In the fierce battle of smart cars, what can support car companies to go further?

The competition for smart cars has begun. Only by seizing the opportunity of intelligent transformation of cars and embracing "software-defined cars" can car companies break through.

The competition between new car-making forces and traditional car companies is fierce.

On the second day of November, the three major new car-making forces all released their new car sales in October: NIO 5055 units, Xpeng 3040 units, and Li Auto 3692 units.

Although there is still some gap compared with the sales of tens of thousands of traditional car companies, the many orders in their hands cannot be underestimated.

On the other hand, established automakers are also accumulating strength to explode. On the evening of November 3, Volkswagen's ID.4 model made its debut in China, challenging the smart car market with a range of 555 kilometers, L2 level autonomous driving and a price of no more than 250,000 yuan.

It is still unknown how big a wave the counterattack of traditional cars can create, but judging from the performance of car companies, it is certain that the trend of smart cars has arrived.

Those players who stand at the forefront of the trend have already tasted the huge dividends brought by the trend.

Tesla's market value has exceeded $400 billion, and it may reach the trillion level in the future;

After NIO's share price dropped to $1 in 2019, it has risen 20 times this year, with a market value of $44.8 billion, surpassing BMW. In October, JPMorgan Chase also raised NIO's target price to $40.

On the surface, these new car companies have repeatedly broken through their market value highs because of the optimism of the market and capital; but in fact, their understanding of smart cars and product polishing are the singularity that allows them to tap into unlimited imagination space.

Compared with the mechanical technology of traditional fuel vehicles, more tangible intelligent performance of automobiles, such as assisted driving and smart cockpit, is becoming a new point of interest for consumers. A competition for smart cars has also begun.

Now that the whistle has sounded and the smoke of gunpowder has begun to rise, what will support the car companies to go further in this fierce battle?

1. The second half is the home court of smart cars

At the Chongqing Smart Expo last year, BYD Chairman Wang Chuanfu said that in the transformation of the automotive industry, electrification is only the first half, and intelligence is the second half.

But no matter what is said, Tesla is the one standing at the forefront.

As many automakers are making inroads into electric vehicles, Tesla's focus has shifted to autonomous driving.

Some time ago, Musk said on Twitter that FSD (Full Self Driving) will have a major update. This is not just a simple functional update, but a complete rebuild of the system infrastructure.

Musk said that vehicles with FSD function updates will be able to achieve zero-intervention driving and have the ability to quickly OTA in 30 hours. Currently, the Beta version of the rewritten FSD has been released to some users. With the update of functions, the price of the FSD kit is also rising.

On the one hand, it reduces the hardware cost of the entire vehicle, and on the other hand, it increases the value of the car software. Relying on this self-consistent cost logic, Tesla has been able to continuously expand its audience.

But this is just the surface that Tesla has left a number of car companies behind. From a deeper perspective, Tesla's understanding of smart cars is the key to supporting the continuous evolution of its vehicles.

Two very important points are: software-defined cars and centralized smart car electronic and electrical architecture. 

It is foreseeable that software will occupy more and more value in future cars. This is why Tesla has to rewrite the underlying code of FSD and reconstruct the deep neural network. Because more perfect software will have a higher premium space.

However, the premise of software-defined cars requires sufficiently excellent hardware support.

Therefore, in addition to continuous improvement at the software level, Tesla is also focusing on the development of vehicle hardware. According to media reports, Tesla is jointly developing a new HW4.0 autonomous driving chip with Broadcom, and its performance will be 3 times more powerful than FSD.

It is understood that this chip can support Tesla's ADAS system, power transmission and vehicle entertainment functions, and help realize the centralized automotive electronic and electrical architecture. The intelligent automotive electronic and electrical architecture has gradually evolved from the traditional distributed ECU architecture to the domain controller architecture, and finally realized the path of the central computing architecture, which has also been generally recognized by the industry.

The ideal situation is that cars of the future will be like computers, with a chip brain that can issue commands to every part of the car.

Therefore, "software-defined cars" and the core hardware of the central computing electronic and electrical architecture, the "central computing platform", have also become the focus of major automobile companies.

On the one hand, many car companies and parts manufacturers are awakening to the software awareness, and are either building their own software teams or expanding their scale, determined to embrace the trend of software-defined cars.

On the other hand, many car companies are also strengthening their hardware awareness and choosing to develop their own underlying chips.

Not long ago, it was revealed that NIO, one of the new car-making forces, is planning to independently develop autonomous driving computing chips.

In October, Ecarx and ARM China jointly invested in the establishment of CoreEngine Technology, planning to carry out research and development and mass production in the fields of automotive chips such as smart cockpits, autonomous driving, and microcontrollers.

Leapmotor is moving even faster. On October 27, Leapmotor officially released   the Lingxin 01 chip , the first domestically produced automotive-grade AI smart driving chip with completely independent intellectual property rights.

2. The rise of domestic chip newcomers

However, self-developed automotive chips is a long process that requires huge investment of time and energy, as well as deep accumulation of talent and experience.

In the urgent tug-of-war, an open chip partnership is, by comparison, a powerful tool for car manufacturers to quickly transform.

For example, the Xpeng Motors P7 model uses Nvidia's Xavier system chip; Li Auto also stated that it will use Nvidia's chips in its next generation of models, but like NIO, it currently still uses Mobileye's assisted driving chip EyeQ4.

In addition to Nvidia and Mobileye, Horizon Robotics, a new unicorn in-vehicle  AI  chip company in China, is also rising. To some extent, the emergence of Horizon Robotics has filled the gap in the domestic automotive  AI  chip industry.

However, automotive AI chips are a field where software and hardware are strongly coupled, requiring players to have a deep understanding of automotive-grade products, as well as to understand both chips and algorithms, which is quite difficult.

Judging from the product implementation, Horizon has already achieved this.

On June 21 this year, Changan Automobile UNI-T model was officially launched on the market. The new car is equipped with Horizon's Journey 2nd generation chip, which is also the first domestic automotive-grade AI chip launched by Horizon.

Based on this chip, Changan Automobile and Horizon jointly developed the "Intelligent Cockpit NPU Computing Platform". In terms of intelligence, the UNI-T intelligent cockpit can deeply integrate multiple perception data such as vision and voice, and provide functions such as voice interaction, human-computer interaction, and fatigue monitoring.

The actual launch of products also proves that Horizon is a leader among domestic automotive chip suppliers with mass production capabilities.

But the smart cockpit is just an appetizer for Horizon's in-vehicle AI chip applications.

Its Journey 2 generation has strong versatility and can be used not only in the field of smart cockpits, but also in ADAS in the field of smart driving.

In fact, in the field of ADAS chips, Journey 2 can already compete with Mobileye's EyeQ4. Horizon Robotics said that the perception computing performance demonstrated by Journey 2 has surpassed Mobileye EyeQ4 in multiple indicators.

At the Beijing Auto Show on September 26 this year, Horizon Robotics also launched the Journey 3 chip. This chip has an AI computing power of 5 TOPS and a typical power consumption of only 2.5W, which can achieve higher-level assisted driving, driver monitoring, automatic parking assistance and other functions.

It is understood that if we look at the average processing speed of AI chips when achieving the highest accuracy in image classification (ImageNet), the AI ​​performance score of Journey 3 is slightly better than that of mainstream chips such as NVIDIA and TI.

Specifically, the power consumption of Journey 3 is 2.5W, and the MAPS performance is 422 FPS; although NVIDIA Xavier's MAPS performance score is as high as 840FPS, its power consumption is not low at 30W; although TI's TDA4 has a power consumption of 12.5W, its MAPS performance is only 181FPS.

In addition, Horizon also stated that the Journey 5 chip with higher computing power will be officially released at the beginning of next year. By then, its performance will surpass Tesla's FSD chip, and it will also be able to meet the mass production needs of car manufacturers for high-level autonomous driving.

At present, Horizon has formed a complete intelligent driving business layout.

Focusing on the two major directions of comprehensive environmental perception and in-vehicle human-computer interaction, Horizon Robotics has created an open empowerment solution of "chip + algorithm + tool chain": it can flexibly provide automotive-grade AI chips, "Heavenly Creations" AI development platform, algorithm model samples, and professional services to help industry partners quickly implement products.