At Google I/O 2026, NVIDIA and Google Cloud announced new initiatives aimed at accelerating AI application development for more than 100,000 developers participating in their joint developer community.

Originally launched at Google I/O last year, the developer community brings together developers, machine learning engineers and data scientists seeking hands-on experience with NVIDIA’s full-stack AI platform on Google Cloud infrastructure. The initiative offers curated learning paths, technical labs, codelabs and live events focused on AI development, deployment and optimisation.

This year, the companies introduced several new resources for community members, including a dedicated learning path for using the JAX machine learning framework on NVIDIA GPUs, a new NVIDIA Dynamo codelab focused on inference optimisation, and monthly developer livestreams designed to support AI experimentation and production deployment.

According to the companies, the community has evolved into a central hub for developers building production-ready AI applications using NVIDIA-accelerated technologies on Google Cloud. Over the past year, developers have leveraged the platform to build retrieval-augmented generation (RAG) applications on Google Kubernetes Engine (GKE), implement observability for AI agent workloads, and experiment with hybrid cloud and on-premises inference architectures.

The collaboration also highlights growing support for open AI frameworks and models. Developers can combine Google DeepMind’s Gemma 4 models, NVIDIA Nemotron open models and Google’s Agent Development Kit with Google Cloud G4 virtual machines powered by NVIDIA RTX PRO 6000 Blackwell GPUs to develop and deploy multi-agent AI applications.

The companies further emphasised their work around JAX optimisation, enabling developers to scale workloads from single-GPU experimentation to multi-rack AI training deployments using NVIDIA AI infrastructure on Google Cloud. These optimisations are also being extended into Google Cloud AI Hypercomputer environments, where the MaxText framework is used for efficient large-scale model training.

Additionally, NVIDIA Dynamo on GKE is being positioned as a tool for optimising large-scale AI inference, including mixture-of-experts models, helping organisations improve efficiency and scalability for enterprise AI workloads running on NVIDIA-accelerated Google Cloud infrastructure.

To further support developers, NVIDIA and Google Cloud announced that new training resources, including the JAX-on-NVIDIA-GPU learning path and the NVIDIA Dynamo inference codelab, will become available next month through the joint developer community platform.

The expanded collaboration reflects the companies’ continued focus on accelerating enterprise AI adoption through open frameworks, optimised infrastructure and developer-focused AI tooling.