Focusing on the design of integrated storage and computing chips, Zhicun Technology completed nearly 100 million yuan in A3 round of financing

On June 10th, Zhicun Technology, the global leader in integrated storage and computing chips, which was continuously reported by Entrepreneur State, announced that it has completed a billion-yuan A3 round of financing. Capital and other investors followed suit, with old shareholders Kexun Ventures, SMIC Juyuan, PricewaterhouseCoopers, and China Merchants Ventures continuing to follow suit, and Index Capital served as the exclusive financial advisor. The financing funds will be used to increase investment in technology research and development, product line expansion and mass production of new products.

Prior to this, Zhicun Technology had received A1 and A2 rounds of financing led by SMIC and SDIC Ventures respectively. Up to now, Zhicun Technology has completed a series of A-round financing of nearly 300 million yuan.

Zhicun Technology was established in October 2017, focusing on the design of integrated storage and computing chips. The team has developed a storage-computing integrated chip for 8 years, and successfully taped out the first international analog storage-computing integrated deep learning chip in 2016, laying the foundation for breaking through the bottleneck of von Neumann’s architecture.

With Moore’s Law approaching its limit, the cost of chip development has increased, and performance improvements have been limited. The problem of “storage walls” is particularly prominent: data transfers between computing units, caches, and memory, occupying more than 90% of power consumption and time, resulting in high power Problems such as power consumption and low efficiency have led to the inability to further improve the computing power of the chip, and the problems in artificial intelligence calculations are more prominent.

Artificial intelligence is implemented on a large scale in some easy-to-implement scenarios, and more scenarios have more stringent requirements on computing power, cost, and power consumption. The integration of storage and computing can effectively solve many restrictions on computing power, power consumption, and cost.

Under this situation, Zhicun Technology has made a breakthrough in using Flash memory to complete the storage and calculation of the neural network, solve the problem of AI storage wall, and improve the efficiency of calculation. Compared with the deep learning chip based on the von Neumann architecture, the computational efficiency is increased by about 10-50 times; a single Flash array can complete 1 million matrix multiplication and addition operations in parallel, and the computational throughput is 100-1000 higher than the bandwidth of DRAM and SRAM. Times. It can exceed the computing efficiency of advanced technology under mature technology, greatly reducing development and tape-out costs. In addition, Zhicun Technology has accumulated experience in the entire process of storage and calculation from design, simulation technology to mass production, and the efficiency of product development has increased by 5 times compared with the past.

After several years of development, Zhicun Technology has launched two chip products of the WTM series, covering many usage scenarios such as smart wearable devices, smart homes, and smart cities. Support low-power multi-command word recognition, noise reduction, voiceprint recognition, etc., built-in 2MB deep learning network parameter storage space, which can store and calculate up to 40 layers of multiple (same or different) deep learning network algorithms at the same time, knowing and storing Science and Technology has launched the first storage and calculation integrated compiler, WITIN Mapper, which supports multiple network structures such as DNN/RNN/LSTM/TDNN.

Wang Shaodi, founder and CEO of Zhicun Technology, graduated from the Department of Microelectronics of Peking University with a bachelor’s degree and obtained a doctorate from UCLA in 2017. He is an expert in memory and automation design. CTO Xinjie Guo graduated from the Department of Microelectronics of Peking University and obtained his Ph.D. in 2017. After 4 years and 6 tapeouts, he captured the floating gate storage and computing integrated deep learning chip technology for the first time. In addition, more than 80% of the Zhicun Technology team are R&D personnel, with professional academic backgrounds from well-known domestic and foreign universities such as University of California, University of Edinburgh, Tsinghua University, Peking University, National University of Defense Technology, Zhejiang University, Beijing University of Aeronautics and Astronautics, as well as IBM, Global Foundries, AMD, Qualcomm, Global Foundries, RDA and other well-known semiconductor companies have a working background and have rich experience and resources in chip design, research and development, mass production and sales.

Based on the development and cutting-edge research of integrated storage and computing technology, Zhicun Technology has successively undertaken a number of government scientific research projects such as the “Cultivation of New Generation Artificial Intelligence Technology” and “Cultivation of New Generation Information and Communication Technology” by the Beijing Municipal Science and Technology Commission; relying on product innovation and broadness In terms of market prospects, Zhicun Technology was rated as the most influential innovative company in the AI ​​chip field in 2019, and KPMG’s 2020 “Core Technology” 50 list of emerging companies and other honorary awards.

Facing the future, Zhicun Technology will continue to optimize the underlying technology to make the storage-computing integrated chip more efficient and reliable, and use it in consumer Electronic devices on a large scale.

Photool Ventures stated that Zhicun Technology is a very good technology company. The founding team is one of the first teams to study the integration of Nor Flash storage and computing. Market demand. The storage-calculation integrated architecture developed by it is hundreds of times that of the traditional architecture in the field of deep learning inference, and it perfectly solves the problems of “storage wall” and “power wall”. Photool Ventures has been deeply involved in the field of science and technology for many years and believes that Zhicun’s products are in line with the investment philosophy of Photool Technology to change the future. Therefore, it is also very fortunate to be able to participate in this financing and accompany Zhicun to grow together on the road in the future.

SDIC Venture stated that Zhicun Technology is an internationally leading company in the field of integrated storage and computing chips, and is an important layout of SDIC Ventures in the frontier field of artificial intelligence chips. The company has successively undertaken a number of government scientific research projects such as “Cultivation of a New Generation of Artificial Intelligence Technology” and “Cultivation of a New Generation of Information and Communication Technology” by the Beijing Municipal Science and Technology Commission. Represents meaning.

Attached

Zhicun Technology’s financing history:

Focusing on the design of integrated storage and computing chips, Zhicun Technology completed nearly 100 million yuan in A3 round of financing

Source: Entrepreneurship-Rui Beast Analysis

Author: Lan Chenxi

The Links:   2RI100E-060 NL6448BC20-30 IGBT-module

Related Posts