DigitalOcean Launches AI‑Native Cloud with Five‑Layer Stack and Inference Router

DigitalOcean has launched its AI‑Native Cloud, a purpose‑built platform that integrates five layers—from silicon to agents—into a single open stack. The stack includes Managed Agents, Data & Learning, Inference Engine, Core Cloud, and Infrastructure, all built on open‑source foundations such as PostgreSQL, MySQL, MongoDB, Valkey, OpenSearch, Kafka, Weaviate, vLLM, SGLang, OpenCode, LangGraph, and CrewAI. Core Cloud now offers a non‑blocking RDMA fabric, RDMA‑enabled NFS, and VPC‑native inference, while new Burstable CPU and MicroVM Droplets (Firecracker‑based) are in private preview for lightweight agent sandboxes. The Inference Engine introduces an Inference Router (public preview) that automatically selects the best model per request, Dedicated Inference and BYOM services (GA), multi‑modal support, batch inference, content safety guardrails, serverless inference, and evaluation tooling. Richmond’s data center is generally available with NVIDIA HGX B300, AMD Instinct MI350X, H100, H200, and MI300/MI325 GPUs, and DigitalOcean now operates 19 data centers and 200+ PoPs, including new liquid‑cooled racks. The Model Catalog has grown by 25 models, including NVIDIA Nemotron 3 Nano Omni, DeepSeek V3.2, Llama 3.3 70B, Qwen 3.5, and MiniMax‑M2.5. Agents can now persist state, use knowledge bases, and capture feedback loops with the new Data & Learning layer, which offers managed knowledge bases, learning & feedback loops, and a private preview of managed Weaviate. This launch positions DigitalOcean as a full‑stack AI infrastructure provider, enabling customers to run agentic workloads without reinventing the stack.