CFO Mike Berry & Core Product Chief Product Officer Ben Cefalo sat down with Bank of America & William Blair this week for two interviews
MongoDB Making the Case for its Document-Oriented Database Over Relational Offerings from Snowflake & Databricks:
Most of the information from both short interviews focused on this topic and MongoDB leadership defending its platform’s AI era positioning. For the last couple years, leadership has argued that their document-oriented storage via JavaScript Object Notation (JSON) objects is better for AI than relational alternatives. JSON-based work is extremely popular for an agent’s coding and work. They also argue that the document-oriented approach fits perfectly with the scaled unstructured data usability required to build valuable models and agents. They argue that relational substitutes, which store data in sheet-like rows, are less capable of handling that data and helping companies in this new age. Why? Because in their minds, competitors can’t scale JSON-based storage or analytics at the same cost efficiency as MDB.
This is based on their data “sharding” architecture that stores information across a distributed system of cheaper and smaller computers vs. a central destination. If you listen to Snowflake or Databricks talk, they’d say the exact opposite. And to be fair, both are more than capable of handling JSON objects and translating them into formats that are usable on their own platforms. And while both are considered online analytical processing (OLAP) data specialists, they’re both getting into MongoDB’s online transaction processing (OLTP) territory via integrated acquisitions.. This approach will provide customers with whichever solution they want and provide a blend of deep analytics and speed in one solution.
Snowflake is beating MDB’s expected 20% annual revenue growth rate by 10 points despite operating at about double the scale. They’ve also (in my opinion) delivered stronger evidence of AI directly contributing to full-year guidance raises. Leadership was specifically asked about this and thinks it’s just a matter of where we are in the AI cycle.
I don’t say this to pick on MongoDB. They should find steady growth for a long time. I just don’t think their arguments that their architecture is superior and others can’t match them are overly compelling. MongoDB is a great company. Snowflake is a great company. Databricks is a great company. All 3 should find tailwinds over the coming years from AI – even if the two relational players have enjoyed the most traction thus far. In terms of when MongoDB really joins the party, they believe the world moving from chatbot and research-based AI work to complex, multi-step agentic enterprise work will favor its platform. This will mean a greater need for OLTP demand, as rapid data processing from agentic task completion takes off.
A Document-Oriented Disadvantage to Overcome:
Because legacy relational systems are much more mature, models are more comfortable recommending them compared to newer document-oriented systems. For this reason, model outputs often will include developer instructions to convert JSON-based model text outputs into a default Postgres-based storage format (Postgres is an open-source object-relational database management system (ORDBMS)). They’re working very hard to educate the world and try to influence these models to overcome this bias and more freely keep outputs in JSON files that can be stored and leveraged by systems like MongoDB’s Atlas. They’re pushing out a ton of information on “AI Engine Optimization” (AEO) and expect this to bear fruit over time as (they think) customers realize this is a better option. And to make sure their platform is as agent-friendly as possible, they’re also investing heavily in Model Context Protocol compatibility so companies can include MDB in their agentic workflows.
More Notes:
- Guidance is conservative, as always, and there could likely be some upside from unpredictable on-premise contract wins. They basically assume a worst-case scenario for this section of the business that rarely manifests.
- MongoDB expects the time from its software platform 8.0 launch to 9.0 will be shorter than the previous development cycle as it uses AI to accelerate innovation.
- MongoDB expects on-premise to move from 25% of its business to 10% in the coming years and stabilize there.
- MongoDB is ahead of schedule on its investor day targets in terms of growth and time to reach a rule 40 level (revenue growth rate + FCF margin = 40+).