Google's NotebookLM Tests Personal AI After 52x Growth

This explosive growth is being driven by enterprise adoption, with Google noting that nearly 350 of its cloud and AI customers are each processing more than 100 billion tokens. This scale, detailed in the company's Q4 2025 earnings call, underscores a massive shift of businesses building on Google's first-party AI infrastructure. The company reported that its direct API, used by customers to access models like Gemini, is now handling over 10 billion tokens every minute, a significant jump from 7 billion per minute just the previous quarter.

As its foundational models achieve unprecedented scale, Google is simultaneously pushing the boundaries of how AI can be personalized. According to reports from Fosstodon’s AI Timeline, the company is now testing a new "Personal Intelligence" feature for its NotebookLM product. This experimental tool represents a move from generic, one-size-fits-all AI toward a more individualized experience, though specific details on the feature's capabilities remain undisclosed.

NotebookLM, originally launched as a project-centric AI notebook, is being positioned as a testbed for this more intimate form of artificial intelligence. The development suggests a strategic focus on creating AI that can learn from a user's own documents, ideas, and preferences to become a tailored assistant. This aligns with an industry-wide race to develop agents that understand context and intent on a deeply personal level, moving beyond simple question-and-answering.

The timing is strategic. The computational firepower demonstrated by Google’s 430 trillion token run rate provides the essential engine required to make personalized AI feasible. Training and running models that adapt to individual users is a resource-intensive endeavor, one that only a handful of companies operating at Google’s scale can attempt. This massive backend processing growth is the unspoken prerequisite for the front-end experiments in personalization now being trialed.

This two-pronged approach—massive scale for enterprise and nuanced personalization for individual users—illustrates Google's broader AI ambition. The company is building out the infrastructure that powers the world’s businesses while simultaneously exploring the future of human-computer interaction. The success of projects like a personalized NotebookLM will depend on their ability to leverage that underlying power without losing the individual focus that makes them useful.

What remains to be seen is how Google will navigate the inherent tension between colossal, impersonal data processing and the promise of a truly personal AI. The company has not disclosed a timeline for a wider rollout of the Personal Intelligence feature, nor how it plans to integrate these experimental tools into its broader ecosystem of products like Google Workspace, which has also been integrating more advanced AI capabilities. For now, the message is clear: Google is building AI at every scale, from the planetary down to the personal.