EN / 中文

Cockpit Agent AI Goes Deeper; Volcano Engine Moves From LLMs to Full-Vehicle Underlying Collaboration

by zhinengchecankao·November 12, 2026

The "big results" of the AI automotive wave may not necessarily be achieved first by players who develop their own technologies—

Shenlong Automobile, a long-established joint venture automaker with a 34-year history, recently announced a comprehensive strategic partnership with Volcano Engine, officially joining the ranks of "Doubao Cars."

In September this year, Shenlong Technology had just signed a global strategic cooperation agreement with Momenta, planning to use Momenta's R7 world model for new mass-produced models under brands such as Peugeot and Jeep.

Intelligent driving found Momenta; intelligent cockpit partnered with Doubao.

More interestingly, Volcano Engine has started touching the underlying layer of the cockpit system and chips in this collaboration.

34-Year Joint Venture Automaker Shenlong Joins the "Doubao Car" Camp

Let's look at the specific content of the cooperation.

The cooperation covers three aspects: enterprise AI transformation, product and technology co-creation, and full-cycle user services.

Among these, product and technology co-creation is the most noteworthy.

Specifically, it includes joint development of smart cockpit systems, co-creation of application software, SoC chip product planning, forward-looking technology research, and mass production vehicle development.

These keywords together go beyond common onboard large model integration.

In the past, when automakers integrated Doubao, they usually introduced large model capabilities into existing car infotainment systems through methods like model APIs and voice interaction solutions.

We have seen many such "Doubao Cars." Going further, the cooperation will delve deeper into connecting Doubao's large model with over 2000 SOA service interfaces of the entire vehicle, allowing AI to call upon vehicle functions such as air conditioning, seats, and sunshades.

Previously, SAIC Roewe Jiayue 07 had already demonstrated a similar technical path.

Joint development of the cockpit system requires both parties to collaborate further around models, software architecture, and whole-vehicle function interfaces.

In this process, besides integrating with the whole-vehicle SOA service interfaces, clarifying calling permissions, execution logic, and feedback mechanisms for different functions, the large model needs more importantly to access lower-level vehicle status parameters, such as speed, inclination angle, acceleration, cabin air quality, passenger status detected by in-car cameras, etc.—

That is to say, Doubao's large model, except for not participating in driving tasks, knows everything it should know, understands everything happening inside and outside the vehicle, and knows what users need most right now.

Another new piece of information is SoC chip product planning.

SoC (System on Chip) bears the computational tasks on the vehicle side, implying that the future scope of collaboration between Shenlong and Volcano Engine may involve requirement definition, computing power configuration, and software-hardware adaptation for the cockpit computing platform.

Who specifically designs the chips and which suppliers are adopted has not yet been disclosed publicly.

At the same time, Volcano Engine's cooperation will also enter Shenlong Technology's R&D, marketing, enterprise management, and overseas business.

Interestingly, a direct procurement announcement disclosed by Dongfeng Motor's procurement platform in October showed that Shenlong Technology plans to procure approximately CNY 23.32 million worth of Volcano Cloud product rental services for marketing system migration, cloud base construction, AI large model applications, and connected vehicle service cluster construction.

This means that the cooperation between the two parties has already involved enterprise cloud infrastructure, and the latest strategic signing in November expanded the scope to cockpit and whole-vehicle product R&D.

What Has Volcano Engine Accumulated Behind Doubao Cars?

Shenlong Technology's choice this time is easier to understand when viewed in the context of Volcano Engine's automotive business development over the past few years.

In its early days entering the automotive industry, Volcano Engine mainly relied on cloud computing, data platforms, and digital services.

Automotive R&D requires massive data storage, computing, and processing resources; autonomous driving teams need cloud computing power to support model training and simulation; and automotive enterprises' marketing, sales, and after-sales services also require data platforms.

Volcano Engine outputs the cloud, data, and algorithm capabilities accumulated by ByteDance in internet businesses to automakers, with cooperation occurring mostly at the level of enterprise digitalization and R&D infrastructure.

At the same time, ByteDance possesses a set of resources that other tech companies find difficult to replicate: content and user platforms such as Douyin (TikTok China), Toutiao, and Dongchedi. Content distribution, user reach, and marketing digitalization have also become components of Volcano Engine's automotive business.

After the emergence of the Doubao large model, Volcano Engine began to further enter the interaction layer of automobiles.

During the 2026 Beijing Auto Show, Volcano Engine released a new generation of automotive AI solutions based on Agentic AI architecture, including two forms: the AI Cockpit Suite and the Doubao Cockpit Assistant.

The former can provide modules such as models, AI engines, interaction tools, and agents according to automakers' needs, allowing them to combine and develop independently; the latter is a relatively complete product-level delivery, with plans to interconnect with the mobile Doubao App.

The core change here lies in Agentic AI.

With working agents becoming so popular, everyone understands this concept without barriers now.

Traditional voice assistants mainly rely on preset intents and functional domain recognition, requiring users to break down needs into individual commands.

Agentic AI aims to understand context, decompose tasks, call onboard tools through a unified model, and continue completing tasks based on execution results: it is equivalent to having your most trusted personal assistant who best understands your needs sitting in the front passenger seat, breaking down your one-sentence command into different steps and then helping you tap the screen or press buttons.

The technical architecture published by Volcano Engine has actually incorporated perception, reasoning, execution, memory, and learning into an overall process.

At the 2026 Beijing Auto Show, Volcano Engine disclosed that cars equipped with the Doubao large model had already exceeded 7 million units, covering more than 50 brands and 145 models, with daily cockpit interactions exceeding 30 million times.

This indicates that Doubao has already established a considerable scale of automotive application foundation.

And it is almost certain that with this basis of 7 million mass-produced vehicles, the cockpit of Shenlong Technology's new cars will likely adopt the latest generation of Agentic AI architecture.

Of course, in this cooperation content, Volcano Engine's scope of participation has extended one step further: joint development of cockpit systems, SoC chip planning, and mass production vehicle development.

This implies that cooperation between model companies and automakers is beginning to involve synergy in whole-vehicle computing resources and underlying software interfaces.

What kind of vehicle-side computing power AI needs, how to call whole-vehicle functions, and how to organize cockpit software may all enter the product R&D phase discussed jointly by both parties.

ByteDance Replicates the "Huawei Moment" in the Auto Circle

From initially providing cloud services and data platforms, to deploying the Doubao large model in cars, to Agentic AI cockpits and joint system development with Shenlong Technology, ByteDance's position in the automotive industry chain continues to extend inward toward products.

AIVA, released in June this year, was a critical turning point node.

The initial popularity of "Doubao Cars" itself demonstrates that the changing role of Volcano Engine or ByteDance is the topic of greatest interest to outsiders.

The product philosophy proposed by AIVA is "AI defines the car; AI comes first, then the car," placing AI upfront in the product definition stage. Volcano Engine begins to participate in experience design during the product development stage, organizing product solutions around model capabilities, interaction methods, and vehicle functions.

Shenlong Technology further extends the scope of cooperation to whole-vehicle Agents, chip planning, and mass production development.

Putting several cases together, the puzzle of ByteDance's automotive business has gradually become clear.

At the bottom layer, there is Volcano Engine's cloud computing and data infrastructure; in the middle, there are Doubao large models, Agents, and smart cockpit solutions; on the vehicle side, there are different automakers' R&D systems and mass production platforms.

It is increasingly gaining trust from automakers and access to underlying permissions, deeply participating in technical R&D and definition.

Cooperation with Shenlong also saw chip-related content appear for the first time.

Compared to Huawei, the ByteDance ecosystem also has its own media platforms...

For automakers, they can choose models, cloud services, and cockpit solutions based on their own R&D capabilities, or further carry out joint development; for Volcano Engine, more mass-produced models mean broader model application scenarios and product implementation opportunities.

Initially, everyone thought this wave of AI automobiles was driven by autonomous driving. However, having experienced the shift from rule-based and map-dominated approaches to today's data-driven dominance, people have gradually recognized that the core power driving automobiles to upgrade from carriers to consumer goods and tech products can only be large models, specifically multimodal large models.

This is the unstoppable power of the underlying paradigm: whether it is the cockpit or driving, only large models enable onboard AI to truly perceive and understand everything happening around it.

After large models took the central position, ByteDance, vertically integrated in the automotive field, is replicating the "Huawei Moment" and "Momenta Moment" in the cockpit sector:

More and more automakers are starting to join Doubao Cars.

Except for AIVA, Shenlong, and SAIC mentioned earlier, brands such as Mercedes-Benz, Audi, Volkswagen, Chery, Hongqi, and Buick already have models or related solutions equipped with the Doubao large model.

On the one hand, the cockpit sector is practically a blue ocean. Except for Qualcomm dominating the chip market, OS and application layers see automakers fighting separately, with virtually no competitors on par with ByteDance in the market.

ByteDance has another special feature: even after vehicle delivery, users may still connect with this ecosystem through products like Doubao, Douyin, and Dongchedi. There is space for further integration between in-car AI, mobile assistants, and content services.

It was unexpected that such a "big result" was achieved first by a veteran joint venture brand:

National-level AI assistant Doubao and the strongest intelligent driving model player Momenta serving Shenlong Automobile—one person receives both blessings; fortune cannot be small.