The AI computing power chip unicorn Yixing Intelligence has secured another 2 billion yuan in financing, accelerating the commercial implementation and ecosystem leadership of RISC-V cloud AI computing power.
Recently, EVAS Intelligence, a domestic unicorn enterprise focused on RISC-V cloud AI computing power chips, has completed a new round of financing. The total financing amount of this round is nearly RMB 20 billion, and its post-money valuation is estimated to be close to RMB 150 billion.
This round of financing boasts a strong lineup, bringing together more than 20 well-known investment institutions including Huatai Innovation, Zhongding Capital, SMIC PE, Andon Health, Renai Capital, Tongfu Microelectronics, SIE Industry Fund, Heli Capital, Xinxin Leasing, Hengxu Capital, Shangqi Capital, Jianguang Zhanlu, Rongyi Capital and others. After obtaining RMB 1.5 billion Series B financing and strategic investment from China Mobile in the first half of this year, EVAS Intelligence has secured another large-scale financing in just a few months, which fully demonstrates the capital market's high recognition of the company's product strength and commercialization progress.
EVAS Intelligence appears at the World Artificial Intelligence Conference (WAIC 2026)
It is understood that the company's first-generation Epoch computing chip series has achieved large-scale mass production and delivery, and its floating-point computing performance, interconnection bandwidth, and Token throughput performance for large-parameter models all rank in the first tier of domestic mass-produced products. This chip has completed mutual certification with relevant super-node solutions, verifying its feasibility and stability in 10,000-card cluster deployment scenarios.
The 64-node orthogonal backplane-free super-node cluster built by EVAS Intelligence together with industrial partners, equipped with its E200-L liquid-cooled OAM module, has now been launched in the operator's data center. It is the industry's first mass-produced RISC-V super-node, as well as the first mass-produced and launched orthogonal backplane-free super-node in the sector. This cluster realizes hundred-nanosecond-level low-latency interconnection of 64 Epoch computing chips, distributes multiple expert routes across 64 chips for collaborative computing, and greatly improves the inference throughput performance of MoE series models.
It is reported that EVAS Intelligence's new generation of cloud computing power chips has completed tape-out, with several times improvement in both computing and interconnection performance compared with the first-generation products. Combined with the capabilities of its second-generation super-node system, the inference performance in large model PD separation scenarios can be increased by more than 10 times. EVAS Intelligence's chips and super-node products follow the iterative rhythm of "one new generation per year". Relying on the full-stack layout of chips, interconnection, software and cluster systems, the company can flexibly deliver solutions according to customers' model workload, deployment environment, power consumption constraints, cluster scale and other requirements, reduce the engineering connection complexity for customers from chip introduction to system deployment, and improve computing power utilization and deployment efficiency.
Recently, the Ministry of Industry and Information Technology and the National Development and Reform Commission have issued the "15th Five-Year Plan for the Development of the Electronic Information Manufacturing Industry", emphasizing the need to accelerate the R&D and industrialization of RISC-V, and support the application of RISC-V chips in the artificial intelligence field. As an open-source reduced instruction set architecture, RISC-V, with its features of openness, flexibility and no licensing fees, has become an important cornerstone of the global computing power ecosystem in the AI era, as well as a key path for China to achieve independent and controllable computing power.
EVAS Intelligence takes the lead in applying the open RISC-V architecture to cloud and data center AI computing, self-develops the RISC-V-based EVAMIND computing core and TPU-like dedicated AI chip hardware architecture, and realizes in-depth optimization at the chip, software and system levels around large model training, inference and related AI workloads. At the 2026 World Artificial Intelligence Conference, the company released the industry's first RISC-V AI computing power super-node, and completed cluster deployment in the operator's data center in just over two months. The implementation rhythm of "delivery upon release" strongly proves EVAS Intelligence's mature engineering implementation and large-scale mass production and delivery capabilities in RISC-V cloud intelligent computing scenarios.
Based on the open-source RISC-V instruction set architecture, EVAS Intelligence continues to promote collaborative innovation of software and hardware, actively participates in the formulation of international standards, and enhances the technical discourse power of Chinese enterprises in the global open-source computing power ecosystem. In terms of software ecosystem, the company promotes the open source of the VISA virtual instruction set, jointly maintains and iterates it with domestic RISC-V open-source organizations, and promotes VISA to become the de facto open source standard for RISC-V to access the upper-layer AI ecosystem. At the same time, it integrates the EVACA software stack into the global AI ecosystem, carries out adaptive optimization in conjunction with domestic AI ecosystems such as FlagOS, natively supports mainstream AI frameworks and domestic large models, and reduces the model migration cost for customers.
Industry observers commented that there are few domestic AI chip enterprises that can realize system-level full-stack capabilities, which has become EVAS Intelligence's core differentiated competitive advantage. As a new benchmark for domestic cloud computing power and a leader in RISC-V cloud computing chips, EVAS Intelligence's products have been deeply applied in new AI infrastructure scenarios such as AI data centers and AI token factories, and are accelerating large-scale shipment in industries including the Internet, finance, communications and energy. Relying on in-depth cooperation with core customers, the company accumulates deployment, operation and maintenance experience from commercial scenarios, feeds back front-line business insights to the iteration of chips, software stacks and super nodes, and continuously consolidates its position in the first tier of domestic cloud computing power chips.