Google bets on agents

- Google pushed AI agents as its enterprise monetisation play at Cloud Next and announced new chips for training and inference. - The company detailed eighth-generation TPUs — TPU 8t for training and TPU 8i for inference — and highlighted customer partnerships. - Google is pairing infrastructure, security and consulting channels to embed agents into workflows, signalling an enterprise-first AI push (reuters.com, ).

Google used its Cloud Next conference on April 22 to pitch AI agents as the core product it wants large companies to buy. (reuters.com) The company paired that software push with new eighth-generation Tensor Processing Units, or TPUs: TPU 8t for training models and TPU 8i for running them in real time. Google said TPU 8t scales to 9,600 chips and 2 petabytes of shared high-bandwidth memory in one superpod. (blog.google) Google said TPU 8t delivers three times the processing power of Ironwood and up to twice the performance per watt, while TPU 8i is tuned for “near-zero latency” inference. The chips are hosted for the first time on Google’s Axion Arm-based processors, which the company said reduces data-preparation bottlenecks. (blog.google, (cloud.google.com), (cloud.google.com)) An AI agent is software that can take a goal, call tools, and complete a chain of tasks with limited human input. Google’s pitch is that companies will need not just models, but also systems to build, monitor, secure and govern thousands of those agents inside everyday work. (cloud.google.com), (blog.google) That is why Google bundled the chip launch with a broader enterprise stack. At Next, it introduced a Gemini Enterprise Agent Platform, expanded Agent Designer, added a unified Inbox for long-running agents, and pushed Projects and reusable Skills for teams managing multi-step workflows. (blog.google), (cloud.google.com), (cloud.google.com)) Google also tied the sales pitch to security and services. Its Cloud unit promoted “Agentic Defense” with Wiz and said consulting partners would get deeper technical roles, including teams of Google forward-deployed engineers to help customers put agents into production. (cloud.google.com), (cloud.google.com)) The customer list was part of the message. Google highlighted companies including Citi, Capcom, Salesforce, SAP, ServiceNow and Palantir as examples of agents being embedded into finance, software, customer service and game-testing workflows. (blog.google), (cloud.google.com)) Google said nearly 75% of Google Cloud customers now use its artificial intelligence products, and that 330 customers processed more than 1 trillion tokens over the past 12 months. Sundar Pichai also said first-party models are now handling more than 16 billion tokens per minute through direct customer API use, up from 10 billion last quarter. (cloud.google.com), (blog.google) The bet is that enterprise buyers will pay for the full bundle: chips, cloud capacity, security controls, partner integrations and consulting help. Google’s closing argument at Next was that agents are moving from demos to managed corporate systems, and it wants to sell the plumbing behind that shift. (reuters.com), (cloud.google.com))

Get your own daily briefing

Scout delivers personalized news, insights, and conversations tailored to your role and industry.

Download on the App Store

Shared from Scout - Be the smartest in the room.