Google Cloud Next ’26, concluding today, showcased the company’s vision for what it calls the agentic enterprise, with a focus on enabling virtualization, cloud infrastructure, and operational teams to manage agents at scale. Key announcements included the unveiling of the Gemini Enterprise Agent Platform, the release of two eighth-generation Tensor Processing Units (TPUs), AI networking enhancements, and updates to cross-cloud compute and connectivity.
The Gemini Enterprise Agent Platform emerged as a focal point, described by Google as a comprehensive solution for building, scaling, governing, and optimizing agents. This platform marks a significant evolution from the company’s previous Vertex AI services, which will now be integrated into the Agent Platform rather than offered as standalone products. Google emphasized that the platform is designed to meet the demands of production agent deployments across various environments, including Google Cloud, other cloud services, on-premises infrastructure, and distributed systems.
Google has structured the Gemini Enterprise Agent Platform around four main areas: build, scale, govern, and optimize. New features include Agent Studio, an enhanced Agent Development Kit, and Agent Runtime, among others. The platform offers access to over 200 models through its Model Garden, which includes Google’s Gemini 3.1 Pro, Gemini 3.1 Flash Image, Lyria 3, and Gemma 4, along with third-party models such as Anthropic Claude Opus and Sonnet. The emphasis here is on treating agents as managed enterprise workloads with robust lifecycle control rather than isolated AI applications.
In a further bid to enhance computational efficiency, Google introduced two specialized TPU chips: TPU 8t, aimed at extensive training, and TPU 8i, optimized for low-latency inference. The TPU 8t is designed for compute-intensive workloads and can scale up to 9,600 chips in a superpod, yielding 121 ExaFlops of computing power. This represents a nearly threefold increase in compute performance per pod when compared to the previous generation. The TPU 8i focuses on inference tasks, boasting enhanced memory and bandwidth capabilities that deliver an 80% improvement in performance-per-dollar relative to its predecessor.
Another significant development is the Virgo Network, introduced as a foundational AI data center fabric designed for scale and efficiency in AI workloads. The architecture segments network functions into three layers to facilitate smoother communication between components. Google claims the network can connect up to 134,000 TPU 8t chips, offering a staggering 47 petabits per second of bi-sectional bandwidth.
Google also announced enhancements to its cross-cloud infrastructure, introducing fluid compute features along with secure cross-cloud connectivity and a unified data layer. The new C4N VM family is designed to handle demanding workloads, while the M4N VM targets large-scale data I/O. Among the most relevant updates for Kubernetes users is the GKE Agent Sandbox, which utilizes trusted gVisor isolation to offer improved price-performance metrics when running AI agents.
The concept of the Agentic Data Cloud was also put forth, aiming to provide agents with governed access to data and a comprehensive business context. Featuring a universal context engine and an AI-native cross-cloud lakehouse, this architecture is poised to facilitate efficient data management across platforms. The Knowledge Catalog, a pivotal component, enables context aggregation across various data platforms, enriching data continuously and supporting access-control-aware searches.
As Google Cloud Next ’26 draws to a close, the announcements reflect a strategic shift towards integrating advanced AI capabilities into enterprise infrastructure. With the Gemini Enterprise Agent Platform at the forefront, Google is positioning itself as a leader in managing complex AI workloads across multiple environments, targeting the growing demand for scalable, secure, and efficient data operations in an increasingly connected world.
See also
Bank of America Warns of Wage Concerns Amid AI Spending Surge
OpenAI Restructures Amid Record Losses, Eyes 2030 Vision
Global Spending on AI Data Centers Surpasses Oil Investments in 2025
Rigetti CEO Signals Caution with $11 Million Stock Sale Amid Quantum Surge
Investors Must Adapt to New Multipolar World Dynamics
