Moffett AI

Moxin Artificial Intelligence Technology (Shenzhen) Co., Ltd., operating as Moffett AI, was founded in 2018 by a team of AI research scientists from Carnegie Mellon University and semiconductor industry veterans. The company is positioned as a leader in sparse AI computing, developing specialized hardware and software to address the high costs and power consumption associated with AI inference. The core of Moffett AI's technology is its proprietary dual sparsity algorithm, which intelligently processes only essential data parameters, significantly reducing redundant calculations. This hardware and software co-design has resulted in a portfolio of over 30 patents and enables order-of-magnitude performance acceleration at a lower cost and with less power consumption than traditional dense computing methods. The company’s product line includes a series of AI accelerator cards, such as the S30 and S40, which are built on its self-developed Antoum chip architecture. These products have gained international recognition by winning top honors in three consecutive MLPerf™ Inference benchmarks, in some cases outperforming competitors like NVIDIA's A100 and H100. Moffett AI is currently scaling its commercial operations, transitioning from single-project validations to the deployment of thousand-card clusters across China. This strategy involves creating a national computing network with regional hubs tailored to specific industries, such as smart manufacturing, green energy computing, bioinformatics, and urban governance. The firm’s business model targets cloud service providers, e-commerce platforms, and other data-intensive enterprises. In May 2026, Moffett AI announced the completion of a Series C funding round of nearly one billion RMB, backed by prominent industrial and state-backed institutions. These funds are earmarked for the mass production and commercialization of its next-generation SparsePrime® computing card, based on the new Antoum2.0 architecture, further solidifying its market position. Keywords: sparse computing, AI inference, AI accelerator cards, semiconductor, hardware acceleration, dual sparsity algorithm, low power AI, AI chips, Moffett AI, Moxin AI, Antoum chip, SparsePrime, MLPerf, data center acceleration, high-performance computing, AI hardware, Shenzhen, edge computing, cloud computing, neural network optimization, inference efficiency, computing cost reduction, TCO reduction, green computing, bioinformatics computing, smart manufacturing AI