Dnotitia, a deep-tech startup, is preparing for its initial public offering (IPO) on the KOSDAQ as it introduces its innovative model for an “AI memory infrastructure.” This move is expected to influence the future of AI semiconductor development in Korea, shifting focus from traditional large language models to more advanced solutions. The company has appointed Korea Investment & Securities and Shinhan Investment Corp. as lead underwriters for the IPO.
Developed by Dnotitia, the Vector Data Processing Unit (VDPU) is designed to enhance vector database operations, while the Seahorse AI agent improves data retrieval efficiency. The completion of the VDPU's design and tape-out occurred in December, with plans to produce the first silicon chips by June 2026 and commence full commercialization by early 2027. Notably, the company has already begun generating revenue through its AI software, creating a robust financial foundation prior to hardware rollout.
Moreover, Dnotitia claims its system can achieve search speeds up to 10 times faster than traditional methods and reduce total ownership costs by up to 80% compared to CPU-based frameworks. Despite facing legal challenges regarding alleged technology leakage linked to its CEO Jung Moo-kyung and former employer Sapeon, the company remains focused on addressing memory bottlenecks, a significant hurdle in modern AI.