DigitalOcean & The Inference Economy
For the past two years, the loudest question in the AI industry has been about demand. Critics wondered if the massive compute spending would ever translate into recurring revenue. Following Cloudflare and Fastly’s strong earnings, DigitalOcean has provided a definitive answer. AI is moving from training to real world deployments.
In their Q4 FY2025 earnings, DigitalOcean introduced a new metric: AI customer revenue. This segment grew 150% YoY to reach $120m in ARR. This represents 12% of the company's total annual run-rate. The most critical detail is that 70% of this revenue did not come from renting bare metal GPUs for training. Instead, it came from inference services and general-purpose cloud resources. Management said do not expect this growth to slow. The most exciting number: Inference services revenue grew by 254% YoY. This number alone makes me believe we will see capacity commitment announcements (20-30 MW is my estimate) in Q1 or Q2 due to come online in 2027. That would represent $440m-$660m of incremental revenue when fully utilized.
The growth model for AI-native companies is shifting. Traditional SaaS relied on headcount and seat counts. AI-native companies scale based on inference requests and token consumption. This turns compute from a capital expense into a recurring operational cost.
DigitalOcean is positioning itself as the agentic inference cloud. They are not chasing the hyperscaler arms race of training foundation models. They are focusing on the application layer. This is where AI startups actually live and breathe.
Key Financial Benchmarks
The financial profile of DigitalOcean shows a significant advantage over other neo-clouds. They are achieving efficiency that competitors cannot match.
ARR per Megawatt (MW)
DOCN: $22m
Neo-Clouds: $9m – $12m
Top 25 Customer Concentration
DOCN: 10%
Neo-Clouds: 70% – 80%
GAAP Operating Margin
DOCN: 17%
Neo-Clouds: -90% to +5%
The company is currently on a path to reach the Rule of 50 in 2027 using only their existing committed capacity. This is a rare combination of 30% revenue growth and 20% free cash flow margins.
The OpenClaw Catalyst
The emergence of OpenClaw signals a shift from centralized chatbots to autonomous agents. DigitalOcean has become the primary infrastructure for the developer ecosystem building these tools. The adoption rate has been unprecedented.
Within days of the OpenClaw launch, users created nearly 30,000 one-click Droplets.
Thousands of additional deployments were activated through the App Platform.
This surge in activity confirms that developers prefer simplified GPU access and predictable pricing. It allows small teams to scale agentic gateways without the complexity of hyperscaler environments.
Looking Ahead to 2026
The market is beginning to differentiate between training and inference. While the focus has been on who owns the most hardware, winners will be whoever powers the most applications. This is why DigitalOcean and Fastly are positioned so well for the coming year. Fastly serves the edge, while DigitalOcean carries the production workloads.
Fund managers will likely see these stocks as a pass-through play on AI revenue. When AI startups succeed, that growth flows directly to the infrastructure provider. DigitalOcean is guiding for 21% growth in 2026 and accelerating to 30% in 2027. Any additional capacity commitments represent pure upside to these numbers.
Management is successfully moving beyond the low-end hobbyist market. DigitalOcean is maturing alongside its users as they scale. The $1m+ customer cohort is now the primary growth engine for the entire company. This group reached $133m in ARR and grew 123% YoY. These are more than just larger accounts, they are structurally stickier. This cohort saw 0% churn in Q4 and carries a Net Dollar Retention (NDR) of 115%. Half of all revenue growth in 2025 came from these larger customers. This proves the platform can support scale as startups turn into enterprises.