Google Cloud Next Sees New AI Chip, GenAI Models, Workspace Features & More

In a bid to lead the artificial intelligence (AI) race, Google has announced a slew of announcements during the company’s Google Cloud Next 25 event.

Google Cloud Next acts as a platform connecting industry leaders from around the world with attendees, allowing them to explore what’s being created globally.

In addition to informative and insightful keynote speeches from Google’s leaders—including Sundar Pichai, CEO of Google and Alphabet, and Thomas Kurian, CEO of Google Cloud—the event showcased Google’s latest innovations, encompassing both hardware and software.

From new chips to accelerate AI applications to new AI features in Google Workspace and new generative AI models, the company sought to push the boundaries of AI capabilities and how they can benefit users.

“AI is transforming the way we work—from boosting productivity and creativity to delivering real business transformation and impact. And at Google Cloud, we’re bringing the best of Google AI to people, organizations and businesses across the globe,” said Google in a blog post.

So, what are some of the biggest reveals Google made through its event? Let’s dive in!

Google’s Ironwood

Google unveiled the seventh generation of Ironwood, its Tensor Processing Unit (TPU).

As per the company, Ironwood is Google’s most performant and scalable custom AI accelerator to date and represents the first Google TPU for the “Age of Inference.”

What is the “Age of Inference?”

“It’s where AI agents will proactively retrieve and generate data to collaboratively deliver insights and answers, not just data,” said Google. “It’s a move from responsive AI models that provide real-time information for people to interpret, to models that provide the proactive generation of insights and interpretation.”


Amid a host of improvements, Ironwood boasts of offering 3,600 times better performance than its first publicly available TPU. It also supports more than 24 times the compute power of the world’s largest supercomputer—El Capitan—which offers 1.7 Exaflops per pod compared to Ironwood’ 42.5 Exaflops per pod, when scaled to 9,216 chips per pod.

As such, Ironwood comes in two sizes—a 256-chip configuration and a 9,216-chip configuration.

When it comes to enhancements, Google’s new TPU sports an enhanced SparseCore (a specialized accelerator for processing ultra-large embeddings) and offers better performance without costing on power efficiency—it’s nearly 30 times more power-efficient than the company’s first Cloud TPU (2018). It also features an increase in High Bandwidth Memory (HBM) capacity, offering 192 GB per chip—6 times that of its predecessor Trillium—and improved bandwidth with 7.2 TBps per chip or 4.5 times Trillium’s capacity.

“Ironwood represents a unique breakthrough in the age of inference with increased computation power, memory capacity, ICI networking advancements (1.5 times of Trillium) and reliability.”

Gemini In Google Workspace

At the event, Google announced a series of updates for Gemini in Workspace, spanning tools such as Docs, Sheets, Meet, Chat, and more, as well as the all-new Google Workspace Flows.

Google Workspace Flows enables users to automate and orchestrate their work across apps using Google’s powerful AI. This capability automates multi-step processes, including researching, analyzing, and generating content. It’s an easy-to-use capability that can design and build sophisticated, logic-driven flows from simple language prompts and without the need for complex coding or configurations. Furthermore, users can develop custom AI agents to execute specialized tasks using Gems.

Google Docs sees new audio features, including the addition of audio overviews inspired by its popularity in NotebookLM. It will also allow users to create full audio versions of documents or choose podcast-style overviews for key highlights.

Google is bringing “Help me refine” to Docs, elevating the application from more than just a text generator to a writing coach. Ahead of rewriting sentences, it offers suggestions to improve content, formatting, structure, consistency, clarity, and more. “The goal isn't just to fix the document, but to help you become a more effective communicator over time,” says Google.

Google Sheets now comes with “Help me analyze” to provide users with an on-demand analyst that offers guidance, points out interesting trends, suggests next steps, digs deeper, and creates interactive charts. This capability will come in later this year.

Google Vids, which is already helping businesses create polished videos, is getting an upgrade—users will now be able to generate high-quality, original video clips directly within Vids, powered by Veo 2. This feature will roll out to alpha customers soon.

Google Meet will offer enhanced meeting summaries, information about specific topics or decisions, the ability to organize thoughts, and more.

Group chats in Chat just got cool! Google’s Gemini can provide a detailed summary, highlighting open questions, key decisions, and next steps, just by typing @gemini in any Chat conversation.

Of course, Google reiterated the importance of data security and privacy—which remains key throughout its AI capabilities.

Lyria In Vertex AI

Google just added Lyria, its text-to-music model, to Vertex AI, making it the only platform with generative media models across all modalities—video, image, speech, and music—allowing users to build complete, production-ready assets.

“Lyria produces high-fidelity audio, meticulously capturing subtle nuances and delivering rich, detailed compositions across a range of musical genres.” It can help businesses enhance experiences and streamline content creation by creating custom music tracks in minutes, aligned to the mood, pacing, and narrative of content pieces.

Veo 2 comes with a host of new editing features. “Inpainting” helps users get professional edits without manual retouching and simplifies the removal of unwanted background images, logos, or distractions.


“Outpainting” enables users to extend the frame of existing video footage, allowing users to easily adapt content for various screen sizes and aspect ratios.


Furthermore, users can implement sophisticated cinematic effects and cohesively connect two existing assets (interpolation) with smooth transitions and maintaining visual continuity.


Chirp 3, Google’s audio understanding and generation model that offers natural and realistic speech in over 35 languages and 8 speaker options, comes with two new features, including enabling users to generate realistic custom voices from 10 seconds of audio input and accurately separating and identifying individual speakers in multi-speaker recordings.

Imagen 3, Google’s highest-quality text-to-image model, provides users with a powerful and user-friendly way to refine and tailor any image and has been applying several improvements over the past year. 

Cloud Wide Area Network

Google brings its global private network to enterprises around the world through the new Cloud Wide Area Network or WAN.

Cloud WAN is a fully managed, reliable, and secure enterprise backbone that leverages Google’s planet-scale network, which spans over 200 countries and territories and is connected by more than two million miles of fiber. It’s designed to transform enterprise-wide area network (WAN) architectures and is optimized for application performance—delivering more than 40% faster performance and reducing TCO by up to 40%. 

Other Announcements

 

• Google highlighted recent advancements made in quantum computing—especially with the release of Willow, its newest quantum chip—and AI infrastructure.

• After releasing Gemini 2.5 a few weeks ago, Google said Gemini 2.5 Pro is available now for everyone in Google AI Studio, Vertex AI, and in the Gemini app.

• At the same time, Google announced Gemini 2.5 Flash, its low latency and most cost-efficient thinking model, which is coming soon in Google AI Studio, Vertex AI, and in the Gemini app.

• Google also boasted about how Gemini is helping it create net-new products and experiences such as NotebookLM, which is used by 100,000 businesses.

• Google brough updates to Agentspace, making it easier for customers to discover, create and adopt AI agents. It also spoke about growing its AI Agent Marketplace, allowing customers to easily browse and purchase AI agents from partners.

What do you think about Google’s announcements? Do you think these moves will help it capture a better position in the AI market?

Let us know in the comments below!