Google Cloud Next 2025: Elevating GenAI Capabilities to the Next Level

Google has been the leading hyperscaler advancing GenAI capabilities for enterprises, and Google Cloud Next 2025 has advanced its mission with infrastructure like the Cloud Wide Area Network, Ironwood chip, and new AI models, like Gemini 2.5 Flash. This article summarizes the key announcements elevating enterprises’ AI capabilities further.

# 1 — Gemini 2.5 As Reasoning Model

Gemini has emerged as a leading frontier LLM, and Gemini 2.5 continued to exceed expectations, establishing itself as a leader in the Chatbot Arena leaderboard (a community-driven evaluation for Best LLM and AI chatbots) — see snapshot below.

Source: https://lmarena.ai/?leaderboard

• Gemini models are now available to be hosted on an air-gapped environment (using the Google Distributed Cloud) — click here to read more.
• Featured Gemini models with key capabilities highlighted as:

Source: https://cloud.google.com/vertex-ai/generative-ai/docs/models

#2 — Vertex AI Capabilities Expanded Further

With 40x growth since last year, Google has continued to add further capabilities to elevate Vertex AI offerings:

• Model Garden expanded the models' lists with **Llama 4 **(Maverick and Scout) — Model Garden has 200+ models from Google, partners, and the open source community like Hugging Face.
• Augmenting **Vertex AI Agent Builder, Agent Garden* has been launched with pre-built and customizable blueprints for building AI Agents along with out-of-the-box Tools integration such as Grounding with Google Search, Vertex AI Search, Salesforce, RAG Engine*, etc.

Source: Agent Garden Screenshot for Vertex AI Console (Preview feature)

• **Vertex AI Agent Engine: a **fully-managed runtime to deploy, manage, and scale AI agents with capabilities to evaluate, monitor, and trace agents’ behavior for continuous improvement and auditing.
• Vertex AI enhanced fine-tuning capabilities for LLMs, including Gemini and other models focused on supervised fine-tuning.

#3 — Multi-agent Systems are the New Frontier

Google is strengthening the trend that — multi-agent systems are the next big thing — launching key capabilities as:

• **Agent Development Kit** (ADK) — An open-source framework launched (currently in preview) to simplify the process of building complex multi-agent systems (see Appendix A for demo).

Image Source: https://cloud.google.com/

• **Agent2Agent Protocol (A2A): Complienting Model Context Protocol (MCP), Google has launched an open protocol — [Agent2Agent (A2A](https://github.com/google/A2A)) — supported by 50+ vendors focusing on Agent interoperability. Key components are — A2A Server (MCP Server), A2A Protocol (MCP Protocol), A2A Client — Remote or Host Agent (MCP Client), and Agent Card** (Metadata for describing Agent’s capabilities).

Agent2Agent Protocol and MCP (Source: Google Cloud Next)

• **Agentspace: **Enterprise version of NotebookLM — a hub powered by Gemini extending multi-agent capabilities to connect enterprise applications (a low-code platform to build AI Agents). It offers pre-built advanced agents:

Pre-built Agents available in Agentspace

Other key announcements related to Generative AI:

• **Gemini Code Assist *with Agentic Capabilities: Continuing the industry trend, Code Assist now has Agentic capabilities to help build applications faster for “vibe coding”*.
• **GKE Inference Gateway: a model-aware enhanced gateway designed for LLM routing and load balancing based on the evolving Kubernetes standard, **Gateway API Inference Extension.

Source: Google Cloud GKE Inference Gateway

• **Pathways: **a distributed runtime behind Gemini for multi-host inferencing.
• vLLM on v6e **TPUs: **supports single and multiple hosts and can serve LLM (e.g., Llama) using TPU with GKE.

#4 – Firebase Studio for Unified AI Apps Development

• Google launched *Firebase Studio*, a cloud-based IDE for full-stack applications development fusing tools like Project IDX, Genkit, and Gemini Code Assistant.
• CodeOSS-based IDE, instant app preview, and one-click deployment with Firebase App Hosting help in unifying end-to-end developers’ experience.

Appendix A — Demo of Google Agent Platform and A2A Protocol

[embed]

References:

• Google Cloud Next 25: New AI capabilities to transform your business
• Google Cloud Next 2025
• Google Agent2Agent Protocol Specification

Disclaimer:

All data and information provided on this blog are for informational purposes only. The author makes no representations as to the accuracy, completeness, correctness, suitability, or validity of any information on this blog and will not be liable for any errors, omissions, or delays in this information or any losses, injuries, or damages arising from its display or use. This is a personal view and the opinions expressed here represent my own and not those of my employer or any other organization.