Article · AI translation
IT Home reported on October 21 that as large language models, multimodal interaction and architectures coordinating on-device and cloud processing continue to mature, AI-native applications such as intelligent voice, AI assistants, navigation Q&A, content generation and personalized recommendations are rapidly making their way into cars.
International Data Corporation (IDC) today, in its China Automotive Cloud Market Tracker report, systematically tracked AI cockpit solutions as a separate add-on market for the first time and released its initial findings on market size, the competitive landscape, technology paths and business models.
IDC defines the AI cockpit solutions market as covering four layers:
AI development platforms/software (model development, Agent orchestration, deployment management, etc.)
Model services, or MaaS (cloud-based, in-vehicle and device-cloud coordinated use of large language models, multimodal models and speech models)
AI applications (intelligent voice, AI assistants, navigation Q&A, content generation, etc.)
AI operations services (model updates, knowledge base management, operational monitoring and continuous optimization)
According to the report, China's AI cockpit solutions market reached RMB 710 million in the first half of 2026. This scale indicates that AI cockpits have entered a phase of growing volume, with market demand and commercialization accelerating in tandem as the market shifts from early exploration to scaled mass production and ongoing operations.
IDC expects China's AI cockpit solutions market to reach RMB 1.66 billion for the full year of 2026 and potentially grow further to RMB 4.71 billion by 2030, with a compound annual growth rate approaching 30%.
IT Home learned from the report that the market shows a pronounced concentration among leading players: in the first half of 2026, Volcano Engine ranked first with a 45% market share, while Alibaba ranked second.
As for the sources of their advantages, leading vendors' positions are built mainly on the number of mass-produced vehicle models they cover and their experience with projects for major customers. These are their current market positions, and the competitive landscape is far from settled. In the second tier:
Baidu remains active in the market, drawing on its existing cockpit business and capabilities such as maps;
iFlytek is steadily upgrading its AI capabilities, drawing on its cockpit-domain hardware and software solutions and existing customer base;
Model startups such as StepFun are also entering the market through close partnerships with automakers and industry-chain partners, and some of their technology has already been deployed in vehicles.
Overall, the market landscape is still evolving, and experience with mass-production projects, the ability to coordinate across ecosystems and consumer brand recognition are becoming key factors in vendor competition.
In terms of business models, the AI cockpit market currently features a variety of billing methods:
License agreements for combinations of features remain the mainstream model, with prices varying considerably depending on which features automakers select;
For customers with strong in-house development capabilities who need only basic model capabilities, a public-cloud model billed by token or number of calls has become an important supplement;
On-device deployment generally uses a one-time activation fee, supplemented by annual fees for technical support and model iteration and maintenance.
Notably, automakers are paying significantly more attention to the return on their AI investment. Pricing negotiations increasingly center on actual performance in use and the value delivered, while the gap in value between projects offering deep integration and licenses for basic features is widening.
Meanwhile, some vendors are using flagship projects to explore a shift from automaker-funded purchases to subscriptions paid by end users, opening up potential new revenue streams for the market.
In terms of technological evolution, coordination between on-device and cloud processing has become an industry consensus: on-device models respond quickly to frequent, simple commands, while large cloud-based models handle complex reasoning and multimodal interaction, balancing user experience, latency and cost.
In terms of interaction patterns, AI cockpits are undergoing a pivotal shift from “passive responses” to “proactive service.” The previous generation of cockpit voice assistants generally used a turn-based architecture combining “intent domains + multi-Agent coordination,” resulting in fragmented interactions and limited frequency of use. Since 2026, several leading vendors have released next-generation cockpit solutions based on an Agentic AI architecture. Their technology paths share an approach: using a unified AI hub to connect vehicle controls, navigation, intelligent driving and other functional domains, moving cockpit interaction from turn-based Q&A toward autonomous, task-driven execution.
According to information publicly disclosed by vendors, leading vendors' large models have been installed in millions of vehicles, and daily cockpit interactions number in the tens of millions. Although vendors use somewhat different statistical measures, it has become an observable industry trend that cockpit AI is gradually shifting from a rarely used “try-it-out feature” to an everyday service that users engage with frequently.