, and a rapid decline in hardware costs fails to bring industry-wide profitability, a new round of reshuffling centered on ecological niches, cash flow, and industrial division of labor has begun.

Part.1 The Truth About the Supply Chain: The Prosperity of NOA and the Ledger

'Available nationwide'—this promotional phrase, loved by automakers and intelligent driving suppliers since 2024, has not only been upgraded to 'available globally' but is also becoming the industry's biggest cognitive bubble.

150,000-yuan-class models now come standard with urban NOA, and 'intelligent driving for all' is shouted loudly... Yet, the industry's power structure has never been more turbulent. Technical barriers are being reconstructed by AI paradigms, mass production delivery capabilities have replaced parameter metrics as the new competitive anchor, and both new and old players stand at a crossroads of reshuffling.

NVIDIA's organizational adjustment is the clearest signal of the global intelligent driving industry's shift toward mass production priority. Nearly three years after Wu Xinzhou took office, he has finally consolidated real power over the autonomous driving business. Veteran Sarah Tariq has faded from core positions, former Qualcomm autonomous driving engineering head Dheeraj Ahuja has joined, and the underlying system software team has been formally transferred to the autonomous driving division.

The figure is Wu Xinzhou

The intent behind this series of moves is clear: break up the original technology-oriented structure, supplement engineering and mass production delivery capabilities, and transform NVIDIA from a computing power supplier into a full-stack intelligent driving solution provider.

Jensen Huang has set a 2026 revenue target of $5 billion for the autonomous driving business, but less than half has been achieved so far. To deliver on this promise, NVIDIA can no longer rely on showcasing computing power and models—it must secure more mass production fixed point (design wins) from automakers and ensure delivery.

The trend of Chinese intelligent driving players going global is also forcing NVIDIA to accelerate engineering iterations. After losing market share to domestic players like Horizon Robotics and Huawei in China, overseas markets have become NVIDIA's foundation. Wu Xinzhou's mass production experience and industry resources are the core assets in this defensive battle.

If NVIDIA's pivot represents a self-revolution by a giant, ByteDance's crossover is a dimensional strike from outside the industry. Led by Seed's Zhou Chang's world model team, ByteDance's autonomous driving exploration has never followed the traditional path of automakers or Tier 1 suppliers. Instead, it targets cost-effective areas like data annotation and simulation testing, leveraging Volcano Engine's automotive industry line to deploy unmanned logistics scenarios.

The essence of this logic is betting on AI paradigms to rewrite the barriers to intelligent driving. When world models become an industry consensus and perception and planning increasingly rely on large models and computing power, the core competitiveness of intelligent driving will no longer be a decade of accumulated engineering experience but data, computing power, and general model capabilities.

ByteDance has ample computing power reserves and a mature large model training system. What it lacks is traffic scenario data and engineering delivery capabilities. With sufficient resources, ByteDance can gradually accumulate the former through unmanned logistics scenarios and quickly supplement the latter by recruiting talent.

If world models can truly compress the infinite scenarios of open roads into a high-fidelity virtual world, testing and verification costs could drop by an order of magnitude. While this remains a laboratory-stage concept today, its mere existence is enough to send chills down the spines of traditional simulation suppliers.

More importantly, autonomous driving is not an isolated business for ByteDance but a training ground for embodied intelligence. Just as Tesla uses FSD data to feed Optimus, ByteDance also needs real-world road data to iterate its world models for broader applications in embodied intelligence.

Momenta's IPO adds a critical capital dimension to this structural shift.

Momenta is a representative player with dual layout (strategies) in mass production solutions and Robotaxi. Its market debut is an industry milestone, signaling that the mass production intelligent driving sector has finally produced a player with scalable revenue. The stories from the primary market are now being tested by the secondary market.

Prospectus data shows that Momenta's revenue surged from 743 million yuan in 2023 to 2.413 billion yuan in 2025, with gross margins improving from 17.5% to 71.6%, directly validating the growth potential of the mass production intelligent driving business model.

As the first independent intelligent driving company to go public, Momenta will also gradually validate the long-term profitability potential of its dual-wheel-drive model under public market scrutiny.

Part.2 The Profitability Dilemma: The Real Ledger Behind the Mass Production Frenzy

The industry consensus is that scaling leads to cost reduction—once volumes rise, hardware costs will be amortized, and profitability will naturally follow.

But the reality is that more automakers are bundling intelligent driving as a free standard feature rather than a high-priced option. While leading intelligent driving companies have doubled their revenue, losses have widened simultaneously. Mass production has not brought a profitability turning point but has dragged the industry into a paradox: the more you sell, the more you lose.

Even though material costs for LiDAR and domain controllers have indeed dropped sharply in recent years, with pure vision highway NOA solutions' BOM costs compressed below 4,000 yuan (a 40% drop in two years), and LiDAR-equipped urban NOA solutions using Horizon Robotics' Journey 6M seeing hardware costs fall to 4,500–5,000 yuan (a 43% drop), the marginal cost of fitting an urban NOA system into a 150,000-yuan family car is only a few percentage points from a material perspective.

However, cost reductions have not translated into synchronized profitability gains, as two critical bottlenecks remain. The first is the trap of high penetration but low utilization rates.

MIIT data shows that since 2026, 70% of new passenger vehicles have been equipped with combined driving assistance functions, with NOA penetration exceeding 30%. But penetration does not equal utilization. Zhou Guang, CEO of Yuanrong leave (likely a typo for Yuanrong Qixing), has publicly stated that real user engagement with urban NOA was only 20–30% in 2025. When features are not must-haves, automakers struggle to charge ongoing premiums for intelligent driving.

The second bottleneck is the rigid pressure of R&D amortization. In 2025, the overall profit margin of the automotive manufacturing industry was only 4.1%, falling further to 3.2% in the first quarter of 2026—razor-thin margins. Even with intelligent driving hardware costs down to the 5,000-yuan level, bundling it as standard directly impacts vehicle gross margins. As a result, most brands still tie advanced intelligent driving to specific trims or 10,000-yuan option packages.

For suppliers, algorithm R&D represents a rigid investment. Annual R&D spending at leading companies universally exceeds 1 billion yuan, with Momenta's R&D expenditure reaching 1.87 billion yuan in 2025 (77.5% of revenue).

High R&D investment is necessary to maintain technological competitiveness, but it also means that a significant portion of the benefits from scaling-driven cost reductions must continuously flow back into technological iteration, making it difficult to directly convert into net profits.

Beyond per-vehicle cost accounting, SAIC Motor's famous metaphor from years ago—'not surrendering the soul to Huawei'—is now evolving into a collective awakening across the industry.

Virtually all automakers selling over 300,000 vehicles annually have initiated or accelerated in-house R&D of core intelligent driving algorithms. BYD's intelligent driving team has expanded to 5,000 people, new energy brands continue to ramp up full-stack self-research, and even traditionally cautious joint ventures have begun establishing independent intelligentization (intelligent) R&D centers in China.

In July 2025, BMW Brilliance established its first and only information technology R&D center in China in Nanjing—its largest such facility in Asia—focusing on core technologies like intelligent cockpits, intelligent driving assistance systems, and AI large model applications. In April 2026, Audi and SAIC Motor jointly established the 'Audi Innovation Technology Center' in Shanghai, prioritizing AI-powered intelligent cockpits and advanced driving assistance systems tailored for Chinese users.

Automakers' demand for intelligent driving suppliers is rapidly devolving from turnkey full-solution packages to 'underlying hardware + basic toolchains + pluggable modular software.'

Behind this shift lies automakers' anxiety over data sovereignty, iteration efficiency, and cost control. Intelligent driving has replaced the three-electric system as the most critical differentiator in the second half of the electric vehicle race. An automaker that does not control the algorithm logic behind driving decisions cedes ultimate authority over user experience.

More critically, intelligent driving capabilities rely heavily on data loops. If core algorithms remain with suppliers, automakers will always be held hostage.

The proliferation of white-box delivery and joint development models essentially represents automakers systematically dismantling suppliers' technical barriers—suppliers provide chips and basic software, while automakers handle upper-layer algorithms and data training.

For suppliers like Horizon Robotics and Momenta, this means sustained pressure on profit margins.

Customized projects yield hard-earned profits, while standardized products lack pricing power. More clients and higher delivery volumes often translate to higher labor costs and lower gross margins. As production scales up, profitability quality declines—a dilemma facing all Tier 1 suppliers.

The rosy vision of a subscription model also struggles against reality. The industry once hailed subscriptions as the ultimate answer to intelligent driving profitability: one-time hardware costs covered by vehicle prices, with software subscriptions generating pure profit. Tesla's FSD subscription model has been worshipped as a blueprint by countless domestic players.

But real-world data from China is far less optimistic. Subscription revenue may contribute incremental growth but cannot yet serve as the pillar of intelligent driving profitability.

Part.3 The Commercial Model Challenge: The Second Half Will Be Won on Cost Control, Business Positioning, and Cash Flow Management

When mass production is no longer a scarce capability and technological experiences rapidly converge, the competitive logic of the autonomous driving industry has changed. The first half was won on technological breakthroughs; the second half will be won on cost control, business positioning, and cash flow management.

Players across different tracks are diverging sharply.

The automaker faction pursuing full-stack self-research wins on closed loops but struggles with investment. Players like Tesla, Xiaomi, XPeng, and Li Auto possess complete data loops and vehicle profit buffers. Intelligent driving can reinforce brand premiums or explore value-added services like subscriptions.

But the risk lies in price wars forcing intelligent driving into standard packages, making R&D investment difficult to recoup through single-product premiums.

For these automakers, intelligent driving is not a standalone profit unit but part of overall vehicle competitiveness. The calculus must consider brand and sales volumes, not just individual features.

The solution provider faction wins on breadth but struggles with pricing power. Players like Horizon Robotics and Momenta serve multiple automakers, scaling rapidly without bearing Vehicle market (vehicle market) volatility risks.

But as automakers accelerate in-house R&D, technological premiums will continue to erode, likely relegating these suppliers to underlying chip and basic software roles—earning hard but stable profits in the middle of the supply chain.

Among them, top players with full-stack delivery capabilities can defend market share through engineering efficiency and cost advantages, while smaller suppliers may be eliminated in the next reshuffling, becoming cost-sensitive contract manufacturers.

The pure Robotaxi faction wins on long-term potential but struggles in the present. Players like Pony.ai and WeRide boast the highest technical barriers and the most imaginative long-term operational models but face the slowest commercialization.

In the domestic market, despite Head platform (leading platforms) launching fully driverless operations in multiple cities and hardware costs continuing to drop, overall profitability remains elusive.

A technical expert in autonomous driving once revealed that domestic Robotaxi vehicles typically generate less than 250 yuan in daily revenue, with annual real revenue of only 70,000–80,000 yuan. Annual per-vehicle labor costs are around 40,000 yuan, and when hardware depreciation and comprehensive operation and maintenance costs are added, total annual costs per vehicle in China reach approximately 140,000 yuan.

From a financial perspective, each domestic Robotaxi incurs an annual real loss of about 50,000–60,000 yuan. The industry's so-called 'per-vehicle UE break-even' does not represent overall scalable profitability but only single-vehicle profitability in select high-quality cases.

A glimmer of breakthrough comes from overseas markets. Robotaxi fares in some regions can reach 2.5–3 times those in China, with only marginal cost increases. As the ratio of cloud-based safety officers to vehicles improves further, stable profitability could be achieved first overseas.

Ultimately, autonomous driving will never have a unified profitability moment. It will not suddenly become universally profitable after a single technological breakthrough. Instead, profitability will gradually materialize across different links and players as industrial divisions are reconstructed.

Just as in the smartphone industry, where some earned profits from brands, others from chips, and others from contract manufacturing, no one can dominate the entire chain. The same will apply to autonomous driving. As technology loses its mystique and the industry matures, every player must find its ecological niche, abandon illusions of dominating the entire chain, and accept the reality of stratification.

Momenta's IPO is a milestone but not the finale. NVIDIA's restructuring is defensive yet opens new offensives. ByteDance's crossover is a variable but also a new opportunity for the entire industry.

The mass production frenzy will eventually fade, and the space for hardware cost reductions will gradually peak. The second half of the competition will no longer be about whose algorithms appear smoother in demo videos but about who can unlock user value in a low-engagement market, who can defend core barriers amid industrial reshuffling, who can reach break-even first in overseas Robotaxi niches, and who can convert technological advantages into sustainable commercial advantages before cash flows dry up.

For all players still at the table, the real battle begins at the moment when costs can no longer be reduced.

END

Wang Qingru

I have been keeping a close eye on major internet companies and leading enterprises in vertical industries. I welcome connections and communication.