Google Cloud unleashed a torrent of updates as part of the hyperscaler’s Next event this week, including updates to its data cloud and AI-focused updates to its Cross-Cloud Network platform.

Google Cloud CEO Thomas Kurian during a media briefing ahead of the Next event noted that the data cloud updates are tied to a “big shift” in how AI models “are being now used to execute tasks on behalf of people.”

“We call that the agent transformation, where a model is now going to be used to execute many tasks on behalf of an agent, of a person,” Kurian said. “These agents, the first step they need is access to context, which is understanding the business context that a company has, which, in turn, it requires them to understand the data that the company has.”

Kurian pointed to “three key elements” Google has now launched toward this goal.

The first is a “multicloud, or cross-cloud lake house” that Kurian said allows enterprises to run analysis from Google’s cloud environment to data sitting in Amazon Web Services (AWS), Microsoft Azure, or any data base or software-as-a-service (SaaS) application that supports the Apache Iceberg format.

“And you can do so without moving any data and you can keep your data where it is, and we can run analysis across these clouds,” Kurian explained.

The data cloud also now includes a “multicloud lake house” that opens up the ability to manage and query data across clouds. Kurian noted that this is enabled through the availability of a new “Lightning engine for Spark processing and a knowledge catalog.”

That Lightning engine is said to provide a two-times better price-performance compared to other Spark processing engines in the market, while the knowledge catalog acts as a central reference point for all of the data within an organization.

“What the knowledge catalog does is it takes data, whether they're in tables as well as unstructured data that sits in storage, process them using Gemini and creates a graph,” Kurian explained. “And the graph is a semantic graph of the underlying data that sits in all these systems. When you then use Gemini to process it, or you use a Spark-based application to run machine learning, this graph then is able to translate the questions you're asking into the underlying data sources you need to go from and retrieve the data with high quality.”

The third element is a “deep” research autonomous agent that can do research and analysis across structured and unstructured data.

Cross-Cloud Network gets Gemini-zed

Google also unveiled a number of AI-focused networking previews for its Cross-Cloud Network offering. These included updated to its Gemini Enterprise Application Platform that is tasked with helping enterprises to build, scale, and manage autonomous agents.

Those specific service previews include an Agent Gateway that taps model context protocol (MCP) and agent-to-agent (A2A) agentic protocols to help manage agentic traffic; Ambient networking that is an integrated data plane for Google Kubernetes Engine (GKE) and Cloud Run to provide network service monitoring and service binding; and plans for a Service Extensions preview that uses cloud-native platforms to manage identity, governance, and runtime security for traffic running between different vendor environments.

Google also continued to extend the reach of its Gemini platform using Nvidia’s Blackwell graphic processing units (GPUs), including a preview of its Gemini Flash models now running in those environments for Google Distributed Cloud (GDC) connected customers. This ability joins the previously offered support for GDC air-gapped environments that was hinted at a year ago and officially entered preview last August.

Google’s Cross-Cloud Network offering is based on its Cross-Cloud Interconnect service initially unveiled in early 2023. That service directly connects Google Cloud with AWS, Azure, and other cloud infrastructures, establishing a secure, high-bandwidth connections that can also include private links.