Google Cloud Next 2025: All the Big AI Announcements, Gemini 2.5, Ironwood TPUs, and More

From new AI models to custom supercomputers, Google is going all-in on the future of intelligent cloud computing

At its Cloud Next 2025 conference in Las Vegas, Google dropped some major updates that confirm its position as a global AI powerhouse. From record-breaking AI models to hypercomputers and supercharged cloud networks, Google is building the infrastructure that could power the next decade of innovation.

Here’s a breakdown of the most important—and exciting—announcements from the event.

Google Pledges $75 Billion to AI and Cloud

A massive investment to lead the AI race

Google isn’t playing small. CEO Thomas Kurian announced that Google Cloud will invest $75 billion in 2025 alone to scale its AI computing power, cloud services, and infrastructure.

He revealed that over 4 million developers are already using Gemini, Google’s flagship AI family, and Vertex AI usage has jumped 20-fold in the past year.

Kurian summed it up simply:

“We want to bring these powerful AI tools to everyone—from solo developers to global enterprises.”

Gemini 2.5 Pro: Google’s Most Advanced AI Yet

A powerful leap in multimodal AI capabilities

Sundar Pichai, Google’s CEO, introduced Gemini 2.5 Pro, which he called the company’s most advanced AI model so far. What’s new?

• Enhanced multimodality: Understands and generates across text, images, and audio
• Real-time processing: Delivers fast and accurate responses
• Native audio/image generation: No third-party models needed

This model will power everything from intelligent document analysis to smarter customer support tools and real-time content generation.

Gemini 2.5 Flash: Smaller, Smarter, and Faster

This lightweight model is 24x more powerful than GPT-4o

Pichai also previewed Gemini 2.5 Flash, a compact but blazing-fast version of the Pro model. It’s designed to run efficiently on edge devices and Ironwood TPUs, yet it delivers:

• 24 times more intelligence than OpenAI’s GPT-4o
• Greater energy efficiency
• Instant responses for AI assistants and agents

Perfect for mobile apps, chatbots, and smart devices.

Introducing Ironwood TPUs: The Next Gen of AI Chips

3,600x performance vs. TPU v1, 29% more energy efficient

Google revealed its seventh-generation TPU (Tensor Processing Unit)—Ironwood. This is the chip that powers Google’s future AI ambitions.

Ironwood TPUs deliver:

• 3,600x more performance than the original TPUs
• 29% better energy efficiency
• Optimized for large AI models and inference at scale

And yes, Gemini Flash was built to run seamlessly on these chips.

Google’s New AI Hypercomputer: Built for the AI Age

A supercomputer designed for intelligent apps

Google announced a new category of machines: AI hypercomputers. These supercomputers are tailor-made to:

• Run massive AI workloads
• Reduce cloud costs
• Improve AI app performance

This could be the backbone for future real-time translation, personalized search, autonomous agents, and more.

Multi-Agent Systems & Open-Source AI on Vertex

Build teams of AI agents using open-source models

Google is embracing open-source with new support in Vertex AI:

• Run Meta’s Llama 4, Gemma, and other models on Google Cloud
• Build multi-agent systems using a new Agent Development Kit (ADK)
• Use open frameworks to connect different models and tasks

Developers and researchers can now create custom AI workflows across models using one flexible ecosystem.

Gemini Now Runs on Google Distributed Cloud

Works even in secure, offline environments

In a big step for enterprise AI, Google announced Gemini will now run on Google Distributed Cloud, both in:

• Connected environments
• Air-gapped/offline environments (for security-focused industries)

That means banks, governments, and healthcare companies can run powerful AI tools locally—without internet access.

Google Workspace Gets Smarter with AI Agents

Docs and Sheets just became intelligent coworkers

Google announced a slew of updates to Workspace, integrating Gemini-powered agents into tools like Docs, Sheets, and Gmail.

New features include:

• Help Me Analyze (Sheets): Instantly interprets and summarizes your data
• Audio Overviews (Docs): Turn written docs into spoken audio summaries
• Workspace Flows: Automate tasks like email follow-ups, calendar bookings, and file organization using AI agents

This is a huge leap for everyday productivity.

Cloud Wide Area Network (WAN): The Private Internet for Enterprises

40% faster speeds and lower costs

Google is opening access to its Cloud WAN, a private cloud network optimized for enterprise performance.

Benefits include:

• 40% faster speeds compared to the public internet
• Up to 40% lower costs for cloud operations
• Global availability for Google Cloud customers

For companies handling sensitive or latency-critical apps, this is a game-changer.

Chirp 3, Lyria, and V2: Next-Gen Creative AI Models

From text to music and video

Google didn’t forget the creatives. The company demoed three new generation models:

• Chirp 3: Converts text into 30-second music clips
• Lyria: A powerful music generation model designed for composition and editing
• V2: Google’s new video generation model—with industry-leading realism and motion accuracy

Expect these to power everything from YouTube tools to music creation platforms.

Google Is Building the AI Future—Now

With these announcements, Google is proving it’s not just participating in the AI race—it’s building the track.

From its $75 billion AI investment to Gemini 2.5 Pro, Ironwood TPUs, and next-gen Workspace tools, Google is designing a world where AI helps everyone: developers, artists, workers, and enterprises.

Whether you want to build custom AI agents, generate music, or automate your daily tasks, Google Cloud is positioning itself as the go-to AI platform of the future.