Analyst: Bert Hochfeld | Date: March 25, 2026 | Price: ~85|* * MarketCap : ** 7.8B | EV: ~$9.1B
DigitalOcean has been reinvented. The company I first recommended in August 2021 at an EV/S of 12.6X as a "lower-volatility growth compounder" serving SMBs is today an entirely different business — an AI inference cloud platform with $120M in AI customer ARR growing 150% YoY, a $970M total ARR run rate, and a credible path to 30% revenue growth with Rule of 50 economics by 2027. I initiated a starter position of 500 shares at $59.79 on March 11 and published a detailed deep-dive on March 17. The shares have since appreciated approximately 42%, which requires me to reassess the valuation case at current levels. My conclusion: the thesis remains intact and the long-term opportunity is compelling, but the near-term risk/reward at $85 is less favorable than at $60. I would not chase the stock here. I would wait for a pullback into the 65−72 range to add meaningfully. The position remains at 2.48% — a starter weight appropriate for the current uncertainty around execution on the capacity ramp and the elevated net leverage.
I haven't covered DigitalOcean in the newsletter since my 2021 initiation except for brief mentions in portfolio commentaries. The company I described then — a simple developer cloud for SMBs — went through a brutal deceleration. Revenue growth fell from 29.7% in Q1 FY23 to 11.0% by Q4 FY23 and bottomed at 12.1% in Q3 FY24. NDR cratered from 107% to 96%. The stock cratered from $100+ to the $20s. The old narrative was dead: customers outgrew the platform, churn was persistent, and growth was structurally impaired.
What happened next is one of the more impressive corporate reinventions I have observed in my career. New CEO Paddy Srinivasan, recruited from a hyperscaler background, assembled an entirely new leadership team: CRO Larry D'Angelo, CPTO Vinay Kumar (founding member of Oracle OCI), and CFO Matt Steinfort. They systematically:
The result: revenue growth reaccelerated from 12.1% (Q3 FY24) to 13.3% (Q4 FY24) to 14.1% (Q1 FY25) to 15.7% (Q3 FY25) to 18.3% (Q4 FY25), with guidance for 25%+ exit rate in Q4 FY26 and 30% in FY27.
| Period | Revenue | YoY Growth | Sequential |
|---|---|---|---|
| FY22 | $576M | -- | -- |
| FY23 | $693M | 20.2% | -- |
| FY24 | $781M | 12.7% | -- |
| FY25 | $901M | 15.5% | -- |
| Q1 FY25 | $210.7M | 14.1% | +2.8% |
| Q2 FY25 | $218.7M | 13.6% | +3.8% |
| Q3 FY25 | $229.6M | 15.7% | +5.0% |
| Q4 FY25 | $242.4M | 18.3% | +5.6% |
| Q1 FY26 (guide) | $249-250M | 18-19% | +3.0% |
| FY26E (guide) | $1,075-1,105M | 21% midpoint | -- |
| FY27E (target) | ~$1,420M | 30% | -- |
The growth acceleration is unmistakable. Sequential revenue additions have expanded from $5.8M in Q1 FY25 to $12.8M in Q4 FY25 — the highest since COVID-era quarters. Incremental organic ARR of $51M in Q4 was a company record. Trailing 12-month incremental ARR of $150M surpassed even peak COVID-era levels.
The physics of the 2026-2027 ramp are straightforward and well-articulated by management: 31 MW of new data center capacity coming online across three facilities in 2026. The smallest (6 MW) ramps revenue in Q2; the larger two ramp in H2. This maps directly to the guided Q4 exit rate of 25%+ and the 30% full-year target for 2027. Importantly, management stated that 30% growth in FY27 is achievable purely on committed capacity — no additional investment decisions required.
This is where the investment case either soars or falls flat. Let me be specific about why DigitalOcean's AI positioning is differentiated, and where it is not.
Revenue per megawatt: $20-22M vs. $9-12M for Neo Clouds. This is the single most important metric. DigitalOcean generates roughly 2X the revenue per unit of infrastructure capacity because it layers managed inference services, GPU Droplets, serverless inferencing, databases, storage, bandwidth, and compute on top of bare metal. Core cloud margins run 70-80% vs. bare metal's ~25%. This produces structurally higher EBITDA margins (40%+ vs. 20-25% for training-focused Neo Clouds) while requiring less capital per dollar of revenue.
Stickiness through full-stack integration. When an AI-native company deploys inference workloads alongside databases, object storage, VPC networking, and observability on DigitalOcean, the switching cost rises materially. Bare metal GPU rental is a commodity; a full-stack inference cloud with data dependencies is not. This is why 70% of AI customer ARR comes from non-bare-metal services and growing.
Open source model orientation. Management's emphasis on open source models — 90% cheaper per token than closed-source alternatives, already 30% of OpenRouter traffic — is strategically sound. As inference demand democratizes and companies optimize unit economics, the open source share of inference workloads will grow. DigitalOcean's platform is purpose-built for this dynamic: multi-model orchestration, intelligent routing between open and closed-source models, and predictable unit economics.
OpenClaw as proof of concept. Nearly 30,000 native one-click OpenClaw Droplets created within days of launch — with zero marketing spend. Jensen Huang's GTC callout transferred the halo. This is precisely the kind of organic, developer-pull demand that characterized DigitalOcean's original rise to a $1B run rate.
Customer concentration is low today but AI could change that. Top 25 customers represent only 10% of revenue — excellent diversification. But as AI-native customers scale (some crossing $1M ARR in weeks rather than years), the revenue mix could tilt. This is a dynamic to monitor, not a current risk.
Competition from 32 reported inference providers. SemiAnalysis benchmarks 32 companies in the inference space. CEO Srinivasan correctly analogized this to the early cloud era ("if SemiAnalysis was around in 2012, there would have been 32 IaaS providers"). Most will not survive. But the winnowing process creates uncertainty in the interim.
Relationship with Nebius. My largest equity investment is Nebius, and both are in the neo-cloud AI infrastructure space. Management says real-world overlap is "quite small" — Nebius serves massive multi-node GPU cluster users; DigitalOcean serves agentic inference deployment. I accept this framing for now. But as both companies grow, the Venn diagram may overlap more.
| Metric | FY23 | FY24 | FY25 | FY26E | FY27E |
|---|---|---|---|---|---|
| Revenue ($M) | $693M | $781M | $901M | $1,090M | $1,420M |
| Adj EBITDA Margin | 40.0% | 41.6% | 41.6% | 36-38% | 40% |
| Adj EBITDA ($M) | $277M | $325M | $375M | $392-414M | $568M |
| Adj FCF Margin | 22.5% | 17.2% | 18.6% | 15-17% | 20%+ |
| Adj FCF ($M) | $156M | $134M | $168M | $163-185M | $284M+ |
| SBC % Revenue | 17.1% | 11.6% | 8.9% | ~8% (est) | ~7% (est) |
| Non-GAAP EPS | $1.58 | $1.92 | $2.12 | $0.75-1.00 | ~$3.50-4.00 (est) |
| Rule of 40 | 32.5 | 29.9 | 34.1 | 36-38 | 50+ |
Several observations:
FY26 is a trough year for margins. Adjusted EBITDA margin compresses 400-600 bps to 36-38% because new data center operating expenses hit the P&L months before revenue ramps. This is purely a timing effect — it happened with prior data center additions but at smaller scale. I am not concerned about this.
Non-GAAP EPS drops sharply in FY26. From $2.12 to 0.75−1.00 — a >50% decline. This reflects not just EBITDA margin compression but higher depreciation on new GPU capacity and higher interest expense from the $606M 2030 convertible notes and credit facility. Investors looking at P/E will be alarmed. Those looking at EBITDA and the revenue trajectory will see through it.
SBC discipline is genuine. SBC fell from 17.1% of revenue in FY23 to 8.9% in FY25 and is still declining. In absolute dollars, quarterly SBC has dropped from $36.4M (Q2 FY23) to $19.4M (Q1 FY25). For a company of this size and growth rate, this is commendable. EBITDA less SBC margin of 33% ranks above the 80th percentile of a broad software comp set.
Rule of 40 is not yet achieved at 34 for FY25, but the path to Rule of 50 by FY27 (30% growth + 20%+ FCF) is clearly articulated. This matters because Rule of 50 companies trade at a premium.
This is the area that demands the most scrutiny.
| Item | Amount |
|---|---|
| Cash & Equivalents | $254.5M |
| 2026 Convertible Notes (due Dec 2026) | $325.1M |
| 2030 Convertible Notes | ~$606M |
| Credit Facility (TLA drawn) | $376M |
| Total Debt | $1,295.8M |
| Net Debt | $1,041.3M |
| Finance Leases | $130.5M |
| Net Leverage (including finance leases) | ~3.2x EBITDA (exit FY25) |
| Projected peak net leverage (FY26) | >4.0x |
The 2026 convertible notes are the near-term headline risk. $325M comes due in December 2026. Management has stated they will repurchase or redeem for cash, funded by drawing the remaining $120M Term Loan A (done in February) plus projected cash generation. This should be manageable but leaves limited balance sheet flexibility in H2 2026.
Net leverage >4x in the near term. Finance lease obligations for GPU and CPU investments increase ahead of revenue. Management expects leverage to decline as utilization ramps. For a company with 40%+ EBITDA margins and growing revenue, 4x leverage is tolerable but not comfortable. An equity raise in 2027 to fund further capacity expansion is a real possibility — management acknowledged as much: "May need equity raise in next year to sustain 30%+ growth."
Share repurchases paused. The $100M buyback authorization is in place through July 2027, but near-term capital priorities are "squarely focused on organic growth and balance sheet flexibility." Appropriate, given the leverage profile.
At ~85pershare, marketcap 7.8B, and EV ~$9.1B:
| Metric | Current (TTM) | FY26E | FY27E |
|---|---|---|---|
| EV/S | 10.1X | 8.4X | 6.4X |
| EV/EBITDA | 24.3X | 22.7X | 16.1X |
| Forward P/E (Non-GAAP) | 40X (TTM) | 97X | ~22X |
| PEG (using FY27) | -- | -- | 0.73X |
| FCF Yield | 2.2% | 2.2% | 3.6% |
| Rule of 40 | 34 | 37 | 50+ |
My 3-year CAGR estimate: I use 25%, roughly the midpoint between FY26's guided 21% and FY27's targeted 30%, declining modestly to ~24-25% in FY28 as the capacity ramp matures. For a 25% CAGR company, the prevailing EV/S range in the current market is 4-8X. At 8.4X forward (FY26), DigitalOcean trades at the upper bound of the cohort range.
However, the trajectory matters. Growth is accelerating, not decelerating. If the 30% FY27 target is achieved, the forward EV/S compresses to 6.4X — squarely in the attractive zone for a 25-30% grower. And if DigitalOcean sustains 30%+ growth into FY28, it migrates into the 30-50% cohort where 7-14X EV/S is the norm.
At 22.7X FY26E EBITDA, DigitalOcean trades at a premium to traditional infrastructure companies but at a discount to high-growth software. The FY27 EV/EBITDA of 16.1X is attractive for a 30% grower with expanding margins. One Seeking Alpha analysis highlighted this as "cheap EBITDA multiples" relative to the growth profile.
P/E analysis is not useful for FY26 due to the trough earnings profile. But on FY27 estimates — if non-GAAP EPS recovers to ~$3.50-4.00 (which requires 30% revenue growth, 40% EBITDA margins, and modestly lower interest expense as the 2026 converts are retired) — the forward P/E drops to ~22X and the PEG is ~0.73X. That is attractive territory.
When I initiated the position at $59.79 on March 11, the forward EV/S was approximately 6.5X — a clear discount to the cohort. At $85, it has expanded to ~8.4X. That is a 29% multiple expansion in two weeks, driven largely by the OpenClaw/Jensen Huang GTC halo and general AI infrastructure enthusiasm. The valuation is no longer compelling on a near-term basis, though it remains reasonable on a FY27 view.
DigitalOcean does not compete head-to-head with the GPU training mega-capex Neo Clouds (CoreWeave, IREN). The real competitive landscape is:
| Competitor | Model | Revenue/MW | EBITDA Margin | Differentiation vs. DOCN |
|---|---|---|---|---|
| Nebius | Full-stack AI cloud | Higher (massive scale) | Growing to 40%+ | Larger users, multi-node clusters, lower latency. Minimal real-world overlap today |
| Vultr | Developer cloud + GPU | ~$9-12M | Lower | Recently valued at $3.5B. Narrower platform |
| Linode (Akamai) | Developer cloud | ~$9-12M | N/A (subsidiary) | Less inference-focused |
| Hyperscaler "lite" | Lightsail, Cloud Run | N/A | N/A | Not optimized for inference |
DigitalOcean's 20 − 22MrevenueperMW—decliningto 20M at full AI mix by end of 2027 — remains structurally superior to bare metal competitors. The 70% non-bare-metal AI revenue mix is a powerful indicator that the platform value proposition is real, not marketing.
Guidance track record: Management has consistently beaten quarterly revenue guidance by 1-2.5% through FY24 and FY25 — a pattern of deliberate sandbagging that I find attractive. The Q4 FY25 beat of 2.1% is consistent with this pattern.
Long-term target acceleration: The original Investor Day target (April 2025) was 18-20% growth by 2027. By Q3 FY25, they pulled it forward to 2026. By Q4 FY25, they achieved the bottom end two full years early and raised the bar to 30% by FY27. This is credibility-building behavior — under-promise, over-deliver.
New leadership team: Most of the C-suite is new (CEO, CRO, CPTO, CFO). The track record is promising but shallow. Vinay Kumar from Oracle OCI is an excellent addition. I want to see 2-3 more quarters of execution before upgrading management credibility from "promising" to "reliable."
Execution risk on 31 MW ramp. Three data centers in one year is the most capacity DigitalOcean has brought online simultaneously. Supply chain delays, GPU procurement issues, or longer-than-expected revenue ramp would compress the growth trajectory. Management acknowledges "supply chain and implementation timing risks."
Balance sheet leverage. Net leverage >4x with a $325M convertible maturity in December 2026. A potential equity raise in 2027. This is the most levered the company has been, and it reduces the margin of safety.
Inference commoditization. If inference unit economics collapse faster than expected (a permanent Jevons paradox with no revenue offset), the $20M/MW revenue assumption degrades. Management's counter-argument — that 70% of AI revenue is from managed services, not bare metal — is valid but untested at scale.
AI native customer durability. Character.AI, Hippocratic AI, and OpenClaw deployments are impressive early wins. But AI-native startups have high mortality rates. The breadth of the customer base (top 25 = 10% of revenue) mitigates this, but the AI cohort specifically could prove more volatile.
Geopolitical. Iran conflict (now in week 3) impacts AI infrastructure sentiment broadly. Oil price spikes and rate uncertainty affect multiples. DigitalOcean's equipment leasing model could face challenges if credit conditions deteriorate.
Valuation at $85. After a ~42% appreciation from the $60 entry point, much of the near-term upside has been captured. A pullback to $65-72 (EV/S ~7X forward) would restore an attractive entry point.
DigitalOcean at $85 is not cheap, but it is not expensive for what the company is becoming. The investment case is a 2027 story: 30% revenue growth, 40% EBITDA margins, 20%+ FCF, and Rule of 50 economics on committed capacity alone. At FY27 estimates, the EV/S is 6.4X, EV/EBITDA is 16X, and the PEG is ~0.73X — all attractive for the growth profile.
The near-term is messier: margin compression from capacity ramp, depressed EPS, elevated leverage, and a convertible maturity to manage. These are known quantities, not surprises, and they represent the cost of building a growth engine.
I maintain my position at 2.48% of the high growth portfolio. I would not add at current levels. The stock has run too far too fast on the OpenClaw euphoria. I would look to add on a pullback into the $65-72 range, which would represent a forward EV/S of ~6.5-7.0X — a meaningful discount to the cohort average for a company accelerating toward 30% growth. A second tranche of 500 shares at that level would bring the weight to approximately 4.5%, which is appropriate for a company still proving out its AI transformation.
The Deploy conference on April 28 and Q1 FY26 earnings will be the next critical data points. I am looking for: (1) confirmation that the 6 MW facility is ramping on schedule, (2) AI customer ARR trajectory (sustaining 100%+ growth), (3) RPO growth continuing to signal committed demand, and (4) any early color on 2027 capacity expansion plans.
Rating: Buy on pullback. Hold at current levels. Do not chase.
| Dimension | Prior (March 11-17, 2026) | Updated (March 25, 2026) |
|---|---|---|
| Growth trajectory | Accelerating to 30% by FY27 | Unchanged — Q4 FY25 confirmed acceleration |
| AI thesis | Differentiated inference-first strategy | Validated by OpenClaw traction (30K deployments) and Jensen Huang callout |
| Valuation | Attractive at EV/S ~6.5-9.25X forward | Less attractive at EV/S ~8.4X after 42% appreciation |
| Management | Promising but new team | Unchanged — need 2-3 more quarters of execution |
| Balance sheet | Manageable leverage | Unchanged — peak leverage >4x is the key risk |
| Position sizing | 2.48% starter appropriate | Appropriate — do not add at current levels |
Analysis based on Q4 FY25 earnings (reported 2026-02-24), scout brief data, and prior coverage (2021 initiation, March 2026 deep-dive and portfolio change). Stock price as of 2026-03-25.