In the near future, it is predicted that every type of database, including graph, relational, document, and key-value databases, will offer some form of vector search. This blurs the line between what constitutes a vector database and what does not. Specialized vector database companies may struggle to compete against incumbent databases that aim to capture the new retrieval augmented generation (RAG) workloads by utilizing their existing data gravity. As a result, the concept of a “vector database” may be seen as more of a feature than a separate category. Larger companies are beginning to enter the vector search market, recognizing the benefits of co-locating vectors and original documents in a single database. Offering vector search is not only a strategic move but also eliminates unnecessary data movement and reduces latency for customers. This convergence trend may lead customers to question the need for specialized vector databases and instead rely on their existing databases. Startups in the vector database space may experience a slowdown in growth as enterprises become more knowledgeable and prefer solutions that integrate seamlessly with their current infrastructure.
https://nextword.substack.com/p/vector-database-is-not-a-separate