AI News Roundup — April 23, 2026
Google Cloud Next 2026 dominated the day with blockbuster announcements including 8th-generation TPUs, a $750 million agentic AI partner fund, Deep Research Max in the Gemini API, and the revelation that 75% of all new Google code is now AI-generated. Meanwhile, Nobel laureate Geoffrey Hinton urged the UN to apply the brakes to runaway AI, and on-chain AI agents gained direct access to frontier LLMs for the first time.
Google Cloud Next 2026: 8th-Gen TPUs, $750M Partner Fund, and Full Agentic Platform
Google's biggest cloud event of the year opened with sweeping AI announcements: 8th-generation TPUs scaling to 9,600 chips and 2 petabytes of shared memory, a $750 million fund to help its 120,000-member partner ecosystem build agentic AI solutions, and a new Gemini Enterprise Agent Platform featuring Agent Designer, long-running agents, Skills, and Projects.
Google Launches Two New Specialized AI Chips to Challenge NVIDIA
For the first time, Google's 8th-gen TPU family includes two distinct chips: the TPU 8t optimized for large-scale training and scaling to 9,600 units in a single superpod, and a second chip purpose-built for agentic-era inference. The move signals Google's intent to offer cloud customers dedicated AI silicon alternatives to NVIDIA.
Google Deep Research Max Arrives in Gemini API with Native MCP Support
Google launched Deep Research and Deep Research Max in the Gemini API, powered by Gemini 3.1 Pro. The autonomous research agents can safely navigate the web and custom data sources to produce fully cited, professional-grade reports, now with native MCP support and built-in visualizations available to developers.
Geoffrey Hinton at UN: 'Time to Apply the Brakes to Runaway AI'
AI pioneer and Nobel laureate Geoffrey Hinton spoke at the UN Digital World Conference, urging world leaders to guide rapid AI advances more carefully to serve societies rather than undermine them. Hinton warned that without meaningful regulation, the pace of AI development risks outrunning humanity's ability to manage its consequences.
75% of Google's New Code Is AI-Generated; Cloud Customers Hit 16B Tokens Per Minute
Sundar Pichai revealed at Google Cloud Next that 75% of all new code written at Google is now AI-generated and reviewed by engineers, up from 50% last fall. Google Cloud customers collectively process over 16 billion tokens per minute via direct API access, nearly doubling from 10 billion last quarter, with 330 customers processing over a trillion tokens each in the past year.
Wipro Launches AI-Native Business Unit with Agent-Native Delivery Platform
Wipro announced a new AI-Native Business & Platforms Unit anchored by its WEGA agent-native delivery platform, which integrates Factory's Droids to delegate software development tasks directly to AI agents. The move positions Wipro at the forefront of a new wave of enterprise IT services built entirely around autonomous agentic workflows.
On-Chain AI Agents Gain Direct Access to Top LLMs for the First Time
A new collaboration between 0G Foundation and Alibaba Cloud enables AI agents operating on-chain to directly access top-tier large language models for the first time, bridging Web3 infrastructure with frontier AI capabilities. The integration opens the door to decentralized agentic applications that can reason and act using state-of-the-art models without centralized intermediaries.