Detailed narrative
AI Native Cloud Strategy
DigitalOcean is building a full-stack AI native cloud, integrating five layers from silicon to inference and agents, with open-source support. This platform aims to eliminate complexity for AI builders, offering an integrated solution for inference and agentic execution, which is critical for production AI applications. The company has shipped over 80 releases across all layers since late April, demonstrating rapid innovation. This integrated approach, running on DigitalOcean's owned infrastructure across 20 global data centers, improves unit economics and differentiates the company from hyperscalers and neoclouds.
Inference Engine Traction
The inference engine, launched in late April, has seen rapid adoption with over 6,000 customers leveraging it. Token volume increased 30x in 60 days, and open-weight models now represent close to 75% of total token volume, up from 15% at launch. This shift is driven by "valuemaxxing," where customers seek the right model at the right cost for business outcomes, a trend benefiting DigitalOcean's offering.
AI Product Innovation & Open-Weight Models
DigitalOcean's inference engine offers advanced features like model synthesis, which orchestrates multiple models in parallel to deliver frontier-grade quality at a fraction of the cost. The platform also provides day-zero access to major open-weight models like Kimi K3, which saw over 400 net new customers in its first week. These models, such as Kimi K3 (140 billion parameters) and upcoming models like Quen3.8 (2.4 trillion parameters), are significant in scale. This focus on open-weight models and advanced serving capabilities is a key differentiator, enabling customers to turn open weights into fast, reliable, economical production tokens.
AI Native Flywheel Effect
An AI-native flywheel is emerging, where customers entering through inference, agents, or core compute are pulled deeper into the platform. This leads to increased adoption of databases, storage, Kubernetes, and observability services. Roughly 70% of AI customers with $100,000+ ARR in Q2 have attached a core cloud product to their AI workloads, showing early evidence of this flywheel in action.
Capacity Expansion and Balance Sheet
DigitalOcean is on track or ahead of schedule for its 2026 data center launches, with Richmond and Kansas City already online and Memphis expected in H2 FY26. The company secured an additional 20 megawatts of capacity, bringing the total committed to 155 megawatts, mostly online by end of 2027. Furthermore, the company strengthened its balance sheet by retiring $472 million of 0% 2030 convertible notes, reducing net leverage to 0.7x and freeing up capacity for future growth.
Customer Growth and Diversification
Growth is increasingly driven by high-spending and AI customers. ARR from $100K+, $500K+, and $1M+ customers grew 98%, 160%, and 214% YoY, respectively. AI customer ARR reached $234 million, growing over 200% YoY, with 85% from non-bare metal services. Despite securing large commitments, the top 25 customers still represent only 20% of ARR, maintaining broad customer diversification.