At its annual cloud conference, Google Cloud Next 2026, Google announced a major push into next-generation AI for businesses. The company introduced the Gemini Enterprise Agent Platform, a system designed to help organizations build AI agents that can handle tasks and workflows with minimal human input. These agents are built to connect across different software tools and manage multi-step operations automatically. Along with this, the Sundar Pichai-led firm also unveiled two new eighth-generation TPU chips aimed at improving both AI model training and real-time performance.
The Gemini Enterprise Agent Platform represents a significant shift from traditional enterprise AI tools toward what Google describes as an ‘agentic model’ of computing. Instead of relying on users to continuously prompt systems, these AI agents are designed to independently plan, execute, and adapt workflows. This means an enterprise could deploy AI systems that not only analyze data but also take action – like updating databases, triggering supply chain operations, generating reports, and even interacting with customers – without constant supervision.
The platform brings together multiple layers of Google’s AI ecosystem into a single unified environment. It integrates model access, development tools, orchestration systems, and governance frameworks. Businesses can use low-code interfaces to quickly build agents for routine operations, while developers can rely on more advanced toolkits to design highly customized systems. One of the defining features of the platform is its ability to support multi-agent collaboration. Rather than functioning as isolated systems, these AI agents can communicate with one another, divide tasks, and coordinate actions.
In terms of security, each agent can be assigned a defined identity with specific permissions, ensuring controlled access to enterprise data and systems. And monitoring tools allow organizations to track agent behaviour, evaluate performance, and enforce compliance standards.
In parallel, complementing the software platform, the tech titan introduced its eighth-generation Tensor Processing Units – TPU 8t for training and TPU 8i for inference. The TPU 8t is optimised for training large-scale AI models and delivers substantial gains in compute capability, with Google claiming around 2.8× to 3× performance gains over its previous generation. It is designed to operate in massive clusters, with superpods scaling up to 9,600 interconnected chips and offering as much as 2 petabytes of high-bandwidth memory, allowing the training of highly complex, trillion-parameter models. At peak scale, these systems can deliver over 100 exaflops-class performance in lower precision formats.
The second chip, TPU 8i, is purpose-built for inference, the stage where trained models are deployed in real-world applications, and focuses on efficiency and responsiveness. It is engineered with large on-chip memory and high-bandwidth capacity, allowing models to run with lower latency and reduced dependency on external systems. The tech giant highlights that TPU 8i delivers around 80% better price-performance and up to 2× improvement in performance per watt, making it well-suited for always-on enterprise AI workloads. These efficiency gains are crucial at scale, where even marginal improvements can significantly lower operational costs.
Along with these, the Mountain View-headquartered company is also expanding the use of AI agents into areas like cybersecurity. It is developing specialized systems capable of detecting threats, analyzing attack patterns, and responding to incidents in real time. The company also revealed that AI is playing an increasingly central role within its own operations, with a significant portion of internal code generation now assisted by AI systems.
The Tech Portal is published by Blue Box Media Private Limited. Our investors have no influence over our reporting. Read our full Ownership and Funding Disclosure →
Ashutosh is a Senior Writer at The Tech Portal, largely reporting on new tech, and intersection of technology and business. Ashutosh’s career spans across nearly a decade of technology writing across multiple platforms and languages.
