Yuanli Semiconductor

Company Profile

原粒（北京）半导体技术有限公司成立于2023年4月，是一家创新的AI Chiplet（芯粒）供应商，致力于通过其独特的“积木式”架构，将Server级别的智能下沉至桌面与边缘设备，为AI Agent(intelligentsia) era to provide cost-effective end-side computing power solutions.

The company was co-founded by Dr. Shaoxia Fang (CEO) and Gang Yuan (COO). The core team is a “special force” from international semiconductor giants (e.g. AMD, Xilinx), with more than 15 years of experience in high-performance processor R&D and completeAI chipEngineering capabilities, having led the mass production delivery of several mainstream AI chips.

Milestones

• April 2023: Prochip Semiconductor was formally established and quickly set up an R&D team consisting of professionals from the world's top chip companies.
• June 2023: Completed tens of millions of RMB seed round financing, with investors including Inno Angel Fund, Zhongguancun Development Group, Tsinghua Venture Capital, Shui Mu Tsinghua Alumni Seed Fund, and Zhongke Chuangxing.
• April 2024: Completed angel round financing amounting to tens of millions of dollars and received strategic investment from investors including CSCI Star, One Dimension Venture Capital, Huafeng Group, Zhongguancun Venture Capital, Inno Angel Fund, and Zero2IPO Venture Capital, among others.
• December 2025The company has completed its Series A round of financing with investors including Shangshi Capital, Capital Science and Technology Development Group, InnoFund, Zhuo Shi Investment, Guangdong Yuan Shi, Leader Ei Fund, and Sanxin Yidi, among others.
• April 2026The company has completed over 500 million yuan of Pre-A round of financing, with a luxurious lineup of capitalists, including international head institutions IDG Capital, semiconductor industry investment institutions Wuyuefeng Science and Technology, and the “national team” of the State New Fund, etc. The company is a member of the State Council of the People's Republic of China.

Products & Services

Yuanli Semiconductor provides energy-efficient, low-cost universal AI Chiplet components and toolchain, allowing customers to flexibly and quickly configure AI chips with different power specifications according to actual business needs, and supporting multi-chip interconnections to expand the power to meet the needs of ultra-large-scale multi-modal model inference and edge-side training fine-tuning. Specific products and services include:

• AI Chiplet: General-purpose high-performance AI processor with adaptive interconnection capability, based on highly efficient training and push integrated architecture, supporting multi-core granule interconnection, which can provide tens to hundreds of TOPS of computing power for edge-side multimodal large-model applications.
• AI chip: Chips co-packed with third-party SoC cores to meet the needs of different application scenarios.
• Edge End Module: Modules designed for edge-side scenarios to provide cost-effective multimodal large model arithmetic support.
• Reasoning Accelerator Card: Multi-core granular large model inference accelerator card to cover the cloud-side end of the large computing power market.
• One-stop solutionIt provides one-stop big model arithmetic solutions including AI Chiplet, AI chip, AI acceleration module, AI acceleration card and other series of products.

core technology

Yuanli Semiconductor's core technologies include a multimodal AI computing core, an AI arithmetic fusion architecture and a framework for rapid AI algorithm deployment:

1. Multimodal AI Computing Core (CalCore™/Calcumex™)::
   • Able to adapt to different AI algorithms, including traditional CNN models and new multimodal macromodels.
   • Provide native arithmetic extension interfaces to support a variety of inference calculation precision to improve algorithm compatibility and adapt to the rapid evolution of algorithms.
   • It flexibly responds to the current development trend of AI algorithm diversity, and is especially optimized for new applications such as multimodal large models.
2. AI Arithmetic Fusion Architecture (CalFusion™/Calcugrid™)::
   • Supports multi-level, flexible and transparent computing core fusion and expansion, enabling stacking and expansion of multiple AI cores.
   • Through the integration of hardware and software innovation, multiple AI cores are presented as a single AI processor, using a transparent and unified programming interface to reduce the difficulty of user development.
   • Provides unique adaptive arithmetic slicing technology to automatically complete arithmetic/task slicing for multi-core systems based on the type of AI algorithms, the connection topology of the cores, and performance goals.
3. Framework for rapid AI algorithm deployment (CalSpeed™/Calcukit™)::
   • Create an efficient and easy-to-use AI algorithm deployment framework and software development environment, providing regular components such as optimizers, compilers, driver and runtime components, performance evaluators, trace debuggers, and more.
   • Supporting a wide range of operating systems and deployment environments, users can use the AI acceleration unit directly through the standard framework API within the popular AI framework for x86 systems, or use the AI acceleration unit in embedded environments through the unified API provided by Original Semiconductor.

development prospect

Yuanli Semiconductor is precisely positioned for the next key node of AI development - end-side intelligence, which has a promising future.

• market opportunityAs AI evolves from a cloud-based tool to a constantly running agent, the need for low-cost, low-latency, high-privacy end-side computing power is growing exponentially. Haragi Semiconductor anticipated this trend three years ago, and its technology path is perfectly aligned with the current market demand for hardware hosts for the “personal digital workforce”.
• Capital Recognition: Founded less than three years ago, the company has completed several rounds of financing, the latest round of Pre-A round of financing amounted to more than 500 million yuan, with investors including IDG Capital, Wuyuefeng Science and Technology, CSCI and other well-known organizations. This fully reflects the capital market's high recognition of its technical strength and business prospects.
• ecological cooperation: The strategic cooperation with industry chain partners such as Chaumo Technology signifies that Original Grain is actively building a localized AI arithmetic ecosystem based on Chiplet, and further broadening the application boundaries of its solutions through the heterogeneous solution of CPU+NPU.

All in all, with its forward-looking strategic vision, experienced core team and disruptive Chiplet technology, Protocell Semiconductor is reshaping the landscape of end-side AI arithmetic, and is expected to become a key arithmetic infrastructure provider in the wave of AI Agent explosion.