Can China's AI chip companies shake industry giant Huida?

Artificial intelligence has become the hottest outlet nowadays. Companies from all walks of life are trying to improve work efficiency and competitive advantage through this technology. In the chip field, Huida as a leading hardware manufacturer, the influence can not be ignored. Earlier, the Bank of America Merrill Lynch Group said in a report that Huida will become the leading supplier of artificial intelligence chips, which is creating standards for the artificial intelligence computing industry. In addition to Huida, there have been a number of artificial intelligence chip startups in the past two years. They all eagerly want to become the next Huida, but the real competition may come from established companies such as AMD and Google, as well as a batch of China AI chip company. In the current China, artificial intelligence has risen to national strategy. According to the "New Generation Artificial Intelligence Development Plan" issued by the State Council in July 2017, by 2020, the strategic goal of China's artificial intelligence is to synchronize technology and application with the world's advanced level, and the artificial intelligence industry has become a new important economic growth point. With a scale of more than 150 billion yuan, the scale of the related industries has exceeded 1 trillion yuan. So, how is China's AI chip company catching up with the world's development? To what extent can they compete with Huida? Huida: AI Chip Market Leader Huida is undoubtedly the undisputed leader in the AI ​​chip market. In 2017, Huida is a fascinating sight. From games and data centers to artificial intelligence, the market demand for Huida chips has been on the rise. Because deep learning requires “violence” for computational speed, Huida GPU chips allow a large number of processors to perform parallel operations at speeds that are ten or even dozens of times faster than CPUs, thus becoming the vast majority of artificial intelligence researchers and developers. Necessary "nuclear bomb". According to the news of the “Financial World Weekly”, almost all domestic AI entrepreneurs are using Huida GPUs and platforms, and it is normal for a small AI startup team to spend hundreds of thousands of yuan on Huida chips. Andreessen Horowitz, Silicon Valley star investor, also revealed: “We have invested in a large number of startup companies in various fields for deep learning, each running on Huida's platform.” Affected by this, Huida's stock price has also soared, rising from 20 US dollars in 2015 to more than 220 U.S. dollars today, an increase of over 1000%, and the market value has exceeded US$130 billion. In addition to being at the forefront of the chip industry, Huida is also active in the field of gaming graphics and automatic driving. At the recent CES 2018 show, Huida released the latest 65-inch high-end gaming monitor. Acer, Asus and Hewlett-Packard will use their technical specifications and standards as a 4K HDR display. This ultra-large display equipped with G-Sync is very suitable for high-end gaming PC products. In terms of autonomous driving, Huida founder Huang Renxun held a press conference in Las Vegas on the evening of January 7th, to announce major cooperation projects with Flowserve and Uber. Huang Renxun also revealed that Huida currently has more than 320 cooperative enterprises and institutions in the field of automated driving. Huida's strength in the field of artificial intelligence is evident. AI Chip Chaser: Intel, AMD, Google Huida is not the only giant in the GPU space. Intel, AMD, and Google also have advantages that cannot be ignored. In the semiconductor market, Intel has been occupying a "leading position" for many years. However, the rise of artificial intelligence has made the world's largest personal computer parts and CPU makers face serious challenges. Under this pressure, Intel chose to fight back through "strong alliances." In December 2017, Intel and AMD, competing for more than 50 years, announced that they will jointly launch a notebook computer chip that combines Intel processors and AMD graphics units. According to Wang Xiaolong, research director of the company, "If Nvidia's revenue from PC GPUs decreases, Nvidia's development in new markets will be hampered, so Intel has chosen to cooperate with AMD." Alan Priestley, head of research at Gartner, a data research organization, also believes that the cooperation between Intel and AMD will have some impact on Nvida. "The current application of this chip jointly launched by Intel and AMD is also the application of Nvidia GPU. It is clear that Intel Processors have infiltrated this area, and through the new chips, AMD will gain a foothold in this area." In addition, the second tier of Google is also trying to trigger a new subversion. In 2016, Google announced that it will independently develop a new processing system called TPU. The system is a dedicated chip designed specifically for machine learning applications. By reducing the computational precision of the chip and reducing the number of transistors required for each computational operation, the number of operations per second that the chip can run can be made higher, allowing a finely tuned machine learning model to run on the chip. Faster. Compared with CPUs and GPUs, TPUs are 15-30 times more efficient and 30-80 times more efficient. Rising domestic AI chip makers With the implementation of major strategies such as “Made in China 2025” and “Internet Power”, Chinese chip companies are also rapidly developing and their strength cannot be ignored. Cambrian Technology The Cambrian is the world's first successful AI chip company with mature products. It has two product lines: terminal AI processor IP and cloud high-performance AI chip. The Cambrian 1A processor (Cambricon-1A), released in 2016, is the world's first commercially available deep learning processor. It was also selected as one of the fifteen "World Network Leading Scientific Achievements" by the Third World Network Conference (Wuzhen). "." The Cambrian team came from the Institute of Computing Technology, Chinese Academy of Sciences. The latter is the first national academic institution in China that specializes in comprehensive research on calculator science and technology. A group of high-tech companies such as Lenovo and Shuguang were born from this institute. At present, it is also an important shareholder of the Cambrian Technology and a long-term partner for production, learning and research. In August 2017, Cambrian Technology completed a round of US$100 million round A financing, which was jointly invested by SDIC Ventures, Alibaba Ventures, Lenovo Ventures, Guoke Investment, Zhongke Turing, Yuanyu Origin, and Yongsheng Investment. After this round of financing, the company became a "unicorn". Huawei According to data from 2017, Huawei has become China's largest and third-largest manufacturer of smart phones in the world. However, while developing mobile phones, Huawei is also developing artificial intelligence processors. In September 2017, at the IFA exhibition in Berlin, Huawei officially released the latest Kirin 970 chip. This is Huawei's first artificial intelligence (AI) chip. The chip uses TSMC's 10nm technology for the first time, and Qualcomm's latest Xiaolong 835 chip is a process, but the integration of 5.5 billion transistors is far more than the 3.1 billion of Qualcomm and the 3.3 billion more of the Apple A10, bringing about a reduction in power consumption. %. It is understood that Kirin 970 inherited the achievements of the past generations, the first integrated NPU (NeuralNetwork Processing Unit) dedicated hardware processing unit, innovative design HiAI mobile computing architecture, its AI performance density is substantially better than the CPU and GPU. In October of the same year, Huawei Mate10 with the Unicorn 970 was launched, becoming the first smart phone equipped with an AI chip on the market. Shen Jian Technology Shen Jian Technology Co., Ltd. was founded in 2016 by four scientists and technicians who graduated from Tsinghua University and Stanford University. It is an artificial intelligence technology company that focuses on deep learning processor solutions and is committed to deep learning of processor and compiler original technologies. R & D, so that all algorithm developers can easily use. In October 2017, Shenzhen Semiconductor announced its chip plan. According to official information, the chips “Kingtao” and “View of the Sea” independently developed by SZK will be launched in the third quarter of 2018. Among them, the “Kentao” series chips adopt TSMC's 28nm manufacturing process, and the core uses their own deep knowledge. Aristotle architecture, peak performance 1.1 watts 4.1 TOPS, is expected to complete the product loading in the first half. In addition, Sage Technology has also announced the Deep Neural Network Development Toolkit “DNNDK”, a deep learning software development kit designed to simplify and accelerate the development of deep learning applications. This product not only greatly reduces the difficulty of DFU platform deep learning application development and deployment, but also greatly shortens the cycle of algorithm deployment to hardware, and achieves the ultimate acceleration of product development process in the AI ​​industry. Shen Jian Technology CEO Yao Song said that DNNDK is the first deep learning and development toolkit in China and can be used to benchmark the Husband TensorRT. Due to its outstanding advantages in the AI ​​field, in October 2017, Shenzhen Shen Technology announced the completion of the latest round of US$40 million in financing, led by Ant Financial, Samsung Electronics, China Merchants Venture Capital, and China Venture Capital. Previously, the company completed the Angel Round, A round of financing in May 2016 and May 2017, respectively. Horizon Robot In 2015, Yu Kai, a former dean of the Baidu Research Institute, founded Horizon Robotics and opened the way to research highly challenging processor chips based on deep neural network algorithms. Two and a half years later, Horizon Robots announced on December 20, 2017 that it will launch two of its embedded artificial intelligence vision chips for smart driving and smart cameras. Among them, the smart driving-based embedded artificial intelligence vision chip is named "Journey 1.0 processor". The chip supports the high-performance L2ADAS system; it can simultaneously be used for pedestrians, motor vehicles, lane lines, traffic signs, and traffic lights. 8 types of targets for real-time detection and identification. The camera-oriented vision chip is named "Sunrize 1.0 processor," which incorporates the horizon deep learning algorithm, which can be widely used in smart security, smart cities and other scenes. The two processors use the data flow model of Attention Engine + Cognition Engine to increase the computation speed of the chip by more than 10 times. Yu Kai also revealed that Horizon will launch the next generation of self-driving journey processor in 2018. In October 2017, Horizon Robots completed nearly a billion U.S. dollars in A+ round of financing. This round of financing was led by Intel Capital. Harvest investment was jointly invested. Other investors include current shareholders Morning Star Capital, Sorghum Capital, Shuanghu Investment, and Linear Capital. . Bit continental According to public information, Bitland was established in 2013 and currently has more than 1,000 employees worldwide, located in Beijing, Taiwan, and North America. The company currently has billions of chip shipments and has established billions of computing data centers worldwide. In November 2017, Bitland announced the company's artificial intelligence brand SOPHON ("Yifeng"), and brought the world's first tensor acceleration calculation chip BM1680, and SC1/SC1+ board, smart audio and video analytics server SS1 and other products. . It is reported that BM1680 is a special custom chip for tensor calculation acceleration processing for deep learning applications, and is suitable for reasoning prediction and training of deep neural networks such as CNN, RNN, and DNN. According to the plan, BitContinent will release the second-generation AI AI chip BM1682 in 2018, and its computing capacity will be greatly enhanced. The third and fourth generations will be successively released. Tang Weiwei said that according to the plan every 9 months or so it will launch a new computing abundance AI chip, and another generation or two, Bitland is confident in deep learning reasoning than the actual performance of the GPU. In September 2017, Bitland was revealed to raise 50 million U.S. dollars from venture capital companies such as IDG Capital and Sequoia Capital to increase its position among mainstream investors. Thinkforce According to Thinkforce's official website information, in 2017, the company was founded by senior experts in chip design, algorithm software, and system development. The company is determined to integrate first-class AI algorithms and advanced manufacturing processes to build an artificial intelligence hardware platform. Provide one-stop industrial application solutions. In December 2017, Thinkforce announced the completion of A-round financing of 450 million yuan from Yitu Technology, Yunfeng Fund, Sequoia Capital, and Gaochun Capital. At the same time, Thinkforce also announced the characteristics of its planned AI chip: based on the industry's advanced semiconductor process technology, the use of self-developed micro-kernel ManyCore architecture, to complete the AI ​​cloud virtualization scheduling at the chip level. This architecture improves the elastic computing and scheduling of AI cloud by an order of magnitude, and CPU-like virtualization brings multiple cost savings to the elastic scheduling of cloud computing. At the same time, this technology, combined with self-developed firmware and TFDL software SDK, can accelerate the calculation of various neural network models, and can achieve 5 times more power consumption and cost savings than Huida's mainstream computing cards. With the dual stimulation of capital and technology, the future of China's AI chip industry will be incalculable.