BAIC is betting on a thousand TOPS computing power chip. How far is the integration of cabin and driving?

There is no traffic jam in Hanamaki

2026.11.20 11:50

· From Beijing

Full text 2426 words

The computing power competition for automotive chips is entering a new stage.

A few years ago, computing power of dozens of TOPS was able to support many assisted driving functions. Nowadays, as urban NOA, end-to-end models and cockpit AI are accelerated to be put into cars, the requirements for computing power of cars continue to increase, and thousands of TOPS-level chips have begun to enter the research and development plans of car companies.

However, the computing power figures continue to be refreshed, and a question also arises: How much computing power does a car need?

On November 12, BAIC Group and the new Unisoc Group announced a deepening of cooperation. BAIC Research Institute and Unisoc Group will jointly develop high-computing cabin driving and control integrated car-grade SoC chips. According to the plan, the computing power target of a single chip is to reach 1,000TOPS, and support flexible expansion to meet the needs of different models. It is planned to achieve mass production of the first model in 2028.

Rather than simply pursuing higher computing power, how to adapt computing resources to different models, reduce system costs, and ultimately form large-scale delivery capabilities may better reflect the actual value of this cooperation.

1. How will the computing power be distributed?

Automotive chips are facing a contradiction: on the one hand, assisted driving and cockpit AI require stronger computing power; on the other hand, car companies must strictly control chip costs, power consumption and the burden on the vehicle's electrical system.

This contradiction is even more obvious for models in different price ranges.

A flagship model that needs to run high-end urban NOA, multi-modal interaction and large terminal-side models at the same time has different computing power requirements from a mainstream model that mainly provides high-speed assisted driving and basic voice interaction. If all models use the same set of high-specification chips, computing power may be idle; developing different chips separately means higher design, verification and software adaptation costs.

The platform design proposed by BAIC and New Ziguang is centered around this demand. According to the plan, the new generation of automotive-grade SoC will have strong computing power and scalability. Through flexible configuration of computing resources, it can meet the needs of different models in terms of intelligent driving, cockpit interaction and on-board AI applications.

The significance of this design is to improve the reuse rate of the underlying technology. Car companies can choose corresponding computing configurations based on vehicle model positioning, and chip manufacturers can also develop differentiated products on a unified technology platform to reduce repeated research and development and adaptation work.

However, computing power configuration is not a simple performance superposition. The chip's computing efficiency, data transmission capabilities, power consumption control, and resource allocation between different tasks will all affect the final performance.

Another direction worthy of attention is cabin-driving integration.

In the past, smart cockpits and assisted driving usually used relatively independent chips and controllers. As intelligent functions increase, the hardware cost, wiring complexity, and resource utilization problems brought by multiple computing platforms gradually become apparent.

If tasks such as environment perception, driving decision-making, and voice interaction can be run on a unified platform, there will be opportunities to reduce duplicate hardware configurations and improve resource utilization efficiency.

However, the cockpit pays more attention to the interactive experience, while assisted driving has stricter requirements for real-time performance, functional safety and stability. After different tasks share the computing platform, resource isolation, task scheduling and fault handling will become new technical challenges.

Therefore, for high computing power chips, peak TOPS is only one of the performance indicators. Effective computing power, energy efficiency and software and hardware synergy capabilities also determine its actual performance on the vehicle.

2. Car companies began to move upstream to chips

Compared with the computing power indicator, BAIC's role in this cooperation is also worthy of attention.

In the past, automobile companies mostly selected chips from suppliers' existing products, and then carried out algorithm transplantation, function development and vehicle adaptation based on vehicle model requirements. This model can use mature products to shorten the development cycle, but it also has limitations. There is a time lag between chip design and vehicle requirements, and the assisted driving algorithm is in a rapid iteration stage, so some hardware designs may not be able to fully adapt to new computing tasks.

For example, urban NOA needs to continuously handle environmental perception and driving decision-making tasks; automatic parking focuses on close range perception and control response; and cockpit multi-modal models put forward different requirements for data bandwidth and reasoning efficiency.

If the chip does not fully consider these needs in the early stages of design, even if it has high theoretical computing power in the later stage, problems such as insufficient algorithm operation efficiency, unreasonable resource allocation, or high power consumption may occur.

An important purpose of BAIC's in-depth participation this time is to introduce vehicle needs into chip design in advance.

According to the plan of both parties, the joint research and development will combine BAIC's existing vehicle electronic and electrical architecture, intelligent driving development experience and mass production verification system to jointly determine the functional requirements and performance targets of the chip.

During the specific research and development process, the two parties will carry out collaborative development in computing performance, energy efficiency, software adaptation and functional safety around typical application scenarios such as urban NOA, automatic parking, and cockpit multi-modal interaction.

For Ziguang Tongxin, obtaining vehicle application scenarios and verification conditions in advance will help reduce the disconnect between chip design and automotive applications. For BAIC, participating in the definition of underlying chips can form a closer collaborative relationship between its software architecture, intelligent driving algorithms and hardware platforms.

This also reflects changes in the automotive industry chain. As intelligence becomes an important factor in product competition, some car companies have begun to expand into areas such as chips, operating systems, and basic software, hoping to have more initiative in core technology routes and product iteration rhythms.

3. Mass production is the real threshold

For automotive chips, there is still a long engineering process from determining the technical architecture to actual installation in the car.

According to the plan of both parties, the joint chip project will achieve mass production of the first model in 2028. Currently, the two parties are promoting joint research and development, and there are still many links to be completed before it is finally put on the car.

Judging from the industry's usual R&D process, a new car-grade SoC needs to go through stages such as chip design, sample verification, software and hardware adaptation, reliability testing, and vehicle verification before it can enter the mass production introduction stage.

The first is the balance between performance and reliability. High computing power chips must not only meet the computing needs of complex AI tasks, but also must withstand the test of temperature changes, long-term operation, and complex usage environments. The performance indicators in the laboratory cannot be directly equated to the vehicle application capabilities.

Second is the software ecosystem. Whether the chip can perform well depends largely on the adaptability of the operating system, development tool chain, AI algorithm and mainstream application frameworks. If the development ecosystem is not mature enough, even if the hardware meets the design goals, it may increase the development burden of car companies.

Another challenge comes from system-level integration. Cabin-driving integration requires the coordination of different tasks on a unified computing platform, which puts forward higher requirements for real-time response, resource isolation and functional safety. At the same time, power consumption, heat dissipation, vehicle power supply and safety redundancy all need to be included in cost calculations.

Therefore, reducing the number of chips or controllers does not necessarily mean a reduction in vehicle costs. Only by completing system-level optimization can the economics of cross-domain integration be realized, which will further test the level of software and hardware integration and supply chain management.

Judging from the direction of cooperation announced by both parties, the scope of research and development is not limited to one SoC, but also includes cabin driving control integrated central computing platform, intelligent chassis domain controller and next-generation intelligent cockpit chip applications.

For domestic automotive chips, in addition to performance, they also need to face market tests on mass production orders, software ecology, cost and reliability.

By 2028, when this chip actually enters mass-produced models, the outside world will have the opportunity to test its computing power utilization, vehicle experience and delivery capabilities. Compared with the planning figure of 1,000 TOPS, these results will be more worthy of attention. (This article was first published on Titanium Media APP, text | Hanamaki)

Respect for the original work, if you are qualified, you should be appreciated.

According to the real-name requirement of the Cybersecurity Law, please bind your mobile phone number before posting a comment.

Express

11:48

Breaking the record, China achieves wireless communication verification with the highest code rate in the 6G frequency band

Malaysian Prime Minister Anwar arrived at the meeting

11:47

Xi Jinping meets with Thai Prime Minister Anutin

Qiyuan Robot signed a strategic cooperation with the Postal and Telecommunications Equipment Consumer Electronics Division

11:46

A-share lunch review: Shenzhen Component Index and ChiNext Index both fell by more than 2%, and the precious metals and liquor sectors performed sluggishly.

11:35

The relevant person in charge of Byte refutes rumors that Xie Xin and Zhang Nan have resigned: the information is inaccurate

11:32

The main domestic futures contracts rose or fell, with palladium and Shanghai silver falling nearly 9%.

11:28

Meta plans to launch Muse Charm, priced at $300

11:27

Spot gold fell below $3,520 per ounce, down nearly 3% on the day

11:23

Over 60% of China's sustainable development indicators are close to or achieving the 2030 goal

11:14

The concept of AI comic drama rebounded during the session, and Guomai Culture rose by more than 10%

11:13

Procter & Gamble plans to cut manufacturing and logistics employees by up to 30% by 2030

11:04

National Energy Group established a hydrogen energy technology company in Ordos with a registered capital of approximately 350 million

11:00

Domestic gold jewelry price falls below 1,100 yuan per gram

10:55

Korean exchange launches SIDECAR mechanism and suspends KOSPI programmed selling

10:54

Pork concept rebounded in intraday shock, Jin Xinnong went out of 3 boards in 6 days

10:51

22 private equity firms appeared in the first batch of 15 active ETFs, holding a total of 524 million shares

10:48

Alibaba Q2 of fiscal year 2027: Xianyu daily activity increased by 23% year-on-year, and order volume increased by 29%

10:47

Thai Prime Minister Anutin arrived at the meeting

10:46

Five departments issued the 2026 version of the "Guidance Catalog" for marine industry investment and financing

Scan to download App