DigitalOcean pivoting to AI-native cloud and agent runtime billing via Paperspace acquisition

📌 Background

In July 2023, DigitalOcean acquired GPU cloud platform Paperspace for approximately $111 million, filling a gap in GPU compute and machine learning toolchains. In its 2026 earnings call, management introduced the concept of an "AI-native cloud," focusing on providing services for building, deploying, and monitoring agentic applications. The value capture focus shifted toward token and agent streams rather than GPU scarcity premiums, differentiating DigitalOcean from large- and mid-sized GPU clouds that primarily rely on bare metal renting.

👤 Target Customers

Small and medium-sized developers, startup teams, and SMBs, providing a computing, training, and deployment platform for AI application builders and machine learning engineers.

💰 Revenue Streams

1) Instance rentals: GPU instances rented hourly or monthly, with single-card instances priced at about $1,343/month or an effective rate of about $1.84/hour; 2) Subscription tiers: Gradient Pro tier at $8/month and Growth tier at $39/month; 3) Software usage: Pay-as-you-go pricing for software services such as agent runtimes, storage, and databases; 4) Ecosystem expansion: Project-based solution replication and integration service fees for new scenario customers (an opportunistic item without verified observable revenue scale yet).

🧮 Cost Structure

Capital expenditures and depreciation for data centers and GPU hardware such as H100, A100, and A6000; power and bandwidth operation and maintenance costs; Gradient platform R&D and product integration costs; and post-acquisition brand and channel investments.

🛡️ Moat

Leveraging DigitalOcean's simple, low-barrier developer brand and existing base of small and medium enterprise customers, differentiating from hyperscalers through transparent pricing and a minimalist user experience; post-acquisition integration forms a one-stop chain combining compute, training tracking, and deployment.

🔑 Keys to Success

• Integrating GPU rentals with training tracking and deployment tools into a closed loop
• Breaking away from low-margin pure compute through software services such as agent runtimes
• Maintaining a mindset of low prices and ease of use for small and medium developers

⚠️ Risks

• Risk of legacy user churn triggered by the August 2026 GPU price adjustment
• Failure of the transformation narrative to translate into revenue if agent traffic falls short of expectations

🏢 Cases

• DigitalOcean acquired Paperspace in July 2023 for approximately $111 million
• Paperspace Gradient covers the full machine learning lifecycle from building and training to deployment, supporting GPUs such as the H100
• DigitalOcean proposed its AI-native cloud strategy during the fiscal year 2026 earnings call, emphasizing token-centric and agent-stream-centric billing

📊 SWOT Analysis