Google has launched a toolkit for AI agents and a $750 million fund to facilitate automation in businesses, along with new AI chips.
Google has unveiled a new suite of tools aimed at helping businesses automate tasks with AI agents, alongside the announcement of a $750 million fund and advanced AI processors.
AI Agent Development Tools
During its annual conference in Las Vegas, Google Cloud introduced a toolkit for developing and managing AI agents within organizations. A notable feature is a dedicated inbox for virtual assistants to report their activities.
The Gemini Enterprise Agent Platform is set to receive enhancements, including two new modules: Memory Bank and Memory Profile, which will allow agents to retain user interaction histories. Additionally, Agent Simulation will let developers rigorously test tools before deployment.
Google highlighted that employees could use Gemini Enterprise to build virtual assistants without needing programming skills, and also launched Projects, a collaboration platform that facilitates employee interaction and aggregates information from various sources like Workspace and Microsoft OneDrive.
$750 Million Fund for AI Implementation
The newly announced fund aims to assist consulting firms such as McKinsey and Deloitte in deploying agent-based AI for their clients. DeepMind will offer select firms early access to the Gemini AI models ahead of their launch.
Funding will support engineer training, AI agent development, and collaborative projects.
“Consulting firms are at the center of some of the largest transformations occurring with clients,” stated Kevin Ichpurani, head of Google Cloud's global partner ecosystem.
New Tensor Processing Units
Google Cloud also revealed its latest generation of custom Tensor Processing Units (TPUs), designed to enhance AI computation efficiency and reduce costs. The new lineup includes:
- TPU 8t, optimized for AI development;
- TPU 8i, tailored for inference tasks.
These processors offer improved information storage, leading to lower response latency. TPU 8t can be clustered up to 9,600 units, achieving 124% greater performance per watt compared to previous versions, while TPU 8i delivers 117% more performance.
Moreover, discussions are underway with the U.S. Department of Defense regarding the potential integration of Gemini into various Pentagon systems.
