Gemini Enterprise Platform
Google Cloud has launched a comprehensive platform aimed at empowering businesses to develop, deploy, and manage autonomous AI agents. This new Enterprise
Agent platform, powered by Gemini, serves as a unified hub for technical teams, allowing them to construct, optimize, and govern AI agents efficiently. It offers flexibility, enabling developers to choose from a selection of pre-built models, including Gemini 3.1 Pro, Gemini 3.1 Flash Image (also known as Nano Banana 2), Lyria 3, and models from partners like Anthropic, such as Claude Opus 4.7, Sonnet, and Haiku. The platform integrates essential features for agent orchestration, DevOps, and robust security, ensuring seamless integration into business workflows while maintaining strict control and scalability. This approach focuses on delivering a fully optimized technological stack rather than disparate services, designed for the demands of production-level AI.
Next-Gen TPUs Revealed
At its annual conference, Google Cloud announced its eighth generation of custom Tensor Processing Units (TPUs), specifically engineered with two distinct architectures: TPU 8t for training and TPU 8i for inference. These specialized chips are built to handle the creation of highly sophisticated AI models, manage complex agent networks, and execute intricate reasoning tasks. The TPU 8t is optimized for training large AI models within weeks, balancing high compute throughput, extensive shared memory, and rapid interchip communication with superior power efficiency. Conversely, the TPU 8i boasts an impressive 288 GB of high-bandwidth memory and 384 MB of on-chip SRAM – a threefold increase from previous generations. This significant memory capacity allows for tasks to be delegated to agents with rapid responses, as the underlying models utilize only the on-chip memory for processing. Both chip generations are slated for general availability later this year, positioning them as competitive alternatives to existing GPU solutions in the AI hardware market.
AI-Powered Security Agents
To bolster business defenses against escalating AI-driven cyber threats, Google Cloud has introduced a suite of new security-focused AI agents. The Dark Web Intelligence agent leverages Gemini models and the expertise of the Google Threat Intelligence Group to construct security profiles for organizations. It meticulously analyzes millions of daily external events on the dark web with 98% accuracy, identifying and flagging only relevant threats to the customer. Complementing this is the Threat Hunting Agent, which utilizes Google's internal threat intelligence to empower teams to proactively seek out novel attack patterns and adversary tactics that might evade conventional security measures. Furthermore, the Detection Engineering Agent automates the creation of persistent detection rules for various threat scenarios, transforming a previously manual process into an automated one. In addition to these agents, the company is integrating reCAPTCHA into its Google Cloud Fraud Defense platform to better distinguish between legitimate users, bots, and malicious agents, enhancing security for digital commerce from account creation through to payment.
