Google is making an enormous financial commitment to AI infrastructure, with planned CapEx for 2026 reaching up to $185 billion, a nearly six-fold increase from 2022. A significant portion of this investment is dedicated to building out ML compute capacity for its cloud business.
The presentation emphasizes a move beyond simple generative AI to "agentic workflows," where autonomous AI agents collaborate to perform complex tasks. Examples include systems of planner, orchestrator, and coder agents for software migration and security agents for threat triage.
Google is extensively using its own AI technologies to rewire internal operations, serving as "customer zero." This has resulted in dramatic, quantifiable improvements in software development (75% of new code is AI-generated), marketing (70% faster campaigns), and cybersecurity (90% reduction in threat mitigation time).
With the announcement of the Gemini Enterprise Agent Platform, Google is addressing the emerging challenge of managing AI agents at scale. The platform provides the "full stack connective tissue" for building, governing, and optimizing agents, moving beyond providing models to offering a comprehensive management solution.
Keep pulling the thread on Sundar Pichai.