Dnotitia wins FMS award as it pitches VDPU to storage partners

Dnotitia, a Seoul-based AI infrastructure developer, won the AI Application Award at the Future of Memory and Storage 2026 Best of Show Awards for its Seahorse AI Storage platform, announced August 4. CEO MK Chung is leveraging the award to pitch the company's VDPU (Vector Data Processing Unit) to storage partners, aiming to integrate software and dedicated silicon for AI retrieval. The platform, which supports cloud, on-premises, and air-gapped deployments, targets RAG weaknesses by combining semantic retrieval with exact term matching and linking passages to original documents.

Dnotitia https://dnotitia.com/en/?ref=runtimewire CEO MK Chung is turning an industry award into a sales opening for a larger bet: storage systems that handle AI retrieval with software and dedicated silicon built together. The Seoul-based AI infrastructure developer, which also operates a U.S. office in San Jose, California, said in an August 4 announcement https://www.prnewswire.com/news-releases/dnotitias-seahorse-ai-storage-wins-ai-application-award-at-fms-2026-302843250.html?ref=runtimewire that its Seahorse AI Storage platform won the AI Application Award at the Future of Memory and Storage 2026 Best of Show Awards. Dnotitia was also a finalist for the event's Startup Business Growth Award. Chung is building that bet with CTO and chief architect Se-Hyun Yang https://www.design-reuse-embedded.com/ipsocdays/2025/korea/bios virtual/bio.jsp?k=Se-Hyun%2BYang&ref=runtimewire , a processor designer with more than 20 years of experience. Yang previously led work at Samsung Electronics on proprietary mobile CPUs and GPUs, along with AI-focused high-performance computing processors for data centers and supercomputers. He holds a computer engineering Ph.D. from Carnegie Mellon University and electrical engineering degrees from KAIST. Dnotitia's leadership page https://dnotitia.com/ko/company/about/?ref=runtimewire also identifies Hongchan Roh as chief data officer, SeungJae Lee as chief AI officer and Junhyuk Son as chief operating officer. That background helps explain Dnotitia's unusual product scope. Founded in 2023, Dnotitia is trying to own the software that prepares and retrieves enterprise knowledge as well as the chip that accelerates the underlying vector operations. Public sources do not establish which members of the leadership team were original co-founders. The award doubles as a partner pitch The product recognized at FMS, Seahorse https://dnotitia.com/en/products/seahorse/?ref=runtimewire , is Dnotitia's retrieval and enterprise AI platform. Dnotitia says Seahorse processes structured and unstructured data, analyzes document layouts, tables and charts, and combines semantic retrieval with methods for finding exact terms, figures and clauses. That combination targets a recurring weakness in retrieval-augmented generation, or RAG. Semantic search may return material related to a question while missing a particular contract clause, account number or financial figure. Dnotitia says Seahorse can link retrieved passages to their original documents, giving users a way to inspect the evidence behind an AI-generated response. Dnotitia also supports cloud, on-premises and air-gapped deployments, according to its announcement. Its Seahorse Cloud product page https://seahorse.dnotitia.ai/en/product/?ref=runtimewire describes a SaaS subscription alongside on-premises deployment and says the platform combines object storage, a vector database, autonomous document parsing and managed agents. The page does not disclose subscription or hardware pricing. The on-premises emphasis puts Seahorse in front of organizations that cannot readily move internal documents into a public cloud because of security, residency or compliance requirements. Dnotitia has disclosed some early traction. A March 2026 report https://en.sedaily.com/technology/2026/03/03/denotisia-to-launch-ai-storage-business-based-on?ref=runtimewire identified KT Cloud as a customer, said Seahorse received Grade 1 certification under South Korea's Good Software program in January 2026, and reported proof-of-concept work with global storage and server manufacturers. The available reporting does not identify those manufacturers or establish how many projects have moved into paid production. The FMS award gives Dnotitia recognition inside the storage industry, where Chung needs partners. At Booth 1144, Dnotitia is showing Seahorse with an accelerator card containing its first publicly unveiled VDPU, or Vector Data Processing Unit. Dnotitia is inviting storage vendors and AI infrastructure providers to pursue integrations, jointly developed systems, enterprise proofs of concept and commercial releases. That invitation defines Dnotitia's immediate objective. An award can create interest, while integrations with established storage suppliers can provide distribution and access to enterprise buyers. Dnotitia's announcement describes a commercialization push and does not demonstrate broad production adoption. A chip built around retrieval The VDPU is designed to process vector search and graph traversal closer to stored data. Dnotitia says this approach reduces work sent to the host CPU and limits unnecessary movement of data between storage and compute resources. Chung summarized the thesis in the announcement https://www.prnewswire.com/news-releases/dnotitias-seahorse-ai-storage-wins-ai-application-award-at-fms-2026-302843250.html?ref=runtimewire : "AI bottlenecks are expanding beyond computing performance to data storage and retrieval." Dnotitia is betting that retrieval will become important enough to justify a specialized processor. Modern AI applications convert documents and other information into vectors, then compare those representations to find data relevant to a user's request. Graph traversal can help follow relationships among entities, records and pieces of knowledge. Both processes can become demanding as data volumes and query traffic grow. Dnotitia's approach ties that processor to Seahorse's document preparation, indexing, synchronization and retrieval software. Vertical integration gives Yang's engineering team tighter control over how workloads move between hardware and software. It also raises the execution burden: Dnotitia must deliver a credible database and RAG product while bringing specialized semiconductor hardware into commercial systems. Dnotitia's product page advertises 10 times faster performance and an 80% reduction in total cost of ownership https://dnotitia.com/en/products/seahorse/?ref=runtimewire , but those comparisons use unnamed competitors and company-supplied testing. The FMS award recognizes the application, but does not establish independent performance, power consumption or price comparisons for the VDPU. The company has also listed two government-backed development programs: a KRW 16 billion vector database and VDPU research project and a separate KRW 3.7 billion AI storage project with FADU and NHN Cloud https://seahorse.dnotitia.ai/en/?ref=runtimewire . Those programs provide development resources and industry partners, though they do not establish commercial demand for the finished products. Investors financed the hardware-software bet Dnotitia entered FMS with substantial financing for a business founded less than three years ago. On April 21, 2026, Dnotitia announced a KRW 90 billion Series A https://www.yna.co.kr/amp/view/AKR20260424105200017?ref=runtimewire , worth approximately $63 million at the time. Elohim Partners led the round. Reported participants included Kiwoom Investment, Starting Line, Maple Investment Partners, Daesung Startup Investment, Shinhan Venture Investment, Ulmus Investment, Kolon Investment, HB Investment, Tony Investment and SJ Investment Partners. Dnotitia did not announce a valuation with the financing. The company said the proceeds would support Seahorse Cloud and on-premises product development, VDPU commercialization, global expansion and hiring. Retrieval is moving into existing infrastructure Dnotitia is entering a crowded market. Pinecone https://www.pinecone.io/enterprise/?utm source=openai&ref=runtimewire , Weaviate https://weaviate.io/platform?ref=runtimewire , Qdrant https://qdrant.tech/?ref=runtimewire , Zilliz's Milvus and Vespa https://vespa.ai/vespa-content/uploads/2025/02/Managers-Guide-to-RAG.pdf?ref=runtimewire address vector retrieval through managed, self-hosted or open-source products. Large infrastructure vendors are also folding retrieval into platforms that already hold enterprise data. On June 30, 2026, MongoDB made Search and Vector Search generally available https://www.mongodb.com/company/newsroom/press-releases/mongodb-delivers-accurate-ai-retrieval-wherever-enterprise-data-lives?ref=runtimewire for its Enterprise Advanced and Community editions, including on-premises, private-cloud and self-managed environments. On June 17, 2026, AWS announced Amazon Bedrock Managed Knowledge Base https://aws.amazon.com/about-aws/whats-new/2026/06/amazon-bedrock-managed-knowledge-base/?ref=runtimewire , a managed RAG service that handles connectors, parsing, storage and retrieval orchestration. Those releases support Chung's view that retrieval is becoming a core infrastructure function. They also give enterprises ways to add AI search through vendors already embedded in their technology stacks. Dnotitia's answer is specialized hardware paired with an integrated knowledge platform. The FMS recognition puts that architecture in front of the storage suppliers Chung wants to recruit. The harder test begins after the conference: turning accelerator cards and proofs of concept into repeatable products whose speed, cost and reliability can be measured against established databases and cloud platforms.