Xiaopeng Motors reported a staggering revenue of 15.81 billion yuan in Q1, marking a 141.5% year-on-year increase—a figure akin to an industry bombshell. Net losses narrowed significantly to 660 million yuan, a near-halving from the 1.37 billion yuan recorded in the same period last year. This achievement transcends mere "cost reduction and efficiency improvement"; it's a "relentless cost-cutting offensive" propelling the company towards profitability. Even more remarkable, automotive sales revenue soared to 14.369 billion yuan, up 159.2% year-on-year, mirroring a money-minting machine in action.

However, beneath these impressive numbers lies a critical truth. Xiaopeng's sales structure mirrors a tale of two models: MONA M03 and P7+ dominate sales, accounting for over 80% of total sales. But the average vehicle price has plummeted from over 200,000 yuan in 2024 to around 170,000 yuan in Q1 2025. This isn't just a price cut; it's a clear "trade-off between price and volume" strategy.

Xiaopeng's Q1 automotive gross margin reached 13.9% (up from 7.6% in the same period last year), with the loss per vehicle shrinking to -7,100 yuan (an 88.9% improvement year-on-year). This validates the effectiveness of its "technology compound interest" model. This diminishing marginal cost effect is fueled by:

• Scale Effect of XNGP Architecture: The BEV perception network, supported by dual Orin-X chips (508 TOPS computing power), enables L3-level intelligent driving functions on the MONA M03 MAX model at just 99,800 yuan, reducing intelligent driving hardware costs from 22% to 14%.
• Neural Network Compression Technology: Through knowledge distillation and quantization pruning, the size of the autonomous driving model is reduced by 40%, compatible with both NVIDIA Orin and Qualcomm Snapdragon Ride platforms.
• Mass Production of 800V SiC Platform: Paired with 480kW ultra-fast charging piles, it achieves 200km of range with a 5-minute charge, reducing the BOM cost of the G6 model by 12% compared to the E platform.

Behind these short-term gains, the broader industry landscape is undergoing seismic shifts.

Historically, the automotive industry was dominated by "electrification," with competition centered on battery power and endurance. But now, electrification feels like a bygone era, and the spotlight has shifted to the "intelligence" feast. Xiaopeng Motors recognizes that hardware stacking alone is no longer viable; software innovation is paramount. Cars have evolved beyond mere transportation tools into mobile intelligent terminals.

However, the reality is harsh. Today's so-called "intelligent cars" are still "pseudo-intelligent." Single-vehicle intelligence has hit a wall, with limited perception range and inadequate decision-making capabilities, akin to children confined to their own worlds. True intelligence necessitates a leap from single-vehicle intelligence to AI networking.

Single-Vehicle Intelligence: The Tipping Point from Quantitative to Qualitative Change

Looking back at the first half of automotive development, electrification was undoubtedly the mainstay. As technology matures, the industry's focus is shifting to the second half—intelligence. Xiaopeng Motors' financial report highlights that its sustained investment in intelligence is bearing fruit. For instance, the launch of the Xiaopeng MONA M03 MAX, powered by dual Orin X chips, not only elevates driving assistance functions among peers but also showcases the vast potential of single-vehicle intelligence.

However, single-vehicle intelligence faces bottlenecks. While vehicles' perception and decision-making capabilities are improving, relying solely on VLM (Visual Language Model) and VLA (Visual Logical Reasoning) is insufficient for widespread L4-level autonomous driving. The industry must now explore a new path: from intelligence to AI networking.

AI Networking: The Future Blueprint for Reshaping the Transportation Ecosystem

In the future, constructing a comprehensive AI-networked transportation ecosystem will be the norm. In this system, vehicles are interconnected intelligent nodes, seamlessly integrating with roads, the cloud, other vehicles, and smart devices.

AI networks' global perception capabilities enable vehicles to obtain real-time traffic information, enhancing decision-making safety and traffic flow efficiency. Imagine a world where all vehicles can predict road conditions and make optimal decisions, eliminating traffic congestion and making travel more efficient and enjoyable.

Furthermore, AI networking's real-time reasoning and decision-making capabilities pave the way for autonomous driving. By continuously refining AI algorithms, vehicles can swiftly make precise decisions in complex scenarios, making autonomous driving more reliable and paving the way for driverless vehicles.

Cost Dilemma: Balancing Intelligence and Cost-Effectiveness

The cost issue of new energy vehicles remains a critical constraint. As intelligent configurations increase, so do manufacturing costs. Despite technological optimizations and supply chain management, intelligent upgrades still impact vehicle prices.

However, these cost increases are not in vain. With safety and experience as priorities, consumers are increasingly accepting intelligent features. For instance, the advanced safety systems and enhanced user experience on the Xiaopeng MONA M03 MAX, despite raising the price, offer higher perceived value. Finding the sweet spot between intelligent upgrades and cost control will be a persistent challenge for automotive enterprises.

Vehicle-Road-Cloud Integration: Ushering in a New Era of Intelligent Transportation

With the evolution of vehicle-road-cloud data integration technology, automotive intelligence is entering a new phase. Deep integration of vehicle and cloud data enhances navigation precision, driving assistance intelligence, and creates a holistic intelligent travel ecosystem for users.

In the future, this technology will extend beyond intelligent connected and autonomous vehicles. Ordinary car users can also access this ecosystem via mobile apps and tablets, enjoying the convenience and safety of intelligent driving. Additionally, connecting AI networks with drones, robots, and other intelligent devices will enable broader interactive and collaborative applications.

As BEV perception networks encounter long-tail bottlenecks and end-to-end models approach physical limits, the automotive intelligent revolution is entering an era of "algorithms, data, and infrastructure." Xiaopeng's financial report not only signifies a financial turnaround but also heralds that, in the next decade, to survive, intelligent driving players must harness both Moore's Law for single-vehicle intelligence and Metcalfe's Law for vehicle-road-cloud collaboration—this is the true "deep water" of AI autos.