According to a report by Electronic Enthusiasts Network (Author: Huang Shanming), against the backdrop of the continuously heating AI toy market, numerous manufacturers are vigorously launching a series of new chips and semiconductor device products, aiming to provide stronger technical support for the intelligent upgrade of AI toys.Among them, SoC is an important part, and many manufacturers have recently introduced SoC products suitable for AI toys, each with its own characteristics.
Zhongke Lanyun AB6003G Wi-Fi Chip
Since its establishment in 2016, Zhongke Lanyun has focused on the research and development of wireless audio SoC chips. By 2023, Zhongke Lanyun has clearly expanded into the IoT and AI edge market, increasing its product line from 8 to 10, adding Wi-Fi chips and video chip directions. The recently launched AB6003G Wi-Fi chip is one of the first results of this strategy.AB6003G can be applied in smart homes, wearable devices, and AI toys, integrating Wi-Fi 6, Bluetooth 5.3, and an audio processing unit, providing a single-chip solution for wireless connectivity, audio codec, and AI interaction needs.This chip integrates IEEE 802.11 b/g/n baseband and RF circuits, including power amplifier (PA), low noise amplifier (LNA), RF balun, antenna switch, and power management module. Among them, the Wi-Fi baseband implements OFDM technology and is backward compatible with DSSS and CCK technologies, supporting IEEE 802.11 b/g/n protocols. It supports a standard bandwidth of 20MHz, providing a maximum physical layer rate of 72.2Mbit/s.At the same time, the chip also integrates a high-performance 32-bit RISC-V processor, providing rich peripheral interfaces; it supports RTOS and third-party components, and provides an open and easy-to-use development and debugging environment.This self-developed AB6003G chip not only features low power consumption and high performance but also achieves localized computing capabilities through deep optimization of the neural network accelerator, allowing toy products to quickly respond to various user interaction needs. Meanwhile, the company has deep cooperation with platforms such as Baidu Smart Cloud and Volcano Engine, bringing the powerful capabilities of large models into toy terminals, endowing products with richer educational content and more natural interaction experiences.AB6003G is a highly integrated, low-power, and strong AI-capable Wi-Fi SoC launched by Zhongke Lanyun for AI toys and IoT scenarios. Through a complete ecosystem of “chip + large model + content,” it helps traditional toy manufacturers quickly achieve intelligent upgrades, making it one of the most competitive chip solutions in the current AI toy track.
Zhuhai Taixin TXM61x
Founded in 2015, Zhuhai Taixin focuses on wireless communication chip design, initially centered on basic connectivity chips such as Wi-Fi 4/5 and Bluetooth 4.2/5.0, gradually accumulating RF, baseband, and protocol stack technologies to lay the foundation for subsequent AIoT chip development.By 2022, Taixin started RISC-V architecture chip development, targeting edge computing and AI at the edge market, planning to launch SoCs that integrate wireless connectivity, audio and video codec, and AI acceleration.The recently released TXM61x is positioned as an edge audio and video AI SoC, integrating a 32-bit RISC-V microprocessor with a clock speed of up to 384MHz, supporting DSP instructions and single-precision floating-point instructions, primarily targeting AI toys, smart locks, and children’s cameras.It has a built-in NPU that supports AI functions such as face recognition, gesture recognition, and voice recognition. It also integrates a high-performance ISP and H.264/MJPEG codec, supporting multi-channel video stream synthesis, layer mixing, and 2D image acceleration to meet high-definition video processing needs.Additionally, it supports built-in/external PSRAM and Flash, providing an open development and debugging environment.The launch of TXM61x also reflects Zhuhai Taixin’s transformation from basic wireless chips to RISC-V + AI SoCs, gradually becoming an important force in domestic edge intelligent chips through technological accumulation, ecological cooperation, and scene implementation, especially showing significant market potential in AI toys and smart home fields.
Feiling Micro A1 Chip
Feiling Micro is a wholly-owned subsidiary announced by Sitwei in 2024, primarily focused on the automotive vision processing chip track, aiming to synergize with its parent company’s CMOS image sensors to create a full-stack visual solution of Sensor + ISP + AI SoC.By this year, Feiling Micro has released its first edge visual AI SoC for AIoT applications—the A1, which can adapt to CMOS image sensors of various resolution specifications, forming a system-level combination of edge AI SoC + Sensor.The A1 is equipped with Feiling Micro’s self-developed AI ISP technology, covering resolutions from 0.3MP to 8MP, with built-in high-performance low-light noise reduction and image enhancement algorithms, supporting up to 3-frame HDR synthesis, with a dynamic range of up to 120dB. It also has a built-in RGB-IR ISP module, capable of synchronous output of RGB and IR image formats, and supports up to 2 image inputs and synchronous processing, adaptable to various specifications and shutter types of image sensors.The A1 AI SoC chip is compatible with 0.3MP-2MP global shutter CMOS image sensors, flexibly forming various monocular/binocular intelligent visual perception module solutions. It can also be paired with 4MP-8MP night vision full-color CMOS image sensors to form high-performance night vision module solutions.In addition, the A1 chip is equipped with a self-developed NPU with 0.8TOPS@INT8 lightweight computing power, supporting mainstream CNN architectures in the industry, and optimizing lightweight neural network structures and visual tasks.In terms of packaging, the A1 offers two small package size versions: 9mm*9mm (QFN) and 7mm*7mm (BGA).As the first edge visual AI SoC launched by Feiling Micro, the A1 focuses on 0.8TOPS NPU, self-developed AI ISP, and low-power small packaging as core selling points, quickly entering the AI toy, smart home, and industrial vision tracks. In the future, with deepening ecological cooperation, the A1 is expected to further expand into high-end vision fields such as automotive and medical.
Current Status of Domestic SoC Chip Development
From the perspective of the core SoC for AI toys, domestic production has basically been achieved. For example, the leading product ByteDance’s “Eye-catching Bag” uses the Espressif ESP32-S3, the Tom Cat AI Companion uses the Allwinner R128-S3, and BubblePal uses the Rockchip RK3576, all of which utilize domestic chips.At the same time, the hardware cost of domestic solutions is 30%-50% lower than that of international competitors, for instance, the unit price of Wi-Fi modules is controlled at 1-2 USD, driving the price of AI toys down to the hundred-yuan market.It is worth noting that many companies have begun to master the full-stack capabilities from instruction set to SoC architecture, such as Allwinner, Taixin, and Saifang Technology, promoting the standardization of the RISC-V instruction set in the AI toy field, reducing dependence on ARM licensing and minimizing reliance on ARM architecture. Moreover, SMIC’s current mature 14nm/28nm production lines are sufficient to support the mass production of domestic SoCs.Looking at future technological development directions, some companies are also exploring near-memory computing technology to improve energy efficiency ratios to over 10 Tops/W, in response to the localization needs of large models. At the same time, emerging demands such as emotional companion robots and educational programming toys will drive customized development of SoCs, with the AI toy-specific SoC market expected to exceed 5 billion yuan by 2025.
Conclusion
In the domestic AI toy field, domestic SoCs have achieved a leap from usable to user-friendly, possessing advantages in cost control and rapid response to market demands, but still need breakthroughs in core technology areas such as high-end computing power and multi-modal integration. In the future, as the RISC-V ecosystem matures and the integration of storage and computing technologies is realized, domestic SoCs are expected to occupy a more important position in the global AI toy industry chain.
Disclaimer：This article is original by Electronic Enthusiasts Network，please indicate the source above when reprinting. For group communication, please add WeChat elecfans999,for submission, exposure, and interview requests, please send an email to [email protected].
More Hot Articles to Read
-
BYD faces “patent trolls” overseas, fearing a complete ban on sales in this region, and publicly disclosing data
-
Inspirational Queen! This semiconductor executive won the Women’s Singles Championship at the US Table Tennis Grand Slam!
-
RK chip exploded! Will the next generation of chips continue the myth?
-
Order surge of 800%! New wind in consumer chips: Northeast air conditioning installation
-
Zongshen names “picking a fight” with Zhijie and Wenjie! The anxiety of the three-wheeled vehicle leader’s electrification transformation
Click to Follow and Star Us
Set us as a star, and don’t miss any updates!