Nvidia and Google Cloud used this week's Google I/O to mark the first anniversary of their joint developer community, which has grown to more than 100,000 members since its launch at the same event last year.
The numbers are a vanity metric on their own. What makes the programme worth watching is what it reveals about where both companies see the bottleneck in the AI buildout shifting. The constraint is no longer chips or cloud capacity. It is people who know how to use them.
From labs to production
The community offers curated learning paths, hands-on labs and monthly livestreams designed to move developers from prototyping to production on Nvidia's full AI stack running on Google Cloud. New additions this year include a learning path for the JAX library on Nvidia GPUs and a codelab focused on inference optimisation using Nvidia Dynamo.
The practical output from the first year's cohort gives a sense of what developers are building: production-ready retrieval-augmented generation applications on Google Kubernetes Engine, observability tooling for agent workloads, hybrid on-premises and cloud inference for enterprise data pipelines, and sports analytics systems. These are not research projects. They are the plumbing of commercial AI deployment.
The platform play
The developer community sits inside a broader partnership that has expanded fast. At Google Cloud Next in April, the two companies announced Nvidia Vera Rubin-powered A5X bare-metal instances, confidential VMs with Nvidia Blackwell GPUs and agentic AI capabilities combining Nvidia's Nemotron open models with Google's Gemini Enterprise Agent Platform.
Nvidia was also the first industry partner to work with Google DeepMind on SynthID, the AI watermarking technology that embeds digital watermarks into AI-generated content. That collaboration now extends to Nvidia's Cosmos world foundation models.
The customer list for the combined platform includes OpenAI, Salesforce, Snap, CrowdStrike, Schrodinger and Thinking Machine Labs. Google named Nvidia its Partner of the Year in both AI Global Technology and Infrastructure Modernisation Compute.
Why this matters beyond the press release
The AI industry has a skills gap that is widening as the technology moves from model training into agent-based production systems. Every major cloud provider is racing to lock developers into its ecosystem before they settle on a workflow. The Nvidia-Google partnership is an attempt to make that choice for them by bundling Nvidia's software tools with Google's infrastructure and distribution.
Related reading
- Nvidia earnings to test Wall Street's AI spending thesis
- Nvidia takes $2.1bn stake option in former bitcoin miner IREN to build 5 gigawatts of AI data centres
- Google wants an AI agent sitting inside every marketer's workflow
For Nvidia, it extends the CUDA moat into cloud-native territory. For Google Cloud, it ensures that developers building on Nvidia hardware are doing so on Google's platform rather than AWS or Azure.
The 100,000 figure is a starting point. The more interesting number will be how many of those developers are still building on the platform in a year, and how many have shipped production workloads. Developer communities are easy to launch and hard to sustain. The learning paths and codelabs are the substance behind the headline, and the focus on inference optimisation and agent workloads tells you exactly where both companies think the commercial demand is heading next.
