$DOCN EXECUTIVE CALL SUMMARY: DigitalOcean Holdings Inc (02/24/26)
Q4 2025 results and 2026 guidance were positioned as an inflection in both growth and strategic narrative, anchored on AI inference demand and a re-acceleration driven by scaled digital-native enterprise customers. Revenue in Q4 was $242 million, up 18% year-over-year, with full-year 2025 revenue of $901 million. Management emphasized record organic incremental ARR of $51 million in Q4 and $150 million over the trailing 12 months, framed as exceeding prior peak periods. Profitability remained high in 2025, with Q4 adjusted EBITDA of $99 million (41% margin) and full-year adjusted EBITDA of $375 million (42% margin), alongside trailing 12-month adjusted free cash flow of $168 million (19% margin). Gross margin was 59% in Q4, down approximately 2.3 percentage points versus the implied prior-year quarter, consistent with mix and capacity ramp commentary.
The call’s primary incremental information content was the magnitude and confidence of the forward growth trajectory, the explicit capacity-to-growth bridge, and expanded disclosure around AI-related demand signals. Full-year 2026 revenue growth guidance increased to 19% to 23% (21% midpoint), above the 18% to 20% growth outlook referenced as having been provided the prior quarter. Exit growth was guided to 25%+ by Q4 2026, with a stated path to 30% growth in 2027 based solely on already committed data center capacity. This growth acceleration was explicitly tied to 31 MW of incremental data center capacity coming online in 2026 across 3 facilities, with revenue ramp beginning in Q2 for the smallest site and in 2H 2026 for the remaining 2 sites. The cost profile was characterized as front-loaded, with near-term gross margin and adjusted EBITDA pressure expected as leases and depreciation are incurred ahead of utilization and revenue.
Customer mix and retention signals were presented as the second major pillar of the investment message. Digital-native enterprise ARR reached $604 million in Q4 (62% of total ARR) and grew 30% year-over-year, while the largest customers were described as the fastest-growing and most durable cohort. $1 million customers reached $133 million in ARR, up 123% year-over-year, with 115% net dollar retention and 0% churn in Q4 (and 0% average churn over the last 12 months). AI customer ARR was disclosed at $120 million in Q4, up 150% year-over-year, and management introduced “AI customer revenue” as a new metric concept to reflect that AI-native customers consume both inference and broader core cloud services. RPO increased to $134 million, up 121% sequentially and “close to 500%” year-over-year, positioned as improved visibility in a business historically associated with shorter-duration consumption patterns.
Key investment implications center on whether this call marks a durable re-rating catalyst via (1) credible, capacity-supported growth re-acceleration to mid-20% exit rates in 2026 and 30% in 2027, (2) evidence that scaled AI-native and cloud-native customers expand and remain sticky on the platform, and (3) an economic model that sustains high EBITDA and free cash flow margins while funding AI infrastructure through leases and selective financing tools. The principal risks highlighted or implied include execution risk on data center delivery timelines and utilization ramps, near-term margin compression and leverage expansion during the buildout, potential volatility of AI-native consumption, and the competitive response from hyperscalers, neo-cloud GPU lessors, and inference-layer providers.
CALL POSITIONING AND STRATEGIC NARRATIVE SHIFT
Prepared remarks were unusually expansive and functioned as an investor update layered on top of quarterly results. The strategic framing asserted that AI is driving a structural shift from “seats to tokens” and from experimentation to production-scale, long-running agentic systems, with DigitalOcean positioning itself as a vertically integrated “agentic inference cloud” rather than a commodity GPU capacity provider. The most repeated thesis statement was that scale customers are no longer a headwind: “Scaling our top customers was once a constraint. Today, it’s our growth engine.” The second thematic anchor was differentiation versus both neo-cloud GPU lessors and hyperscalers: “We don’t just rent GPUs. We run production AI.”
This narrative matters because it attempts to redefine the company’s category positioning and thus the valuation anchor. DigitalOcean historically has been associated with simpler developer-centric cloud infrastructure with SMB characteristics (high breadth, smaller accounts, and shorter customer decision cycles). The call asserted that the business is now defined by production AI workloads and scaled digital-native enterprises, supported by enterprise SLAs, compliance use cases, and increasing contractual visibility (RPO). If the market accepts this repositioning, the relevant comp set and expected terminal growth rate can shift, but the burden of proof rises sharply and places more weight on execution against the 2026 to 2027 growth and margin roadmap.
The transcript includes multiple “Technical Difficulty” interruptions, and several examples were described without complete detail. The core financial and guidance statements were sufficiently repeated and consistent across CEO and CFO remarks to remain interpretable, but certain product examples and slide references were not fully captured in the text.
Q4 2025 PERFORMANCE: WHAT IMPROVED AND WHAT DID NOT
Top-line growth accelerated to 18% year-over-year in Q4 on revenue of $242 million, implying approximately $205 million of revenue in the prior-year quarter. Management described this as a 500 basis point acceleration versus the same period a year ago, implying prior-year Q4 growth was approximately 13% and highlighting a multi-quarter re-acceleration pattern rather than a single-quarter anomaly.
Profitability remained strong, but the margin trajectory implied near-term pressure already present. Q4 gross profit was $142 million (59% gross margin), up 13% year-over-year, implying prior-year gross profit of approximately $126 million and an implied prior-year gross margin of approximately 61%. The delta is directionally consistent with management’s repeated emphasis that AI mix and capacity-related costs can pressure gross margins, even while adjusted EBITDA margins remain resilient. Adjusted EBITDA in Q4 was $99 million (41% margin), and full-year adjusted EBITDA was $375 million (42% margin), a level that remains unusually high for an infrastructure-oriented model and suggests the operating expense structure has not expanded proportionally with the growth investments being made.
A central proof point cited for the “quality” of growth was organic incremental ARR. Management stated Q4 delivered $51 million of incremental organic ARR (a company record) and $150 million over the trailing 12 months, described as balanced across AI and cloud customers. This is important because it suggests net new demand plus expansion is not isolated to AI-only GPU consumption but is also being expressed across the broader platform. However, the precise definition of “incremental organic ARR” and how it bridges to revenue is not fully specified in the transcript, limiting the ability to independently validate sustainability without subsequent disclosure.
Customer mix and cohort dynamics were positioned as the strongest internal validation of durability. Digital-native enterprise ARR was reported at $604 million in Q4, 62% of total ARR, up 30% year-over-year. This implies total ARR of approximately $974 million, consistent with the “$1 billion revenue run rate” milestone referenced for December. Cohort growth rates were reported as extremely high at the top end: $100,000 customers growing 58%, $500,000 customers growing 97%, and $1 million customers growing 123% year-over-year. The retention profile improved with size: 102% NDR at $100,000, 106% at $500,000, and 115% at $1 million. “Churn for our $1 million customers was zero in Q4” and was stated to have averaged 0% over the last 12 months. The analytical takeaway is that the highest-value cohort is expanding rapidly with very low loss rates, which can create a compounding effect on growth if the cohort can be expanded in count while preserving retention characteristics.
The AI dimension was quantified more explicitly than in many prior periods, at least within this transcript. AI customer ARR was stated to be $120 million in Q4, growing 150% year-over-year (implying approximately $48 million in the prior year). The CEO emphasized that 70% of AI customer ARR was already coming from inference services and general-purpose cloud products rather than bare-metal GPU rentals, reinforcing the thesis that AI customers increase wallet share across the integrated stack and may become sticky on higher-margin services over time. The CFO also characterized the $1 million customer base as approximately 50/50 AI versus core cloud on an account basis, and “a little bit more AI” on an ARR basis, suggesting that AI is already material in the largest cohort without being the sole driver.
Visibility was materially emphasized through RPO, which was disclosed at $134 million, up 121% sequentially and close to 500% year-over-year. In a business historically associated with monthly consumption and limited contractual duration, this signal matters as a bridge between management’s medium-term growth confidence and the typical skepticism around forecasting consumption-based infrastructure demand. The CFO stated that “a decent chunk” of RPO is AI, but not all, which implies contractualization is not purely an AI phenomenon and may also reflect the scaled DNE cohort. The principal analytical caveat is that RPO quality depends on cancellation terms, ramp schedules, and the extent to which obligations are tied to specific capacity provisioning; those details are not provided in the transcript.
GUIDANCE AND OUTLOOK: WHAT CHANGED VERSUS PRIOR GUIDANCE AND WHY
The most concrete change versus prior guidance was the upward revision to 2026 revenue growth. Full-year 2026 revenue growth was guided to 19% to 23% (21% midpoint), explicitly described as above the 18% to 20% growth outlook shared the prior quarter. Management also described a longer-term evolution of the outlook: an original plan to reach 18% to 20% growth by 2027 was stated to have been pulled forward by a year on the prior earnings call (to 2026), and the company now stated it has already reached the bottom end of that range in Q4 2025 at 18% growth. The new path is 21% growth for full-year 2026, 25%+ exit growth by Q4 2026, and 30% growth in 2027.
Short-term quarterly guidance provided additional phasing detail. Q1 2026 revenue was guided to $249 million to $250 million (18% to 19% year-over-year), implying modest sequential growth off Q4 and a continued mid-to-high teens year-over-year run rate. Q2 2026 growth was guided to remain around 18% to 19%, with acceleration beginning in Q3 and exiting at 25%+ in Q4. This implied shape is critical: the model is being positioned as supply-enabled and capacity-gated in 1H 2026, with revenue acceleration following the ramp of new facilities.
A notable “clean-up” item was the planned sunset of a “small, legacy, dedicated bare metal CPU” offering, with approximately $13 million of ARR expected to roll off by end of Q1 2026. The CFO stated that this revenue is non-core and has been excluded from customer-specific year-over-year growth metrics, and the company provided an “ex-legacy” view of 2026 growth: 21% to 24% projected growth excluding the discontinuation impact, versus 19% to 23% reported. This is analytically relevant for 2 reasons. First, it implies underlying demand is strong enough that management is willing to absorb a near-term headline growth headwind to concentrate investment on higher-return growth levers. Second, it introduces a normalization issue for investors and models, as headline growth in early 2026 will be mechanically depressed and comparisons will require explicit adjustments.
Margin guidance signaled a deliberate near-term trade-off: growth acceleration at the cost of near-term margin compression. Full-year 2026 adjusted EBITDA margin was guided to 36% to 38%, a material step down from the 42% achieved in 2025, with Q1 2026 adjusted EBITDA margin guided to 36% to 37% versus 41% in Q4. The CEO described this as “measured near-term pressure on gross margin and adjusted EBITDA” driven by 31 MW of incremental capacity ramping, and characterized the pressure as “a physics problem” tied to the timing mismatch between startup costs and revenue ramp. The CFO reinforced that lease and depreciation costs hit “several months before we generate our first revenue in these facilities,” leading to an “upfront drops in gross margin and net income” that are larger than prior cycles due to the magnitude of capacity being turned up at once.
Despite EBITDA margin compression, unlevered adjusted free cash flow margin guidance was reiterated at 18% to 20% for 2026 (with $207 million at midpoint). This combination implies that working capital, capex timing, and non-cash expenses (especially depreciation) are expected to remain supportive of cash generation even as accounting profitability compresses. However, the call made clear that cash flow interpretation is complicated by equipment leasing and the distinction between unlevered, levered, and “all-in” cash metrics. The CFO emphasized that principal payments on finance leases are financing transactions and are excluded from adjusted free cash flow measures, and stated that even if “the kitchen sink” is included, the company expects to remain cash generative while accelerating growth.
Non-GAAP EPS guidance for 2026 was $0.75 to $1.00 on 111 million to 112 million fully diluted shares, which is directionally low versus 2025 non-GAAP EPS of $2.12 (on 2025 weighted average shares of approximately 105 million, per CFO commentary). The transcript does not provide a full bridge, but the combination of (1) EBITDA margin compression, (2) higher depreciation and lease-related expense timing, (3) higher interest expense from expanded financing toolkit, and (4) a higher diluted share count creates a plausible framework for EPS compression even as revenue grows. The company repeatedly directed attention to adjusted EBITDA and free cash flow margins as the more relevant profit measures during the infrastructure ramp.
CAPACITY, UNIT ECONOMICS, AND THE “ARR PER MW” FRAMEWORK
The call anchored the multi-year growth outlook to committed physical capacity. The CFO stated the existing estate is approximately 43 MW to 44 MW, and that 31 MW will be added in 2026, taking the footprint to approximately 75 MW when fully online. This is a substantial capacity expansion relative to the base and is the explicit driver of the forecasted acceleration in 2H 2026 and continued growth in 2027 through utilization increases.
Management and analysts also discussed an internal efficiency metric: ARR per MW. The CFO stated this is currently $22 million and is expected to be approximately $20 million by end of 2027, with the decline attributed to mix as AI becomes a larger share of ARR. Neo-cloud peers were referenced as having public data indicating $9 million to $12 million ARR per MW, and DigitalOcean positioned its superior efficiency as evidence that a full-stack platform captures more wallet share per unit of power than a bare-metal model. This metric matters because it frames return on incremental capacity and helps translate MW additions into an ARR and revenue trajectory. The critical analytical question is whether the ARR per MW advantage persists as AI scales (where COGS are higher) and as competition intensifies. Management’s thesis is that the advantage is structural because 70% of AI customer ARR is already non-bare-metal and is expected to rise, increasing attachment of higher-margin services (databases, storage, networking) over time.
The call also offered insight into investment discipline and obsolescence management. Management stated a preference not to “go all in on 1 generation of GPU technology” and to “stripe out the generations,” implying a portfolio approach to GPU deployment across AMD and NVIDIA generations. This mitigates risk of rapid hardware obsolescence and potential pricing resets on older GPUs, but it can also introduce operational complexity and supply constraints if customer demand is concentrated in a specific new generation.
FINANCING, FREE CASH FLOW QUALITY, AND BALANCE SHEET IMPLICATIONS
The CFO highlighted an expanded financial toolkit to support infrastructure investment, including equipment financing and leasing, with the explicit goal of aligning investment timing with revenue. This approach can reduce upfront cash capex volatility and improve near-term reported free cash flow, but it increases fixed payment obligations and raises the importance of utilization and demand durability. The CFO framed the payback discussion as less relevant under leasing mechanics: “we’re already paying our gear back within a month or 2 because we’re earning more cash than we’ve spent on that gear,” and stated that the company still targets approximately 3-year paybacks on most investments, with flexibility to extend for strategic customer wins. The analytical interpretation is that leasing transforms capex into an operating leverage and liquidity management problem: the economic success of the model becomes more sensitive to sustained utilization, pricing power, and churn, because the cost base becomes more fixed through lease obligations.
Leverage and maturity profile were directly addressed. The company described having proactively addressed 2026 convertible note maturity through an $800 million bank facility, issuance of $625 million of 2030 convertible notes, and repurchase of the majority of the 2026 notes. The remaining 2026 note balance was stated at $312 million, with an intention to redeem or repurchase for cash before or at maturity in December 2026. Net leverage was stated at approximately 3.2x, with a projection that net leverage will move above 4x in the short term as finance lease obligations are added ahead of revenue and EBITDA ramp. This is an explicit near-term risk factor, partially mitigated by the claim that utilization and earnings growth should bring leverage back below 4x over the medium to long term.
Capital allocation messaging remained growth-first. 2.4 million shares were repurchased in 2025 for $82 million at an average price of approximately $35, and a $100 million authorization remains in place through 07/31/27. SBC declined to 9% of revenue in 2025 from 12% in the prior year. Management stated share repurchases remain a long-term tool, but near-term priorities are organic growth and balance sheet flexibility, consistent with the intensity of the 2026 capacity expansion.
PRODUCT STRATEGY AND COMPETITIVE POSITIONING: WHY MANAGEMENT BELIEVES THE PLATFORM WINS
The CEO’s strategic claims centered on vertical integration for production inference and agentic applications. The thesis was that inference workloads require more than GPU access: “They need compute, storage, databases, networking, observability, security, all working seamlessly together with predictable and transparent unit economics.” This is a direct critique of both neo-cloud GPU lessors (commodity infrastructure) and inference wrapper providers (API layer without full cloud primitives), and implicitly a critique of hyperscalers (complexity and cost structures aimed at large enterprises rather than AI-native builders).
Several customer and ecosystem examples were used to substantiate the platform’s production-grade claim:
https://t.co/j6DFFriv2O was cited as a scaled AI-native customer where DigitalOcean achieved “100% throughput increase” and “roughly 50% lower cost per token” on production-scale inference using AMD Instinct GPUs, described as live traffic across “tens of millions” of customers. This example supports the narrative that performance engineering and hardware/software co-optimization can create differentiation beyond raw GPU access. The limitation is that this is a company-provided case study without disclosed methodology, and repeatability across customer workloads and across GPU generations remains a key question.
Hippocratic AI was cited as selecting DigitalOcean for HIPAA-compliant clinical AI workloads, with optimization of a multimodal deployment on NVIDIA hardware. This supports the claim that compliance, security, and reliability attributes are competitive differentiators for certain vertical AI-native workloads, and that the platform supports heterogeneous GPU ecosystems.
OpenClaw was cited as an open-source agent framework that became viral, with “nearly 30,000” agents running within days of launch, with no marketing claimed. The implication is that DigitalOcean’s developer-centric distribution and simplicity can translate into rapid adoption of emerging AI frameworks, creating a demand funnel that can later convert into production-scale consumption. Monetization, retention, and durability of these deployments were not quantified, and the example is better interpreted as an early indicator of developer mindshare rather than a direct revenue driver at this stage.
A key strategic bet embedded in the call was the centrality of open-source models for inference unit economics. The CEO stated, “the cost per token for the open-source models is about 90% cheaper,” and cited external routing ecosystem data indicating “30% of the traffic already today is served by open-source.” This underpins a product roadmap that includes multi-model orchestration and “intelligent routing” across open and closed models to optimize cost, latency, and accuracy. The strategic upside is significant: if open-source inference adoption rises, providers that offer performant, low-cost deployment of open models with integrated cloud services can gain share from closed-model-only workflows. The strategic risk is commoditization: as open-source models become widely available and model-serving frameworks standardize, differentiation may shift to distribution, reliability, compliance, and economics at scale. The call’s emphasis on being “an integrated cloud” rather than “a token provider” is an attempt to defend against that commoditization vector.
The competitive landscape commentary also included a concentration argument. Management stated the top 25 customers represent only 10% of revenue, contrasting with neo-cloud models characterized as having extremely high concentration. This matters for risk management, because a growth model centered on scaling AI-native customers can otherwise imply heavy concentration and customer-specific demand volatility. The counterpoint is that if the largest cohort continues to grow at triple-digit rates, concentration can rise mechanically even if the current base is diversified, and monitoring concentration trends becomes important.
