Thomas Kurian’s keynote at Google Cloud Next painted an ambitious picture of an “agentic era” in which AI agents, not just chatbots and copilots, are embedded across enterprise workflows.
While this vision isn’t unique to Google, what was distinctive about Kurian’s keynote was how concretely he tied it to a full-stack blueprint — linking silicon, data, security, and the Gemini Enterprise agent platform into a single, opinionated architecture that makes that agentic future operational in the enterprise.
Taken together, five themes stood out in particular for IT leaders.
1. From pilots to the ‘era of the agent’
Kurian opened by declaring that customers have moved beyond experiments:
“The era of the pilot is over. The era of the agent is here.”
He emphasized that most Google Cloud customers already use AI products and framed the main challenge as scaling from isolated use cases to enterprise-wide impact.
The shift in language, from models and copilots to “agents” and “digital task forces,” reflects the mindset of most business and IT leaders today. Kurian wants organizations to view AI as a set of coordinated workers that can orchestrate complex workflows, not merely as a tool for answering questions.
For many enterprises, this keynote sets a directional goal: move from scattered AI pilots to a more systematic approach in which agents are designed, governed, and managed as first-class assets.
2. Gemini Enterprise as the agent control plane
The keynote’s centerpiece was Gemini Enterprise, introduced as a “mission control for the agentic enterprise” and “the environment where your business logic, your data, and your models converge to drive autonomous action.” Kurian positioned it as the evolution of Vertex AI into a broader agent platform.
Key elements included a low-code agent studio for building natural-language agents, an agent registry to track and govern agents across the organization, a skills and tools registry to surface reusable capabilities, and an agent gateway with “agent identity” for policy enforcement and traceability.
The vision is that enterprises will be able to build, secure, and scale agents with the same rigor applied to mission-critical applications.
For IT leaders, the appeal is the holistic nature of Google’s offering. AI agents are no longer one-off projects but part of a unified platform with built-in governance, observability, and lifecycle management. The practical work ahead will be deciding how to integrate Gemini Enterprise with existing integrations, APIs, and low-code investments… and where to standardize on Google’s patterns versus those already in place.
3. AI hypercomputer: Designing for agent-scale workloads
On infrastructure, Google introduced its AI “hypercomputer” concept, with SVP Amin Vahdat noting that “in the agentic era, compute is no longer defined by chip.
Compute is the entire data center.” The keynote highlighted new generations of TPUs optimized separately for training, inference, and reinforcement learning, a custom Axion CPU for general-purpose workloads, and the early availability of Nvidia’s latest GPUs.
These announcements are about more than raw performance; they are clearly shaped by the workloads Kurian is pushing — large numbers of concurrent agents, long-context reasoning, and increasingly complex orchestration. The message to enterprises is that Google’s infrastructure is being optimized not only for foundation model training but also for AI operations at scale in production.
For IT professionals, the key takeaway is that Google is trying to abstract away this complexity behind higher-level platforms like Gemini Enterprise. The details of chip selection, interconnects, and storage throughput matter, but the intent is that most teams consume them via managed services and agent platforms rather than by tuning infrastructure.
4. Agentic data cloud: Putting context at the center
Kurian and his team repeatedly emphasized that “intelligence plus automation must deliver value” and that context is essential to move beyond “intelligent guesses.” The Agentic Data Cloud was introduced to address this, combining:
- A knowledge catalog that automatically enriches both structured and unstructured data, extracting entities and relationships so agents understand business semantics.
- A data agent kit that embeds AI skills into familiar environments like IDEs and notebooks, allowing developers and data practitioners to describe outcomes (“predict churn”) and have pipelines and models scaffolded for them.
- Cross-cloud capabilities, built on open table formats, to query data across clouds with less data movement.
The live demo used the knowledge catalog to discover that a specific ingredient contained soy, then used cross-cloud data to identify affected customers and forecast the impact on demand, showing how these concepts come together in a realistic scenario. For many enterprises, this will be the most relevant part of the story: using AI to turn fragmented data into a trusted context that agents can act on.
The opportunity is substantial. The work for IT and data teams will be to map existing data estates, governance frameworks, and analytics platforms into this new model in a way that adds intelligence without creating another silo.
5. Security, governance, and an ‘open’ agentic stack
Security and trust received significant attention.
Google’s security leadership underscored that “your security must operate at machine speed” and showcased a Gemini-native approach to SecOps in which agents triage, investigate, and help remediate incidents faster. A notable example was the integration with Wiz to discover AI assets, validate risks, and streamline remediation down to specific code changes.
Kurian also articulated a broader stance on openness: support for multiple model providers (including partners like Anthropic), integration standards such as the Model Context Protocol, cross-cloud data capabilities, and a partner ecosystem of specialized agents and tools.
The underlying message to customers is that while Google offers an end-to-end stack spanning everything from silicon to agents, it expects them to bring heterogeneous models, tools, and clouds into that environment.
For businesses, this combination of strong governance, multimodel support, and cross-cloud data access is encouraging. It suggests that adopting Gemini Enterprise and the Agentic Data Cloud does not require abandoning existing investments. The strategic decision will be how central to make Google’s agentic blueprint within your overall AI strategy and how to balance it with other platforms you already rely on.
What this means for the IP practitioner
Taken as a whole, Kurian’s keynote presents a coherent thesis: the next phase of enterprise AI will be driven by agents, not standalone models; by context-rich data platforms, not disconnected silos; and by infrastructure, security, and governance designed with autonomy in mind from the start.
For IT pros, the keynote is less a checklist and more a roadmap. It provides visibility into where Google Cloud is heading, and where IT architecture may need to evolve over the next several years: from pilots to platforms, from copilots to agents, and from fragmented tooling to more unified control planes for AI.
Also read: Google’s $40 billion Anthropic deal shows how cloud infrastructure is becoming the backbone of the AI race.
Read the full article here
