news

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

Text|Things about the Internet

Recently, Li Bin became the focus of controversy for showing off the world's first automotive-grade 5nm smart driving chip, with the following three voices being the strongest:

Is this a slap in the face to the domestic mobile phone manufacturers who are still competing to be the first to launch Qualcomm and MediaTek SoCs?

With an annual loss of over 20 billion yuan, the company still dares to spend money crazily on smart driving chips that have not been sanctioned. Is there a strategic risk?

The second-generation NIOPhone was unveiled as scheduled. Why does a new force focusing on car manufacturing have such an obsession with smartphones?

Did the money in NIO’s account really come from the wind?

01. Crush Nvidia! Is NIO really far ahead this time?

Whether it is the smartphone industry or the new energy vehicle industry, whether a company has the ability to develop its own core chips has become the key to the industry's and the outside world's evaluation of whether a company is hardcore. Huawei, Apple, Qualcomm, and Nvidia have already proved this to the world. Xiaomi, OPPO, vivo, Yaoxingmei and other domestic mobile phone manufacturers have not been able to get rid of the labels of "comprador factory" and "assembly factory" given by netizens because they have mainly purchased chips from Qualcomm and MediaTek so far. How come NIO has managed to tape out its self-developed 5nm chip?

There is an obvious difference in design difficulty or technical threshold between mobile phone SoC and automotive-grade SoC. Let’s first exclude the 5G baseband part with patent barriers as high as Mount Everest and only talk about the competition of AP. For example, Huawei’s HiSilicon Kirin 9000, which was launched four years ago, has an incredible 15.3 billion transistors in an area of ​​100mm2. Although the number of transistors of NIO NX9031 reaches 50 billion per day, the area is as high as 826mm2. The transistor density (MTr/mm2) of Kirin 9000 is far ahead of NIO NX9031.

Considering the almost stringent requirements on technical indicators such as power consumption, performance and energy efficiency, the design difficulty of advanced process SoC on mobile phones, where every inch of space is valuable, is much higher than that of automotive-grade chips that are not sensitive to advanced processes. Currently, there are still a number of Tier 1 companies that provide automotive-grade chips with mature processes of 7nm or even 28nm.

Of course, due to the special nature of automobiles, automotive-grade chips have extremely stringent requirements in terms of safety and stability. They are far ahead of consumer-grade chips (including chips for smart terminals such as mobile phones and PCs) in terms of chip life, error rate, continuous supply time, and testing standards. Therefore, it is not objective to say that there is no threshold for automotive-grade chips.

So at this point, we can first draw the first conclusion that NIO’s self-developed 5nm automotive-grade chip did not slap Xiaomi, OV and others in the face, but it may embarrass a number of new energy vehicle companies such as Ideal, Xiaopeng, BYD, Xiaomi, etc. that have not yet launched their own intelligent driving SoC chips!

Let’s continue to look at what Li Bin said: Shenji NX9031 is the world’s first 5nm automotive-grade high-performance intelligent driving chip. Does the industry have any opinions?

From: Li Bin's video

According to public information, in February 2023, Samsung Electronics and AI semiconductor company Ambarella jointly announced that Samsung will provide Ambarella with 5nm process for foundry use of the newly launched CV3-AD685 SoC chip. CV3-AD685 is the first mass-produced model of Ambarella's CV3-AD automotive AI domain controller. Many Tier-1 companies have announced that they will adopt Ambarella's series of SoC solutions.

In September 2023, ADTechnology, a South Korean semiconductor design solution provider, announced that it will work with Samsung Electronics' foundry business and BOS Semiconductor to develop a high-performance SoC based on a 5nm automotive-grade process, and aims to apply the NPU accelerator SoC to various perception and decision-making task scenarios for intelligent driving.

In June 2024, semiconductor giant NXP released the S32 Core Ride central computing solution (taking into account both smart cockpit and smart driving), which has a built-in 5nm automotive-grade SoC chip S32N55.

It is not difficult to find that before Weilai's Shenji, many giants have released 5nm automotive SoC chips, and they are more or less closely related to smart cockpits and smart driving. Of course, Li Bin is also very "cunning" because he only advertised that he had developed his own 5nm smart driving, and the chips that came out before Shenji generally came from semiconductor suppliers, not OEMs, and there was not much motivation to compete for the "first 5nm" because these SoC solutions naturally have a lot of car customers willing to pay for them, and it is up to the OEMs to brag about them after buying them back...

Therefore, for Li Bin, the greatest value of this chip is not (at present) its ability to enhance the performance of NIO cars, but rather it provides an explanation to shareholders and investors. At the same time, NIO also has a "world's first" marketing gimmick and can significantly reduce costs in the future, which means killing two birds with one stone.

Can it significantly improve its differentiated competitiveness? We will have to wait until next year when it is installed in vehicles for actual testing, and when it is compared with Ideal and Xiaopeng's 5nm automotive chips that are still waiting to be taped out after they are officially mass-produced and installed in factories, to determine which one is better.

Ranking of installed capacity of domain control chips in China's intelligent driving market in 2023

It is worth mentioning that in the field of intelligent driving SoC domain control chips in the domestic market, Tesla FSD occupies the largest market share, and Nvidia Orin-X has also made a lot of money thanks to the favor of a large number of new forces (including NIO). Horizon, Mobileye, Black Sesame Intelligence, Huawei and others have all got a piece of the pie.

From the perspective of the global market, Qualcomm, the "king of cockpits", has growing ambitions. Its second-generation Ride SoC SA8560P and SA8620P have already won orders from famous cars such as BBA and Porsche, and have successfully entered the Tier 1 supply chains of Bosch, Valeo and other companies. The high-end integrated cockpit solution based on Qualcomm SA8775P chip jointly released by Qualcomm and DJI is also expected to be launched within the year.

Li Bin's courage to take the initiative and invest a lot of R&D resources in the research and development of smart driving chips is undoubtedly extremely courageous and wise. Only when this kind of underlying core technology is in our own hands can we stop looking at the "face" of Nvidia, Qualcomm, and Intel, and can we say that we can ensure safety through autonomy and control. Of course, this also involves huge financial risks.

According to a report by Zoss Automotive Research, if the annual shipment volume of an automotive SoC chip does not reach more than 1 million, it will be difficult to support the continuous investment in chip research and development. If car companies do not have sufficient shipment volume, it will be difficult to recover costs, and reducing the chip cost of the entire vehicle will be a luxury.

NIO's sales in 2023 will exceed 160,000 vehicles, which is still nearly 35% away from the annual target of 245,000 vehicles. Does it seem that annual sales of 1 million vehicles have become an unrealistic fantasy? According to NIO's financial report, in 2023, NIO's net loss will reach 20.72 billion yuan.

Under such circumstances, Li Bin is still betting on self-developed chips and still spending money to make mobile phones. Is he "crazy"?

02. How long can Li Bin’s money last?

On July 27, at the NIO IN 2024 Innovation Day press conference, in addition to the Dimensity NX0931, Li Bin also released the second generation of NIO Phone. Of course, NIO officially called it the new AI NIO Phone. The phone also uses the performance combination of Qualcomm Snapdragon 8Gen3+LPDDR5X+UFS4.0. The price of 6499 yuan to 7499 yuan is also in line with the current Android high-end flagships.

Even the screen follows the trend, adopting the widely acclaimed equal-depth four-micro-curved screen. However, unlike Huawei, Xiaomi, and OPPO, NIO's screen is customized from Samsung with QHD+ specifications at a high price. A supply chain source said that with the negligible sales volume of NIO mobile phones, the customization cost alone for customizing such a top-level OLED screen may be as high as nearly 10 million.

On the domestic supply chain side, there are obviously more cost-effective options such as BOE, TCL Huaxing, and Visionox. Why didn't NIO choose it? It can only be understood that it is rich and willful, or Li Bin wants to bring NIO owners a higher quality viewing experience. After all, NIO's cars are high-end products, and the components of mobile phones must also be high-end.

In addition, in terms of camera modules, NIO also provides flagship specifications, 48MP main camera + 48MP ultra-wide angle + 50MP periscope telephoto, and CIS is also flagship specifications. However, NIO's achievements in mobile phone imaging algorithms are still inferior to those of the Chinese OV camp.

In addition, it has dual-mode 13-band 5G, 2-meter-deep IP68 waterproof, USB 3.1, X-axis linear motor, UWB ultra-wideband communication (smart car key)... and it also features differentiated selling points of "zero system advertising" and "zero commercial pre-installation". For NIO owners or users who like a pure system experience, it is indeed a sincere work. Does NIO Phone 2 have a chance to replace the iPhone and become Li Bin's main phone?

According to NIO's financial report and other public information, NIO's R&D investment in 2023 will reach 13.43 billion yuan, a year-on-year increase of 23.9%, much higher than Xpeng's 5.28 billion yuan and higher than Ideal's 10.59 billion yuan. This naturally includes the R&D of intelligent driving chips and mobile phone projects. According to Li Bin's explanation, he believes that the money spent on mobile phones is within reasonable R&D expenses.

"We don't expect to sell tens of millions of units a year, and of course we won't invest heavily in marketing. Our investment in mobile phone research and development is fixed, and we only release one phone a year. The company can bear the related costs, and as time goes by, the costs will gradually narrow, and the losses will not increase."

What this means is that the concerns of the media and self-media friends may be unnecessary.

In addition to continuing to spend money on smart driving chips and mobile phone product lines, Li Bin has also spared no effort to invest heavily in self-developed global operating systems. It is precisely because of the existence of these big money-burners that NIO's R&D expenditure is as high as 3 billion each quarter. In other words, if NIO's annual sales cannot grow rapidly, it will continue to bear the pressure of profitability.

From 2021 to 2023, NIO's gross profit margins were 18.9%, 10.4% and 5.5% respectively, showing a gradual downward trend. However, from Q1 to Q4 of 2023, NIO's gross profit margin of complete vehicles increased significantly, from 5.1% in the first quarter to 11.9% in the fourth quarter. This is considered a positive sign.

Judging from the new forces sales list released last week by market institutions, Ideal still sits firmly in the top spot with a performance of 12,800 units, and Wenjie's sales also successfully exceeded 10,000 units. NIO sold 5,400 units, showing obvious progress. The fourth and fifth places were Leapmotor and Xiaomi respectively, and Xpeng ranked eighth.

Of course, the originator of this kind of weekly list, which is widely circulated in the automotive industry, is Ideal. Li Bin and Ma Lin have repeatedly expressed their "dissatisfaction" with the list. Just now, even Ma Lin of NIO directly challenged Li Xiang: "The weekly report of Volkswagen has a bit of low-level involution. Brother Xiang, please stop it. The current mission of Chinese auto companies is to achieve technological breakthroughs through research and development, isn't it? Let's power up together!"

As of the end of 2023, NIO's cash reserves are 57.3 billion yuan. Combined with the support of the Middle East consortium, NIO still has enough money to burn for a while, but when can it achieve a commercial closed loop for its self-developed chips? When can it feed back into research and development? There are still many uncertainties...

Li Bin dared to take the step of self-developed chips and self-developed OS, which is more radical than "Beijing Cash King" Lei Jun. This spirit of revolutionary optimism is quite cool.

Can the serious NIO win the future?

Stay tuned!

References:

DJI Car and Qualcomm jointly release intelligent driving solution based on Qualcomm intelligent driving chip SA8650P

Samsung, ADTechnology and BOS Semiconductors collaborate on 5-nanometer automotive semiconductors - Gasgoo

Samsung Electronics' 5nm process is used by Ambarella in new automotive AI domain controller chips - China Business News

The self-developed chips of leading car companies have exploded! BYD is the most mysterious, and NIO exceeds expectations - Electric Car News

Ideal Chip's Latest Progress: 5nm SoC and SiC Chip-52RD

NIO Ma Lin, Sina Technology Weibo, etc.