BlueRun Ventures Leads Oversubscribed RMB 100M+ Round in Weina Core Chip | BlueRun Portfolio

Recently, Hangzhou Winicore Electronics Technology Co., Ltd. (微纳核芯), a global leader in compute-in-memory AI chips, completed a strategic Series B financing round exceeding RMB 100 million. The round was led by BlueRun Ventures, with participation from China Fortune-Tech Capital, Jinqiu Fund, Junke Danmu, Shaoheng Investment, and other institutions. This not only demonstrates strong market consensus around the disruptive potential of 3D Compute-in-Memory (3D-CIMTM) — integrating "3D near-memory computing + in-memory computing + RISC-V compute-in-memory" — for AI computing applications, but also injects powerful momentum into Winicore's push to become the world's fastest-to-mass-produce 3D edge AI chip, opening a new chapter in its journey.

BlueRun Ventures stated: "We have made extensive investments in edge AI products across robotics, automotive, consumer electronics, and other sectors, so we are very bullish on the rapid scaling of edge AI hardware. Going forward, edge-cloud collaborative AI infrastructure is an inevitable trend — the hard requirements of edge-side latency and privacy protection are driving demand for edge AI chips. The Winicore team has deep expertise in low-power novel architectures and has now introduced an innovative architecture combining RISC-V with compute-in-memory, earning recognition from numerous industry players and establishing deep partnerships with key suppliers and customers. We are very optimistic that the Winicore team will become a leading enterprise in edge AI chips."

As large models evolve, AI agents will transform from execution tools to decision-making partners, driving an industrial revolution. China's State Council recently issued Opinions on Deepening Implementation of the "AI+" Initiative (State Council Document [2025] No. 11), explicitly setting a target of 70% penetration for intelligent agent terminals by 2027. Leading global tech companies also project 900 billion intelligent agents worldwide by 2035, with computing demand increasing 100,000-fold. Compared to cloud deployment, edge deployment offers advantages in privacy, latency, cost, and reliability, making edge-device-cloud collaboration increasingly essential for large model inference. Research firm Omdia predicts the global edge AI chipset market will grow from $2 billion in 2024 to $16.7 billion by 2028. Large model inference chips face three critical demands: high performance, low power consumption, and low cost. First, high performance means high TPS (Tokens Per Second), which is crucial for inference speed; second, low power enables longer standby time and lower heat dissipation requirements on the edge side, and lower electricity and cooling costs on the cloud side; finally, low cost is foundational for democratizing and equalizing access to AI agent applications. In the era of large AI models, three-dimensional (3D) stacking is the inevitable choice for achieving high TPS performance and has become industry consensus, yet it still cannot solve the "impossible triangle" of high performance + low power + low cost, urgently requiring further technological integration and optimization.

Incubated at the Zhejiang Peking University Information Technology Research Institute, Winicore's team has been researching compute-in-memory (CIM) technology since 2018. Building on "computing in memory (CIM)," they creatively integrated "3D near-memory computing" and "RISC-V and compute-in-memory heterogeneous architecture (RV-CIMTM)," pioneering the 3D compute-in-memory (3D-CIMTM) architecture in 2023 to crack the "impossible triangle" of high performance + low power + low cost. First, the combination of "3D near-memory computing + in-memory computing" achieves chip-level PPA (Performance-Power-Area) advantages. Whether in convolutional networks or large model networks, tensor computation accounts for over 99% of operations, and tensors can be efficiently parallel-accelerated on CIM arrays without simplification. By fusing storage and compute units, this significantly improves compute density (cost-related) and computational energy efficiency (power-related). Multiple tape-outs and testing iterations have shown that compared to traditional von Neumann architectures, Winicore's CIM technology achieves over 4x improvement in compute density (equivalent cost improvement) and over 10x reduction in power consumption. Meanwhile, Winicore's RV-CIMTM full-stack technology (ISA architecture, instruction set, operator library, compiler, toolchain, etc.) addresses the "chip usability" challenge. Leveraging the RISC-V open-source ecosystem and dozens of leading upstream and downstream enterprises, Winicore has achieved end-to-end integration of "instruction set + operator library + simulator + compiler + algorithm adaptation," solving the computational completeness and software-hardware ecosystem barriers of compute-in-memory. Winicore's pioneering 3D-CIMTM chip targets the 3D edge AI chips under development by international leading manufacturers, delivering substantial TPS performance improvements alongside significant power and cost advantages. Not relying on advanced process nodes, it ensures supply chain autonomy and controllability, and will become the world's fastest-to-mass-produce 3D edge AI chip. It provides high-performance, low-power, and extreme cost-performance chip solutions for large model inference applications including AI phones, AI PCs, IoT, and all-in-one devices, with extensibility to servers, AI robots, and other scenarios.

With a world-class compute-in-memory innovation team and industry-leading engineering team, Winicore has built a highly efficient, resilient, and powerful industrial team driven by continuous innovation. The company already has dozens of PhDs/postdocs from renowned universities and numerous veteran engineers from major tech companies, with top-tier talent density ranking in the industry's first echelon. The team has published 14 chip measurement results at the "Olympics of Chip Design" ISSCC over the past six consecutive years, breaking current world records and maintaining a position in the global first tier. The company's exceptional talent pipeline and highly imaginative industrial vision have earned continued recognition from capital markets, having previously received investment from prominent investors and strategic capital including HSG, Peking University Technology Transfer Fund, Founder Heson, Xiaomi, Luxshare Precision Industry Co., Ltd. Industrial Investment, AVIC Lianchuang, Yida Capital, and Lenovo Capital and Incubator Group.

Appointed by the China RISC-V Industry Alliance, Winicore serves as the lead unit of the "RISC-V Compute-in-Memory Application Group" and formally initiated the development of the world's first RISC-V compute-in-memory standard and ecosystem building in September 2025. The company is collaborating with dozens of upstream and downstream enterprises including Xiaomi Mobile, vivo, Inspur Computer, GigaDevice, UNISOC, Shanghai Zhaoxin, H3C, Anlogic, InnoGrit, KylinSoft, xFusion, SOPHGO, TriEye, Tsingmicro, SpacemiT, PerfXLab, Sipeed, Chipintelli, Lingchuan Technology, Yuanstone Intelligence, Xiong'an Ansuan, and Chendian Technology. Using standards to drive ecosystem development and ecosystem feedback to strengthen industry, they are jointly building an autonomous and controllable compute-in-memory AI chip ecosystem and preemptively securing international discourse power in compute-in-memory. Winicore is committed to being an ecosystem-friendly enterprise. It currently maintains deep cooperation with domestic leading memory manufacturers and multiple top-tier terminal companies, with multi-dimensional deep partnerships in product definition, collaborative R&D, and supply chain. It is the only 3D AI chip company simultaneously working closely with multiple leading smartphone companies and coordinating with mobile phone main chip vendors. Winicore, together with its industry partners, is jointly advancing the application of 3D-CIMTM chips to large model inference, opening up a hundred-billion-level blue ocean market for large model inference, and realizing the grand vision of ubiquitous AI democratization!

BlueRun Ventures Leads Yuanli AI's $8 Million Seed Round | BlueRun Family ZeroG Aircraft Completes Nearly RMB 300 Million A+++ Round | BlueRun Family DeepMaterial Raises Tens of Millions in Series A Funding, Using AI to Compress Metal Materials R&D Cycle from 10 Years to 2 Months | BlueRun Family : When Embodied Intelligence Meets Reality | Path to AGI