Google Cloud has launched two new TPUs, TPU 8t and TPU 8i, optimised for AI training and inference, respectively. It has designed the TPUs in-house alongside Google DeepMind models, the company said, so its infrastructure offerings keep pace with what future AI models need.
These innovations form part of Google’s wider AI chip expansion strategy in a bid to be more competitive in the technology race.
Google Cloud CEO Thomas Kurian shared that Google Cloud is one of the first clouds to introduce the Nvidia VR200, the Vera Rubin next-generation AI GPU platform. The goal is to harness its AI infrastructure to power the ‘engine’ of agentic AI, the company said.
“The Agentic Enterprise demands a fundamental shift in data architecture. For a decade, ‘systems of intelligence’ functioned as reactive data archives built for human scale,” Kurian explained. “The Agentic Data Cloud delivers an AI-native architecture – a unified foundation that frees your data from the silos of the last generation – allowing agents to perceive, reason and act on your behalf in real-time.”
Accelerating the agentic era
AI is the cloud’s new foundation, according to Google Cloud, which has propelled itself into the agentic AI era to offer businesses unified AI-enabled solutions.
During a virtual announcement, Kurian shared the top announcements and innovations ahead of Google Cloud Next. He introduced the Gemini Enterprise Agent Platform, a platform for agent development, orchestration and governance.
Kurian described it as an end-to-end system for the agentic era – a comprehensive platform to build, scale, govern and optimise agents.
“Built on top of our leading infrastructure, and deeply integrated with our data and security capabilities, it is the foundation for the Agentic Enterprise,” Kurian explained. “We are announcing everything you need to make your organisation an Agentic Enterprise – anchored by a unified stack designed to turn intelligence into a growth engine for your business.”
The AI platform combines Vertex AI with a range of features to support technical teams with a unified platform for managing and delivering AI agents through the Gemini Enterprise app.
Harnessing an agentic data cloud, Google Cloud said it provides agents with clear, organised data as context for reasoning and analysis. There is also a deep research agent, acting autonomously across structured and unstructured data, in addition to an Agentic Security Operations Center that involves agents that can triage and investigate cyber threats.
Kurian explained how Google Cloud has the capability to build and deploy these agents, but also to scale them – using company data to feed the agents business context. With these tools, he advocated for businesses using their capabilities as a starting point to turbocharge their own innovation
“Gemini Enterprise is now the end-to-end system for the agentic era – the connective tissue between your data, your people and all of your apps and agents that transforms all of your processes into a single, intelligent flow,” Kurian added.
“Transformation to an Agentic Enterprise is the future of every organisation, and Google Cloud uniquely offers a unified system that can take you there. While technology can provide transformational power, you must still provide the vision.”
Infrastructure powering innovation
In the past, customers have needed clouds for storage, application hosting and compute on demand. Now, Google Cloud said customers – regardless of size – are using the cloud to orchestrate the entire AI stack, encompassing chips, data, agents and apps, into autonomous workflows.
To that end, Kurian also announced that Google Cloud would be advancing its AI Hypercompute architecture, with updates designed to handle large-scale and low-latency requirements of the agentic era.
He explained that to ‘do’ AI well, customers need a strong foundation in both data and security – and that Google Cloud can offer both.
“Google Cloud provides a platform to feed AI the data necessary to deliver genuine insights, action and outcomes across every piece of a customer’s business,” the company added in its announcement. “Protecting it all requires security built for the AI era. Not only is our platform secure by default, we also offer leading intel-driven and AI-powered cybersecurity solutions that integrate development and security operations across hybrid and multi-cloud environments to effectively prevent, detect and respond to threats up and down the entire AI stack.”
How Google plans to scale
It seems Google is eager to build on momentum from last year, with the company reporting that 75% of Google Cloud customers were already using its AI in 2025. Kurian added that the company had more $1 billion deals in 2025 than the past three years combined.
The new announcements have arrived shortly ahead of Google’s earnings announcement next week. Parent company Alphabet is expected to announce continued AI and cloud momentum as its main growth catalysts, with analysts predicting a quarterly revenue of $106.5986, as reported by MarketBeat this morning.
Google Cloud has also announced an expanded partnership with BICS, a Proximus Global company, to strengthen cloud-based communications for enterprises, to support BICS in positioning itself as a global provider of PSTN connectivity across more than 40 countries.
Related stories
Google Cloud’s journey towards the ‘agentic telco’ to reshape network AI
Liberty Global, Google Cloud partnership designed to scale AI
