MDB · Research
MongoDB's CEO Left for Meta After 10 Months. The Returning CEO Inherits a 30% Grower That Still Pays Its Owners Almost Nothing
CJ Desai resigned on Sep 28, 2026 and Dev Ittycheria returned as interim CEO with $15.0M of fast-vesting stock. Atlas grew 28.9%, but stock pay took 84.7% of trailing operating cash flow.
- Stock pay as a share of operating cash flow
- 84.7%
- Owner cash after stripping interest income on corporate cash
- $9.2M
- Months the outside CEO served before leaving for Meta
- 10.5
1. Verdict and what this piece answers
Verdict: watch. There is no positive anchor to value.
Owner cash, meaning operating cash flow less all capex and finance lease payments, less stock pay, was $48.3M for the six months to Jul 31, 2026, and only $9.2M once the $39.1M of interest earned on corporate cash is removed, our calculation [1]. In every one of the last five full fiscal years owner cash was negative, from -$252.2M in fiscal 2022 to -$50.4M in fiscal 2026 [7].
Under our standard, a base value needs three positive years, so none is set.
On Sep 28, 2026, MongoDB said its chief executive, Chirantan "CJ" Desai, had stepped down effective immediately to pursue a senior role at Meta Platforms, and that the board had named Dev Ittycheria, chief executive from 2014 to 2025, interim President and CEO [3]. Desai had held the job since Nov 10, 2025, about 10.5 months [6].
The announcement came the day before a scheduled Investor Day, and the company reaffirmed its guidance for the third quarter and the full fiscal year [3].
The market read this as a loss of strategy. Our view is that it is closer to the opposite.
The business Desai inherited, and the one Ittycheria takes back, was built by Ittycheria over eleven years: a consumption database where Atlas grew 28.9% in the second quarter of fiscal 2027 and net ARR expansion is 122% [1]
2. Rating metrics
| Metric | Value |
|---|---|
| R&D share of revenue | 28.4%, research and development expense, six months to Jul 31, 2026, our calculation [1] |
| GAAP net margin | 3.1%, six months to Jul 31, 2026; 5.3% in the second quarter of fiscal 2027, the third straight profitable quarter, our calculation [1][2]. Six-month net income includes $16.6M of gains on financial instruments and $39.1M of interest income, so the operating business alone earned $3.6M [1] |
| API quality | 5 of 5, our scoring: a documented Atlas administration API with self-serve keys; public pricing and published limits; a hosted Atlas Managed MCP Server for coding agents [2]; official drivers in many languages with public release notes; and a large partner and integration ecosystem [1] |
| Toll-gate score | 2 of 4 pass, 2 partial (section 2b) |
2b. Thesis and the toll-gate test
A superintelligence agent that runs part of a business still has to read and write live operational data: orders, claims, user profiles, game state. That data lives in a database of record, and moving it is the most expensive migration in enterprise software.
MongoDB's claim is that its document database, with search and vector retrieval built in, is where agents will keep that state. The test is whether that makes MongoDB a toll gate or one of several interchangeable stores that an agent's tools choose on price.
| Test | Result | Evidence |
|---|---|---|
| Owns the record | Pass | Atlas and Enterprise Advanced hold primary operational data for more than 70,600 customers, and growth in the quarter came mainly from more consumption by large existing customers [1][2]. Net ARR expansion of 122% at Jul 31, 2026 shows workloads, once landed, keep growing [1] |
| Owns the connection | Partial | The Atlas Managed MCP Server connects coding agents directly to live Atlas data, and Investor Day showed 8x growth in its use from week 1 to week 7 [2][5]. But open alternatives, PostgreSQL above all, are one tool call away for a new application |
| Prices close to outcomes | Partial | Atlas is billed on usage, so revenue rises with the work an application does, not with seats [1]. It is still a meter on compute and storage, not a share of the value the application produces |
| R&D on the gate vs features | Pass | Spending goes into the engine itself: MongoDB 9.0, search and vector search now in self-managed Enterprise Advanced, automated embeddings and reranking inside Atlas, and vector search in stream processing [2][5] |
Two passes and two partials. MongoDB owns the record and builds on it, but it does not own the agent's choice of database for new work, and it charges for resources rather than outcomes.
Leadership matters here because the open test, winning new agent workloads before they default elsewhere, is a sales and product-priority question, not an engineering one.
3. Growth
| Measure | Value | Change |
|---|---|---|
| Q2 fiscal 2027 total revenue | $771.8M | +30.5% [2] |
| Q2 fiscal 2027 Atlas-related revenue | $565.9M | +28.9% [2] |
| Q2 fiscal 2027 Enterprise Advanced and other revenue | $181.2M | +35.9% [2] |
| Q2 fiscal 2027 services revenue | $24.6M | +29.3% [2] |
| Six-month revenue to Jul 31, 2026 | $1,459.4M | +28.0% [1] |
| Trailing twelve-month revenue | $2,782.8M | our calculation [1][7] |
| Fiscal 2026 revenue | $2,463.8M | +22.8% [1] |
| Q3 fiscal 2027 revenue guide | $756M to $761M | +20.3% to +21.1% against $628.3M [2][7], our calculation |
| Full-year fiscal 2027 revenue guide | $2.99B to $3.03B | +21.4% to +23.0%, reaffirmed Sep 28, 2026 [2][3] |
The quarter added $180.4M of revenue, our calculation [2]. Atlas supplied $126.9M of it, or 70.4%, and Enterprise Advanced and other supplied $47.8M [2].
The Enterprise Advanced line deserves care: it includes term licenses whose license portion is recognized upfront when a contract starts, so a few large multi-year deals can lift a single quarter [1]. Remaining performance obligations rose 91% to $1,519.2M and the current portion 73% to $797.3M, which says those deals were signed, not that the run rate has stepped up by the same amount [2].
The guide shows the shape. Third-quarter revenue of $756M to $761M would be below the second quarter's $771.8M, and the full-year range implies fourth-quarter revenue of about $774.6M to $809.6M, growth of 11.4% to 16.5% against $695.1M a year earlier, our calculation [2][7].
Part of that is a tough comparison and part is conservatism, but it means the 30.5% quarter is not the new trend. Atlas, at 73% of revenue, is the line to follow; the company's own Investor Day commitment was Atlas growth of 20% or more, against 28.7% delivered in fiscal 2026
4. What MongoDB sells and how fast it ships
Atlas is the managed cloud service: MongoDB runs the database on AWS, Google Cloud or Microsoft Azure, bills by usage, and bundles search, vector search, time series, stream processing and analytics on the same data [1]. It accounted for 73% of second-quarter revenue [1].
Its cost of revenue is mostly what MongoDB pays the cloud providers; subscription cost of revenue rose $34.3M in the quarter, of which $31.4M was third-party cloud infrastructure, while subscription gross margin improved to 77% [1]. The economics are a reseller's margin on compute, plus software that customers would struggle to replace.
Enterprise Advanced is the self-managed server for customers who run their own infrastructure or hybrid estates, sold mostly as one-year subscriptions with some multi-year term licenses [1]. It has been a steady 21% of subscription revenue for the last five quarters [2].
That steadiness matters for the superintelligence case. Regulated customers that will not put data in a public cloud can now run the same search and vector retrieval on their own hardware, which MongoDB made generally available in Enterprise Advanced this quarter [2].
The shipping pace is fast. In the second quarter alone MongoDB launched the Atlas Managed MCP Server, connecting coding agents including Claude Code, Codex, Grok Build and Devin to live Atlas data without new infrastructure; made automated embeddings, an embedding and reranking API, a code retrieval model and vector search in stream processing generally available; and brought search and vector search to Enterprise Advanced
5. Net revenue retention
MongoDB discloses net ARR expansion: ARR from customers who were customers a year earlier, divided by all ARR a year earlier, including churned customers. It was 122% at Jul 31, 2026, and the company attributes subscription growth mainly to more Atlas consumption by large existing customers [1].
The proxy for breadth is customers with $100,000 or more of ARR, which rose 17.0% to 2,999 from 2,564 a year earlier [2]. Total customers rose 17.9% to more than 70,600, and Atlas customers to more than 69,300 [2].
Because Atlas ARR for direct customers annualizes the last 90 days of usage, retention here is a consumption measure and can move quickly in either direction [2].
6. Cash flow and owner cash
| Period | Operating cash flow | Capex | Stock pay | Owner cash |
|---|---|---|---|---|
| Fiscal 2022 | $7.0M | $8.1M | $251.1M | -$252.2M [7] |
| Fiscal 2023 | -$13.0M | $7.2M | $381.5M | -$401.7M [7] |
| Fiscal 2024 | $121.5M | $6.1M | $456.9M | -$341.5M [7] |
| Fiscal 2025 | $150.2M | $29.6M | $493.9M | -$373.3M [7] |
| Fiscal 2026 | $505.1M | $5.0M | $550.5M | -$50.4M [7] |
| Trailing twelve months to Jul 31, 2026 | $666.6M | $7.7M | $564.5M | $94.4M, before finance leases [1][7] |
| Six months to Jul 31, 2026 | $343.5M | $8.3M including $3.6M of finance leases | $286.8M | $48.3M [1] |
| Q2 fiscal 2027 | $141.9M | $4.3M including $1.8M of finance leases | $148.9M | -$11.3M [1] |
MongoDB reported free cash flow of $137.6M in the second quarter, nearly double a year earlier [2]. Counting stock pay as the cost it is, owners lost $11.3M in that quarter, our calculation [1].
Over six months, owner cash of $48.3M sits beside $39.1M of interest income on $2.4B of cash and short-term investments; strip that interest out and the operating business produced $9.2M for owners, our calculation [1]. Deferred revenue fell $24.2M over the six months, so the cash flow was not flattered by prepayments [1].
Two other cash lines show where the money goes. MongoDB paid $117.7M in the six months to cover employee taxes on vesting shares through net share settlement, and spent $200.3M on buybacks [1].
Together that is $318.0M, or 92.6% of operating cash flow, spent on managing the share count that stock pay creates, our calculation [1]. The company frames buybacks as a way to manage dilution
7. Acquisitions in the last 24 months
Voyage, a maker of embedding and reranking models, was acquired in February 2025; part of the consideration was restricted stock that vests over time [1][6]. Its models now power automated embeddings and reranking in Atlas [2].
The full price and its effect on owner cash are in the fiscal 2026 Form 10-K, which is not reproduced here. On May 8, 2026, MongoDB bought Clarity Business Holdings, a services firm serving U.S. government workloads, for $16.7M, recording $12.8M of goodwill [1].
Neither deal used debt, and both are small against $2.4B of cash and investments [1].
8. Risks and catalysts
Risks. Leadership: the company has had an interim CFO, a new CFO from May 2025, a new CEO from November 2025, the departure of its field operations head in April 2026, and an interim CEO since September 2026 [3][6].
A permanent search is under way with an executive search firm, and the interim CEO's term is open-ended [3]. Consumption: Atlas revenue tracks customer usage and can slow without warning when customers optimize [1].
Lumpiness: multi-year Enterprise Advanced licenses booked upfront lifted the second quarter and set a hard comparison [1]. Competition: open-source PostgreSQL and the cloud providers' own databases compete for every new agent workload.
Cost of retention: the interim CEO and CFO awards add $25.0M of new stock pay, most of it vesting within eighteen months [4].
Catalysts: interim CEO restricted stock vests in quarterly installments from Jan 1, 2027 to Oct 1, 2027 (confirmed) [4]; third-quarter fiscal 2027 results and the Form 10-Q that will carry the interim CEO's employment agreement, expected in early December 2026 (date unconfirmed) [4]; the CFO's retention grant vests Apr 1, 2027, Oct 1, 2027 and Apr 1, 2028 (confirmed)
9. Valuation anchor and balance sheet
Because owner cash has been negative in each of the last five full years, there is no valuation band and no base value [7]. The anchor to watch is the first full fiscal year of positive owner cash after interest income is stripped out.
The balance sheet gives the company time to get there.
MongoDB has no debt and $2,411.3M of cash and short-term investments, our calculation [2]. Remaining buyback authorization was $399.7M at Jul 31, 2026 [1].
Among infrastructure software peers, MongoDB ranks high on growth and retention and near the bottom on owner cash conversion: owner cash was 3.4% of trailing revenue before finance leases, our calculation [1][7].
10. Superintelligence angle
Tailwind, with one reason: agents that act on live data need an operational database of record with retrieval built in, and MongoDB has moved its engine, vector search and embeddings into exactly that position, with the hosted MCP server as the front door [2][5]. The risk inside the tailwind is that agents choose databases by default and price, which favors whatever the coding tools reach for first.
The interim CEO has spent the last year at Sequoia Capital, and his own framing on returning was that the shift is from software that helps people manage work to software that performs it [8]. A month later he made a point that applies to MongoDB's own customers: fast-growing companies built on heavy compute only work if customers spend enough to cover it [10].
Atlas is the meter on that compute.
11. Leadership and control
Dev Ittycheria is interim President and CEO from Sep 28, 2026; he led the company from September 2014 to November 2025, has been a director since 2014, and since September 2026 is also a Partner and Senior Advisor at Sequoia Capital [3]. His terms: $500,000 base salary, a $500,000 sign-on cash bonus, a 70% target bonus, and $15.0M of restricted stock vesting in four quarterly installments through Oct 1, 2027 [4].
Michael Berry has been CFO since May 2025 and received a $10.0M retention grant on Oct 12, 2026 [4][6]. Tom Killalea is board chair, and the guidelines keep chair and CEO separate [3][6].
Desai joined from Cloudflare, after ServiceNow, and was hired with a $15.0M sign-on restricted stock award and a $17.5M performance award tied to sustained stock price thresholds over five years [6]. The 8-K says his resignation was not due to any disagreement with the company [3].
His Meta title and terms come from news coverage, not from a MongoDB filing. On X, Ittycheria wrote that he "never really left" and "Today, I'm returning as interim CEO" [8]; the company account posted "Welcome back"
12. Questions for the owner and predictions
Questions for the owner. First, will the interim CEO commit to a stock pay budget that falls in dollars, not only as a share of revenue, while he holds the job?
Second, how much of the second quarter's 35.9% Enterprise Advanced growth was upfront term licenses, and what is Atlas growth excluding the largest new workloads? Third, will MongoDB disclose agent workloads, such as Managed MCP Server usage or revenue, so investors can see whether the connection test is turning into a pass?
Fourth, conviction on a 1 to 5 scale: our view is 3. The record is real and the product is shipping fast, but owners have not yet received cash, leadership is temporary, and the guide points to a sharp slowdown in the fourth quarter. The trigger to revisit is a permanent CEO named together with two straight quarters of positive owner cash after interest income.
Predictions to register: third-quarter fiscal 2027 revenue lands at or above $761M, the top of the guide (70%); net ARR expansion at Oct 31, 2026 stays at or above 120% (65%); third-quarter owner cash after interest income is negative (60%); a permanent CEO is named by Apr 30, 2027 (45%); fiscal 2027 stock-based costs exceed $628.9M, the guided figure, after the new leadership grants (60%).
Stock Profiles & Comparable Companies
MDB Stock Profile →More Published Research
View All →Tenable Maps Every Weakness Agents Will Hunt. It Bought Back $516M of Stock on $81M of Owner Cash, and Borrowed the Difference
Tenable owns the exposure record and owner cash is finally positive, but growth is slowing toward 7%, the share count shrank 11.5% on convertible debt, and stockholders' equity turned negative.
SentinelOne Guards the Device Where Agents Run. Owners Lost $291M Last Year, and the Gap Is Widening
Revenue grows near 20% and the products target agent security, but stock pay was 5.6 times operating cash flow, owner cash worsened to -$291.0M, and a new convertible could add 46.1M shares.