DigitalOcean (NYSE: DOCN) announced the launch of its AI-Native Cloud platform, designed for artificial intelligence inference and agent workloads. The integrated platform combines five layers: infrastructure, core cloud services, inference capabilities, data management, and managed agents.

The platform is currently running production workloads for customers including Higgsfield AI, Hippocratic AI, ISMG, Bright Data, and LawVo. The company unveiled the platform at its Deploy 2026 conference and made it available to customers.

DigitalOcean's analysis shows the AI-Native Cloud costs $67,727 per month for a representative 1 million bookings per month corporate travel agent workload, compared to $84,827 on Baseten plus AWS and $110,337 on AWS AgentCore. The platform includes no egress fees between layers and uses consumption-based pricing.

The platform supports open-source technologies including OpenCode and LangGraph for agent frameworks, PostgreSQL and pgvector for data, and models from DeepSeek, Llama, Qwen, and NVIDIA Nemotron alongside commercial models like Claude and GPT. The infrastructure spans 20 global data centers with NVIDIA H100, H200, and HGX B300 GPUs, plus AMD Instinct MI300X, MI350X, and MI355X processors.

Key features include an Inference Router that uses a mixture-of-experts model to optimize requests for cost and latency, with early customer LawVo reporting a 42% inference cost reduction. The platform offers over 70 open-source and commercial models through a centralized catalog, plus new services including Knowledge Bases for retrieval-augmented generation and Managed Weaviate vector database.

DigitalOcean serves more than 640,000 customers and projects the global market will process over 500 trillion inference tokens daily by 2030, up from approximately 50 trillion currently.