MoffettAI Completes Nearly RMB 1 Billion Series C Financing to Accelerate Commercialization of Sparse Computing

According to a May 28 report from the investment community, Moxin Artificial Intelligence Technology (Shenzhen) Co., Ltd. (hereinafter referred to as 'Moxin AI') has recently officially completeda Series C funding round of nearly RMB 1 billion.This round brought togetherindustry investors and market-oriented institutions including Shenzhen Capital Group, Yanshan Technology, Greater Bay Area Common Home Investment, Leader Capital, and Yunsheng Capital,as well asmultiple returning shareholders such as Kaisa Venture Partners, Chuangxiang Investment, and SummitView Capital.

It is reported that the proceeds will be primarily allocated toward the mass production and commercialization of the next-generation accelerator card SparsePrime® (hereinafter referred to as 'SparsePrime®'), as well as further expansion of the company’s nationwide computing power network footprint.

SparsePrime® is Moxin AI's flagship product and will officially launch later this year.

SparsePrime® is a high-performance, general-purpose AI inference accelerator card designed for intelligent computing centers and data centers. Built on Moxin AI's self-developed Antoum 2.0 chip architecture, it is specifically optimized for large models and complex inference scenarios. The product adopts a holistic, top-down design philosophy, offering broad compatibility with mainstream Transformer models and enhanced general adaptability. It comes with a comprehensive toolchain that enables customers to achieve immediate sparse acceleration with zero adoption cost. Developers can migrate and deploy existing models built on PyTorch, TensorFlow, and efficient inference frameworks like vLLM with near-zero code modifications. Additionally, SparsePrime® supports custom operator development using the Triton language, significantly lowering the barrier to entry.

Moxin AI has continuously accumulated technical expertise in sparse computing. Previously, Moxin AI’s S30 and S40 accelerator cards have consecutively won first place in three successive rounds of the MLPerf™ Inference AI benchmark, demonstrating industry-leading performance-per-watt and inference throughput per unit of compute across mainstream tasks including computer vision, natural language processing, and large models—delivering superior inference performance at lower power consumption than competing flagship products.

In terms of industrial adoption, Moxin AI has progressed from single-project validation to the stage of 'nationwide deployment of multi-regional clusters with thousands of accelerator cards.' Its inference clusters, built upon proprietary sparse computing technology, are now serving as the core computational foundation for intelligent computing centers in multiple key regions.

Strategically, Moxin AI has established a presence across four major regions—Northwest, Southwest, East China, and North China—achieving large-scale application across diverse industry scenarios and aligning closely with national macro strategies, particularly the 'East Data West Compute' initiative and 'computing-power and electricity coordination.'

In the Northwest region, Moxin AI has deployed a thousand-card-scale inference cluster to support the intelligent transformation of traditional industries, implementing multiple factory security projects in sectors such as electronics manufacturing and consumer goods production, enabling efficient real-time AI analytics at the edge. In the Southwest, leveraging abundant local green energy resources, Moxin AI has built a low-power, eco-friendly computing pool. In East China, the company has deployed computing clusters tailored for high-end service sectors like bioinformatics and healthcare, significantly accelerating genomic sequencing data analysis workflows; it is already collaborating with industry leaders to provide high-performance AI computing support for computationally intensive tasks such as high-throughput sequencing and protein structure prediction. In North China, Moxin AI empowers urban governance and community intelligence upgrades by deploying multimodal vision applications—including facial recognition and pose estimation—to enable real-time intelligent monitoring and early warning of anomalous behavior.

This nationwide computing network also serves the foundational large model training and inference needs of internet cloud service providers (CSPs).

Moxin AI has already established partnerships with leading telecommunications operators, integrating its sparse computing inference solutions into their computing service portfolios. It is also collaborating with a major business travel and hotel group to explore applications of sparse computing in smart hotel management. In the intelligent mobility sector, Moxin AI is jointly developing solutions with leading automotive manufacturers.

Shang Yong, Vice President of Commercialization at Moxin AI, stated,: “Our deployment of thousand-card clusters is not merely about building raw computing capacity. Instead, by placing high-performance, low-TCO inference nodes close to industrial clusters, we are embedding the technological advantages of sparse computing directly into real-world applications across countless industries. Whether it’s accelerating genomic sequencing in bioinformatics, enabling real-time video analytics for urban governance, or performing visual inspection on smart manufacturing lines, every cluster we deploy is strategically positioned to support large-scale inference demands nearby, efficiently, and cost-effectively—making AI computing as accessible as electricity and water.”

Moxin embeds technological innovation at the source. In international academic collaborations, Moxin works with research teams at Carnegie Mellon University on key technologies such as inference acceleration, long-context serving, and sparse training. Its LLM sparse training initiative has already achieved phased results. Domestically, Moxin has launched a joint research project with Fudan University’s Institute for Trustworthy Embodied Intelligence on 'semi-structured sparsity'; is advancing collaboration with Tsinghua University’s CCNI Lab and SparseMind on frontier topics in sparse computing; and has established a joint sparse computing laboratory with Hangzhou Dianzi University.

Wang Shuayu, Secretary of the Board and General Manager of Corporate Development and Capital Markets at Moxin, stated: "Inference cost is a critical bottleneck to AI adoption, and sparse computing is providing a fundamental solution. From an investment perspective, the value of an AI chip company should not be judged solely by theoretical compute performance per card, but more importantly by its effective compute throughput and energy efficiency when running equivalent AI workloads in real-world cluster environments. Moxin’s multi-site deployments and continuous customer scaling serve as robust validation of our product strength and commercial value. We aim to become an indispensable green computing foundation within the AI infrastructure layer through our combination of in-house chips and computing networks."

156 Views