NEUCHIPS

NEUCHIPS is a Taiwanese semiconductor company established in January 2019 by Dr. Youn-Long Lin, who also serves as the company's chairman. The company was founded in the incubation center of National Tsing Hua University, spurred by a call from Meta (then Facebook) for the industry to develop hardware acceleration for recommendation inference. Dr. Lin, a professor at National Tsing Hua University with a Ph.D. in Computer Science from the University of Illinois, Urbana-Champaign, brings over 35 years of experience in the IC design industry, including co-founding Global Unichip Corp., a TSMC subsidiary. In August 2023, Ken Lau, former General Manager of Intel Taiwan, was appointed CEO to lead the company's global expansion.

The firm specializes in designing and providing Application-Specific Integrated Circuit (ASIC) solutions for the artificial intelligence market, focusing on purpose-built accelerators for both data centers and edge computing. NEUCHIPS operates on a hardware-centric business model, generating revenue by selling its specialized AI chips and accelerator cards to enterprise clients, cloud service providers, and various industries that require high-performance AI computation. The company also provides comprehensive software support, including software development kits (SDKs), drivers, and data integration services. Since its inception, NEUCHIPS has secured significant funding, including a Series B2 round of $20 million in October 2022, with strategic investors like Wistron, JAFCO Asia, and Powerchip Group.

The company's product portfolio is centered around its high-performance, low-power AI accelerator chips. Its first major product was the RecAccel™ N3000, an accelerator chip manufactured on TSMC's 7nm process, specifically designed for Deep Learning Recommendation Models (DLRM). This chip features a patented 8-bit number format (FFP8) that maintains high accuracy while significantly reducing energy consumption. Building on this, NEUCHIPS has expanded into the generative AI market with its Raptor Gen AI inference chip and the Viper series of PCIe accelerator cards. These products are engineered for power efficiency, capable of running large language models offline to ensure data privacy and security. They are compatible with both Intel and AMD CPUs and support various operating systems and LLMs. Through partnerships, NEUCHIPS is integrating its hardware into compact AI systems and delivering its solutions to cloud-native customers via Taiwan's National Center for High-performance Computing (NCHC).

Keywords: AI accelerator, ASIC solutions, deep learning recommendation models, DLRM, generative AI, inference chip, semiconductor, data center acceleration, edge computing, low-power AI, RecAccel N3000, Viper PCIe card, Youn-Long Lin, Ken Lau, hardware acceleration, MLPerf, power-efficient AI, AI inference, cloud service providers, LLM inference, neural networks, custom silicon