2026-04-22 · Google

Cloud Next '26: Momentum and innovation at Google scale

securityagentsmodelsenterprisecapitalinfrastructure

read at source ↗ blog.google

Cloud Next ‘26: Momentum and innovation at Google scale

Source: Google Date: 2026-04-22 URL: https://blog.google/innovation-and-ai/infrastructure-and-cloud/google-cloud/cloud-next-2026-sundar-pichai/

Summary

Sundar Pichai’s keynote at Google Cloud Next ‘26 announced four headline items: the Gemini Enterprise Agent Platform for orchestrating AI agents at scale, new AI-powered security tools built on the Wiz acquisition, eighth-generation TPUs (TPU 8t for training, TPU 8i for inference), and the internal AI adoption milestone that 75% of all new code at Google is now AI-generated and approved by engineers.

Implications

The 75% code-generation figure is the most consequential number in this post — not because it’s verifiable, but because it’s a deliberate benchmark-setting move. Google is normalizing AI-authored code as the standard operational state, which repositions every vendor selling coding tools (GitHub Copilot, Cursor, Anthropic’s Claude for code) against a self-reported floor of near-total adoption at one of the world’s largest software organizations.

Three threads this signal feeds:

Cloud agent platform thread. The Gemini Enterprise Agent Platform isn’t a product — it’s a platform-of-platforms pitch. Google is trying to be the orchestration layer above individual model vendors, including Anthropic and OpenAI. That’s a direct collision course with AWS Bedrock Agents, Azure AI Foundry, and the emerging open-agent infrastructure layer.

Infrastructure moat thread. TPU 8t and 8i shipping as Cloud Next announcements means Google is using custom silicon as a cloud differentiation argument, not just a cost advantage. Watch whether next-gen pricing tiers are tied to TPU access.

Enterprise positioning vs. OpenAI/Anthropic. The Wiz integration as a security layer is Google’s answer to the enterprise objection that multi-agent systems are attack surface. This is the governance pitch OpenAI doesn’t yet have a matching answer for.

Watch:

  • Whether the 75% code-gen figure becomes a recurring metric in Google’s earnings calls
  • How Microsoft responds at Build — the Copilot/Azure equivalent is the obvious counter
  • Pricing structure for Gemini Enterprise Agent Platform relative to Bedrock/Azure AI Foundry

← all signals