Cambricon Technologies' AI chips are ubiquitous, used in 100 million smartphones and servers across Asia.
In March 2016, Google's artificial intelligence (AI) program caught the attention of the world by defeating world Go Champion Lee Se Dol. What is not well known is that to achieve this achievement, AlphaGo requires 2,000 CPUs and 3,000 GPUs. The electricity bill also cost $ 3,000 per game at the time. This was revealed by Han Song University of Technology (MIT) Professor Han Song at Tshinghua University.
image.png

In the same month, two Chinese brothers working at the China Science Institute (CAS) research office opened their own company in Beijing. Their mission is to develop a computer chip dedicated to AI applications, allowing for smaller, more energy-efficient, cheaper machines.
Two brothers, Chen Yunji, 36 and Chen Tianshi, 34, grew up in a family with an educator's mother and an electrical engineer father. Both entered university at the age of 14 and 16, and at the age of 24 they received a PhD in computer science before joining CAS as research assistants.
While most users have never heard of their names, Cambricon Technologies' AI chips are everywhere. According to CAS, they are used in nearly 100 million smartphones and servers, including equipment from Huawei and Alibaba.
In June 2018, Cambricon was valued at $ 2.5 billion, becoming one of the most valuable AI startups in China after raising hundreds of millions of dollars in a B-calling round. Alibaba, iFlyTek.
The CAS nominates Cambricon's scientists for the Outstanding Science and Technology Achievement Award in 2019 with the Cambricon A1 chip. The institute believes that this is the first commercial deep learning processor in the world. Deep learning is a branch of machine learning that can adapt to new data and system training to self-study and identify patterns. Its applications range from disease detection through medical imaging to self-driving cars on the road.
Cambricon A1 completely beats traditional CPU and GPU. Chen Yunji compared traditional processors with Swiss versatile knives and deep learning processors like kitchen knives. Although the versatile knife has many uses, when it comes to cooking, kitchen knives are more convenient.
The past few years have seen a surge in the number of Chinese companies producing AI kitchen knives, especially after Beijing announced plans to turn the country into an AI leader before 2030. The escalating US-China tension in which President Trump cut supply of chips to many Chinese technology firms has further increased their determination to become independent of AI chips.
Just 10 years ago, AI was a minority option, without causing a big effect. Many things have changed since then: AI is no longer a niche, an AI chip designed by giants like Baidu, Alibaba, Huawei. Faced with increasing competition, Chen Tianshi said that Cambricon has an advantage over the "big companies" because they provide a more comprehensive product suite, applicable in many situations.
What Cambricon really wants to achieve is to be a "stepping stone" for the popular AI application in the future. People don't need to know the name Cambricon, as long as they contribute to AI.