Intel Inside, now involved in smart cars

The CT-Y1 launched this time has the ability to schedule large AI models with 7 billion parameters. In the running score data released by AnTuTu, it is on par with Qualcomm Snapdragon 8295, reaching the top flagship level.

The CT-X1 has the ability to schedule large AI models with 13 billion parameters and supports multi-modal generative large models, with performance about 30% better than Snapdragon 8295.

Let’s take a look at the position of Intel’s chip.

It can be judged from the disclosed data or information that whether it is storage, visual support, or CPU and GPU performance, the Arc A760-A is comparable to the other two flagship products, but Intel is better in terms of AI computing power required after the large end-side model is installed.

Compared with Qualcomm Snapdragon 8295 chip, with AI computing power of 30TOPS, Intel Arc A760-A is more than 7 times that.

Compared with MediaTek CT-X1, the Arc A760-A chip can support large models with more than 14 billion parameters running in the car cabin, exceeding the 13 billion of CT-X1.

In fact, if you look at the way Intel, Qualcomm, and MediaTek make cockpit chips from the initial end, you can clearly see the similarities and differences.

The common point is that the three parties are all "destocking" and slightly adjusting the chip products for mobile and PC ends and applying them to the cockpit field. Qualcomm Snapdragon 8295 and mobile phone chip Snapdragon 888 are the same source, MediaTek CT-X1 chip and Dimensity 9300 chip also have certain similarities, and Intel's Arc A760-A launched this time is the same source as the PC chip A750.

And this is where the difference comes from. Qualcomm and MediaTek emphasize the connection between cars and mobile phones, following the traditional perception of cars as large-sized mobile phones.

However, Intel emphasizes another expression of the future AI cockpit, which is to bring cars closer to the concept of PCs and data centers. From the previous emphasis on multi-screen interaction, business office, and support for 3A games, we can also see a strong PC color.

Jack's argument for data centers is that mobile devices such as mobile phones are good at single-user, single-operating system environments; data centers cover diverse operating systems and massive users, and the future direction of car cockpits will be closer to the latter, which means that different operating systems are needed and compatibility is maintained.

Whether it is PC or data center, this is what Intel is very good at. Therefore, Intel can reasonably transplant the experience of AI PC and data center to the smart cockpit.

For example, in taking over the software ecosystem shaped based on the X86 architecture, Intel has gathered more than 100 independent software vendor (ISV) partners, who have created more than 500 functions and AI applications that can shape a powerful ecosystem for vehicles.

And apply data center solutions to the vehicle side, including optimizing task allocation, reducing task interference, improving system efficiency and stability, etc. For example, when the car is charging, the ADAS system is automatically turned off; when there are people in the car, the high-power car sentry mode is controlled.

Obviously, Intel is emphasizing its compatibility with the future cockpit form factor.

Li Ying, Intel vice president and general manager of Intel China Software and Advanced Technology Group, said that in addition to data center and PC capabilities, Intel can also graft mobile capabilities to the x86 platform through IBT (Intel Bridge Technology).

This ecological integration capability is also an advantage over Qualcomm, MediaTek and other chip companies.

03. Intel’s opportunity in the future cockpit ecosystem

The most direct competition among chip manufacturers is still in the installed capacity, that is, obtaining the support of the OEM and being installed in the vehicle at a fixed point.

Both MediaTek and Intel are latecomers, and it is not easy to catch up directly.

On the one hand, the barriers to replacing automotive-grade chips are very high, and automotive certification and safety testing bring time and cost pressures. It is reported that it costs about 10 million yuan for the OEM to re-certify the entire process, so chips are generally not replaced easily.

On the other hand, Qualcomm's advantage as a "big company" in the cockpit chip field is difficult to shake for the time being. Qualcomm can provide OEMs with more than just chips, but also technical support, quality control, and multiple solutions. The stability, long-term nature, and convenience of this kind of cooperation are also attractive.

Therefore, companies including MediaTek and Intel need to amplify the differentiated advantages of their products and transform demand from homogeneity to specialization.

Now they are betting on computing power.

Both MediaTek and Intel hope that, as AI defines cockpits, the demand for large models on the edge will tilt the market toward cockpit chips with advanced processes and high computing power. Although this computing power currently has a large redundancy, both companies are focusing on the 3-5 year preset growth.

Intel expert Cloud Li said that the R&D cycle of a car by an OEM is 18-24 months. This long cycle cannot guarantee that there will not be applications with higher standards and higher requirements. Before this, the chip needs to be sufficiently preset, and computing power is the core link.

At present, only Zeekr has publicly voted in favor of Intel's cockpit chip. In January this year, Zeekr announced its cooperation with Qualcomm's first-generation Soc chip, and Zeekr 009 will become the first mass-produced car model.

But Intel urgently needs to be recognized by more OEMs, especially those that previously chose Qualcomm.

In response to this, Intel came up with another idea: openness.

Openness can be understood as the AI ​​cockpit ecosystem built by Intel, which gives OEMs and main manufacturers extremely high degrees of freedom.

Jack believes that “customers need solutions that are more scalable (in terms of performance), can reuse software, and cover all aspects from high-end to entry-level models. They need a more open ecosystem to pursue differentiation, rather than everyone using the same solution.”

Therefore, at the software level, Intel has created a future AI cockpit architecture model. Based on this architecture, the future cockpit will be able to efficiently deploy AI software and conveniently build customized AI solutions, ultimately improving development results and reducing costs.

Intel believes that the future development of AI cockpits is not limited to a single technology or product, but requires the construction of a complete ecological experience. Therefore, openness is a matter of course. From this perspective, Intel, which has integrated openness into its development strategy for more than 20 years, can use it as an advantage in the competition.

Intel actually wants to win the battle for cockpit chips.

It is eager to launch its global debut in the Chinese market, targeting the trend of large language models being first implemented on the domestic end.

Intel even placed its automotive business headquarters in China, staffed with over 2,500 engineers to focus on local innovation and research and development.

No one doubts Intel's determination for the automotive business anymore. As Jack once said, "Intel has 100 billion reasons to stay in the automotive industry for a long time, because we have invested $100 billion in chip production capacity."

At present, whether Intel and MediaTek can shake Qualcomm's position can only be seen after the landing, but what is certain is that the cockpit chip landscape monopolized by Qualcomm will be further rewritten. When the wave of AI-defined cockpits hits, the battle between new and old players will be