Centaur, a subsidiary of IC design house VIA, has officially launched an artificial intelligence (AI) processor, which is expected to be mass-produced using TSMC’s 16nm process. According to the corporate circle, the AI processor is expected to start mass production in the second half of 2020, and the shipment of VIA AI-related product lines has the opportunity to flourish simultaneously.
Centaur launched the world’s first AI processor with x86 architecture, which supports Deep Computing Accelerator (DLA), and introduced TSMC’s 16nm process. According to the Linley Group, a well-known American semiconductor media, after testing the product, the product’s neural network computing performance is quite excellent, and it does not require additional graphics processing unit (GPU) assistance.
According to Linley Group, Centaur’s AI processor is equivalent to 23 Intel x86 architecture processors with vector neural network instruction set (VNNI) in AI performance processing, and this product can be used in combination with x86 architecture CPUs to give full play to AI efficacy.
It is understood that Centaur’s AI processor was reported as early as 2019 that the product had entered the testing and verification stage, and the official announcement was made in early December. It is also rumored in the legal circle that Centaur has the opportunity to start mass production in the second half of 2020, making every effort to seize this wave of AI business opportunities.
Centaur was originally held by IDT, a major American communication chip manufacturer, and was subsequently acquired by VIA in 1999. Centaur has been under VIA for more than 20 years, and has jointly developed many PC-related chips with VIA, which occupies an important position in the VIA subsidiary group. . The re-launch of new products with the x86 architecture with Centaur is equivalent to a symbol that VIA is expected to re-enter the x86 market with AI products and seize this wave of AI orders brought by 5G.
In addition, VIA not only has an AI processor product line, but also has previously built AI embedded IoT, advanced driver assistance systems (ADAS), etc. in place, and has begun mass production and shipments to enter Chinese buses and the Internet of Things. Manufacturer supply chain.
Driven by the resumption of work by mainland customers and the return of orders, VIA’s consolidated revenue in March increased by 60.9% to 629 million yuan, a record high in 70 months. The highest in a single quarter in more than five years. According to the analysis of the legal person, China is actively carrying out de-beautification. VIA has been in China for many years. It is expected to take advantage of this wave of AI business opportunities and the effect of de-beautification to take advantage of AI orders from mainland customers, and the subsequent consolidated revenue will be expected to grow steadily.
Further reading: VIA from fame to extinction
VIA started in Taiwan, China, and like many Taiwanese IT companies that later became famous, its current achievements should be attributed to the transfer of the global chip industry center in the late 1970s, and this transfer is also catering to the Semiconductor industry. The direction of development towards a global network.
In the 1980s, the information technology industry of Taiwan, China, took advantage of the new trend of the US semiconductor industry to undertake the work of producing complete sets of equipment for the other party, and accumulated technology in the process. ON the other hand, Taiwanese enterprises are mainly small and medium-sized enterprises and owner-oriented enterprises. Because of this, they turn quickly and move quickly. As long as they see an opportunity, they can start immediately.
Therefore, the information technology industry in Taiwan, China, has made full use of this advantage, so that it has developed into a production base for information technology products in the world.
Wang Xuehong officially saw this opportunity and decisively founded VIA.
From agent start to motherboard first
Like other Taiwanese manufacturers, VIA started out as an agency factory.
Almost everything on the motherboards that VIA supplies to other companies is purchased from foreign manufacturers, and VIA itself has no technology or ability to produce it independently.
In Wang Xuehong’s own words, “At that time, we were just an assembly workshop, no different from assembling toys.” Despite this, Wang Xuehong also understands that if the machine is installed well, it has great potential for development.
Wang Xuehong hopes that VIA can start from assembly like Dell, and can also accumulate certain experience and funds in the case of shortage of funds and lack of talents.
But the fact is very different from Wang Xuehong. Due to market differences and product direction issues, VIA’s profit margins in motherboards are very small. What makes Wang Xuehong even more annoyed is that not many customers buy even the motherboard.
Wang Xuehong clearly realized that without changing its strategy and development direction, the company would not be able to survive.
After a period of planning, Wang Xuehong began to recruit talents to independently design and produce motherboards.
This process is arduous, but the result is gratifying, VIA has finally successfully used the ability to design and produce motherboards in-house, and continue to strengthen this ability.
Soon, VIA occupied the market and achieved the first place in the motherboard market.
At this time, Wang Xuehong had a new idea. Since he can make the motherboard himself, why can’t he make the chip himself?
Talent: The Rise of VIA
“What is the most expensive in the 21st century, talent!”
In a sense, talent is the core and future of the entire enterprise.
Wang Xuehong realized something as early as the end of the 20th century, so after she made up her mind to make chips, the first thing she had to do was to recruit talents.
To serve, Wang Xuehong recruited two partners, Chen Wenqi and Lin Zimu.
Chen Wenqi has a strong professional ability in the field of design, while Lin Zimu has a solid foundation in semiconductor academic theory and practical design. At that time, when looking at the whole of Taiwan, China, very few people could surpass him.
Since then, Wang Xuehong is in charge of funding and planning, Chen Wenqi is in charge of strategy and marketing, and the “VIA Iron Triangle” relationship where Lin Zimu specializes in technology research and development has been formed.
One of the main things that Chen Wenqi did at the very beginning was to carry out drastic reforms for VIA and to redefine VIA’s strategy.
In fact, although the number of chip set design companies in the early 1990s was staggering, reaching a hundred at one time, when VIA really started shipping, there were only seven chip design companies left, both in terms of product quality and business scale. See, VIA is at a disadvantage.
In order to restore the disadvantage, Chen Wenqi decided to move VIA’s headquarters from Silicon Valley, California, USA back to Taiwan, China, in order to solve customer needs nearby and make products that best suit market needs.
In addition, Chen Wenqi also made good use of the foundry trend, and handed over 90% of the chipsets to TSMC to avoid the financial burden of having to invest more than one billion US dollars to build a fab.
As a result, VIA opened the low-cost market, and eventually won a place.
At its peak: The second-largest chipset giant
“As long as VIA’s idiots continue to operate for a long time, it is unpredictable who will die in the future.”
Just when Wang Xuehong was delighted that VIA had a place in the chipset market, Intel’s siege had begun quietly.
At that time, VIA had just seized the low-cost market in chipsets, with a global workforce of less than 3,000 yuan and a market value of less than $2.5 billion.
This is a power disparity that even elementary school students can see.
Wang Xuehong knew that since he chose to enter the chipset, it would be a matter of time to challenge Intel.
To this end, Wang Xuehong decisively formulated measures and implemented them effectively.
Marketing is important, but technological innovation is even more urgent.
In July 1999, VIA officially launched its self-developed PC 133 chipset. Its main memory specification was more cost-effective than the memory specification of the Intel Pentium 3 processor at that time, and VIA made its name.
Since VIA’s “red”, orders from many international manufacturers have poured in, and the share of chipset products once reached 70%.
It is unrealistic to surpass technology in a short period of time. Intel has to change its strategy and choose to sanction VIA through patent litigation on a global scale.
In June 1999, Intel successively filed a lawsuit against VIA in the United States, the United Kingdom, Singapore and other places, claiming that VIA infringed on Intel’s patented technology.
Intel’s strategy has really worked, VIA’s orders have plummeted, customer confidence in VIA has declined, and the most difficult thing to make up for is that VIA’s hard-earned reputation has received unprecedented challenges.
The most direct manifestation is that during the lawsuit, the global market share of VIA chipsets once fell from the highest 70% to about 20%.
After a long-term tug-of-war, in April 2003, Intel and VIA withdrew all ongoing lawsuits and signed a 10-year cross-licensing agreement for their product lines as Intel’s competitive strategy shifted.
So far, VIA can be regarded as the second-largest chipset giant’s position.
Overexpansion: A precursor to decay
Even among Taiwanese companies, VIA is known for its diversity. In addition to VIA Electronics, VIA Group also includes more than 30 companies including HTC, Quanda, Jianda, Dopod, Weihong, Weihan, etc. Its business covers chip design, production and manufacturing of PCs, mobile phones, and PDAs. , sales, agents.
At that time, Wang Xuehong had been guiding the development of VIA with the “diversified management strategy”, not only in the field of chipsets, but also in the fields of CPU, GPU and communication chips.
She said that in addition to its global layout in the chip field, VIA also has chipsets, set-top box chips and NC chips, as well as terminal manufacturing and sales.
Even while litigating with Intel, VIA is actively expanding its footprint.
In 1999, VIA successively acquired three CPU companies including Cyrix, IDT, and S3. Among them, S3 has signed a cross-licensing agreement with Intel. In other words, VIA also owns some of Intel’s patents through the acquisition of S3.
But it is precisely because of the problem of technology licensing that VIA’s chipsets have been in the process of declining gross profit margins. In 2001, VIA’s revenue was NT$34.104 billion, the overall gross profit margin dropped to 37.1%, and the net profit was also lower than the level in 2000.
Based on the above considerations, VIA urgently needs new profit points, and optical storage just fills this vacancy.
At the end of 2001, VIA first released 60 times CD-ROM single-chip products, and by April 2002, VIA began to fully provide DVD-ROM and CD-RW chips. S
As for the embedded platform field, VIA released the system platform code-named Eden, once again a solution known for its speed, which has made a great contribution to VIA in the embedded reference market.
VIA is also actively exploring the graphics chip market, releasing the first-generation Zoetrope core Savage XP desktop computer graphics chip.
However, whether it is a rapid or slow expansion, VIA encountered many problems in the process of expansion, and also bought a foreshadowing for the subsequent decline.
Because VIA has invested most of its manpower and material resources in areas other than the main business chipset, VIA’s market share in chipsets has dropped significantly, and the gross profit level has continued to be sluggish. Spreading too much and being a novice at the same time makes it difficult to compete with other competitors.
Over time, VIA felt more and more powerless.
In May 2009, VIA lost NT$6.6 billion for two consecutive years. It was even entered into a full-delivery share. At the shareholders’ meeting in June, the stock price was still only 12 yuan, and the general manager Chen Wenqi bowed and apologized to the shareholders again and again.
Fortunately, VIA’s hardware product division was already working on smartphones before 1997. HTC was spun off from VIA in 1997 and went public independently in 2002 when VIA and Intel were entangled with each other.
After the independent listing, HTC’s performance can be said to be better than the blue. On February 12, 2005, HTC surpassed MediaTek for the first time with a stock price of NT$232 to become the stock king, opening the legend of HTC’s stock price. In May of the same year, HTC launched the world’s first 3G mobile phone equipped with Microsoft’s Windows Mobile 5.0 operating system. On April 26, 2006, HTC’s closing price was NT$1,020, breaking the 1,000-yuan mark, becoming the second Taiwanese stock to break the 1,000-yuan mark in 16 years after Yitong Solar, which was better than the original performance of VIA’s NT$629 per share. Much excellent.
In 2011, HTC mobile phone sales peaked, and the stock price peaked at NT$1,300 per share.
However, just like VIA, a patent war once again extinguished the hopes of the stock king.
In addition to his unique investment vision, Wang Xuehong’s ability to revitalize assets is also top-notch. VIA, which has been losing money for years, is facing a crisis of delisting. However, VIA still holds many important patents. In August 2015, VIA disposes of VIA’s CDMA intellectual property assets to Intel, a manufacturer that owns the CDMA silicon intellectual property rights of important communication technologies.