Google Cloud Next 25: Comprehensive Overview of New AI Models, Tools, and Protocols

Today Google wrapped up Cloud Next 25, announcing nearly 20 AI‑related models, applications, developer tools, and hardware. The information is scattered across many announcements, so this article consolidates the most important updates.

Key Highlights

Google released many impressive AI models and tools, though many are still in preview and eagerly awaited.

Gemini 2.5 Flash is arriving soon, a cost‑effective inference model that builds on the already powerful Gemini 2.5 Pro.

A new Agent‑to‑Agent (A2A) protocol enables seamless collaboration between agents, shaping future AI coordination standards.

Google disclosed 601 real‑world AI use‑case examples, providing clear guidance for entrepreneurs.

5 New AI Model Updates

Gemini 2.5 Flash – Fast, Cheap Inference

The Gemini 2.5 Pro model already supports 1M‑token context, multimodal prompts, and tops benchmark rankings. Gemini 2.5 Flash will retain the speed and low cost advantages, dynamically adjusting inference depth based on task complexity and allowing developers to control costs. It will be available in Vertex AI soon.

Veo 2 – High‑Quality Video Generation (waitlist)

Veo 2 is a top‑tier video generation model supporting P‑video, key‑frame generation, scene expansion, and camera control. It powers Google VideoFX and can remove unwanted elements, generate stable videos from start/end frames, and adapt to vertical formats for platforms like TikTok.

Chirp 3 – Custom Voice Generation from 10‑Second Samples

Chirp 3 offers over 35 languages and eight voice styles. With just a 10‑second recording, developers can create realistic custom voices, though the API currently limits direct experimentation. It also distinguishes speakers in multi‑speaker recordings, improving transcription usability.

Lyria – Text‑to‑Music Generation (waitlist)

Lyria creates full music pieces from simple text prompts, expanding generative media capabilities alongside Veo 2 and Chirp 3.

Imagen 3 – Improved Image Generation and Editing

Imagen 3 enhances object removal and image restoration, delivering higher‑quality edits and stable generation for product, game, and interior design visuals.

1 Future‑Oriented AI Protocol

Agent to Agent (A2A)

A2A is an open protocol for agent collaboration, complementing Anthropic’s MCP. It lets a client agent define tasks while a remote agent executes them, enabling long‑running, context‑rich interactions between agents.

Other AI Applications, Developer Tools, and 601 Cases

Firebase Studio – End‑to‑End AI‑Powered Cloud Development

Firebase Studio integrates AI coding, build, and deployment, allowing developers to create, test, and launch applications entirely in the cloud with Gemini 2.5 Pro assistance.

Google Workspace – AI‑Enhanced Productivity Suite

AI features are added to Docs, Sheets, Meet, Chat, and more: audio versions of documents, "Help me improve" suggestions, AI‑driven data insights, and Veo 2 video generation in Vids.

Google AI Studio UI Refresh

The AI Studio interface now mirrors Gemini Web design, offering a cleaner, more intuitive experience.

Ironwood TPU – 7th‑Gen Inference‑Optimized Accelerator

Google’s upcoming Ironwood TPU delivers up to five times the peak compute of the previous Trillium generation, dramatically accelerating AI inference workloads.

601 Real‑World Generative AI Use Cases

The expanded case library covers customer agents, employee agents, creative agents, code agents, data agents, and security agents, providing valuable inspiration for B2B AI solutions.

References

Official Cloud Next 25 announcement: https://blog.google/products/google-cloud/next-2025/

Gemini 2.5 Flash details: https://cloud.google.com/blog/products/ai-machine-learning/gemini-2-5-pro-flash-on-vertex-ai

Vertex AI generative media (Veo 2, Chirp 3, Lyria, Imagen 3): https://cloud.google.com/blog/products/ai-machine-learning/expanding-generative-media-for-enterprise-on-vertex-ai

A2A protocol announcement: https://developers.googleblog.com/en/a2a-a-new-era-of-agent-interoperability/

Agent Development Kit (ADK): https://developers.googleblog.com/en/agent-development-kit-easy-to-build-multi-agent-applications/

Google Workspace AI updates: https://blog.google/products/workspace/cloud-next-2025-workspace-gemini/

601 AI case study collection: https://cloud.google.com/transform/101-real-world-generative-ai-use-cases-from-industry-leaders

Ironwood TPU announcement: https://blog.google/products/google-cloud/ironwood-tpu-age-of-inference/

Overall, these updates provide a rich set of tools and models for AI developers and signal Google’s intent to reclaim a leading position in the generative AI space.