Text | Angel Lee
Edit | Su Jianxun
Mobileye is trying to tear off the label of "hash rate laggard, ecological closure and dragging its legs".
Recently, the annual CES (International Consumer Electronics Exhibition) opened in Las Vegas, USA. At the conference, Intel's Israeli chip company Mobileye released three intelligent driving chips in one go, covering the field of L4-L2 intelligent driving.
At the same time, it also announced that it will cooperate with Geely Automobile's Extreme Kr Automobile to develop L4 level autonomous driving consumer pure electric vehicles, which is expected to be the first to be launched in China in 2024.
A few months ago, Intel announced that Mobileye announced that it would be available in the United States in 2022. It is not difficult to see that this company hopes to come up with a batch of chip releases and future plans before going public to give capital confidence.
However, the current competition for smart car chips is fierce. On the opening day of CES, Nvidia brought the eighth generation of autonomous driving platforms, and also gained the new car-making force Jidu Automobile as a partner.
The mass production time of the new products released by Mobileye is after 2023, and there will be no cooperation between car companies that will announce the use of its new chips. Friends are fiercely attacking, can Mobileye still block the attack?
Can the computing power catch-up and open ecology bring a turnaround for Mobileye?
As an Israeli tech company founded in 1999, Mobileye is not young. In the past 20 years, because of the forward-looking view and grasp of the first wave of automobiles and artificial intelligence combined with the outlet, Mobileye used algorithm + EyeQ series chip solutions to help car companies achieve primary auxiliary driving functions, and once established the industry position of "global intelligent driving hegemony".
Even in 2021, Mobileye still holds 41 orders from more than 30 car companies, accumulating 100 million shipments of EyeQ chips.
However, under the aura, the transformation of smart cars is changing, and Mobileye is facing a severe market crisis. With the fierce attacks launched by chip manufacturers such as NVIDIA, Qualcomm, Huawei, and Horizon in terms of product computing power and customer orders, Mobileye's "falling behind" has become a fact.
Taking computing power as an example, the same is for the L4 level automatic driving scenario, NVIDIA's automatic driving Orin chip hash rate of 254 TOPS, and the dual-chip computing power reaches 400 TOPS. At present, the car companies that have determined to use Orin chips include: Weilai, Xiaopeng, Ideal, Weima, Zhiji, Mercedes-Benz, Volvo, Jidu, etc., which can be called "car harvester".
Another chip giant, Qualcomm Snapdragon Ride platform, has a single chip hash rate of up to 60 TOPS, and an AI accelerator equipped with 300 TOPS can reach 700 TOPS. With this chip platform, Qualcomm successfully intercepted BMW's new car orders for 2025 from Mobileye.
In addition, the relatively closed algorithm + chip solution has also dragged Mobileye's hind legs to a certain extent. Mobileye's historical claim is that through the deep coupling of chip hardware and minimalist algorithms, chips achieve a high energy efficiency ratio. Therefore, when selling products to the outside world, Mobileye is also packaged with hardware and software, and sold to customers in the form of black boxes, making it difficult for car companies to flexibly adjust according to their own needs in the process of using products.
In contrast, chip companies that can provide flexible product modules and tool chains, such as Huawei and Horizon, will be more favored by car companies. Under the guidance of a variety of factors, car companies that originally belonged to Mobileye, such as Weilai, Ideal, Tesla, Great Wall, BMW, Volvo, etc., have thrown themselves into the arms of other chip companies.
Apparently, Mobileye is also aware of the crisis.
At CES 2022, Mobileye not only unveiled the EyeQ Ultra system integration chip specifically for autonomous driving, but also launched two new EyeQ 6L and EyeQ 6H for assisted driving.

EyeQ Ultra System Integration Chip
Among them, the EyeQ Ultra hash rate of 176 TOPS, using a 5-nanometer process. Mobileye said the chip's performance is equivalent to the combined performance of 10 existing EyeQ 5 chips. It will be delivered at the end of 2023, and the full realization of vehicle-grade mass production will not be until 2025. The EyeQ 6L and EyeQ 6H are for entry-level assisted driving and high-end assisted driving, respectively, with mass production times in mid-2023 and 2024 respectively
It can be seen that Mobileye is trying to catch up with the computing power and complete the product matrix.
However, in Mobileye's view, hash rate is not the only measure of chips. "176 TOPS sounds like a small number, about one-fifth of the power that competitors claim to have. But the key is not only computing power, but efficiency. Mobileye President and CEO Amnon Shashua said in a CES speech.
Mobileye values the efficiency with which the chip handles problems during actual operation. This is evidenced by the fact that the SuperVision, a driver assistance system mounted on a krypton car, uses only two EyeQ5 chips to keep the entire system running.
Similar to the "focus on efficiency rather than computing power" statement, is also the view of the horizon of domestic chip companies. For this reason, Horizon proposes an AI chip evaluation method to evaluate the average processing speed of the chip within a reasonable accuracy range.
The view of "only computing power theory" is indeed undesirable, but it cannot be denied that the desire of car companies for high computing power is still there. After all, the production and development cycle of automobiles often lasts for several years. Car companies hope to use the highest possible computing power to ensure that there is enough computing power space to support the arrival of automatic driving after the vehicle is launched. Mobileye did not give more information on whether higher hashrate deployments will be required in the future.
Compared with computing power, Mobileye is more anxious at the moment, I am afraid that how to retain more car company partners.
Because from the perspective of the product matrix, the mass production time of Mobileye's three new chip products is after 2023, while the mass production of the same level of chips of NVIDIA and Qualcomm is in 2022, and the two-year time difference will make the opponent have more opportunities to exert power.
So, Mobileye is urgently adjusting its black-box delivery model. Erez Dagan, executive vice president of product and strategy at Mobileye, said in an interview with 36Kr and other media outlets: "We have released an SDK (Software Development Kit) for this system-on-chip, and the software technologies related to partner development. ”
This means that some of the computing power of the EyeQ5 can be used to meet the R&D needs of partners, and the OEM or Tier1 has room for third-party programming. This is already reflected in the EyeQ5 chip.
It is true that under the tide of intelligent cars, the transformation anxiety of car companies is spreading, and they are eager to seek a more certain intelligent route. The dimensions considered by OEMs are always many and complex, including technology, products, costs, deliveries, services, etc. Under such a time window, who can solve problems for the main engine factory faster and reach a mass production goal with the main engine factory faster, who has a higher chance of surviving.
It remains to be seen whether Mobileye's computing power catch-up and open ecological measures can bring about a turnaround for them.
Deeply bound extreme krypton, Mobileye intends to fight back
So, in addition to future products, does Mobileye have no other opportunities to fight back? Nor is it.
At CES, Mobileye announced that it has partnered with Extreme Krypton To develop a new consumer-grade pure electric vehicle with L4 autonomous driving capabilities. Amnon Shashua, president and CEO of Mobileye, said the consumer-grade self-driving car will cost well less than $5,000.
This is more like Mobileye's helpless move, after all, three years later, the mass-produced chip will not solve the current urgency. Recently, autonomous driving and OEMs have gathered into an alliance, aiming at the large-scale landing battle of Robotaxi (self-driving taxi) in 2024-2025.
Previously, Extreme Krypton just announced a partnership with Waymo, an international autonomous driving company, to provide exclusive vehicles for Waymo's driverless fleet. Domestic self-driving players Xiaoma Zhixing and Wenyuan Zhixing have also reached cooperation with FAW Nanjing and GAC respectively to jointly promote the scale of Robotaxi. Xiaopeng Automobile said that it will explore the robotaxi business of advanced autonomous driving in Guangzhou starting in the second half of next year. Tesla often talks about fully autonomous driving.
Therefore, Mobileye is also trying to use the current chip EyeQ5, grasping Robotaxi on the one hand and consumer-grade autonomous vehicles on the other.
Mobileye's self-driving application landing path
Amnon Shashua said in his speech at the conference that there will be two directions for autonomous driving applications in the future, Robotaxi (self-driving taxi) and consumer-grade autonomous vehicles. The former can be used to carry people and goods, but is expensive and can only be driven in limited areas, while the latter is cheap and does not travel geographically.
"Robotaxi will be available earlier, probably in 2022 to 2023. For consumer-level autonomous vehicles, the more reasonable time to market is in 2025. ”
According to previous plans, Mobileye and NIO have cooperated to develop the Robotaxi vehicle, equipped with eight EyeQ5 chips, and are expected to provide Robotaxi services in Tel Aviv, Israel, and Munich, Germany, in 2022.
"We can learn a lot by launching Robotaxi, which is critical for [building] the next phase of consumer-level autonomous vehicles." Erez Dagan, mobileye's executive vice president of product and strategy, said in an interview.
As for consumer-grade autonomous vehicles, Mobileye's cooperation with Extreme Kr will be based on the Geely SEA Haohan architecture, and equipped with Mobileye's existing product EyeQ5 chips, the number of chips will also drop from 8 to 6 required by Robotaxi. The cost of the vehicle can be further reduced, "[the chip] price is only a few hundred dollars." ”
Mobileye's already accumulated Experience in the EyeQ5 on existing models can be reused in consumer-grade autonomous vehicles. Previously, the Extreme Kr 001 model has been equipped with Mobileye's SuperVison system, which also uses two EyeQ 5 chips.
In response to the cooperation with Extreme Kr, Mobileye said it will strengthen its R&D capabilities in China, establish local data centers and grow its China team to support the fast-growing business in China. "We've transferred people from within Intel and technicians from Israel to China."
All in all, in the case of the time difference between high-power chips and fierce rivals in the industry, Mobileye is using existing products to try first. Wait until the hardware is mass-produced and then do the algorithm change, then there will be a chance to catch up.
After more than a decade of development, the butterfly effect that has stirred up from the other side of the Atlantic has just swept the world, and the power of the storm is still increasing, and Mobileye is shaking. How to take back and stabilize the market, Mobileye may have to give more answers.