On the occasion of its annual event Google Cloud Next ’26 in Las Vegas (April 22–24), Google Cloud will champion, just like its competitors, the advent of “Agentic Enterprise.” Behind this concept lies a technical promise: AI agents capable of perceiving, reasoning, and acting autonomously to run end-to-end business processes.
Gemini Enterprise, the Pivot Platform
The centerpiece of the 2026 edition is the rebranding and substantial expansion of Vertex AI under the name Gemini Enterprise Agent Platform.
The platform is structured around four core pillars: create (via Agent Studio, a low-code interface, and the Agent Developer Kit), scale (runtime under a second, long-running agents that can operate for up to seven days), govern (Agent Gateway paired with Model Armor), and optimize (OpenTelemetry observability and automated simulations).
The Gemini Enterprise application itself is enhanced with an Agent Designer, a dedicated inbox for agents, “Skills” for automating repetitive tasks, a multi-application Canvas, and native connectors to Salesforce and Slack.
8th-Generation TPUs: Google Challenges NVIDIA on Agentic Inference
On the infrastructure front, Google is making a bold move with two new TPU chips.
The TPU 8t, designed for training, delivers 121 FP4 exaflops, i.e., 2.8 times the power of the TPU Ironwood (7th generation). The TPU 8i, optimized for low-latency inference, promises an 80% reduction in latency for agent swarms. Together, the two chips double the performance per watt compared with the prior generation.
These TPUs are complemented by the Lustre storage system (10 TB/s), the Virgo network, and the Boardfly architecture. Google also maintains an open strategy by continuing to offer NVIDIA GPUs and its Axion processors alongside its own silicon.
Data, Security, Productivity: The Three Other Fronts
Agentic Data Cloud refers to the data layer redesigned for AI-native operations: a Knowledge Catalog to contextualize enterprise data, a Data Agent Kit, and a multi-cloud Lakehouse architecture enabling agents to act directly on data.
On the security front, Google integrates Wiz for threat intelligence, complemented by detection and remediation agents and an AI-APP approach tailored for multi-cloud and hybrid environments.
In Google Workspace, the innovations focus on Drive Projects and shopping agents capable of managing a complete purchasing journey, from discovery to payment, in natural language.
Large-Scale Deployments at Customer Sites
The testimonials presented illustrate an already industrial-scale adoption. Tata Steel operates more than 300 agents in production.
The Home Depot reports voice agents four times faster than previous solutions. Macy’s deployed its “Ask Macy’s” assistant in four weeks. Virgin Voyages shows a 40% reduction in the production time of marketing campaigns.
Other major accounts such as Bosch (AskBosch), Mars, Merck (which adopts an “agentic-first” strategy in R&D), and NASA for the Artemis II mission are also cited among the references. In France, Valeo uses Gemini for Workspace for general productivity as well as for business AI-agent use cases.
Google Cloud claims 14% market share targeted by the end of 2025 and positions this conference as a direct response to OpenAI, Anthropic (a partner as well as a competitor), AWS and Azure.
The differentiated stance emphasizes governance of agents and openness of the stack more than pure code-generation capabilities (a topic reserved for Google I/O in May).
