ZOZO Auto Research Releases the "2026 OEM and Tier 1 Intelligent Cockpit Platform (Hardware & Software) Innovation Strategy Research Report"
Reconstruction of Intelligent Cockpit Software System Architecture to Build an AI (Artificial Intelligence) Native Cockpit System
Driven by the evolution of the centralized integrated architecture in automotive EEA (Electrical/Electronic Architecture) and the advancement of intelligent technologies, the automotive intelligent cockpit has advanced from the initial distributed signal-oriented cockpit system architecture to a service-oriented software system architecture.
Driven by factors such as AI large models and the integration of Agents into vehicles, the intelligent cockpit system architecture has undergone further transformation, giving rise to the AI-native cockpit system architecture. This refers to a cockpit system centered on AI large models and Agents, featuring a full-chain reconstruction from underlying computing power to upper-level interaction. It breaks through the limitations of traditional intelligent cockpits characterized by "function stacking" and "passive interaction," achieving a proactive intelligent evolution where "services seek out users."
Currently, multiple OEMs are gradually building an AI-native cockpit system architecture and realizing a proactive service intelligent cockpit experience through layouts that include underlying OS AI integration, AI foundation large models, application ecosystems, and multi-Agent collaboration.
Layout of AI-Native Cockpit System Architecture by Major OEMs
Source: ZOZO Auto Research "2026 OEM and Tier 1 Intelligent Cockpit Platform (Hardware & Software) Innovation Strategy Research Report"
In May 2026, BYD newly launched the super intelligent agent "Didi Xia," an in-vehicle AI secretary integrating intelligent interaction, proactive services, and an open ecosystem. It empowers the DiLink AI intelligent cockpit and possesses three core capabilities: "engaging in conversation, executing tasks, and providing entertainment." Behind Didi Xia is a full-chain agent architecture:
The AIOS operating system serves as the underlying capability platform;
The super agent engine handles interaction and task scheduling;
The Agent open platform integrates with the third-party service ecosystem, thereby achieving linkage and collaboration among different services.
In April 2026, GAC newly launched the Xinghe Intelligent Cockpit ADiGO Intelligence edge-cloud integrated architecture. Relying on high-computing-power chips and its self-developed AIOS underlying system, it utilizes three self-developed core engines (multimodal perception engine, memory-based multi-agent collaboration engine, and emotional expression engine) to drive the cockpit's evolution into a caring partner that "understands you," realizing a proactive service experience.
Reconstruction of Cockpit Hardware Platforms: Beyond Centralized Integrated Computing Platforms, New Solutions like External AI BOX Empower the Scaled Deployment of AI Cockpits
Traditional cockpit hardware only meets the basic display and computing needs of the central control and instrument panels, and cannot support high-level multimodal large models and multi-Agent parallel inference capabilities. In response, the intelligent cockpit hardware platform is undergoing transformations in directions such as centralized computing platforms, high-computing-power AI cockpit domain controller platforms, and external AI BOX elastic computing power.
Key Direction 1: Accelerated Mass Production Layout of Centralized Computing Platforms. Driven by multiple factors such as the evolution of the "centralized computing + zonal control" vehicle EEA and the continuous mass production and deployment of high-computing-power automotive SoCs (System on Chip), intelligent cockpit computing hardware is accelerating its iterative upgrade toward centralized computing platforms. Multiple suppliers in the industry chain are increasing their investments in the layout of integrated cockpit and driving computing platform products.
Riding the wave of industry transformation, Neusoft Intelligent Mobility, a subsidiary under Neusoft Group, has completed a strategic upgrade. Its business focuses on the full-domain scenarios of upper body electronics, building a full-stack automotive electronic product matrix covering centralized computing platforms, integrated cockpit-driving-parking systems, intelligent cockpits, vehicle communication, intelligent audio, and zonal control units (ZCU), and taking the lead in deploying edge-side large model applications.
Radar Chart of Neusoft Intelligent Cockpit Platform Capabilities
Source: ZOZO Auto Research "2026 OEM and Tier 1 Intelligent Cockpit Platform (Hardware & Software) Innovation Strategy Research Report"
Neusoft Intelligent Mobility's key A³ integrated cockpit-driving-parking product platform adopts a OneChip single-chip architecture. Leveraging hardware virtualization technology, it connects all upper body business domains such as the cockpit, autonomous driving, vehicle communication, and body control, building a new-generation vehicle centralized computing foundation and achieving intelligent integration across all vehicle domains. This platform can adapt to edge-side AI large models with parameter scales of 4B/7B, partnering with ecosystem partners to create a highly intelligent, multimodal human-vehicle interaction cockpit. Meanwhile, it integrates the E2E (End-to-End) autonomous driving model and the VLA (Vision-Language-Action) world model, realizing all-scenario, high-safety advanced autonomous driving capabilities.
In April 2026, Hangsheng Electronics, facing the era of AI vehicles, officially released a full set of technology foundation + domain controller + electronic control product matrix, and launched the Mozi 3.0 five-domain fusion centralized computing platform. It integrates five major capabilities: cockpit, intelligent driving, connected vehicle, vehicle control, and edge-side AI. It replaces traditional distributed ECUs with 1-2 high-performance computing units, with computing power ranging from 100+ TOPS to 1000+ TOPS. Equipped with the Moyuan AI OS, it features cross-screen 3D interaction and proactive service agents, achieving a closed loop of edge-side AI computing power.
Key Direction 2: External AI BOX Enters the Stage of Large-Scale Design-in and Mass Production. For existing stock models and mid-end models, the native cockpit SoC computing power is insufficient to support high-level large models and multi-Agents, and replacing the vehicle domain controller involves extremely high costs and long certification cycles. The AI BOX, by connecting an independent computing unit via a high-speed Ethernet interface, requires no changes to the vehicle EEA architecture, enabling independent iterative upgrades of hardware computing power while balancing cost and intelligent experience. In 2026, it enters a cycle of large-scale design-in and mass production. Multiple suppliers have proposed related AI BOX product solutions for this purpose.
In 2026, ThunderSoft launched the AIBOX-N1 product. Built on the NVIDIA DRIVE AGX Orin, it features 200 TOPS of AI computing power and supports the real-time in-vehicle operation of 7B large models. Paired with its self-developed Dishui AI OS 2.0, it natively adapts to the multi-Agent scheduling framework and has been deployed in mass production in models such as the Geely Galaxy M9 Black Gold Smart Edition.
In June 2026, Desay SV exhibited the EA01S + AI BOX solution. By adding an external AI BOX to the basic cockpit domain controller platform, it assists in the mass production deployment of mainstream AI cockpits. The basic cockpit domain controller platform adopts the Qualcomm QCM8838 chip, providing basic cockpit function support such as multimedia playback, navigation, and basic voice interaction, with a stable hardware foundation and software ecosystem. The independent AI computing unit AI BOX is built with the integrated Qualcomm QCS9075 chip, connecting to the EA01S via a standardized interface, and is responsible for running edge-side AI large models and complex AI tasks.
In April 2026, Neusoft Group released the NAGIC.AI cockpit software platform product, adopting a parallel dual-architecture design of "distributed + centralized," covering both existing stock models and new models. For existing models, it introduces an AIBOX solution to supplement computing power via an external AI computing unit. This solution possesses three core characteristics: easy expansion, easy iteration, and easy deployment. Without subverting the existing cockpit architecture, it can quickly and cost-effectively make up for computing power shortcomings simply by adding an external high-performance AI computing unit.
For new models facing the next-generation centralized computing platform: with the centralized computing platform as the core foundation of the next-generation intelligent cockpit, Neusoft provides full-stack AI empowerment capabilities from underlying operating system optimization to upper-level AI application deployment. It fully releases the powerful computing power and data throughput capabilities of the centralized architecture, ensuring the efficient operation, stability, reliability, and full-domain controllability of the intelligent experience, meeting the strict requirements of automakers for system safety and long-term management.
Reconstruction of the Cockpit Ecosystem: Evolution from APPs to a Multi-Agent Collaborative Ecosystem to Build a Proactive Service Intelligent Cockpit
In 2026, the competition logic of the intelligent cockpit ecosystem has been completely iterated. The industry officially bids farewell to the traditional model centered on the in-vehicle APP application store ecosystem and enters a brand-new stage of multi-agent collaborative ecosystem.
An increasing number of automakers, suppliers, and technology platforms are no longer satisfied with simply putting a "voice large model" into a car. Instead, they enable multiple AI agents to collaborate with division of labor, understanding intentions, planning tasks, invoking tools, and executing across domains, providing users with a proactive intelligent service experience.
Latest Intelligent Cockpit Agent Ecosystem Solutions from Major Suppliers
Source: ZOZO Auto Research "2026 OEM and Tier 1 Intelligent Cockpit Platform (Hardware & Software) Innovation Strategy Research Report"
In April 2026, Huawei released the HarmonyOS Cockpit HarmonySpace 6. Its core technical variable is the upgrade of the Xiaoyi intelligent agent from the MoLA architecture to MoLA 2.0. Through the architectural logic of "brain + cerebellum," 2.0 enables a single model to achieve optimal performance in all scenarios. The System Agent is the "brain," upgraded to a hundred-billion-level end-to-end large model, responsible for cross-domain collaboration and intention judgment. The Navigation Agent, Vehicle Control Agent, and Chat Agent are the "cerebellum," each managing its own domain, with efficient execution in their respective fields handed over to them.
Empowered by the brand-new MoLA 2.0 architecture, the Xiaoyi intelligent agent achieves cross-service invocation across all scenarios, completing fuzzy semantics and complex travel tasks in one go. At the ecosystem level, it integrates massive third-party Agents in travel, catering, audio, and video, building a closed-loop proactive service ecosystem.
In April 2026, Neusoft officially launched OneCoreGo® Global In-Vehicle Intelligent Mobility Solution 7.0. This upgrade focuses on AI integration and globalization. Utilizing the cross-Agent collaboration capability of One Mate AI, and relying on product matrices and systems such as One Map, One Sight, One Cloud, One Pay, One Store, One Link, and One Guard, it forms a mobility solution that is deeply empowered by AI, widely covered globally, rich in ecosystem, and precise in scenarios.
Among them, the newly upgraded One Mate (AI mobility companion) breaks free from rigid one-question-one-answer commands. Through deep semantic understanding and situational prediction, it accurately reads the real needs behind users' fuzzy expressions. Meanwhile, One Mate possesses cross-Agent collaboration capabilities, which can break down the functional silos of all modules such as navigation, AR, payment, cloud, security, ecosystem, and vehicle connectivity, achieving cross-agent collaboration.
[Contact Information] Mobile number is the same as WeChat number
Industry Research Department | Mr. Fu 15810027571
Mr. Zhao 18702148304
Data Service Department | Ms. Zhang 13716037793
Strategic Consulting Department | Ms. Han 15810133447
Promotion and Communication Department | Ms. Liao 13718845418
Mr. Du 13910162318