Seventy-five percent of all new code written at Google is now generated by AI and reviewed by engineers. Sundar Pichai shared that figure at Google Cloud Next in Las Vegas this week, and it landed as a benchmark for where the industry is heading. It is also the context for everything else Google announced at the conference, built around a single argument: the shift from AI as a tool to AI as infrastructure is already underway inside the company that built much of the internet.
A platform for managing agents, not just building them
The most substantial announcement was the Gemini Enterprise Agent Platform, which Google is positioning not as a way to build your first agent but as a way to govern hundreds of them once you have. The platform includes an Agent Registry for cataloguing internal agents and their capabilities, an Agent Gateway that acts as an access-control layer across the whole estate, and support for long-running agents that can work autonomously for days on complex problems without needing to be restarted or monitored at every step.
Google framed the core problem as a shift in what enterprises are asking. Early AI projects have produced a sprawl of chatbots and automations that no single team fully tracks. A governance layer that treats agents the way IT treats software systems is the natural next step, and this platform is an attempt to supply it. Paid monthly active users of Gemini Enterprise grew 40% quarter-on-quarter in Q1 2026, and the platform now processes more than 16 billion tokens per minute through direct API use, up from 10 billion the previous quarter.
New chips built for inference
Google also introduced its eighth-generation Tensor Processing Units in two variants: the TPU 8t for training and the TPU 8i for inference. The 8i connects 1,152 chips in a single pod with 3x more on-chip memory than its predecessor, designed to run many agents simultaneously at low latency. The architecture reflects where production pressure actually falls: once a model is trained, inference is typically the bottleneck, and the 8i addresses that directly.
$750 million for partners
Alongside the platform launch, Google committed $750 million to its partner ecosystem of 120,000 firms, covering AI prototyping, deployment, upskilling, and embedded Google engineers working directly with partner teams. The company also announced a deeper partnership with Salesforce that lets agents from Agentforce and Gemini Enterprise share context and run end-to-end workflows across both platforms, surfacing results inside Slack and Google Workspace.
One concrete internal result worth noting: Google's Security Operations Center now uses agents to triage tens of thousands of unstructured threat reports each month, reducing threat mitigation time by more than 90%. That kind of specific, measurable outcome tends to move enterprise buyers, and Google's decision to lead with its own internal numbers was a notable choice at a conference full of abstractions.
Commentarii · 0