DigitalOcean Holdings, Inc. (DOCN) | The Buildout — AI Infrastructure
The Verdict
DigitalOcean is a cloud-infrastructure provider that has repositioned itself around AI inference. It leases data centers, fills them with GPU and CPU compute, and sells the network and software stack above them, up through an inference engine and a managed-agents platform. Its customers are AI-native builders and digital-native enterprises rather than frontier model labs. In the build-out it sits downstream of chip and data-center suppliers and upstream of the companies building AI applications.
| Market Cap | — |
| Revenue (TTM) | $1.0B |
| Revenue Growth | +21.4% |
| EBITDA Margin (TTM) | 41.3% |
| Net Debt | $732M |
| Earnings Beats | 7 of 7 |
| P/E (TTM) | — |
| EV/EBITDA (TTM) | — |
What We Like
- Revenue grew 29% year over year to $281M in Q2 2026, more than double the 14% growth of a year earlier.
- Contracted visibility inflected: RPO of $894M, up more than 12x year over year, with average life moving from 1.6 to 3.7 years.
- AI customer ARR reached $234M, up 212% year over year; inference services grew almost 800% and are now over 70% of AI ARR.
- The largest cohorts are growing fastest: $1M-plus ARR grew 214% and is now 23% of total ARR, up from 9% a year ago.
- The balance sheet de-risked: 0.7x pro forma net leverage after equitizing $472M of 2030 converts.
What We’re Watching
- Capacity timing: ~155 MW committed, majority online by end-2027, and management ties revenue growth to data-center turn-up it calls lumpy.
- Gross margin fell to 55.0% in Q2 2026 from 59.9% a year earlier, as data-center costs land before the revenue they carry.
- FY2027 adjusted FCF margin is guided to "high teens," lowered from an earlier "20%+" framing, with ~$100M of 2026 FCF absorbed by startup costs.
- Concentration risk: management deflected Q&A on Cursor and the reported SpaceX option; top 25 customers are 20% of ARR.
On the top line the thesis is strengthening: two consecutive guidance raises, a contracted book that grew more than twelvefold year over year, and an inference business scaling near 800%. The counterweights are the cash line and the clock. Gross margin is compressing while capacity is built ahead of revenue, the FY2027 FCF margin was marked down to high teens, and growth depends on a schedule the company itself calls lumpy. The open question is whether ARR per megawatt rises enough for the added capacity to earn more than commodity GPU rental would.
Earnings Beat
Q2 2026 revenue was $281M, up 29% year over year and above the high end of guidance. Gross margin was 55.0%, down from 59.9% a year earlier, as data-center expansion costs landed ahead of the revenue they carry. AI customer ARR reached $234M, up 212% year over year. Management said the acceleration came from its highest-spending customers and AI-native builders, and that a ~30% GPU list-price increase had "very little impact" on the quarter.
| Metric | Q2 FY2026 | Q1 FY2026 | Q2 FY2025 | YoY |
|---|---|---|---|---|
| Revenue | $281M | $258M | $219M | +28.6% |
| Gross margin | 55.0% | 56.1% | 59.9% | -490bps |
| EBITDA | $126M | $82M | $68M | +84.2% |
| EPS | $0.28 | $0.14 | $0.37 | −23.9% |
| AI customer ARR | $234M | $170M | n/a | +212% YoY |
| Remaining performance obligations (RPO) | $894M | $243M | n/a | >12x YoY |
Revenue grew 14% year-over-year in the second quarter of last year. In a single year, we have doubled our growth rate to 29%.— Paddy Srinivasan, CEO, 2026-08-04
Management tone: Across the Q1 and Q2 2026 calls, management raised guidance twice and, on the Q2 call, said it is "even more confident" in the FY2027 50%-plus target while calling a formal 2027 number premature. They declined to let the ~30% GPU list-price increase take credit for the quarter, saying it had "very little impact," and disclosed the 0.7x pro forma net leverage directly. They flagged RPO lumpiness and capacity-timing dependence without being asked. Two lines of questioning were declined: AMD-versus-NVIDIA hardware mix and per-megawatt unit economics.
Management Guidance
FY2026 revenue is guided to $1.17–$1.18B (~30.5% growth), with Q4 2026 exit growth of 35% or more; Q3 2026 revenue is guided to $304–$307M (32–34% growth). FY2026 adjusted EBITDA margin is ~39% (Q3 38–39%, a step-down from Q2's 40%), and adjusted FCF margin is 11–13%. FY2027 revenue growth is held at 50% plus, which management says does not embed all the upside it can see; formal 2027 guidance is expected later this year. Management ties the growth to data-center turn-up timing, which it calls lumpy, and says ~$100M of 2026 FCF is absorbed by startup costs for 2027 capacity.
Trajectory
Revenue has accelerated through five straight quarters, from $219M in Q2 2025 to $281M in Q2 2026, with sequential growth rising each period. The mass is shifting to the largest customers: the $1M-plus ARR cohort grew 214% and moved from 9% to 23% of total ARR in a year. The offset is gross margin, which fell from 59.9% to 55.0% as data-center expansion costs are incurred ahead of the revenue from new data centers. Management says capacity, not demand, is the gate on growth.
The Model
The model's locked median projections are FY+1 revenue of $1,180M and EBITDA of $354M (a 30.0% margin), and FY+2 revenue of $1,825M with EBITDA of $589M (a 32.3% margin). The FY+1 revenue figure is in line with management's FY2026 guide. The FY+2 step-up requires the $894M RPO and the ~155 MW of committed capacity to convert on the schedule management describes, with the majority online by end-2027.
| Metric | FY2025 | Next FY (E) | Following FY (E) |
|---|---|---|---|
| Revenue | $901M | $1.2B | $1.8B |
| YoY Growth | — | +30.9% | +54.7% |
| EBITDA | $345M | $354M | $589M |
| EBITDA Margin | 38.3% | 30.0% | 32.3% |
Projections are the median of 5 independent model runs. The model’s revenue sits 0.6% above analyst consensus.
FY2026 revenue is guided to $1.17–$1.18B (~30.5% growth), with Q4 2026 exit growth of 35% or more; Q3 2026 revenue is guided to $304–$307M (32–34% growth). FY2026 adjusted EBITDA margin is ~39% (Q3 38–39%, a step-down from Q2's 40%), and adjusted FCF margin is 11–13%. FY2027 revenue growth is held at 50% plus, which management says does not embed all the upside it can see; formal 2027 guidance is expected later this year. Management ties the growth to data-center turn-up timing, which it calls lumpy, and says ~$100M of 2026 FCF is absorbed by startup costs for 2027 capacity.
What Could Go Right — and Wrong
- Data centers turn up on schedule, converting the ~155 MW of committed capacity into revenue as planned.
- Core-cloud attach on $100k-plus AI customers keeps climbing above the ~70% disclosed in Q2.
- More large commitments land without the top-25 share of ARR rising much.
- Open-weight models stay close to frontier quality, keeping the inference cost wedge and the ~75% open-weight token mix.
- Formal FY2027 guidance lands above 50%, or ARR per megawatt is quantified.
- A capacity slip in Memphis, the 60 MW, or the 20 MW pushes revenue into a later period and puts the 50%-plus FY2027 guide at risk.
- Gross margin keeps falling below 55.0% if data-center costs stay ahead of revenue.
- Adjusted FCF margin settles below the "high teens" FY2027 floor, given ~$100M of 2026 startup costs and higher capex per megawatt.
- A marquee AI customer changes owner, in-sources, or reprices; management deflected Q&A on Cursor and the reported SpaceX option.
- A neocloud or hyperscaler matches the integrated stack cheaply, compressing DigitalOcean toward commodity capacity pricing.
Looking Ahead
The next twelve months are about converting the build into revenue. Management guides FY2026 revenue to $1.17–$1.18B and a Q4 exit growth rate of 35% or more, then 50%-plus growth for FY2027. The AI Builder Summit on October 13, 2026 is where management says it will detail the managed-agents layer, and formal FY2027 guidance is expected later this year. The variable management names is data-center turn-up timing, which it calls lumpy.
- 2026-10-13AI Builder Summit — Management to detail the managed-agents layer.
- 2H 2026Memphis data center launch — On track for second-half launch; ~15 MW implied.
- Q4 2026Exit-growth test — Company targets 35%+ revenue growth exiting FY2026.
- Later in 2026Formal FY2027 guidance — FY2027 held at 50%+ growth; precise number pending.
- Throughout 202760 MW revenue ramp — Committed 60 MW ramps revenue through 2027.
- Late 2027 into 2028Incremental 20 MW online — Most recent 20 MW targeted online late 2027 into 2028.
Financials
Annual Summary
| Metric | FY2024 | FY2025 | TTM | YoY |
|---|---|---|---|---|
| Revenue | $781M | $901M | $1.0B | +15.5% |
| Gross Margin | 58.9% | 59.9% | 57.2% | +98bps |
| EBITDA | $221M | $345M | $418M | +56.2% |
| EBITDA Margin | 28.3% | 38.3% | 41.3% | +998bps |
| Net Income | $84M | $259M | $235M | +207.2% |
| Free Cash Flow | $99M | $8M | $15M | — |
| Net Cash | — | — | — | — |
Key Ratios (Trailing)
- P/E TTM—
- EV/EBITDA TTM—
- EV/Revenue TTM—
- Price/FCF TTM—
- Gross Margin (TTM)57.2%
- EBITDA Margin (TTM)41.3%
- Net Margin (TTM)23.3%
- ROIC7.1%
- FCF Conversion3.6%
- SBC / Revenue11.6%
The Company
DigitalOcean provides cloud infrastructure. Its reported products fall into three groupings: IaaS (compute, storage, networking, such as Droplets, Spaces, Volumes and Kubernetes), PaaS/SaaS (managed hosting, managed databases, managed Kubernetes, App Platform and Functions), and AI/ML (GPU Droplets, the Gradient AI Platform and Bare Metal GPUs). The company reports one operating segment. Its identity has moved within one operating year, from "agentic inference cloud platform" in the FY2025 10-K to "the AI-Native Cloud, purpose-built for inference and agentic workloads."
DigitalOcean runs on leased data centers rather than owned facilities. The 10-K lists leases in the New York, San Francisco and Atlanta metropolitan areas plus Australia, Canada, Germany, India, the Netherlands, Singapore and the United Kingdom, with plans for new 2026 sites in Memphis, Richmond and Kansas City. It cites 680,000-plus customers and references 20 global data centers. Its model is "largely consumption-based," with contract durations that management says may be only 3 months, 6 months or a year, which it contrasts with neocloud take-or-pay contracts.
Business Segments
Competitive Landscape
Management lines up three competitor sets on the Q2 2026 call. Hyperscalers "optimize for frontier labs and large enterprises." Neoclouds "have built strong GPU rental businesses for model training and are adding software mostly through acquisitions." Inference providers "serve tokens well, but rent their GPUs with margins stacked on margins." DigitalOcean pitches itself as an "open, integrated, full-stack AI-native cloud with data gravity." CEO Paddy Srinivasan's line: "Models and GPUs are not sticky, data is."
- Amazon (AWS)Named in the 10-K as a primary cloud competitor; not otherwise discussed.
- Microsoft (Azure)Named in the 10-K as a primary cloud competitor; not otherwise discussed.
- Google (GCP)Named in the 10-K as a primary cloud competitor; not otherwise discussed.
- Named in the 10-K as AI/ML infrastructure competition. Management says neoclouds are "adding software mostly through acquisitions."
- Named in the 10-K as a smaller/niche cloud competitor; not otherwise discussed.
Supply Chain
DigitalOcean buys GPUs, CPUs and servers and leases data-center space, then sells cloud and inference services to AI application builders and digital-native companies. NVIDIA is named on both earnings calls, with AMD confirmed as a second accelerator source.
More on DOCN: Earnings recap