The enterprise AI race is heating up, with Google Cloud introducing a new Gemini-powered Enterprise Agent platform alongside next-generation TPU chips and advanced AI agent-driven security tools for businesses.The Gemini Enterprise Agent Platform is designed to serve as a one-stop platform for technical teams of businesses to build, scale, govern, and optimise autonomous agents. Similar to Vertex AI, it allows developers to select their own models as well as build their own models and AI agents, the tech giant said at its annual Next conference in Las Vegas, Nevada, United States on Wednesday, April 22.Some of the models available on the platform include Gemini 3.1 Pro, Gemini 3.1 Flash Image, also known as Nano Banana 2, and Lyria 3 as well as Anthropic’s Claude Opus 4.7 and other Sonnet and Haiku models from the AI startup.
The Enterprise Agent Platform is also equipped with new features for agent integration, DevOps, orchestration, and security so that employees can seamlessly deploy agents through the Gemini Enterprise app while ensuring tight control, governance, and security at scale, the company said.
“This isn’t about offering individual services that can be cobbled together; we are offering a vertically optimized stack where everything is co-developed to deliver the scale and efficiency required for this new era of production AI,” Thomas Kurian, CEO, Google Cloud, said in a statement.
“Our AI Hypercomputer provides the scalable foundation, our Data Cloud provides the context, our Agentic Defense provides the protection, and our Agent Platform provides the orchestration. The technology is here—now it’s time to build your growth engine,” Kurian added.
Google Cloud’s latest enterprise-focused AI product arrives at a time when rivals such as OpenAI, Microsoft, Amazon, and Anthropic are vying to capture the robust demand from businesses for AI agents as well as AI coding tools.Story continues below this ad
In the past few weeks, OpenAI has pulled the plug on several consumer-facing bets such as Sora and Instant Checkout while doubling down on enterprise, an area where Anthropic has historically been stronger with tools such as Claude Code and Claude Cowork witnessing a surge in popularity.
Amid the escalating turf war, Google appears to have no intention of staying on the sidelines. Nearly 75 per cent of Google Cloud customers are using its AI products to power their businesses, with 330 customers having processed more than one trillion tokens each. “Our first-party models now process more than 16 billion tokens per minute via direct API use by our customers, up from 10 billion last quarter,” the company said.
New lineup of TPUs
Google on Wednesday also unveiled its latest eighth generation of custom Tensor Processor Unit (TPU), with two distinct, purpose-built architectures for training and inference: TPU 8t and TPU 8i.
These specialised chips have been designed to build the most capable AI models as well as orchestrate swarms of AI agents and handle complex reasoning tasks, Google said.Story continues below this ad
The TPU 8t can be used to train a frontier AI model in weeks by balancing the highest possible compute throughput, shared memory, and interchip bandwidth with the best possible power efficiency and productive compute time, according to the company.
TPU 8i, on the other hand, pairs 288 GB of high-bandwidth memory with 384 MB of on-chip SRAM – three times more than the previous generation of chips. This allows users to delegate a task to an agent and get a response in less than five seconds with the underlying model using only the memory that sits directly on the chip.
Both chips will be generally available later this year, as per Google. They come at a time when the search giant has been pushing to make its TPUs a viable alternative to Nvidia’s dominant GPUs.
AI-powered cybersecurity tools
Google Cloud further introduced the following new security-focused AI agents to help businesses defend against AI-driven attacks:Story continues below this ad
-Dark Web Intelligence: An AI agent that can create a security profile of a customer’s organisation using the latest Gemini models, based on the Google Threat Intelligence Group’s dark web expertise. It can further analyse millions of daily external events on the dark web with 98 per cent accuracy to elevate only the threats that are relevant to a customer, Google claimed.
-Threat Hunting Agent: It uses Google’s internal threat intelligence knowledge to “allow teams to proactively hunt for novel attack patterns and adversary behaviors that bypass traditional defenses.”
-Detection Engineering Agent: It can be used to generate “persistent detection rules for threat scenarios, transforming detection creation from a manual craft into an automated process.”
To help customers against the menace of unauthorised bots, the company is folding reCAPTCHA into Google Cloud Fraud Defense – a platform designed to help customers discern the legitimacy and authorisation of bots, humans, and agents. Fraud Defense will also soon come with agentic AI capabilities that help secure the digital commerce journey, from account creation and login to payment and checkout, Google added.