DigitalOcean Holdings, Inc. (DOCN) Earnings

Management highlights

### Acceleration of Growth and Profitability - Q2 delivered 29% YoY revenue growth, more than double the 14% growth rate in Q2 2025, with a record $93 million in incremental ARR, nearly tripling incremental ARR from the year prior quarter. - Profitability remains strong: 40% adjusted EBITDA margin, 24% adjusted operating income margin, and 17% trailing 12-month adjusted free cash flow margin. - The highest-spending customer cohorts have been the fastest growing for 8 consecutive quarters, demonstrating a clear, sustained upward trend in deal size. ### Inference Engine Traction - The managed inference engine, launched in late April 2026, has acquired over 6,000 customers, with average monthly customer growth of ~60% and a 30x increase in token volume over 60 days. - Open-weight model token volume grew from 15% of total token volume post-launch to 75% as of Q2 end, driven by customer demand for cost-effective, customizable production AI. The company was an exclusive launch partner for Kimi K3, the largest open-weight model released to date, acquiring 400 net new customers in the first week of launch. - The inference engine includes full production capabilities beyond basic model endpoints, including intelligent routing, model synthesis, prompt caching, batch inference, and native agent tooling, all optimized on DigitalOcean's owned infrastructure. ### AI-Native Growth Flywheel - A self-reinforcing growth flywheel is emerging: customers enter via inference, agent or compute entry points, then adopt deeper platform layers including databases, storage, observability, and agent runtimes, increasing ARR per customer and margin. - Early evidence confirms the flywheel works: more than half of all new AI customers added year-to-date have attached core cloud products, and leading AI builders including OpenCode, Daytona, Vercel, and OpenRouter have already expanded from initial use cases to full platform adoption. - The full-stack integrated platform is differentiated from competitors: hyperscalers focus on large enterprises, neoclouds rely on acquired software capabilities, and pure-play inference providers rent third-party GPUs, while DigitalOcean builds and owns the entire integrated stack from data center silicon to inference runtime. ### Operational and Balance Sheet Discipline - All planned 2026 data center capacity launches are on or ahead of schedule: Richmond (Q1 2026) and Kansas City (Q2 2026) launched ahead of target, with Memphis on track for H2 2026 launch. - Secured an incremental 20 megawatts of new capacity this quarter, bringing total committed capacity to 155 megawatts, majority online by end of 2027, with new capacity coming online in late 2027 through 2028. - Proactively strengthened the balance sheet by equitizing $472 million of 2030 convertible senior notes, reducing net leverage to 0.7x LTM EBITDA with effectively no shareholder dilution and minimal cash use, freeing up capacity for future growth investments. - Landed the company's first nine-figure annual revenue commitment this quarter, pushing remaining performance obligations to $894 million, up more than 12x YoY with a 3.7-year average contract term.

Guidance

- Q3 2026 revenue guidance is $304 million to $307 million, representing 32% to 34% YoY growth, with adjusted EBITDA margins of 38% to 39%, and non-GAAP diluted EPS of $0.28 to $0.30. - Full year 2026 guidance was raised from prior levels, now calling for $1.17 billion to $1.18 billion in total revenue, representing ~30.5% YoY full year growth, with a Q4 2026 exit growth rate of 35% or higher. Adjusted EBITDA margin is expected to be ~39%, non-GAAP diluted EPS is $1.35 to $1.40, and adjusted free cash flow margin is 11% to 13%, an increase from prior guidance. - While formal 2027 guidance is not yet provided, management increased confidence in its prior estimate of 50%+ YoY revenue growth for full year 2027, with all current operating momentum and higher 2026 exit growth indicating potential further upside to this estimate. - 15 megawatts of planned 2026 capacity remains on track to launch in H2 2026, as previously scheduled.

Segment performance

DigitalOcean reports Q2 2026 total revenue of $281 million, with 29% year-over-year (YoY) growth. AI customer segment ARR reached $234 million, growing 212% YoY, and representing approximately 23% of total company ARR. Within the AI segment, Inference Services (all non-bare metal AI inferencing capabilities) grew nearly 800% YoY, and now accounts for over 70% of total AI customer ARR, and 85% of total AI ARR comes from non-bare metal inference and core cloud services. ARR from customers with $100K+ annual spend grew 98% YoY, $500K+ customer ARR grew 160% YoY, and $1M+ customer ARR grew 214% YoY. These high-spending cohorts collectively grew from 9% of total ARR one year ago to 23% of total ARR in Q2 2026. Core cloud services have a 50%+ attach rate to new AI customers, and over 70% of AI customers with $100K+ ARR attach core cloud products to their AI workloads.

Risks & headwinds

Management did not explicitly outline new material risks in this earnings call. The only risk referenced was the standard forward-looking statement disclaimer that actual results may differ materially from projected guidance due to industry-wide supply chain challenges for GPU and data center capacity, as well as timing uncertainty for new data center implementation that can impact full year 2027 revenue recognition.

Analyst Q&A

• Q: How is DigitalOcean adapting its sales, engineering, and infrastructure operations to serve larger, higher-spending customers after historically focusing on smaller customers, and how is it managing new capacity delivery? /

  A: DigitalOcean has over a dozen years of experience operating a global cloud platform with hyperscaler SLAs, so it already has the core operational scale to serve larger customers. The company added an experienced new CRO and CMO to scale its go-to-market motion for larger customers, and standing up a forward-deployed engineering team to support sophisticated large workloads, with most core platform software requiring minimal customization for these customers. On the infrastructure side, DigitalOcean has strong, diversified partnerships with top data center operators, chip manufacturers, and OEMs, allowing it to deliver planned capacity on or ahead of schedule despite industry-wide supply chain challenges. (357 words)

• Q: How does the recent 30% GPU list price increase impact Q2 growth and 2026 guidance, and what is the expected year-end 2026 net leverage? /

  A: The 30% GPU price increase has already been mostly implemented via renewals and capacity reallocation to higher-value use cases, and its impact was modest on Q2 2026 incremental ARR. The full impact of the price increase is already baked into the updated 2026 guidance, and it contributed to the increase in the expected 2026 exit growth rate. After the $472 million debt conversion, pro forma net leverage stands at 0.7x LTM EBITDA, well below the 4x guidance cap, giving the company significant flexible capacity to fund future growth, and the company expects to remain free cash flow positive on all metrics in 2026. (251 words)

• Q: What is the adoption pattern for large nine-figure customer commitments across DigitalOcean's five-layer AI stack, and what are the go-to-market investment plans for larger deals? /

  A: Nearly all large AI customers enter through one of three entry points (inference, agents, compute) and already attach core cloud services like databases, storage, and orchestration, with over 70% of large-scale AI customers already using multiple platform layers. Agentic AI workloads natively require multiple platform capabilities beyond just tokens or GPUs, so the natural expansion across layers is expected to continue. The current go-to-market priority is nailing a high-quality, engineering-focused sales motion for AI-native enterprise customers with forward-deployed engineering support, rather than rapid scaling of the sales team; product-led growth via ecosystem partnerships remains the primary customer acquisition channel for the inference business. (228 words)

• Q: Is there upside to the prior 50%+ 2027 growth estimate given strong Q2 performance, and is open-weight model token volume mostly recurring production traffic? /

  A: The strong Q2 performance, higher 2026 exit growth rate, and large RPO backlog do indicate potential upside to the prior 2027 growth estimate, but formal guidance is premature due to lingering uncertainty around the timing of new data center capacity launches. Most of the 75% open-weight token volume is recurring production traffic, not discounted batch inference. Open-weight and closed models have similar core cloud attach rates, and most production workloads now use a mix of both model types, a use case DigitalOcean supports with its model synthesis feature. (168 words)

• Q: How will ARR per megawatt evolve going forward, given falling bare metal mix and the growth of inference and attached core cloud? /

  A: Management expects incremental ARR per megawatt to increase over time, reversing the prior temporary decline from the shift to AI capacity. Higher attach rates of higher-margin core cloud and inference services, compared to just bare metal GPU rental, along with higher token throughput from newer GPU generations, will drive higher ARR per megawatt. DigitalOcean already achieves higher ARR per megawatt than neocloud peers due to its multi-layer full stack offering and diversified customer base that avoids discounted single-customer long-term commitments. (134 words)