Yixing Intelligence Raises RMB 1.5B for RISC-V AI Chips

A Chinese AI-chip startup just bagged RMB 1.5 billion to push RISC-V into the data center.

Yixing Intelligence, founded in 2022, designs RISC-V-based AI chips and computing platforms. Its flagship Epoch series is built for large language models and deep learning workloads, and it supports FP8 precision while staying compatible with lower-precision formats to squeeze efficiency. The company touts a full-stack approach: chips, PCIe accelerator cards, and server clusters, backed by a software ecosystem that includes compilers and runtime systems. A key differentiator is a proprietary high-speed interconnect that enables scalable multi-node deployments for demanding AI compute tasks.

The Series B round, disclosed on April 22, 2026, totaled RMB 1.5 billion (about USD 210 million) and was led by several Beijing-based industrial investment funds, with participation from multiple institutional investors. Chinese-language reporting indicates the round places Yixing in an ecosystem of domestically financed chip startups that are aiming to shorten China’s path to self-sufficient AI compute. The funding will be used to accelerate mass production, push next-generation products, and deepen the company’s ecosystem.

The Epoch family embodies China’s push to domesticate AI hardware via RISC-V, an open instruction set architecture that Chinese policymakers and investors see as a way to reduce reliance on foreign IP while accelerating local design and manufacturing capabilities. Yixing’s claim to offer a “full-stack” solution—chips, PCIe accelerators, and server-scale clusters—is a conscious bet on the idea that toolchains and runtimes matter as much as silicon, especially for large-model inference and training workloads. The software layer, including compilers and runtimes, is positioned to ease integration into data-center deployments, a critical step when hardware is still maturing in a fast-moving AI landscape.

This moment matters beyond one startup. Proving mass production and an end-to-end stack for RISC-V in China signals a deeper commitment from both capital and policy-aligned investors to build domestic AI compute capabilities. The emphasis on FP8 and cross-precision compatibility indicates a deliberate effort to balance performance with silicon yield and software practicality—a tradeoff many Chinese chipmakers stress as they scale.

Two practitioner takeaways stand out. First, the move underscores the market's belief that an open ISA like RISC-V, when paired with a robust domestic software ecosystem, can compete for both training and inference workloads at scale, provided the supply chain and manufacturing capacity keep pace. Second, the funding pattern—Beijing-based industrial funds leading rounds—suggests access to local procurement channels and policy-aligned pilot programs, which could shorten the path from lab to a near-production stack in domestic data centers. Yet execution risk remains high: turning a multi-component, hardware-software stack into reliable, mass-produced products depends on mature packaging, testing, and supply-chain reliability that Chinese fabs and partners are still building out for AI-grade accelerators.

Industry watchers will want to monitor Yixing’s roadmap: how quickly mass production begins, how the Epoch line scales beyond LLM and DL workloads, and how the software ecosystem advances in compiler and runtime maturity to actually unlock performance gains in real-world deployments. If successful, the company could become a reference for a growing cohort of RISC-V players aiming to translate policy-driven capital into practical, scalable AI compute.

Sources

Yixing Intelligence Raises RMB 1.5 Billion in Series B for RISC-V AI Chips