DigitalOcean announced at NVIDIA GTC 2026 a major expansion of its AI inference capabilities, positioning itself as an 'AI Factory' for agentic workloads. Key announcements include a new Richmond data center built exclusively for AI featuring NVIDIA HGX B300 systems with 400 Gbps RDMA fabric, serverless endpoints accessible via build.nvidia.com for models like GLM-5 and Kimi-K2.5, adoption of NVIDIA Dynamo 1.0 delivering up to 3x lower inference costs, and availability of NVIDIA Nemotron 3 Nano with a 1-million-token context window. The partnership also includes NVIDIA NemoClaw, an open-source stack for running autonomous agents. DigitalOcean reports 43,000+ deployments of the OpenClaw agent on its platform.
Table of contents
The Proven Home for AI AgentsInvesting in the “AI Factory”: Deep Cloud & Inference IntegrationPurpose-Built AI Infrastructure: The Richmond Data CenterA Frictionless NVIDIA Ecosystem PartnershipExpanding the Open Source Model CatalogSort: