Sacra Logo
Sign In
Your Feed Community
Q&A Ask a question

How is Neo4J doing as of June 2024?

Jan-Erik Asplund

Co-Founder at Sacra

We estimate that Neo4j decelerated sharply in 2023, growing 18% year-over-year compared to 38% in 2022, ending 2023 with roughly $163M ARR. 

That’s borne out by company disclosures about hitting $100M ARR at the end of 2021 and $150M ARR circa July 2023, as well as db-engines data showing graph databases like Neo4j losing significant ground to vector databases over the last 18 months and a litany of Glassdoor complaints about slowing sales growth.

None

One of the recurring notes we’ve read from Neo4j employees is that the company was late to recognize and act on the AI market opportunity. However, Neo4j does appear to be getting more traction around AI use cases in 2024. 

In the last 6 months, Neo4j has established partnerships with Langchain, Google Cloud and Microsoft around using their graph database tech for generative AI.

Graph database technology is becoming a more common topic of conversation with respect to more accurate and less hallucinatory retrieval augmented generation (RAG)—especially important for enterprise use cases where there’s less tolerance for LLMs making mistakes.

Companies like Writer, Deutsche Telekom and ServiceNow have written publicly about their internal use of graph-based retrieval for building more reliable generative AI systems for customer support and content creation.

Given this (emergent) possible tailwind, we modeled a slight revenue acceleration for 2024, with Neo4j on target to grow 21% year-over-year.

The two key risks we see for Neo4j are:

  • Lack of sales DNA to successfully transition into selling a generative AI solution vs. their core graph database technology. Most of Neo4j’s recent negative Glassdoor come from members of the sales team, and the high volume of churn in upper management around marketing and sales creates doubt around their ability to execute here.
  • There’s a risk that graph databases become more widely used in generative AI but that Neo4j, as a non generative AI native company, doesn’t see the upside. This has already played out with Pinecone in the vector database space—as soon as vector embeddings became a core part of the generative AI stack, we saw a proliferation of startups doing their own vector DB implementations, as well as existing tools like Postgres adding vector support.