Author 丨 Zhang Zhidong
Responsible editor 丨 Cui Liwen
Edit 丨 Chic
There is not much time left for Chinese cores.
"Automakers don't really take action, they just shout."
At the 2022 China Electric Vehicle 100 Forum, Miao Wei, deputy director of the Economic Committee of the National Committee of the Chinese People's Political Consultative Conference and former minister of industry and information technology, spoke fiercely when talking about chip supply.
Not only bluntly saying that China's automotive industry lacks cores and souls, but also expresses dissatisfaction with the "inaction" of domestic car companies that throw card neck technology such as vehicle regulation chips and operating systems to first-level suppliers.
To be fair, like Tesla, there are not many car companies that develop FSD chips for self-driving. However, director Miao can be allowed to say such a tiger and wolf word as "light calling", which shows that there are indeed some deficiencies in the way the automobile circle deals with the shortage of semiconductor chip supply.
As far as the current situation is concerned, China's new energy automobile industry may be at the forefront of the times.
However, from the real "curve overtaking", there are still many key links that have not been opened, such as the intelligent functions of the car that are promoted by car companies with great fanfare.
On the whole, AI+ functions such as intelligent driving, intelligent cockpit, and intelligent networking all belong to the category of automotive intelligence.
But now the situation is that this series of intelligent functions is inseparable from the blessing of advanced AI chips, which in turn has led to the emergence of card neck problems.
Director Miao said to the point, in fact, what he wants to explain is a problem: the domestic automotive semiconductor supply chain, only pinning on Tier 1, will put all the weight on international giants such as Nvidia and Qualcomm, which is not wise.
And it is worth noting that NVIDIA's "kidnapping" of China's autonomous driving has reached a very profound level.
01
Closed or open, the battle of the kings of chips
Without hesitation, NVIDIA is a behemoth that always stands on the cusp of the trend of the times.
From supercomputers and parallel computing in data centers, down to personal PC graphics cards, digital currency "mining", etc., Nvidia can be seen in various scenarios. And on the other hand, NVIDIA's role as a pillar in the field of AI has grown to the point where no one can shake it.
In 2012, geoffrey Hinton, the "giant" of deep learning, and his student Alex won the competition in the ImageNet competition to accelerate the training of deep neural networks with the help of GPUs.
The influence of this competition, directly penetrating from academia to industry, not only promoted the arrival of the third wave of artificial intelligence, but also helped Nvidia's GPUs open up new incremental markets.
Further down, as one of the most clear directions in the field of AI, the autonomous driving market in a blue ocean has naturally become the pocket of NVIDIA.
However, the transformation of the pattern of the industry market will not be simply completed by dividing by two because of a "academic opportunity". As far as the rules of the market are concerned, if Nvidia wants to come in and be the "king", it needs to remove the roadblocks set by the old forces.
As a result, the battle for the king of autonomous driving chips between NVIDIA and Mobileye is on the verge of erupting.
Looking back on those days, for a long time, "packaging" has become the basic coordination method of the automotive industry. Tier 2, Tier 1, and car companies have collectively combined to form a pyramid-style industrial chain.
Car companies are responsible for product function definition and technology integration at the top; Tier 2 is responsible for the production of the most basic components, such as chip design, manufacturing and packaging; Tier 1 is responsible for packaging the mastered technology and "packaging" into a module that can be used by car companies with only a small amount of development, testing and verification.
Obviously, Mobileye is the epitome of this model. However, it is precisely for this reason that its unique and tried-and-true "closed package" has become Mobileye's hardest ceiling now.
Tesla is a good example, although the reasons for Tesla's "breakup" with Mobileye are complicated, but Mobileye's technology is too closed, making the aggressive Tesla very constrained. And this shortcoming, in the current market environment, is constantly being amplified.
In fact, it is precisely because it is "open" enough that NVIDIA has been able to become stronger step by step. And in addition to not restricting car companies from developing their own autonomous driving software algorithms and trying their best to create corresponding tool chains for them, NVIDIA also has the advantage of "computing power is justice". The dimensionality reduction strike from computing power is tantamount to a "yang conspiracy".
It is understood that Nvidia's autopilot chip, from Parker in 2018 to today's Orin, has achieved 254 times the growth of computing power; the latest Atlan, the computing power has reached 1000TOPS, "rolling" the entire industry is powerless.
Although people always say that chips should not be "only computing power"; but on the other hand, if there is no computing power as a stepping stone, it may not even be qualified to talk.
02
Latecomers, move forward
As a result, with the "combination of soft and hard", as well as the dual advantages of computing power, NVIDIA has gained a large wave of supporters.
Whether it is Robotaxi companies such as GM Cruise, Amazon Zoox, AutoX, and Xiaoma Zhixing, or globally renowned car companies such as Mercedes-Benz, Jaguar Land Rover and Volvo, they have established cooperation with Nvidia.
And it is worth mentioning that those hot Chinese "new forces" have also chosen NVIDIA.
Not long ago, at the 2022 NVIDIA GTC conference, the "Yellow Sect Lord" officially announced the partnership with BYD. According to the plan, from the first half of 2023, BYD will be equipped with NVIDIA's Orin chip and supporting software and hardware systems on some models.
Coincidentally, the new models of weilai, Xiaopeng and ideal three strong new forces this year are equipped with NVIDIA's automatic driving chips. Players like Gaohe, Zhiji L7, Jidu, etc. who gradually began to emerge were invariably "captured" by NVIDIA.
Could it be that really foreign monks chant the sutras? I'm afraid not.
On the one hand, it is because NVIDIA's products have been tested by the market, the technology is mature, and the quality is guaranteed.
On the other hand, car companies choose NVIDIA's chips, which can greatly shorten the run-in cycle, straightforward and effective.
As far as the current situation is concerned, most of the new forces of new energy vehicles have already put the bet on NVIDIA.
And this trend of China's self-driving chasing Nvidia chips, there are many other reasons behind it.
On the one hand, the current mainstream autonomous driving technology is deep learning.
With the help of big data feeding, the AI neural network model will be further trained and constructed, until the whole vehicle is on the road, the model is gradually improved, in order to achieve accurate and silky control of the vehicle under the blessing of the large computing chip.
On the other hand, domestic self-driving chip suppliers are still relatively "young" and the technology is not perfect. In this way, the "fast-paced" new energy vehicle war will not leave too much time for car companies.
In other words, car companies choose NVIDIA, although it may not be able to help themselves outperform other friends in the automatic driving function; but this method of "stacking materials" with computing power means that they will not fall too far behind in the competition.
The good news is that domestic smart chip manufacturers are not without the slightest ability to fight back.
For example, the horizon favored by the ideal car, the 2021 new ideal ONE has been equipped with the Journey 3 chip; intelligent driving, central gateway, intelligent cockpit all-round layout of the core Chi Technology, the product has covered more than 70% of the domestic car manufacturers.
In addition, chip manufacturers such as Black Sesame Intelligence, Cambrian Xingge, and Huawei, these latecomers, will become the potential force to replace NVIDIA's self-driving chips.
Although that day is relatively far away, it can be seen that the domestic new energy vehicle market is now creating opportunities for the supply chain restructuring of the core and the localization of substitution.
03
Chinese core, to the virtual to the real
The cluster of capital has opened a door for the development of core manufacturers. But at the same time, it must be made clear that behind this door is not necessarily a flower blooming and rising step by step, but more likely to be dark and lonely.
In fact, in the face of a series of uncertainties such as the turmoil of the international environment, the ban on technology, the recurrence of the epidemic, and the rise of raw materials, Chinese cores urgently need to "go from virtual to real", rather than self-flaunting, mouthing cannons, and making headlines.
What is virtual?
In recent years, the state attaches great importance to the development of the semiconductor industry and has introduced a series of policy measures, and under the background of such policies, chip projects have been carried out in various places.
The original intention was to "concentrate on doing big things", but did not want to "virtual fire is too strong", there have been a large number of cases like Wuhan Hongxin," "high-profile factory construction, follow-up weakness".
Second, the rush of capital creates a thriving scene, but whether the tide will suddenly fade, no one can guarantee.
In fact, when the relevant products cannot be tested by the market, but only shouting "hanging" NVIDIA, piling up some awkward professional parameters to block the mouth of the public, is not the so-called face project?
It is not difficult to find that the long-term and low-return characteristics of the semiconductor industry have made many capitals who want to see "fast money" cry bitterly; and the once-hot financing tide is also slowly "fevering".
Taking the core as an example, the entire design and manufacturing process not only requires vehicle-level certification, but also the accumulation of semiconductor technology and the mutual support of the entire core industry chain. Even if chip manufacturers design and produce products, Tier 1 is not adopted, the supply chain is cut off, and it is difficult to be configured by the whole vehicle for market inspection and technology iteration.
There is no doubt that the current automobile industry is a highly divided labor, layer by layer of the industry.
Car companies' mastery of technology is often concentrated in the parts they think are the most critical, such as engines, gearboxes, and chassis. However, with the strong rise of new energy vehicles, car companies have to start paying more attention to "new components" such as algorithms, smart chips, and power batteries.
In fact, in order not to lose their soul and heart, many car companies have begun to improve their right to speak and develop freedom through investment and internal incubation. BYD Semiconductor, Great Wall Honeycomb Energy, Geely Core Qing Technology, etc., can be used as the most direct proof.
However, the question now is: are car companies really "spending money to make money", the name of the picture? Or do you really want to fundamentally solve the "kidnapping" problem in order to cut costs and make profits?
There is not much time left for Chinese cores.
Zhang Zhidong
What a great writer,
It's just writing
One-sidedness of their own.
Read
recommend
In April, will Tesla's sales in China be "waist chopped"?
Gao He, more and more "magic"
BYD's sales volume in March is second, surpassing SAIC Volkswagen as the norm?
- END -
Micro signal | C-dimensional