Eight major automakers are venturing into chip - making. Elon Musk, Li Bin, and He Xiaopeng are fiercely competing in high - computing power, while Lei Jun has issued a solemn pledge.
Eight major automakers are collectively "making chips", and the industry competition has reached a white - hot stage.
According to CheDongXi on June 20th, foreign media reported recently that Tesla's next - generation Full Self - Driving (FSD) chip, AI 5, has entered the mass - production stage. Its computing power may reach 2500 TOPS, which is 4 - 5 times higher than that of AI 4 and even higher than that of NVIDIA's most powerful Thor - X. It is expected to be put into use as early as the end of this year.
▲ Tesla's self - driving chip AI 5 (Image source: Internet)
Meanwhile, the chip business of Chinese local new car - making enterprises has also made new progress.
According to LatePost Auto, NIO is planning to introduce strategic investors to its in - house chip R & D department and sell a small stake. Almost at the same time, a company named "Anhui Shenji Technology Co., Ltd." was registered and established in Hefei, Anhui Province.
NIO's self - developed Shenji NX9031, the world's first mass - produced 5 - nanometer intelligent driving chip, was installed in vehicles during the Shanghai Auto Show. The computing power of a single chip exceeds 1000 TOPS.
▲ NIO's Shenji NX9031
In addition to these two companies, XPeng officially announced on June 11th that three Turing chips have been installed in the G7, with a comprehensive computing power exceeding 2200 TOPS.
▲ XPeng's Turing AI chip
Lei Jun also revealed at the Xiaomi Investor Conference on June 3rd that Xiaomi's automotive chips are under development and will be launched soon.
......
In the past three weeks, the wave of automakers "making chips" has been surging, with frequent R & D achievements and continuous technological breakthroughs. A chip - making revolution is in full swing.
According to the incomplete statistics of CheDongXi, since the mass production of Tesla's first - generation self - driving chip HW 3.0, eight automakers have been reported to have self - developed intelligent driving chip plans in the past six years.
▲ Information sorting of automakers' self - developed intelligent driving chips
Eight major automakers, including BYD, Tesla, Xiaomi, Li Auto, NIO, XPeng, and Leapmotor, are firmly focused on the self - developed chip track. A chip "war" that will affect the future of the intelligent vehicle industry has entered a white - hot stage.
As the battle reaches 2025, the self - developed chips of automakers have also undergone a process from quantitative to qualitative change. The final outcome of the industry in the future is still unpredictable, but in these major changes, we can glimpse the historical evolution, the current pattern, and the future trend.
01. Eight automakers are reported to be self - developing chips with computing power reaching 1000 TOPS
In 2025, more and more automakers have appeared on the battlefield of intelligent driving domain control chips, and the self - developed chips of automakers have entered a new stage of mass - production verification.
In the past year, NVIDIA has still dominated the global intelligent driving chip market share.
According to Gasgoo data, in 2024, among the global intelligent driving chips, the installed capacity of NVIDIA's Drive Orin - X series reached 2.1 million units, accounting for 39.8% of the market share and ranking first in the market.
▲ The top three global intelligent driving domain control chip shipments in 2024
In 2025, some changes have quietly taken place among automakers.
With the upgrade of the intelligent assisted driving technology competition driven by the end - to - end large model, in order to achieve a smoother intelligent driving level, the demand of automakers for high - performance large - computing - power chips has become increasingly prominent.
More and more Chinese automakers are accelerating the mass - production and implementation of self - developed chips.
On June 11th, XPeng Motors officially launched its new mid - size SUV, the XPeng G7. The G7 Ultra version is equipped with three self - developed Turing AI chips by XPeng Motors, and its comprehensive computing power is nearly four times that of the G7 Max version, which is standard with two Orin - X chips.
▲ XPeng G7 is equipped with three Turing AI chips
On June 10th, the patent for "Chip startup method, system - on - chip, and vehicle" applied by Xiaomi Automobile Technology Co., Ltd. was authorized. One week ago, Lei Jun, the founder of Xiaomi, announced that Xiaomi's automotive chips would be launched soon.
There is also new news about Li Auto's chip - making. It is reported that Li Auto's chip - making plan, codenamed "Schumacher", was launched in 2021. In mid - May, the Mahe 100 chip was successfully taped out and will be officially mass - produced in 2026.
On April 23rd, NIO announced that the world's first mass - produced 5 - nanometer intelligent driving chip, Shenji NX9031, was officially installed in vehicles with the delivery of the NIO ET9. A single chip can achieve a computing power of over 1000 TOPS, which is at the same level as NVIDIA's Thor X.
▲ Shenji NX9031
In addition, some media reported that BYD launched a project related to self - developed intelligent driving chips in April 2024. The NPU computing power of this chip is about 8 TOPS, mainly targeting TI's TDA4VM, and it will cover BYD's mainstream models priced between 100,000 and 200,000 yuan.
The Lingxin 01 chip, the first domestic AI self - driving chip jointly developed by Leapmotor and Dahua Technology, had an installed capacity of over 100,000 units in 2024 and is applied to models such as the Leapmotor C11 and B10.
▲ Lingxin 01 chip
Overall, there are two technical routes for automakers' self - developed chips. One is represented by automakers such as BYD and Leapmotor, which focus on self - developed and self - supplied intelligent driving chips. After meeting their own needs, they expand outward, targeting the mainstream automobile market and covering more models.
The other is represented by automakers such as NIO and XPeng, which self - develop chips for the mid - to - high - end market to solve the performance problems caused by insufficient chip computing power and reduce costs.
02. Future directions: Competing in performance, cost, and talent
Based on this pattern, the future competition directions of automakers' self - developed chips are gradually becoming clear, mainly reflected in the following aspects.
1. From hundreds of TOPS to thousands: Automakers' chip computing power is surging forward
The first is to compete in performance, such as computing power and process.
NVIDIA, the domestic leader in intelligent driving chips, laid out large - computing - power chips earlier. The Thor chip it released in 2022 can reach a maximum computing power of 2000 TOPS.
▲ NVIDIA released the DRIVE Thor computing chip
Currently, NVIDIA's Thor has also been officially installed in vehicles. Since May this year, the first batch of new cars equipped with NVIDIA's Thor U have started to be delivered, including models such as the Li L8 and Lynk & Co 900. The computing power of a single chip of this chip is as high as 700 TOPS, more than twice that of the current mainstream NVIDIA Orin X (254 TOPS).
Higher computing power means being able to process massive amounts of sensor data more quickly, providing stronger computing support for the self - driving system. Therefore, large - computing - power chips will still be an important direction for automakers' self - developed chips.
2. One million chips to break even?! Making chips while reducing costs
In addition to competing in computing power, another major trend for automakers' self - developed chips is to compete in cost reduction.
Research by the China (Wuxi) Internet of Things Research Institute shows that the capital threshold for self - developed automotive - grade chips is very high. Taking the investment in basic chips as an example, the starting threshold is 1 billion yuan. If we add the overall R & D and operation, the investment scale of general chips generally exceeds 10 billion yuan.
▲ Chip - making requires tens of billions of funds (Image generated by AI)
In addition, according to data from the semiconductor technology research institution Semiengingeering, the development cost of a 7 - nm chip has reached 297 million US dollars (about 2.1 billion yuan), and that of a 5 - nm chip has soared directly to 540 million US dollars (about 3.9 billion yuan).
This puts higher requirements on the profitability of automakers. Installing chips in vehicles requires a huge amortization.
He Xiaopeng, the chairman of XPeng Motors, once asked someone in the industry how many chips need to be sold to make a profit, and the answer was one million.
▲ He Xiaopeng said that the computing power of one Turing chip is equivalent to three
More and more automakers will face the cost survival line of self - developed chips, and annual shipments of one million pieces are the profit threshold.
In order to further balance costs, automakers are showing two different self - developed chip paths. BYD spreads the cost through the order volume in the mainstream market, while automakers such as NIO and XPeng spread the cost through technological premium and only install self - developed chips in high - end models.
3. Talent war: The invisible smoke behind automakers' chip - making
Looking back at the process of automakers' self - developed chips from their start to their rise, it is not difficult to find that the deep - seated game in this chip competition has always revolved around "people". The competition in technology, production capacity, and market is essentially a competition in talent strategy.
Behind the autom