Yixing Intelligence Raises 2 Billion Yuan for RISC-V AI Chips
Yixing Intelligence closed a 2 billion yuan round to expand its open-architecture AI chip ecosystem and data center deployments.
Key points
- Yixing Intelligence raised 2 billion yuan in a new financing round, reaching a valuation near 150 billion yuan.
- The company deployed the industry's first mass-produced RISC-V super-node cluster in a commercial data center.
- Next-generation chips are in development, targeting a ten-fold increase in inference performance for large models.
Yixing Intelligence has secured a new round of financing totaling 2 billion yuan, pushing its post-money valuation close to 150 billion yuan. The capital raise reflects strong investor confidence in the company’s ability to scale its RISC-V based computing infrastructure for artificial intelligence workloads.
The company’s first-generation Epoch chips are already in mass production, offering top-tier domestic performance for large model inference. By leveraging an open-source architecture, the firm aims to reduce licensing costs and engineering complexity for customers deploying large-scale AI clusters.
Open Architecture Reduces Costs
Unlike proprietary instruction sets that require expensive licensing fees, RISC-V is an open standard. This allows Yixing to offer flexible hardware designs without the financial burden of intellectual property royalties. For data centers, this translates to lower long-term operating expenses and greater freedom in system customization.
The company’s approach also simplifies integration. By handling the full stack from chip to cluster software, Yixing reduces the engineering effort required to connect hardware with specific AI models. This is particularly useful for enterprises that need to balance power consumption, cluster size, and deployment speed.
New Cluster Boosts Inference Speed
A key achievement is the deployment of a 64-node super-node cluster in an operator’s data center. This system uses liquid-cooled modules and a backplane-free design to achieve low-latency communication between chips. It is the first mass-produced RISC-V super-node in the industry, marking a significant step in commercial viability.
The cluster improves the throughput of mixture-of-experts models by distributing calculations across multiple chips. This architecture allows for more efficient inference, which is critical for handling the growing demand for real-time AI responses in cloud environments.
Next Generation Chips In Development
Yixing’s next-generation chips have completed tape-out, promising several times the computing and interconnection performance of the current generation. When paired with the second-generation super-node system, the company expects to increase inference performance by more than ten times in specific large model scenarios.
The company follows a yearly iteration cycle, continuously updating its hardware and software stack. This rapid development pace is supported by recent government policies that encourage the industrialization of RISC-V technology. According to reports from 36kr.com, this alignment with national electronic information manufacturing goals strengthens the company’s strategic position in the domestic market.
