The quarter by the numbers
MongoDB reported total revenue growth exceeding 30 percent for fiscal Q2 and added approximately 2,900 customers in the quarter, the highest quarterly customer addition total in the company's history. Net revenue retention held at 122 percent, meaning existing customers are expanding their spend at a healthy clip even as the company brings on new logos at record volume. CEO CJ Desai called the growth level a return to a pace the company has not sustained in a long time, and pointed to two distinct forces driving it rather than one, a combination that matters because it means the growth is not dependent on a single customer segment staying strong for the trend to hold up next quarter.
The more telling number sits underneath the headline growth figure: 99 percent of new customer additions chose Atlas, MongoDB's fully managed cloud service, over self-hosted deployment. That ratio has been trending toward Atlas for years, but a figure this close to saturation confirms that for nearly every new MongoDB customer today, the decision to adopt the database and the decision to adopt the managed service are effectively the same decision.
Two very different customer profiles, one platform
Desai described two distinct growth engines feeding the same number. The first is enterprise modernization: financial services and insurance companies migrating workloads off legacy relational databases, a customer profile that tends to be slow-moving, compliance-heavy, and typically the last segment to adopt newer infrastructure rather than the first. Their arrival in meaningful volume suggests document databases have crossed a credibility threshold with exactly the buyers who were historically most conservative about database choice, the same buyers whose procurement and risk teams spent the last decade treating anything outside a traditional relational engine as an audit liability rather than a viable option for core systems.
The second engine is AI-native companies choosing MongoDB as their foundational data layer from day one, with Desai naming Mercor, ElevenLabs, and Harvey specifically as examples. That is a notably different sales motion than the enterprise modernization story, selling to fast-moving startups building AI products rather than large regulated institutions replacing something old. MongoDB positioning itself as credible with both ends of that spectrum simultaneously is unusual, and it is worth asking which motion is actually driving more of the 2,900 net customer additions before assuming the growth is broad-based.
What actually changed in the product
The AI-native pitch is backed by specific product work, not just positioning. Atlas now offers automatic embeddings through Voyage AI, removing a step that previously required teams to run their own embedding pipeline before storing vector data alongside operational data. More significantly, MongoDB shipped a fully managed Model Context Protocol offering that went live on August 13, giving AI agents a standardized, managed way to query MongoDB-backed applications without custom integration work for every agent framework a team adopts.
MongoDB also highlighted developer tool integration with Claude Code and Codex specifically, an explicit bet that the next wave of database adoption decisions will increasingly be influenced by which platforms are easiest for an AI coding agent to work with correctly, alongside which platforms are easiest for a human engineer to learn by hand. That is a real shift in how infrastructure vendors are competing, and one every data platform buyer should expect to see more of across the category over the next year, as query generation quality inside an AI coding tool becomes a genuine input into which database a team defaults to for a new project.
Why the legacy migration story deserves scrutiny
Financial services and insurance firms migrating off relational databases is the more consequential claim in this announcement, because that segment does not move for marginal reasons. A regulated institution rarely undertakes a core system migration without a specific forcing function: a costly license renewal, a compliance requirement the old system cannot meet, or a genuine architectural need that a document model solves better than a decades-old relational schema. Any CDO in a similar position should ask what actually triggered these particular migrations rather than assuming the underlying motivation applies equally to their own environment.
It is also worth remembering that MongoDB is the vendor telling this story, and the examples it chooses to name are, by definition, its best cases. A rigorous evaluation means finding an independent reference in the same regulatory environment and asking pointed questions about total migration cost, downtime during cutover, and how long it actually took from decision to production, not just the headline growth figures a public company reports to investors on an earnings call.
The decision this creates for data leaders
For any organization currently evaluating a database for a new AI-native workload, MongoDB's combination of automatic embeddings and managed Model Context Protocol support removes real integration work that would otherwise fall on an engineering team, and that convenience is a legitimate factor in a build-versus-buy calculation, not just a marketing checkbox. The 99 percent Atlas adoption rate among new customers also means self-hosted MongoDB is increasingly a niche choice, worth doing only when a specific regulatory or data residency requirement genuinely demands it.
For teams sitting on an aging relational system and wondering whether now is the moment to move, the growing volume of financial services and insurance migrations is a useful data point but not a mandate. The right test is still the same one it has always been: does the workload's access pattern genuinely fit a document model better than a relational one, and does the AI tooling advantage MongoDB is now offering change that calculus enough to justify the migration cost on its own merits.
