At the 2024 CES, AI, chips, and automobiles are among the three major keywords, and Intel's latest "Automotive Plan" spans across these three areas.
During this CES, Intel first announced an acquisition agreement with Silicon Mobility SAS to upgrade its electric vehicle energy management technology. In addition, Intel also established a committee with the International Society of Automotive Engineers (SAE International) to provide automotive standards for vehicle platform power management.
At the same time, Intel has launched a new AI enhanced software defined automotive system level chip (SoC), which will be launched in the second half of the year. Extreme Krypton will become the first automotive manufacturer to adopt the new SoC, which also means that Intel is bringing the performance of AI PCs into the automotive scene.
In addition, Intel has launched the industry's first open automotive chip platform, allowing customers to integrate their chips into Intel's automotive products.
It can be seen that from standard setting, to chip platforms, and then to ecosystems, Intel is gradually advancing and rapidly entering the automotive market.
"We hope to drive the transformation of the entire industry and solve some major challenges. The current automotive industry is in a crucial period of transformation. I firmly believe that the development prospects of this industry are very promising, and by 2030, the entire industry will be transformed," said Jack West, Intel Academician, Vice President of Intel Corporation, and General Manager of the Automotive Business Unit, in an interview with 21st Century Business Herald and other media outlets
At the same time, he said, "In the future, it is very likely to witness one or two Chinese car manufacturers enter the world's top five, and some well-known brands will have to bid farewell to the automotive industry. We see a rare opportunity to help and ensure that the automotive industry becomes stronger, and to ensure that car manufacturers can smoothly overcome this difficult transformation period."
Bringing AI PC into the car cockpit
At the CES booth, Mercedes Benz, BMW, and Volkswagen all emphasized the application of generative AI in car cabins, some directly introducing ChatGPT, and others based on other large language models to enhance the intelligence and human-machine interaction of voice assistants.
Although the current experience is still limited, the trend towards AI is very obvious. And the upgrading of these experiences requires the support of underlying hardware, and chip computing power is the cornerstone of intelligence. Therefore, the intelligent cockpit has become a new battlefield for global chip manufacturers to compete.
For Intel, this is not the first cockpit chip. During an interview with reporters, Gao Yu, General Manager of Intel's China Technology Department, stated: "Previously, Intel also had the Apollo Lake cockpit chip platform, but in the future, Intel's cockpit chips will be mid to high end PC grade chips, using the latest hybrid architecture of P-core and E-core. The entire CPU and integrated GPU computing power can be greatly improved. On the basis of computing power improvement, we can make more innovations in software."
He also pointed out that it is difficult to differentiate the software and hardware architectures of some existing cockpit platforms after binding, so Intel wants to support customers with more innovation through the powerful performance of hardware and the open software architecture as much as possible.
From a vertical perspective, Intel's own chips are iterating, while from a horizontal perspective, the underlying chips involve competition in chip architectures such as x86 and Arm. Therefore, the outside world is also paying attention to Intel's differentiation in using its x86 architecture for intelligent cockpit chips.
Jack West talked about the two major advantages of x86 architecture. The first is the software ecosystem supported by x86, such as PC gaming experience in the car, and the software ecosystem based on Intel PC, which can provide an in car experience that integrates entertainment, multimedia, and conference functions; The second advantage is to provide better basic support based on the experience accumulated in the data center field.
It can be said that Intel is planning to move the PC ecosystem into the car cabin and integrate AI for empowerment, but AI in cars has its own characteristics.
Gao Yu further analyzed: We will divide AI into two levels. One is traditional AI in the cabin, such as OMS and DMS, which require AI computing power to enable voice interaction and gesture control. These usage scenarios have become standard features for each vehicle. The second development direction is generative AI. Recently, many car manufacturers have provided some large language models as a service to the cabin, but now large language models are done in the cabin through cloud based methods, and there is a challenge Some limitations, including latency issues, strong dependence on networks, security issues, and important cost issues, as cloud computing power is becoming increasingly expensive
Therefore, car manufacturers are all concerned about whether large language models can be implemented in a closed-loop manner inside the cabin, Gao Yu stated: "This is what Intel is talking about as AI PCs, an important focal point for AI in the car. With certain combinations of x86 cockpit chips, powerful and localized large language models can be provided to enhance the user experience in the cabin. Large models require more powerful CPU and GPU computing power, software architecture changes, and many require the coordination of toolchains, which is also a direction that Intel is currently designing and promoting."
It is understood that Intel has already planned a series of cockpit chips and is accelerating updates. Jack West revealed, "In addition to publicly releasing the first generation of our products, there are two other generations in the roadmap, and Extreme Krypton will be using our products on a large scale in the second half of 2024."
Regarding Intel's new automotive chips, Neil Shah, Vice President of Research at Counterpoint, said, "Intel is not only introducing generative AI into cars, but more importantly, it can help car manufacturers transform their vehicles and processes into software centric and bring industry players into the era of intelligence."
The competition for cockpit chips
Under the trend of automotive intelligence, the growth trend of intelligent cockpit chips is also optimistic. A report from China International Capital Corporation (CICC) states, "We believe that the penetration rate of cockpit SoC is expected to double in the next five years. The evolution of the automotive E/E architecture is expected to drive the penetration of cockpit SoC." According to Roland Berger's estimation, the penetration rate of multi-core cockpit SoC chips in the Chinese market reached 24% in 2020, and is expected to reach 59% by 2025, higher than the global average.
Faced with a growing market, various manufacturers have entered, and there are many competitors on the battlefield of cockpit chips. According to previous research by Zhongtai Securities Research Institute, cabin SoC chip players are mainly divided into three camps, namely consumer grade chip manufacturers, traditional automotive chip manufacturers, and domestic chip manufacturers.
Among them, traditional automotive chip manufacturers include Renesas, NXP, Texas Instruments, etc., which previously dominated the traditional automotive MCU and ECU chip market and have experience in automotive grade chips; Consumer grade chip manufacturers include Qualcomm, Samsung, Intel, and others, who have advantages in the field of high computing power and advanced process automotive chips. Domestic chip manufacturers include not only major companies such as Huawei, Quanzhi Technology, Ruixin Micro, and Jingchen Shares, but also start-up automotive chip companies such as Horizon, Xinchi Technology, and Xinqing Technology.
It is worth noting that Qualcomm has a significant market share in the automotive intelligent cockpit chip market, and in 2023, Nvidia, which focuses on intelligent driving chips, also teamed up with MediaTek to jointly develop automotive SoC system level chips. MediaTek will develop an automotive SoC that integrates Nvidia GPU chips (Chiplets), equipped with Nvidia AI and graphics computing IPs.
This means that Nvidia and MediaTek are fully entering the fields of intelligent cockpit chips, automotive SoC, and directly entering Qualcomm's hinterland. Now, Intel is also further investing in the cockpit chip market. Is Intel entering the fiercely competitive automotive market too late?
Jack West denied this: "I don't think it's too late for us. Looking back at the history of the smartphone market, the industry only accounted for 20% of the market in 2010 and has grown to 80% in 2020. We believe that by 2035, 80% of cars will be software defined or electric vehicles. Therefore, we believe that now is the best time to enter the market, and we have the technology and support capabilities needed by the industry."
On the other hand, many Chinese car companies are currently developing their own chips, and Intel has its own views and solutions on this trend and future competition.
Gao Yu stated that Intel is embracing the trend of customized chips. Intel has strong OEM capabilities, advanced processes, and the Chiplet UCIE architecture, "This can achieve a mix and match of Intel's process chiplets and customers' self-developed chiplets, which is our advantage. Intel embraces this direction. In the future, Intel will use chiplets to find partners, design or entrust them to Intel to manufacture their own chiplets, and package them with other Intel chiplets into a final chip, which may adopt the customer's brand." Gao Yu said.
"This is a very important direction for Intel, and it is also a direction that we have communicated with many car manufacturers, especially top car manufacturers, who are very concerned about." Gao Yu further said, "The advantage of making chips is that the yield of each chip can be greatly improved, and it is also very easy to combine different functions of the final chip, greatly reducing the development cycle. It can also achieve various mixing and matching of different processes, which can improve the scalability of Intel products and accommodate customers' customized needs."