Author: Zihao
Recently, AgiBot, a leading enterprise in embodied AI, has made a strategic investment in ChipV Intelligent Technology, a Tsinghua-affiliated RISC-V AI chip company. In the current industry context, this investment is not merely a capital transaction between two companies, but rather a deep binding between a leading humanoid robot OEM and a domestic underlying computing power enterprise. It also positions ChipV Intelligent Technology, which has secured a foothold in both automotive-grade chips and humanoid robot computing, at the intersection of two high-growth tracks.
01. ChipV Intelligent Technology Forges a 'Chip' Path for Embodied AI
As the humanoid robot industry advances rapidly and the pace of intelligent iteration accelerates, the robot's demand for computing power continues to surge. A humanoid robot is equipped with dozens of joints, requiring not only millisecond-level synchronous motion control but also simultaneous environmental perception and edge large model inference tasks. This imposes extremely stringent comprehensive requirements on the chip's real-time response capability, computing power scale, power consumption performance, and data security and privacy.
Globally, embodied AI chips have mainly evolved into two mature routes. One is Tesla's self-developed FSD chip, which serves only its own robot products; the other is the general-purpose computing platform represented by NVIDIA Jetson, which is widely adopted by numerous robot manufacturers. However, general-purpose platforms are not tailor-made for humanoid robot scenarios and face objective practical limitations in terms of cost and supply chain. Against this backdrop, ChipV Intelligent Technology provides a third viable route, relying on full-stack self-developed RISC-V technology to build a dedicated computing foundation for humanoid robots.
ChipV Intelligent Technology fully leverages the inherent advantages of the RISC-V architecture, enabling on-demand expansion of AI and vector computing-related instruction sets, and flexible hardware modular tailoring. Combined with advanced Chiplet packaging technology, it breaks through the physical ceiling of single-chip computing power, achieving a balance among performance, cost, and usage flexibility. In contrast, for other architectures, if ARM is to simultaneously accommodate hard real-time motion control and AI computing power, it will lead to a significant increase in cost; the x86 architecture has high power consumption and is also difficult to adapt to the portable usage conditions of robots.
AgiBot's increased investment in ChipV Intelligent Technology provides realistic soil for the implementation of this brand-new RISC-V technology route. Founded by former Huawei 'Genius Youth' Peng Zhihui and his partners, the company is a first-tier player in the humanoid robot track, with its two major product lines, Yuanzheng and Lingxi, having achieved large-scale mass production and delivery. Backed by top-tier capital such as Hillhouse and Tencent, it initiated the process of listing in Hong Kong in July this year, targeting a valuation of CNY 40 billion to CNY 50 billion. AgiBot is not satisfied with merely focusing on the iteration of whole-machine products, but chooses to take root downwards and layout the entire embodied AI industry chain, with large models, whole robots, and underlying computing chips all within its layout. This strategic investment in ChipV Intelligent Technology is a key step to complete the underlying computing puzzle, seeking a deeply customizable computing foundation for the next generation of humanoid robots.
For ChipV Intelligent Technology, what AgiBot brings is not just capital. Real business scenarios at the robot system level and massive operational test data can drive the rapid iteration and optimization of chip products in return; it also opens up channels for industrial implementation, accelerating the transition of robot chips from laboratory samples to mass production verification. Coupled with the continuous supply of technical talents from Tsinghua University and the automotive industry cluster resources brought by Anhui state-owned capital, the company holds dual momentum: a solid foundation in automotive-grade business and a second growth curve in robotics.
02. Building a Complete Product Matrix for Robot Computing Power with Integrated Hardware and Software
Backed by Tsinghua University's rich scientific research resources, ChipV Intelligent Technology is also one of the few RISC-V chip solution providers in the industry capable of covering all levels of humanoid robots, from end-effector execution, motion hub, cerebellum control, to brain decision-making. At the product level, the XC64AE series is tailor-made for embodied AI scenarios. The XC64AE, serving as the 'cerebellum chip,' adopts advanced processes, is equipped with a 12-core RISC-V core, and paired with a self-developed NPU, balancing robot motion control and mid-tier environmental perception computing power. The high-end positioned V1000 'brain chip' achieves 1000 TOPS computing power relying on a 2.5D Chiplet architecture. At this stage, robot joint control chips have achieved mass production, and the XC64AE is also conducting joint verification with several leading domestic embodied AI companies. Currently, the first cerebellum chip has completed tape-out and bring-up, simultaneously launching a standardized development solution; in 2027, the second-generation XC64AE full series will complete tape-out, further compressing motion control latency to meet the stringent conditions of millisecond-level coordinated operation of multiple joints across the whole body. Subsequent products will also extend outward, covering more scenarios such as industrial collaborative robots and medical rehabilitation robots.
Beyond chip hardware, ChipV Intelligent Technology also attaches great importance to the construction of the software ecosystem. The company has independently developed a full set of toolchains, SDKs, and real-time operating systems, while simultaneously laying out the embodied AI model ecosystem. In collaboration with a Tsinghua Ph.D. team, it has incubated the 'Xuantong' vertical large model, forming an integrated solution of 'chip + software + data + algorithm' to make up for the supporting shortcomings of hardware and software collaboration.
03. Mastering the Embodied AI Track Stems from the Expertise in Automotive-Grade Chip Mass Production
Notably, ChipV Intelligent Technology's high-reliability engineering capability for robot chips was not explored from scratch; a large part of it originates from its already commercialized automotive-grade chip business. Automotive electronics is also an important track for RISC-V to demonstrate its advantages. The comprehensive licensing cost of traditional ARM automotive-grade chips can be as high as 15%-20% of the product cost, which is a fixed rigid expenditure. In contrast, the characteristics of RISC-V—open-source, license-free, modular, and flexibly adjustable—are highly adaptable to the needs of numerous in-vehicle scenarios such as body control and chassis domain control.
ChipV Intelligent Technology's XC32 series automotive-grade chips fully cover various in-vehicle scenarios including body, powertrain, chassis, and centralized domain control. The XC32MC body motor control MCU has achieved mass installation in vehicles, targeting low-voltage motor control for body components such as windows, seats, and lights, holding total orders of over 5 million units. The XC32LS complies with the highest ASIL-D functional safety level, adapting to high-safety scenarios such as powertrain and chassis braking and steering. It has recently completed tape-out and bring-up, and is entering the testing phase of multiple automakers and Tier 1 customers, with plans for official mass production in 2027. The XC32DC central domain control chip, built with a self-developed NPU, targets the new generation of centralized electronic and electrical architecture for whole vehicles, and is expected to achieve designated mass production in 2028. According to the company's plan, the annual shipment volume of automotive chips will advance towards the tens of millions level in 2028. From the perspective of product installation, automotive-grade chips have currently entered the supply chains of multiple mainstream automakers and component enterprises, completely running through the entire process from IP definition, functional safety certification, tape-out, testing to whole-vehicle delivery, and product stability has been tested in real industrial scenarios.
The high-reliability chip design capability forged through the stringent requirements of the automotive industry has also become a unique advantage in exploring the robot track. The underlying logic of vehicle chassis domain control and robot joint control is identical, both requiring hard real-time response and multi-node synchronous control. Mature automotive-grade design experience can be directly reused in robot chip R&D, significantly reducing trial-and-error costs. The company's performance will enter a stage of rapid growth, and the automotive-grade business, as a stable foundation, also provides sufficient cash flow support for the long-term R&D of robot chips.
04. Ten Years of Accumulation Rooted in Tsinghua's Technical Soil
The confidence to advance the robot business is rooted in the profound Tsinghua industry-university-research system behind ChipV Intelligent Technology. ChipV Intelligent Technology is the only RISC-V AI chip enterprise with technology transfer and shareholding from Tsinghua University, and its technical foundation originates from Tsinghua laboratories. As early as 2016, the core team started research related to the RISC-V architecture. Relying on three major platforms—the High-Performance Processor Laboratory of Tsinghua University, the Automotive-Grade Chip Laboratory of the Shanghai Tsinghua International Innovation Center, and the RISC-V Intelligent Chip R&D Center of the Tsinghua University Shenzhen Research Institute—it has completed the accumulation of full-stack capabilities from underlying IP and chip architecture design to solution output, undergoing ten years of polishing.
In 2026, Tsinghua Huakong Technology Transfer Co., Ltd. made a strategic investment in ChipV Intelligent Technology by valuing core RISC-V intellectual property, becoming a key move for Tsinghua to layout the RISC-V automotive-grade and robot computing tracks. The enterprise has also officially joined the Tsinghua AI hard-tech industry matrix composed of Zhipu AI, Dongfang Suanxin, Qingwei Intelligent, and Infinigence AI. Universities and enterprises are no longer simply engaged in one-time technology licensing, but have formed a two-way symbiotic model: universities continuously output core patents, R&D talents, and industry standard resources; enterprises complete the engineering implementation of technology, and then feed back the data from real working conditions of whole vehicles and robots to the laboratories, feeding back the continuous iteration of the next-generation architecture.
He Hu, Chief Scientist of ChipV Intelligent Technology, comes from the School of Integrated Circuits at Tsinghua University and is also an expert who has deeply engaged in systematic RISC-V research in China at an early stage. He has led major provincial and ministerial projects on high-computing-power intelligent chips, spearheaded the country's first open-source GPGPU 'Chengying' project, and deeply participated in the compilation of multiple group standards for RISC-V and automotive-grade chips, accumulating rich scientific research and engineering experience in processor architecture and high-reliability chips. After joining ChipV Intelligent Technology, he introduced his years of technical accumulation into the enterprise, leading the technical route and product definition of full-stack RISC-V chips, and promoting the migration of verified automotive-grade high-reliability technology to embodied AI scenarios.
Beyond scientific research capabilities, the enterprise has also equipped itself with a mature industrial management team. Founder and CEO Cheng Baozhong, who has a dual background in semiconductor industry operation and investment, is paired with CTO Wei Jian, who formerly worked at Qualcomm. Wei Jian led Qualcomm's first-generation self-developed GPU and completed the tape-out and mass production of ASIL-D level RISC-V cores and automotive-grade cockpit chips. The core management team, composed of scientific research experts, industrial operation talents, and chip engineering experts, coupled with the continuous R&D power delivered by Tsinghua, enables the enterprise to balance the height of academic innovation with the strength of industrial mass production.
05. Conclusion
Looking back at the development context of the domestic RISC-V industry, the period from 2018 to 2023 belonged to the industry incubation period, with the core goal of verifying the feasibility of the architecture; from 2024 to 2025, IoT and home appliance MCUs achieved batch shipments; stepping into 2026, industry differentiation has become increasingly obvious, with involution intensifying in low-end tracks, while high-value tracks such as automotive-grade and embodied AI still face supply gaps.
Looking ahead to the next few years, intelligent vehicles and humanoid robots will become the markets with the most robust growth momentum. The differentiated competitiveness of ChipV Intelligent Technology lies in the mutual reuse of the technical foundations of these two major tracks. The high-reliability capability accumulated in automotive-grade empowers robot chips, while the frontier exploration of computing power and real-time performance in robot scenarios, in turn, drives the iterative upgrade of the new generation of vehicle domain control chips.
Throughout the process of tracking the enterprise, we have also witnessed ChipV Intelligent Technology's hardcore technical capabilities and industrial connection capabilities. The enterprise has deeply participated in the formulation of multiple industry standards, including the China Automotive Technology and Research Center automotive RISC-V chip standard, the China Electronics Standardization Institute RISC-V GPGPU extended instruction set, and the China Automotive Chip Innovation Alliance RISC-V automotive-grade chip technology ecosystem.
The PC era bred the x86 ecosystem, the mobile era bred the ARM ecosystem, and the AI and intelligent terminal era is breeding a new RISC-V ecosystem. Since its establishment in August 2023, ChipV Intelligent Technology has focused on the high-speed development of the RISC-V track. From the product side, product R&D, large-scale mass production, and application implementation make the technology visible; from the application side, taking the lead in cutting into top customers makes the strength visible; from the financing side, from early investments by Tsinghua-affiliated capital such as Shuimu Tsinghua Alumni Fund and Qidi Zhixing Venture Capital, to the strategic investment by Tsinghua Huakong Technology Transfer Co., Ltd. this year valuing core RISC-V intellectual property, and then to AgiBot's strategic investment in October, the potential is visible. Relying on the source innovation of Tsinghua's industry-university-research system, this young enterprise has been moving forward pragmatically all the way, and has also been favored by the industry and investment circles! A great journey begins with a single step, and a boundless future is in sight!