The enterprise AI race is heating up, with Google Cloud introducing a brand new Gemini-powered Enterprise Agent platform alongside next-generation TPU chips and superior AI agent-driven safety instruments for companies.
The Gemini Enterprise Agent Platform is designed to function a one-stop platform for technical groups of companies to construct, scale, govern, and optimise autonomous brokers. Just like Vertex AI, it permits builders to pick out their very own fashions in addition to construct their very own fashions and AI brokers, the tech large stated at its annual Subsequent convention in Las Vegas, Nevada, United States on Wednesday, April 22.
A number of the fashions obtainable on the platform embrace Gemini 3.1 Professional, Gemini 3.1 Flash Picture, also called Nano Banana 2, and Lyria 3 in addition to Anthropic’s Claude Opus 4.7 and different Sonnet and Haiku fashions from the AI startup.
The Enterprise Agent Platform can also be geared up with new options for agent integration, DevOps, orchestration, and safety in order that workers can seamlessly deploy brokers via the Gemini Enterprise app whereas making certain tight management, governance, and safety at scale, the corporate stated.
“This isn’t about providing particular person companies that may be cobbled collectively; we’re providing a vertically optimized stack the place every part is co-developed to ship the size and effectivity required for this new period of manufacturing AI,” Thomas Kurian, CEO, Google Cloud, stated in an announcement.
“Our AI Hypercomputer supplies the scalable basis, our Information Cloud supplies the context, our Agentic Protection supplies the safety, and our Agent Platform supplies the orchestration. The know-how is right here—now it’s time to construct your progress engine,” Kurian added.
Google Cloud’s newest enterprise-focused AI product arrives at a time when rivals reminiscent of OpenAI, Microsoft, Amazon, and Anthropic are vying to seize the sturdy demand from companies for AI brokers in addition to AI coding instruments.
Story continues under this advert
Up to now few weeks, OpenAI has pulled the plug on a number of consumer-facing bets reminiscent of Sora and Immediate Checkout whereas doubling down on enterprise, an space the place Anthropic has traditionally been stronger with instruments reminiscent of Claude Code and Claude Cowork witnessing a surge in recognition.
Amid the escalating turf conflict, Google seems to haven’t any intention of staying on the sidelines. Almost 75 per cent of Google Cloud prospects are utilizing its AI merchandise to energy their companies, with 330 prospects having processed a couple of trillion tokens every. “Our first-party fashions now course of greater than 16 billion tokens per minute by way of direct API use by our prospects, up from 10 billion final quarter,” the corporate stated.
New lineup of TPUs
Google on Wednesday additionally unveiled its newest eighth technology of customized Tensor Processor Unit (TPU), with two distinct, purpose-built architectures for coaching and inference: TPU 8t and TPU 8i.
These specialised chips have been designed to construct probably the most succesful AI fashions in addition to orchestrate swarms of AI brokers and deal with complicated reasoning duties, Google stated.
Story continues under this advert
The TPU 8t can be utilized to coach a frontier AI mannequin in weeks by balancing the best potential compute throughput, shared reminiscence, and interchip bandwidth with the very best energy effectivity and productive compute time, based on the corporate.
TPU 8i, alternatively, pairs 288 GB of high-bandwidth reminiscence with 384 MB of on-chip SRAM – 3 times greater than the earlier technology of chips. This enables customers to delegate a process to an agent and get a response in lower than 5 seconds with the underlying mannequin utilizing solely the reminiscence that sits instantly on the chip.
Each chips will probably be usually obtainable later this yr, as per Google. They arrive at a time when the search large has been pushing to make its TPUs a viable different to Nvidia’s dominant GPUs.
AI-powered cybersecurity instruments
Google Cloud additional launched the next new security-focused AI brokers to assist companies defend towards AI-driven assaults:
Story continues under this advert
-Darkish Internet Intelligence: An AI agent that may create a safety profile of a buyer’s organisation utilizing the most recent Gemini fashions, primarily based on the Google Risk Intelligence Group’s darkish internet experience. It could actually additional analyse thousands and thousands of each day exterior occasions on the darkish internet with 98 per cent accuracy to raise solely the threats which might be related to a buyer, Google claimed.
-Risk Looking Agent: It makes use of Google’s inside menace intelligence information to “enable groups to proactively hunt for novel assault patterns and adversary behaviors that bypass conventional defenses.”
-Detection Engineering Agent: It may be used to generate “persistent detection guidelines for menace situations, remodeling detection creation from a handbook craft into an automatic course of.”
To assist prospects towards the menace of unauthorised bots, the corporate is folding reCAPTCHA into Google Cloud Fraud Protection – a platform designed to assist prospects discern the legitimacy and authorisation of bots, people, and brokers. Fraud Protection may even quickly include agentic AI capabilities that assist safe the digital commerce journey, from account creation and login to fee and checkout, Google added.