Gasgoo Munich-On September 18, at the Gasgoo 4th AI-Defined Vehicle Forum 2026, Richard Lam, Cockpit Division, Product Director Huizhou Foryou General Electronics Co., Ltd, delivered a speech titled "Sharing AI Cockpit Intelligent Evolution Solutions." He provided a systematic overview of the AI cockpit industry’s current landscape, application pain points, and Foryou’s edge-cloud collaborative solution.
Richard Lam began by sharing industry data: intelligence has become a decisive factor for car buyers, accounting for roughly 73% of purchasing decisions—a figure climbing annually. The AI cockpit market is projected to surge from around 2 billion yuan to over 20 billion yuan, with growth exceeding 20 times between 2026 and 2035, signaling immense potential. He traced the cockpit’s evolution from early hardware-driven days to the rise of software experience 8 to 10 years ago, and finally to the current "AI-defined car" stage, driven by iterations in VLM and VLA technology over the past three years. According to SAE standards, multi-modal intelligent cockpits are classified into five levels; the industry currently sits at a stage of partial cognition, with the long-term goal being a full-domain AI ecosystem. Automakers like BYD, Geely, and Li Auto have all initiated AI cockpit deployments, accelerating the industry’s rollout pace. However, Lin pointed out a critical bottleneck: large models inside the cabin currently cover only basic scenarios like voice interaction and intelligent assistants, offering limited practical utility and failing to fully unlock AI’s value.
On the technical front, Richard Lam outlined three implementation paths for AI cockpits: pure cloud solutions, brand-new cockpit edge-side AI solutions, and retrofitting existing cockpits with AI coprocessors. He compared their core metrics: the pure cloud approach requires no additional development cycle but incurs token costs of about 160 yuan per vehicle annually, suffering from high latency and a lack of personalization. The new edge-side AI solution demands a roughly 20-month development cycle with high R&D investment but eliminates long-term token fees while supporting personalization and low latency. Retrofitting existing cockpits with an AI coprocessor takes about 10 months—saving roughly 12 months compared to a full redevelopment—while also avoiding token fees and delivering an experience close to native edge-side solutions. Lin argued that adding an AI coprocessor to existing architectures is currently the optimal path to balance experience, cost, and development efficiency.
Richard Lam introduced Foryou’s full-domain AI technical solution, launched last year and continuously iterated upon via the company’s open platform. Built on an SOA cross-service architecture, the solution uses the NCP protocol to bridge upper and lower model services and relies on A2A for multi-agent collaboration. On the device side, the architecture ensures user data privacy remains inside the vehicle, executes local tasks with low latency, and offers a more cost-effective lifecycle solution. He illustrated the edge-cloud logic using voice command processing: voice input is converted to text via ASR, parsed by LM and MLU to determine intent, and then scheduled between local and cloud models based on user profiles and environmental parameters. After verification and filtering at the edge, the system finally executes the intent and updates the interface. Lin noted that edge-side AI enables differentiated results for different demographics, scenarios, and regions. The system identifies the speaker’s identity, age, gender, and preferences, combining contextual memory to deliver personalized responses rather than generic replies. Furthermore, the solution eliminates the need for multi-level user selection; it directly fulfills requests by synthesizing parameters like battery level, time, road conditions, and user habits.
Regarding hardware compatibility, Richard Lam noted that Foryou’s solution adapts to mainstream cockpit chips such as the 8392 and 8797. When paired with an AI coprocessor, it delivers superior cost-performance ratios. The AI coprocessor currently in pre-research boasts computing power of 37D and has already been adapted to mainstream general large models like Qwen. Lin emphasized Foryou’s differentiated value as a Tier 1 supplier with deep automotive-grade expertise. He pointed out that most model developers and solution providers avoid automotive-grade control, focusing instead on specific scenarios, while OEMs remain cautious about opening external permissions and defining liability. Foryou brings mature experience in system integration, functional safety, information security, and mass production engineering, covering the full chain of needs including intent bridging, permission control, multi-sensor scheduling, SOA services, OTA deployment, and multi-agent cross-domain scheduling.
Richard Lam also detailed Foryou’s comprehensive product matrix, which has formed mature cockpit solutions across multiple dimensions: one-chip-multi-screen, one-domain-multi-chip, SOA distributed architecture, edge-cloud agents, and central computing. With cockpit domain control as the core hub, the company covers a full chain of in-cabin hardware and system support services, including visual systems like HUD and PHUD, as well as auditory interaction and audio processing. Currently, Foryou’s AI cockpit R&D roadmap is clear: its edge-cloud software architecture based on an open platform has achieved mass production, the AI coprocessor with 37D computing power is in pre-research, and a multi-modal full-domain AI cockpit experience is in the planning stages.
In his closing remarks, Richard Lam likened the human pursuit of AI to the vastness of the starry sky—boundless. Foryou remains committed to crafting full-domain AI cockpit experiences for its clients, providing OEMs with high-performance, cost-effective, full-stack solutions for all-scenario AI intelligence.
