DBX › analyze
DBX · Analyze
2026-08-11 · $34.20 · Mkt cap $7.44B · EV $10.38B · Technology / Software Infrastructure
Verdict up front
WATCH — conviction 5.0 / 10.
Dropbox is a public leveraged buyout that has not been taken private yet. Revenue is flat, the product is not winning, and none of that is the thesis — the thesis is that ~$900M of levered free cash flow retires ~10% of the share count every year against a shrinking denominator. That works, arithmetically, and it has worked: shares are down from 363M to ~226M in four years. It works right up until either the user base resumes declining or the debt stack stops being refinanceable, and the company has now added a new CEO, a founder heading for the exit, and $3.4B of debt against negative $2.2B of book equity to that sentence. At $34.20 the stock has already priced the good version. Fair, not cheap.
1. Fundamentals
| $M | FY22 | FY23 | FY24 | FY25 |
|---|---|---|---|---|
| Revenue | 2,320 | 2,500 | 2,550 | 2,520 |
| Gross profit | 1,880 | 2,020 | 2,100 | 2,020 |
| Operating income | 357 | 384 | 486 | 688 |
| Net income | 553 | 454 | 452 | 508 |
| EPS (dil) | 1.52 | 1.31 | 1.40 | 1.86 |
| R&D | 892 | 937 | 915 | 732 |
| OCF | 797 | 784 | 894 | 952 |
| Capex | (35) | (25) | (23) | (21) |
| FCF | 762 | 759 | 871 | 930 |
| Buybacks | (795) | (540) | (1,240) | (1,710) |
| Diluted shares (M) | 363.3 | 345.6 | 323.4 | 272.8 |
CAGRs (3yr): Revenue +2.7% · Net income −2.8% · FCF +6.9% · OCF +6.1% · Diluted shares −9.1%/yr
Read the table properly. Revenue is flat. Operating income nearly doubled (357 → 688) — and look at the line above it: R&D fell from $937M to $732M, a 22% cut. The margin expansion is the 2023 and 2024 layoffs (16% then ~20% of staff; headcount 2,113 at year-end, the smallest since 2017 and −32% from the 2022 peak). That is real and it is durable, but it is a one-time harvest, not an engine. There is no second round of it available.
Per-share. FY25 FCF/share on the FY25 diluted count is $3.41. On the current ~225.7M shares outstanding (150.6M Class A + 75.1M Class B, as of 2026-08-03) it is $4.12 → 8.3× P/FCF. That is the number the bulls quote, and it is the right number to start from.
But it is unlevered thinking. Interest expense is now $50.0M per quarter and rising. FY26 guidance is unlevered FCF ≥ $1.070B; subtract ~$200M of cash interest and levered FCF is ~$870M, or $3.85/share → 8.9× levered P/FCF. Use that one.
Balance sheet — the part that decides everything. Assets $2.84B against liabilities $4.64B → stockholders' deficit of $(1.80)B at FY25, now $(2.19)B. This is not a distress signal by itself; it is the arithmetic consequence of buying back more stock than you have retained earnings. But it removes every cushion.
Current debt stack: term loan $2,674.8M principal; a $700M delayed draw was taken to retire $695.8M of 2026 converts in cash; $693.3M of 2028 converts remain; a new $400M revolver (undrawn) closed 2026-06-01. Cash $1,056.2M plus $57.6M short-term investments. Covenant: consolidated leverage ≤ 5.0×, tested quarterly.
Where does that leave leverage? EBITDA implied by EV/EBITDA 12.51 on EV $10.38B is ~$830M; net debt ~$2.3B → ~2.8× net leverage. Comfortable against a 5.0× covenant, with room for several more years of debt-funded buyback. Debt/assets of 126% looks alarming and is economically meaningless here — the assets are a rounding error because the value is contractual, not recorded.
Capital allocation. This is the entire investment case, so be precise. Authorizations: $1.5B (Aug 2025) plus a new $900M (June 2026), with $1.385B remaining — 18.6% of the market cap. FY25 repurchases $1.714B. H1 2026 $697.1M, of which Q2 alone was 12.6M shares for $315M. Capex is negligible ($21M). No dividend. No meaningful M&A ($13M).
Every dollar of free cash flow, plus borrowed dollars on top, goes into the share count. Management re-authorized in June 2026 — the clearest available signal that the pace continues.
Scorecard: cash generation A · capital-return discipline A · balance-sheet safety D · revenue growth F · reinvestment F (by design). Flagged gaps: FY21 absent from the feed; the vendor P/B (−3.51) and ROE (blank) are meaningless with negative equity — do not report them.
2. Q2 2026 — the premise that changed
Reported 2026-08-06.
| Q2 2026 | |
|---|---|
| Revenue | $631.5M, +0.9% YoY |
| ARR | $2.566B, +1.0% |
| Paying users | 18.19M, +96K QoQ — the strongest net add in ~3 years, third straight up quarter |
| ARPU | $139.68 (vs $138.32) |
| GAAP operating margin | 26.1% (non-GAAP 39.7%) |
| Unlevered FCF | $283.5M (H1 OCF $443.0M) |
| Net income | $95.8M, down from $125.6M — interest expense $50.0M vs $18.6M |
The paying-user decline has stopped. Eight-quarter trend: −15K, −60K, ~−30K, −64K, +~10K, +~14K, +96K. ⚠️ Two of the middle points are derived from disclosed sequential deltas rather than cited directly; vendor sources disagree slightly on Q1'26 (18.09 vs 18.10M).
This matters more than any other datapoint in the file, because the bear case was never "flat revenue" — it was "shrinking user base plus rising leverage equals a countdown." Three quarters of stabilisation weakens that.
But be honest about the cause: management attributes it to removing onboarding friction (12 steps → 4) and retention work. That is self-help, not demand. Onboarding-funnel fixes are non-repeatable. And ARPU is guided modestly down through 2026 on FX and mix shift to monthly plans.
FY26 guidance (raised): revenue $2.513–2.523B — flat to slightly down versus FY25's $2.521B; non-GAAP operating margin 40.0–40.5% (+50bps); unlevered FCF ≥$1.070B (+$15M).
The stock fell after hours on a beat. The market is not paying for margin without growth.
Note the earnings-quality tell: net income fell 24% YoY while operating margin expanded. The entire gap is interest expense. Every buyback dollar borrowed raises the hurdle the operating business must clear.
3. Moat
Quantitative base. Gross margin 80.2% — exceptional and stable. ROE is undefined (negative equity). ROIC on invested capital of $1.04B against NOPAT ~$540M is nominally >50%, which is an artefact of a book value hollowed out by buybacks rather than evidence of a moat. Ignore it.
| Moat source | Verdict |
|---|---|
| Switching costs | Moderate and weakening — file sync is portable; the lock-in is habit and shared-folder inertia |
| Brand | Real for the "cross-platform, not-Google, not-Microsoft" buyer |
| Network effects | Weak — shared folders create some, but nothing like a true network business |
| Cost advantage | None |
| Efficient scale | None |
Adversarial stress-test — and the incumbent already answered it. Drew Houston named the problem himself: Dropbox competes against giants who give storage away as ecosystem glue. Google Workspace and Microsoft 365 bundle it into a seat licence the customer already pays for. Box owns the compliance/governance niche. AI-native workspaces (Notion, Glean) attack the "find your stuff" layer that Dash targets. A well-funded rival does not need to attack Dropbox — it needs to keep bundling, and the erosion is automatic.
Who still pays, and why: cross-platform users who refuse to live inside one ecosystem; SMB/prosumer/creative teams needing large-file transfer, e-sign and the FormSwift legacy; and inertia — deferred revenue is $748.9M.
Dash — the growth product, downgraded to a feature. This is the most important change since the last look. Dash was sold as a standalone AI universal-search SKU (reportedly up to ~$30/user/month ⚠️ third-party aggregators, not company disclosure). Revenue contribution has never been disclosed — no seats, no ARR — and management said explicitly they are "not baking major AI monetization into this year's outlook." New CEO Ashraf Alkarmi has now folded it in: "We're embedding Dash intelligence directly into Dropbox itself." The only hard adoption figure given is >150,000 users connected via ChatGPT/Claude integrations.
A year ago Houston framed Dash as "both as a standalone product and embedded." The standalone ambition is gone. Read Dash as a retention patch, not a growth restart — a product that were winning would have a seat count attached to it. Houston, on Dropbox's product history generally: "Build it, launch it, nobody uses it, and then shut it down."
Evergreen rating: 3/10. The job-to-be-done (sync files across devices) is permanent. That Dropbox specifically is paid for it is not.
4. Sentiment
Leadership — the largest change and the brief's second stale premise. Drew Houston is no longer CEO (announced 2026-05-26). Ashraf Alkarmi takes over; Houston moves to executive chairman and is explicit that he is leaving to build in AI elsewhere. He retains supermajority voting control through Class B, so this changes the operator without changing who decides.
Analysts: neutral-to-negative with extreme dispersion — roughly 1 Buy / 7 Hold / 3 Sell, median target ~$29, i.e. below the current $34.20. BofA double-downgraded to Underperform. JMP raised to $37 (Outperform). JPMorgan cut to $25. Yahoo's mean target of $30.67 is consistent. The sell side thinks this is expensive here.
Short interest: ~28.6M shares, ~12.9% of float, up ~59% since July 2025. Note the float is only ~42% of shares outstanding under the dual-class structure, so the buyback mechanically tightens the borrow — a squeeze dynamic that is a feature of the capital structure, not a thesis.
Insider selling. Houston sells continuously under a 10b5-1 adopted March 2025 — 164,502 shares (2026-02-02), then trust sales of 109,498 (3/02), 111,166 (4/01), 37,498 (5/14), 30,332 at $27.50 (5/18). Programmatic and pre-scheduled, so not a signal in isolation — but persistent, and now paired with his exit from the operating role.
Activist / LBO. Half Moon Capital (small fund, public since March 2025) is pushing to collapse the dual-class structure. LBO chatter is live in mid-2026 — a ~32% FCF margin and already-negative book equity read as a company engineered for private ownership. Houston's supermajority vote is the blocker, and moving to executive chairman does not remove it.
5. Valuation
Company type: no-growth, high-FCF, levered, no dividend. DDM and DYT are N/A. Graham is N/A — book value is negative, which is itself the finding: there is no asset floor here at all. Weight levered FCF per share and EV/FCF.
Multiples: P/E(ttm) 18.9× · P/E(fwd) 10.4× · P/FCF (unlevered, current share count) 8.3× · P/FCF (levered) 8.9× · EV/levered FCF 11.9× · EV/Revenue 4.1× · EV/EBITDA 12.5×
The EV-based multiple is the honest one. Market cap flatters because $2.3B of net debt is excluded, and that debt is a permanent feature of the strategy, not a temporary financing.
Bogle expected return — the cleanest way to see the thesis: 0% dividend + 0% revenue growth + ~10%/yr share retirement = ~10% annual growth in FCF per share, before any multiple change. Hold the multiple flat at ~9× and you compound at about 10% a year owning a business that grows not at all. That is the entire bull case, and it is a legitimate one.
The fragility is that both terms are assumptions. Revenue growth of 0% requires the user stabilisation to hold. The 10% retirement requires the buyback to continue at a pace that is not contractual and can be paused the moment leverage or covenants bite.
Reverse-DCF — what is priced in? At EV $10.38B and ~$870M of levered FCF, a 10% discount rate implies the market expects roughly flat FCF in perpetuity. Not growth, not decline. That is a reasonable expectation — which is exactly why the stock is not cheap. You are being asked to pay for the base case with no discount for the terminal risk.
Fair value range: $27.00 – $40.00, centre ~$33. - Bear $27 — 7× levered FCF; net adds turn negative again in 2027, ARPU keeps sliding, the market applies a terminal discount to a levered melting asset. - Base $33 — 8.5–9× levered FCF/share $3.85; revenue flat, buyback continues at pace, leverage stable. - Bull $40 — 10–11×; user stabilisation proves durable, Dash-embedded retention lifts ARPU, leverage falls, or an LBO clears at a premium.
At $34.20 the stock sits just above base. The sell side's median target is $29.
6. Tensions and the debate round
Fundamentals vs. Moat. Fundamentals presents 8.9× levered FCF, 80% gross margin, and a share count falling 9%/yr — a machine. Moat replies that the machine is bolted to an asset being bundled away for free by two of the largest companies on earth, and that the one product meant to fix it was just demoted from a SKU to a feature. Resolution: both stand, and the verdict is the tension itself. This is not a business you own for a decade; it is a capital-return vehicle you own while the arithmetic holds and exit when it stops. That is a legitimate position, but it is not the buy-and-hold fundamental lens this framework defaults to, and it should be sized accordingly.
Fundamentals vs. Sentiment. Fundamentals highlights the +96K net adds as evidence the melt has stopped. Sentiment replies that the cause was a one-time onboarding-funnel fix, ARPU is guided down, and the stock fell on the beat. Resolution: Sentiment wins on durability. One quarter of self-help does not establish a demand trend. Two more quarters of positive net adds would; that is precisely the recheck trigger.
The debate that decides the name. Bull: levering a no-growth business is standard, rational financial engineering. Bear: levering a shrinking business is a countdown clock, and every buyback dollar borrowed raises the operating hurdle. Both are right, and which one applies depends entirely on whether the last three quarters were a floor or a pause. Nobody can currently know. That caps conviction at 5.
7. Risks
- Terminal value. ~$3.37B of debt against a flat-to-declining cash stream with zero equity cushion. If net adds revert negative in 2027, there is nothing underneath.
- Interest expense compounding — $50.0M/quarter and rising; already cut net income 24% YoY.
- Covenant and refinancing — 5.0× leverage tested quarterly; $693.3M of 2028 converts need refinancing at whatever rates then exist.
- The +96K may be one-time — funnel fixes are non-repeatable and ARPU is guided down.
- New CEO with the founder halfway out, and the growth product folded into the core.
- AI-native disintermediation of the "where is my file" job entirely.
- The buyback is not contractual — pause it and the thesis has nothing left.
- Dual-class control — Houston decides, and public holders cannot force an outcome (the activist's entire complaint).
A portfolio-specific passage was removed from the public build.
9. Verdict
WATCH — conviction 5.0 / 10. Fair value $27–40, centre $33. Entry $26–29 (where you buy at ~7× levered FCF and are paid for the terminal risk). Trim at 11× levered free cash flow.
The machine works. ~$870M of levered FCF against a $7.4B market cap, $1.385B of remaining authorization equal to 18.6% of the company, a 9%/yr share retirement, and — new this quarter — a user base that has stopped shrinking for three consecutive quarters. Compounding at ~10% a year from a business with no growth is a real and underappreciated outcome.
What stops it being a buy at $34.20 is that this is precisely the outcome already priced. The reverse-DCF says the market expects flat FCF forever, which is the base case, so there is no discount for the risk that the melt resumes against $3.4B of debt and negative book equity. The sell-side median sits at $29. Buy the arithmetic when it is offered at a discount, not at fair value.
What would change the rating: two more quarters of positive paying-user net adds · ARPU stabilising rather than declining · leverage falling below 2.5× · a credible LBO bid · the price below $29.
Next check: Q3 2026 print, early November.