Google has introduced a suite of tools designed to create AI agents that assist companies in automating tasks. Concurrently, a $750 million fund has been launched, along with new AI processors.

Have a great #GoogleCloudNext!

Watch select sessions and your favorite moments from the opening keynote on demand → https://t.co/Ss9VorZkOvhttps://t.co/VnbGXrIjvi

— Google Cloud (@googlecloud) April 22, 2026

AI Agent Software

At its annual conference in Las Vegas, Google's cloud division showcased a toolkit for creating AI agents and managing their operations within companies. Among the new features is a dedicated inbox where virtual assistants can send reports on their activities.

The Gemini Enterprise Agent Platform will receive two new modules: Memory Bank and Memory Profile, allowing agents to retain interaction histories with users. Agent Simulation will enable developers to test tools more thoroughly before deployment.

Google also announced updates to its Workspace application suite and described a scenario where AI agents significantly transform the daily lives of regular employees.

The company stated that employees will be able to use Gemini Enterprise to create virtual assistants without needing to write code.

Additionally, Google introduced Projects, a collaboration platform designed for employee interactions with each other or with support operators.

This tool aggregates information from various sources, such as Workspace, Microsoft OneDrive, and corporate chats, allowing users to work with the necessary context in mind.

New Fund

Google Cloud announced a $750 million fund aimed at assisting consulting firms like McKinsey, Accenture, and Deloitte in implementing agent-based AI for their clients.

DeepMind will provide select firms with early access to the Gemini AI models before their official release.

The capital will be used to support engineer training, develop AI agents through the corporate platform, and co-fund projects and pre-sales activities.

“Consulting firms are at the center of some of the largest transformations occurring with clients. They understand the situation and bring unique expertise in specific industries and knowledge of business processes,” said Kevin Ichpurani, head of Google Cloud's global partner ecosystem.

New Chips

Google Cloud has unveiled a new generation of its custom Tensor Processing Units (TPUs). These chips are designed to accelerate and reduce the cost of AI computations.

The lineup includes two versions:

  • TPU 8t — designed for AI development;
  • TPU 8i — better suited for inference.

Google has established a strong position among manufacturers of custom AI chips, competing with Nvidia. In recent months, demand for TPUs has surged in Silicon Valley.

The new processors store more information, which reduces response latency.

“It's about ensuring the lowest possible response latency at the lowest possible cost per operation,” noted Mark Lomayer, Google’s vice president of computing infrastructure.

AI services are created and launched using systems capable of quickly processing large data sets to identify connections and patterns that are then mathematically expressed. Calculations, program launches, and services are executed on processors with substantial built-in memory.

This approach allows AI responses to be nearly instantaneous, as the component does not need to fetch data from external sources.

TPU 8t can be clustered in groups of up to 9,600 units. At such scales, energy consumption becomes a key factor. The new chips deliver 124% greater performance per watt compared to the previous generation, while TPU 8i offers 117% more.

It’s worth noting that Google is discussing with the U.S. Department of Defense the potential integration of Gemini into Pentagon systems across all information access categories, from open to highly classified.