On November 3, at the 2021 Tencent Digital Ecosystem Conference, Tencent officially disclosed its actions in chip development for the first time: the "Zixiao" chip for AI computing, the "Canghai" chip for video processing, and the "Xuanling" chip for high-performance networks have all made substantial progress. Among them, the "Zixiao" has now been successfully fabricated and successfully powered on.
What kind of chips does Tencent want to build? Why does Tencent want to build chips? Li Qing, director of the Frost & Sullivan and LeadLeo research institute, was interviewed by The Economic Daily to discuss Tencent's "chip journey" with everyone.
The Economic Daily
On November 3, at the 2021 Tencent Digital Ecosystem Conference, Tencent officially disclosed its actions in chip development for the first time: the "Zixiao" chip for AI computing, the "Canghai" chip for video processing, and the "Xuanling" chip for high-performance networks have all made substantial progress. Among them, the "Zixiao" has now been successfully fabricated and successfully powered on.
Tencent started from QQ and has been expanding around the software ecosystem since its establishment. Entering the chip industry, moving from "software" to "hardware," can be regarded as an important exploration in business layout. What kind of chips does Tencent want to build? Why does Tencent want to build chips?
What kind of chips to build
Tencent's intention to build chips has long been evident.
In May 2018, Pony Ma, chairman and CEO of Tencent, revealed some ideas about chips during a forum speech.
He said at that time that when doing software and services, it seemed a bit distant from chips. But in fact, Tencent has built many data centers, including the cloud, and has a certain influence on many server-side needs, even including chips. The application services are massive, with a particularly large user base, which can force the chip design industry to make targeted designs. "If we can also get involved in supporting some chip research and development, it might be better."
Soon after, Tencent took the investment approach and initially entered the chip field: In August 2018, the AI chip startup SinoFoundry received 340 million yuan in Pre-A financing, led by Tencent.
In June or July this year, the outside world suddenly noticed that Tencent had posted multiple job openings for chip research and development on the Internet, involving chip architects, chip verification engineers, etc. So people speculated that Tencent's layout in the chip field had been strengthened.
On November 3, at the 2021 Tencent Digital Ecosystem Conference, the veil was lifted over Tencent's chips, and three types of chips all belong to dedicated chips. Among them, the "Zixiao" is mainly used for AI, the "Canghai" is mainly used for video, and the "Xuanling" is mainly used for intelligent networks. Regarding the characteristics of these chips, Tencent has made professional explanations, but it is still not easy for non-industry insiders to understand. In summary, the performance of these three types of chips has been significantly improved compared with similar products.
It is understood that in the future, Tencent's strategy in chip development will adopt a two-pronged approach, one is self-development and the other is cooperation. Tang Daosheng, senior executive vice president and CEO of the Cloud and Smart Industries Group, said that Tencent will always seek to maintain in-depth strategic cooperation with domestic and foreign chip companies through the model of ecological co-construction, combining the customization ability of chips with that of software.
Why build chips
Behind Tencent's chip-building efforts, there are two questions: one is why it wants to do it, and the other is why it can do it.
First, let's look at the question of "why it wants to do it".
Qiu Yuepeng, vice president of Tencent, COO of the Cloud and Smart Industries Group, and president of Tencent Cloud, introduced at the conference that with the development of the cloud computing industry, the trend of integrating software and hardware is becoming more and more obvious, and software-defined hardware is irresistible. Chips are the most core part of hardware. Tencent has been actively exploring in the chip field and has long-term investments. And AI computing, video processing, and high-performance networks are three scenarios with strong demand.
Tang Daosheng also responded during a media group interview. He said that Tencent's investment in this area is driven by demand, and it is also because some problems need to be solved, either to reduce costs or to use Tencent's infrastructure more efficiently.
“For example, the scale of AI services is very large. How to reduce energy consumption through self-developed chips, how to optimize costs, these are all areas that drive us to keep investing in this field. Another example is that as audio and video usage scenarios become more and more common, as an underlying technology and a more natural interactive mode for communication between people, it is applied in many scenarios of Tencent to meet needs such as healthcare, education, and live streaming, and it has also become an important place that consumes computing resources. We have also made a lot of investments in audio and video coding technology and hardware chips, hoping to achieve integrated software and hardware to support audio and video encoding and decoding more efficiently. Another example is the Xingxinghai server, which is built for the cloud and is a multi-core chip that meets the needs of cloud application scenarios.” Tang Daosheng said.
Li Qing, director of the Frost & Sullivan and LeadLeo research institute, told The Economic Daily reporter that integrating software and hardware is an AI trend. The support of algorithms must rely on the capabilities of underlying chips, and the existing ones are not sufficient to perfectly match the application of AI algorithm models. Not only Tencent, but Huawei and Alibaba are also building AI chips, which is the inevitable direction for enterprises building AI platforms.
Now let's look at the question of "why it can do it".
As we all know, the threshold for chips is very high. But compared with general-purpose chips, there are more opportunities for dedicated chips. Tencent's entry point is dedicated chips.
Li Qing believes that Tencent's choice is wise. This part of the chips is not meant to be open for all industries, but has specific uses. From the perspective of AI algorithm applications, it also tends to be in a professional direction. Video processing and audio and video technology are areas where Tencent is strong. Matching this ability with AI is very consistent with Tencent's current development strategy. R & D follows actual needs, not blindly following trends or hype concepts, which is very realistic.
Of course, Tencent's chip-building efforts are not without challenges.Li Qing said that after all, Tencent did not start from hardware. The key lies in the degree of self-development and cooperation with the outside world. Establishing an ecosystem has always been a principle that Tencent adheres to, and this part will definitely be opened up to partners in the future.
*This article is reprinted from The Economic Daily, with reporter Guo Rongcun. The original article is titled "A Detailed Explanation of Tencent's Chips: What Kind of Chips Are Being Built and Why".
