Oracle, Microsoft SQL Server 2025, MongoDB Atlas and Postgres now offer vector embeddings, indexes and search, shifting enterprise AI toward existing data platforms.
The report argues vector search is usually a feature, not a separate persistence layer, because production AI depends on freshness, permissions, lineage and reducing siloed data.
Standalone providers such as Pinecone, Weaviate and Milvus may still suit specialised retrieval, recommendation or multitenant RAG workloads, but must compete on quality, scale, latency and operational simplicity.
As giants like Oracle add vector search, is your specialized vector database now a liability?
Can new 'knowledge engines' save vector startups from becoming a mere database feature?