Jiwei Network News, a few days ago, Intel announced that it plans to promote Mobileye's listing in the United States in mid-2022 through an initial public offering (IPO), what are the deep reasons behind it? In this regard, Robert Castellano, senior analyst of Wall Street stock market analysis and destruction, Madekingalpha, made an analysis.
Mobileye currently offers Advanced Driver Assistance Systems (ADAS) and corresponding products to more than 30 automakers that have chosen EyeQ as their driver assistance technology, including Audi, BMW, Fiat Chrysler, General Motors, Honda, Hyundai, Kia, Nissan and Volkswagen.
According to Intel, Mobileye's annual revenue was $967 million in 2020, compared to $879 million in 2019, up 10% year-over-year. The company expects to achieve record annual revenue this year, which is expected to be more than 40% higher than in 2020. Mobileye's EyeQ chip shipments were 19.3 million units in 2020, compared to 2.7 million in 2014, representing a compound annual growth rate of 39%.
While Mobileye attracted numerous customers, Qualcomm announced a partnership with BMW at the 2021 Investor Conference. BMW's new car will use Qualcomm's self-driving platform Snapdragon Ride (including chips) from 2025.
In addition to the performance gap discussed later in this article, Intel's 3940 chip belongs to the E3900 series and uses a 14nm process when it was released in 2016. At first, it was mainly used for consumer-grade chips, and then it was separately classified as a vehicle-grade chip. The automotive grade chip is the A3900 series, which costs about $40. Qualcomm's S8155 is a mass-produced 7nm automotive grade chip that costs about $100.
Most self-driving cars use systems with cameras, radar, laser sensors, and other technological systems to assess road conditions and adjust driving behavior. These vehicles may have adaptive cruise control, lane adjustment and automatic braking, steering, and acceleration systems.
The level of autonomous driving is inextricably linked to the computing power of the chip. The autonomous driving industry generally believes that the chip computing power required for level 2 of autonomous driving (Trillion Operations Per Second, "TOPS") is less than 10 TOPS, level 3 requires 30 to 60 TOPS, level 4 requires more than 100 TOPS, and level 5 requires more than 1000 TOPS. It is precisely because of this that chip computing power has become the core force of competition between various chips.
If a car company uses multiple chips to set up autonomous driving domain controllers, they can reach up to 1024 TOPS, which can support level 4 autonomous driving.
While the computing power of a single chip TOPS is a key indicator, it is not the only indicator. Autonomous driving is a complex system that requires vehicle-to-cloud collaboration. Therefore, in the competition of autonomous driving chips, in addition to the core, there is the synergy of software and hardware, as well as the platform and tool chain.
At present, there are many competitors in the automotive AI chip market that have adversely affected Intel and Mobileye, including Qualcomm and Nvidia, Tesla and Chinese companies Huawei, Horizon, Black Sesame and Xinchi Technology.
Seekingalpha also cited details of Mobileye's competitors' products and plans.
Qualcomm
The Intel A3900 series is based on the X86 architecture design, released in 2016, using a 14nm process. At first, it was mainly used for consumer-grade chips, and then it was separately classified as a vehicle-grade chip. The automotive grade chip is the A3900 series, which costs about $40. EyeQ4's upgraded version, EyeQ5, was released in 2020. The EyeQ5 was only installed on Geely's Geely 001 model for the first time in the fourth quarter of this year. The EyeQ5 uses a 7nm FinFET process with a chip computing power of 24 TOPS.
The SA8155P is an integrated, next-generation automotive cockpit platform. It is a 7nm System-on-Chip with a custom hardware module design, including an octa-core CPU subsystem powered by a fourth-generation Qualcomm Kryo CPU based on the ARMv8 architecture. The system-level chip uses an efficient machine learning architecture, consumes less than 7 watts of power, and has up to 10 TOPS of chip computing power.
Qualcomm's fourth-generation Snapdragon digital cockpit platform SA8295P uses 5nm process technology and chip computing power of up to 30 TOPS.
The SA8155P is an integrated, next-generation automotive cockpit platform. It is a 7nm System-on-Chip with a custom hardware module design, including an octa-core CPU subsystem powered by a fourth-generation Qualcomm Kryo CPU based on the ARMv8 architecture. The system-level chip uses an efficient machine learning architecture, consumes less than 7 watts of power, and has up to 10 TOPS of chip computing power. Qualcomm's fourth-generation Snapdragon digital cockpit platform SA8295P uses 5nm process technology and chip computing power of up to 30 TOPS.
Nvidia
Xavier processors feature programmable CPUs, GPUs, and deep learning accelerators with up to 30 TOPS of on-chip computing power. NVIDIA's next-generation autopilot chip, the Orin chip, will also begin mass production in 2022. The Orin chip has a single computing power of 254 TOPS, which is 10 times more than the computing power of the EyeQ5. Nvidia Atlan has up to 1,000 TOPS of single system-on-chip computing power (samples expected to be made available to developers in 2023).
Huawei
Huawei is positioning it as a Tier 1 supplier and building a "5G automotive ecosystem" targeting the high-end autonomous driving market, which cannot be ignored.
Huawei's Ascend 310 has a chip computing power of up to 16 TOPS and consumes only 8 watts of power. The Ascend 610 has up to 160 TOPS of on-chip computing power for Level 3 and Level 4 autonomous driving. The 610 processor features a 64-bit quad-core CPU architecture and an advanced 4G LTE system that balances power consumption and performance in high-end smartphones.
horizon
Horizon Robotics was founded in 2015 to make AI chips for autonomous vehicles and machines. It also customizes software for these chips, which can be installed on devices such as cars and smart speakers. Horizon Robotics announced its third automotive-grade AI chip, The Nineney 5, and TogetherOS, a real-time in-car operating system. Journey 5's single-chip AI computing power can reach 128 TOPS. Journey 5's debut target partners are leading equipment OEMs, including SAIC, Great Wall Motors, JAC Motors, Ideal Motors, Changan Automobile and BYD.
Black sesame seeds
Black Sesame Smart Technologies' new self-driving chip, the A1000 Pro, is popular with record computing performance among such chips produced by local companies. According to Black Sesame Smart Technology, the A1000 Pro is based on the company's A1000 chip, which is optimized for up to 106 TOPS in standard computing power and up to 196 TOPS in accelerated mode. Black Sesame Intelligent Technology cooperates with COMPANIES SUCH ASI, SAIC, BYD, Dongfeng Motor, FAW Group and Bosch to focus on L2/L3 advanced driver assistance systems and automatic driving sensing system solutions.
Chip Technology
ChipCom is a China-based semiconductor company focused on next-generation high-performance automotive chip solutions. The V9T chip, introduced in 2021, has a driving capacity of up to 1 TOPS. In 2022, Chipcom will launch the V9P/U autonomous driving chip, which has a computing power of 10 to 200 TOPS. The product has a higher degree of integration of computing power and can support Level 3 autonomous driving. In 2023, Chipcom will launch a V9S autonomous driving chip with higher computing power. The chip was developed for the central computing platform architecture. With computing power of up to 500-1000 TOPS, it can support self-driving taxis for level 4 or level 5 autonomous driving. At present, the V9 (automatic driving) of Xinchi Technology adopts a 16nm process. This level of process has lagged behind in consumer electronics, but in the automotive industry, the 16nm process is still the mainstream trend.
tesla
The Fully Automated Driving Chip (FSD chip, previously the Autopilot Hardware 3.0) is a Tesla-designed autopilot chip that was launched in early 2019 for use in the company's internal cars. Tesla claims that the chip is aimed at Level 4 and Level 5 autonomous driving. It is manufactured using Samsung's 14nm process technology. The FSD chip consists of three quad-core Cortex-A72 clusters with a total of 12 CPUs operating at 2.2 GHz, one Mali G71 MP12 GPU operating at 1 GHz, two neural processing units operating at 2 GHz, and various other hardware accelerators.
Tesla's self-driving cars will be powered by a computer based on two of its new AI chips, each equipped with a CPU, GPU and deep learning accelerator. The computer's computing power is up to 144 TOPS, enabling it to collect data from a range of surround camera, radar, ultrasonic, and power deep neural network algorithms.
The following table shows the chip versions (current and planned versions) of these companies, as well as the TOPS of the chips. According to our report titled "Hot Integrated Circuits: Artificial Intelligence ("AI") Market Analysis, 5G, CMOS Image Sensors and Memory Chips", NVIDIA ORIN chips have up to 200 TOPS of computing power, more than 9 times the computing power of EyeQ5. Horizon Robotics' Tourney 5 and Huawei's Ascend 610 both have more computing power than EyeQ5.
ADAS is defined by one of the following six feature levels (L0-L5):
In 2020, the penetration rate of Level 2 autonomous driving in China will reach 15%. This means that nearly 4 million new cars are equipped with Level 2 autonomous driving systems. It is expected that by 2030, self-driving cars will account for more than 40% of the total mileage, and the penetration rate of fully autonomous new cars will reach 10%.
Wall Street expects that from 2019 to 2030, the penetration rate of Chinese self-owned brand passenger cars from ADAS to Level 3 will increase from 20% to 75%. This means that the chip market size will increase from about $50 million to more than $1.5 billion. Chinese cars are less affected by the "shortage" of semiconductors, and likewise, Chinese independent brands will use domestic in-vehicle chips.
Strong demand in the Chinese market is crucial for Mobileye and its competitors, as Huawei and Chinese startups will weaken Mobileye's market leadership. In summary, China Horizon Robotics works with leading equipment OEMs, including SAIC Motor, Great Wall Motors, Jacques Automobile, Ideal Automobile, Changan Automobile and BYD – all of which are based in China.
SAIC motor is both an investor in Horizon Robotics and black sesame technology. In August 2020, SAIC motor said it had invested in more than a dozen leading chip companies in China, including Horizon Robotics and Black Sesame Technology.
Globally, by 2025, only 15% of the world's vehicles will not have an ADAS autonomous driving system, compared with 42% in 2020, while 40% of vehicles will be equipped with a Level 1 autonomous driving system. Importantly, by 2025, 36% of vehicles will be equipped with L2 autonomous driving systems, an increase compared to 10% in 2020. Only 9% of vehicles have autonomous driving systems with Level 3 or higher features. Intel announced on December 7, 2021, that the company is planning to publicly launch Mobileye's self-driving cars. Frankly, this is not a "notorious" moment for Intel.
Autonomous driving technology is still in the early stages of development and is far from reaching the mature stage. This has made autopilot chips highly competitive, and Mobileye has been building partnerships with automotive companies for years. Mobileye's system-level chip shipments were 19.3 million, while Horizon's chip shipments were nearly 500,000.
The real difference is in computing power, and while Mobileye's computing power is only 24 TOPS, Horizon Robotics' Journey 5 chip has 128 TOPS. Level 2 autopilot requires less than 10 TOPS of chip computing power, Level 3 autopilot requires 30 to 60 TOPS, and Level 4 requires more than 100 TOPS. This means that compared to the Horizon Robot, which was established in 2015, and mobileye, which was founded in 1999, horizon robots can achieve Level 4 autonomous driving, while Mobileye stops at Level 2 autonomous driving.
Mobileye and self-driving cars are not the strategic direction CEO Gelsinger has set for Intel. The later it goes public, the more resistance Mobileye will face, and he is concerned about the IPO valuation.
The split is an important decision made by Intel, as Intel is in the chip business and the company will use the IPO funds to build new fabs. In addition, STmicroelectronics designed EyeQ chips, TSMC produced EyeQ chips, and in recent years, Intel's CPU share has been lost to AMD and thus fallen into a trough, which in any case will not affect Intel's goal of getting out of the trough.
(Proofreading/Jenny)