Google Cloud accelerates AI infrastructure with Gemini Enterprise and new TPU chips at Next 2026

Google Cloud’s Next 2026 conference unveiled major advances in AI deployment, featuring Gemini Enterprise control layer, eighth-generation TPU chips, and security-driven autonomous workflows, signalling a shift from pilots to scalable enterprise AI operations.

Google Cloud used its Next 2026 conference to argue that enterprise AI has moved beyond experiments and into production at scale, with chief executive Thomas Kurian framing the moment as the end of pilots and the beginning of agent-driven operations. The company’s central pitch was Gemini Enterprise, presented as a control layer for building, deploying and governing autonomous agents that can carry out whole workflows rather than isolated tasks. TechRadar said the event drew about 32,000 attendees, underlining the scale of Google’s push to position itself at the centre of the AI infrastructure race.

A major part of that message was a fresh emphasis on infrastructure. Google unveiled its eighth-generation TPU family, splitting the line into a training chip and an inference chip to better match the different demands of model building and real-time responses. Axios reported that the move reflects the broader competition among cloud giants to supply AI workloads, while ITPro said the TPU 8t is aimed at accelerating training and the TPU 8i at lowering latency for inference. Google also highlighted its wider “AI hypercomputer” stack, including the Virgo network and custom Axion processors, as part of an effort to make large-scale agent deployment more efficient and more power-conscious.

Google also pushed harder on the data layer that feeds these systems. The company introduced the Knowledge Catalog, which it said is designed to surface and enrich previously buried content so files can be made ready for agent use. TechRadar reported that the Cross-Cloud Lakehouse is built on Apache Iceberg and aims to let customers analyse data across cloud providers without moving it, while further announcements around Smart Storage and related tools were presented as ways to make enterprise information easier for AI systems to understand and act on.

Security was given equal billing. Google described an agentic defence model in which AI triages incidents and shortens investigation times dramatically, and it used the stage to spotlight its closer relationship with Wiz following the security company’s acquisition. According to the event coverage, the live demonstrations were meant to show how “red” agents could probe for weaknesses while “green” agents proposed fixes, reflecting Google’s attempt to fold security into the same autonomous workflow it is selling to customers.

The company also widened the product story beyond infrastructure and security into day-to-day work. ITPro reported that Gemini Enterprise now brings together tools for low-code and code-first agent development, orchestration, observability and evaluation, while Workspace Intelligence is being positioned as a more context-aware layer across Gmail, Docs, Slides and Sheets. Google said the update is intended to reduce fragmentation between apps and speed up tasks such as summarising messages, drafting presentations and resolving comments, and TechRadar said the event also featured demonstrations of customer-facing agents and sector-specific use cases.

Source Reference Map

Inspired by headline at: ^[1]

Sources by paragraph:

Source: Fuse Wire Services