After HiSilicon's market share slipped, a gap has been left in the video encoding chip market: Boya Vision, with the world's first AVS3 end-side chip, seizes the window period for independent standards.
Boyar Vision enters the video data storage and domestic substitution market with its independent AVS3 visual intelligent computing chip, and is currently advancing a new round of financing
Storage cost of video data and pressure from coding standards
Video surveillance data is becoming one of the fastest growing digital assets, with the costs of storage, transmission, security compliance rising accordingly. However, the encoding and decoding chips supporting this link have long relied on overseas standard systems and a small number of suppliers. In December 2024, Beijing Boyar Vision Technology Co., Ltd. released SPARK RE3200, the world's first end-side visual intelligent computing SoC chip supporting China's independent public security audio and video intelligent coding standard AVS3-S/SAVIC, which attempts to respond to the pressure of both cost and supply chain with independent standards and self-developed chips. This enterprise incubated from the National Engineering Research Center of Video and Visual Technology of Peking University is currently pushing forward a new round of financing.
The video surveillance industry has gone through the stages of analogization, networking and high-definition since 1999, and has now entered the stage of intelligent upgrading. After integrating AI modules into SoC chips, terminals can carry out structured processing and real-time analysis on video data, enabling cameras to shift from passive recording to active identification, but the expansion speed of data volume far exceeds the expansion pace of storage and bandwidth. According to IDC's forecast, the total global data volume will increase to 181ZB by 2025, among which video surveillance data is one of the fastest growing fields; superimposed with the requirements of full coverage of video surveillance, high definition and long-cycle storage, the industry is generally faced with practical problems of high storage cost, high bandwidth pressure and high threshold for intelligent transformation.
Compliance and authorization constitute another layer of pressure. Public security video surveillance needs to meet national standards such as GB35114 and GB/T28181, which involves national cryptographic algorithms and full-process data security, while overseas coding standards have complex authorization rules and uncertain supply chain stability. Public information shows that H.265 (HEVC) has formed a multi-party patent pool pattern including MPEG LA, HEVC Advance and Velos Media. In June 2025, Velos Media also filed a lawsuit against TikTok for H.265 patents. In contrast, China's independent AVS3 standard only charges a symbolic patent fee of 1 yuan per unit for terminal equipment in terms of patent policy, and does not charge content providers; tests by the International Digital Video Broadcasting Organization (DVB) show that AVS3 saves more than 40% of the bit rate than H.265 under the same image quality.
Changes in policy and industrial pattern further expand the opportunity window. AVS3 has achieved large-scale implementation in the radio and television field, and its extension to industrial standards such as security is advancing, while pilots are also being carried out in fields such as water conservancy and emergency response. At the same time, after Huawei was sanctioned, HiSilicon's original share of more than 90% in China's video encoding chip market dropped to less than 30% in 2025, and the market gap continues to expand.
From standards to chips, the key capabilities and business landing points of RE3200
Boyar Vision has participated in the technical verification of AVS series standards since 2017, providing a number of core coding algorithm proposals for the standards; in 2019, it launched the AVS3-based video transcoding prototype system to support the initial test of CCTV 8K trial channel; in 2021, its Realscene Live video transcoding system became the first device in China to pass the AVS3 network access certification of the National Radio and Television Administration. Long-term accumulation of standards and algorithms has become the foundation for the company to extend to the chip sector.
SPARK RE3200 adopts the self-developed R-VICA visual intelligent computing architecture, integrating CPU, video, audio, security subsystem and AI subsystem based on RISC-V architecture, supporting multi-core heterogeneous parallel processing. The chip has a built-in NPU with 1.5 TOPS computing power and 512 MAC units, supports encoding standards such as AVS3, AVS3-S, SAVIC, H.264 and H.265, supports national cryptographic SM2/SM3/SM4 algorithms, and is compatible with GB35114 and GB/T28181 public security video surveillance standards. In terms of power consumption, the peak power consumption of the chip is less than 1.5W, the standby power consumption is less than 50μW, and the deep sleep mode is less than 25μW, which can adapt to passive video terminals sensitive to battery life. The chip design was finalized in August 2024, after 128 days of tape-out, it won the "Product Gold Award" at the 2024 Global Artificial Intelligence Application Expo.
Around this chip, Boyar Vision has formed three business lines. For the ultra-high-definition video sector, it provides broadcast-level transcoding, decoding, scheduling and quality detection equipment, serving the main station 8K channel and event live broadcasts such as Spring Festival Gala, Hangzhou Asian Games, Paris Olympics, 2026 FIFA World Cup in USA, Canada and Mexico. According to public information of the enterprise, the bandwidth cost has been reduced by about 40% on platforms such as CCTV.com and Migu Video. For the intelligent management of video data sector, with intelligent compression as the core, it achieves 5 to 50 times compression on the premise of unchanged coding format, resolution, frame rate and duration, lossless human visual perception and lossless machine vision, which has been applied to projects such as Hangzhou Public Security, Beijing Subway and Jinan Subway. According to the disclosure of the enterprise, it can reduce the storage cost by about 89% and the energy consumption of the machine room by about 50%. For the visual intelligent computing chip and intelligent agent sector, with RE3200 and its modules as the carrier, it exports end-side solutions to scenarios such as security surveillance terminals, intelligent compression gateways, law enforcement recorders and AI toys.
Team background, implementation progress and market space
The core team of Boyar Vision is built relying on the National Engineering Research Center of Video and Visual Technology of Peking University, led by Gao Wen, academician of the Chinese Academy of Engineering and head of the AVS Working Group. Wang Shanshe, founder and chairman of the company, is an associate researcher of the National Engineering Research Center of Video and Visual Technology. The key technology of multi-dimensional primitive encoding and decoding for ultra-high-definition video he participated in won the first prize of 2020 National Technology Invention Award, and in July 2026, he won the second prize of National Science and Technology Progress Award for the project of "Key Technology of Visual Perception Encoding and Decoding and Industrialization Application of Ultra-high-definition Video". Han Wei, general manager of the company, graduated from Guanghua School of Management of Peking University and has been starting businesses continuously in the audio and video field for more than 20 years. The R&D personnel of the company account for nearly 70%, with its headquarters located in Zhongguancun Software Park, Beijing, and R&D centers set up in Shanghai, Shenzhen, Hangzhou and Chengdu.
In terms of industrial cooperation and verification, Boyar Vision has established cooperation with institutions such as China Institute of Water Resources and Hydropower Research, Information Research Institute of the Ministry of Emergency Management, China Digital Culture Industry Group, Shandong Port Group and other institutions. In January 2026, Shandong Port, together with the National Engineering Research Center of Video and Visual Technology of Peking University, Boyar Vision and other units, released the world's first port AI chip "Shangang Smart Chip · Xingyu SA5200", targeting scenarios such as remote control of container cranes and offshore ship navigation, to alleviate the problems of limited transmission bandwidth and poor image quality; in June 2026, the "Half-megabit High-definition" intelligent surveillance terminal solution equipped with RE3200 and its modules passed the GB35114-2017 public security video surveillance networking information security certification. In terms of production capacity and ecological layout, the company is building an ultra-high-definition video and visual intelligent computing chip module project in Mianzhu, Sichuan, which is promoted in two phases. The first phase targets equipment for the radio and television market, and the second phase targets new generation intelligent terminals and visual intelligent computing chip modules, and the company has jointly built the "Joint Innovation Laboratory of Visual Intelligent Computing and Embodied Intelligence" with Peking University.
According to the company's previously public plan, the upgraded chip SPARK RE5200, which targets the mid-to-high-end market and realizes the integration of encoding and decoding, is planned to be launched in 2026. In terms of financing, the company completed the 100-million-yuan A-round financing in April 2024, with the investor being Jiuzi Capital. The funds are used for the construction of AVS industrial ecology and the R&D of encoding and decoding chips; currently, this project is advancing a new round of financing.