Artificial intelligence, particularly deep learning, has experienced explosive growth in recent years, largely due to the long-term accumulation of chip technology. If it weren't for the advancement of chip technology that provided sufficient processing power for large-scale machine learning, there would be no AlphaGo that defeated top human players. Over the past decade, the development of chip technology mainly focused on applications such as communications, multimedia, and smartphones. With the slowdown in these markets, chip technology has gradually shifted towards AI, and the impact of AI on chip development will become even more pronounced.
The era of artificial intelligence is here. Everyone sees the potential and promise of AI. Since 2016, chip companies and internet giants have fully entered the AI chip field. In the highly competitive global AI research landscape, China's presence is growing stronger, and AI has quickly become the hottest area for Chinese startups. From 2012 to the third quarter of 2017, investors poured over $45 billion into more than 200 AI companies. So, in this unprecedented wave, which Chinese chip companies will stand out and become the next big names?
Cambrian, founded in 2016 by Professor Chen Tianshi, focuses on developing core processor chips for intelligent cloud servers, terminals, and robots. The company pioneered a deep learning processor and created the world's first dedicated deep learning chip, significantly improving computational efficiency in AI. Unlike Google’s general-purpose processors, Cambrian chips target deep learning. Simulation experiments show that "Cambrian" achieves performance improvements of two orders of magnitude compared to conventional x86 implementations.
In August, Cambrian completed a $1 billion A-round financing and became the first AI chip unicorn globally. Currently, the Cambrian series includes three prototype processor architectures: Cambrian 1 (Dian Nao) for neural networks, Cambrian 2 (DaDianNao) for large-scale neural networks, and Cambrian 3 (PuDianNao) for various machine learning algorithms. Compared to Google’s benchmarks, Cambrian aims to make devices like cameras, smartphones, and even smartwatches with 1 watt of power as "smart" as AlphaGo, offering better performance and lower power consumption. Industrialization of Cambrian chips is expected to begin soon, with an anticipated listing in 2018. Huawei and the Institute of Computing Technology of the Chinese Academy of Sciences have established a joint lab, and Huawei’s Kirin 970 uses Cambrian technology.
On November 6, Cambrian released three smart processor IP products: Cambrian 1H8 for high-performance visual applications in low-power scenarios, Cambrian 1H16 with high versatility, and Cambrian 1M for terminal AI. It also launched the CambriconNeuWare AI software platform for developers.
China Star Microelectronics introduced China’s first embedded neural network processor (NPU) chip, Star Smart One, which is the world’s first embedded video compression system with deep learning capabilities. It achieves a 98% accuracy rate in face recognition, surpassing human performance. Developed over three years, Star Smart One uses advanced billion-gate IC design and ultra-submicron manufacturing processes, achieving mass production at TSMC.
Horizon Robotics, founded by Yu Kai, former director of Baidu’s Deep Learning Institute, focuses on creating an AI “brain†platform based on deep neural networks, including software and chips for low-power, localized solutions in environment perception, human-computer interaction, and decision-making. Its hardware side will design an NPU (Neural Processing Unit) to support its OS, boosting performance by 2-3 orders of magnitude. Horizon has secured funding from Sequoia and Yuri Milner and is working on a one-stop AI solution to make life easier and safer.
Bitmain, established in 2013, specializes in ultra-high-performance computing and has developed various ASIC chips for Bitcoin mining. It recently launched the SOPHON brand, introducing the BM1680 tensor acceleration chip. The company is now entering the AI industry, aiming to provide cost-effective and power-efficient AI computing through its chip engineering and system design capabilities.
Shen Jian Technology, founded by Tsinghua University researchers, develops Deep Processing Units (DPUs) that offer better performance than GPUs with lower power consumption. Its embedded-side products exceed the Nvidia TK1 in performance while costing about 1/4 of the latter. Server-side products match the Nvidia K40 GPU but consume only 35 watts and cost less than $300.
Yu Sai Technology, based in Shanghai, develops smart driving laser radars and other AI-based sensors. It has partnered with major OEMs and Tier 1 suppliers, aiming to release its PandarGT solid-state lidar in early 2018.
Codelong Technology specializes in computer vision and deep learning, providing AI platforms for product recognition and image analysis. It won the WebVision International Contest and secured a $220 million B-round investment from Softbank China.
Vizum, founded in 2016, focuses on intelligent visual perception, offering SDKs, IPs, and chips for robotics and mobile devices. Its ViEye technology surpasses international competitors in 3D vision and gesture recognition.
Baidu has integrated AI with hardware through DuerOS smart chips, collaborating with Spreadtrum and others. It also introduced the XPU, a 256-core FPGA-based cloud acceleration chip.
Yun Zhisheng is a leading voice interaction provider, offering speech recognition services across multiple industries. It secured a $300 million strategic investment and plans to expand its IoT and AI chip initiatives.
Deep Thinking, founded in 2015, focuses on AI algorithms and hardware, offering FPGA+ARM integration platforms and deep learning ASIC solutions. It has developed the iDeepWise AI robot and continues to innovate in AI chip design.
Westwell Technology, based on neuromorphic engineering, has developed a brain-inspired AI chip with a simulated neural network capable of mimicking millions of neurons. It has raised A-round funding and is expanding its research in AI chip development.
As the AI era progresses, chip innovation becomes crucial. Companies like Cambrian, China Star, Horizon, Bitmain, and many others are leading the charge, driving the semiconductor industry forward with AI as the engine. This collaboration between AI and chip technology is shaping the future of intelligent systems, making them more efficient, powerful, and accessible.
Optical Fiber Fusion Splicer
Optical Fiber Fusion Splicer,Water-Proof Fusion Splicer,Six Motors Fusion Splicer,Fusion Splicer For Trunk Project
Guangdong Tumtec Communication Technology Co., Ltd , https://www.gdtumtec.com