An energetic welcome to Google Cloud AI includes the question Why Not? Why not transform organizations with AI?

Thomas Kurian shares a vision for how AI can transform organizations.

Today, that vision is not just a possibility – it’s the vibrant reality we are collectively building. 

Thomas kurian

Google delivered more than 3,000 product advancements across Google Cloud and Workspace in 2024. There are now over 4 million developers building with the power of Gemini, the most advanced AI model family.

All of this is powered by global infrastructure, which has grown to 42 regions.

Known as Cloud Wide Area Network (or Cloud WAN). It makes Google’s global private network available to all Google Cloud customers. Cloud WAN is a fully managed, reliable, and secure enterprise backbone to transform enterprise wide area network (WAN) architectures.

The Accelerating Momentum of Google AI

The true measure of Google success lies in the impact on customers.

Customers are choosing Google Cloud for three fundamental reasons: 

1. AI-Optimized Platform: Only Google Cloud offers an AI-optimized platform with leading price, performance and precision.
2. Open and Multi-Cloud Capabilities: Google Cloud allows customers to adopt AI agents while connecting them with their existing IT landscapes.
3. Interoperability: Google Cloud offers an enterprise-ready AI platform, built for interoperability. This enables customers to adopt AI, while addressing evolving sovereignty, security, privacy, and regulatory requirements.

Google is proud to announce innovations, including: seventh-generation TPU: Ironwood, that delivers levels of efficiency. Advancements in Google Distributed Cloud enable customers to bring Gemini models on-premise. They also support a full suite of generative media models and Gemini 2.5. Innovations in Vertex, like Agent Development Kit and Agent2Agent Protocol enable a multi-agent ecosystem.

AI Hypercomputer: Unleashing Unprecedented Computational Power

AI Hypercomputer is a revolutionary supercomputing system meticulously designed to simplify AI deployment, dramatically improve performance, and optimize costs. It includes hardware, software, and consumption models. These are all optimized to deliver more intelligence at a consistently low price. This is essential for training, tuning, and serving AI workloads.

Google introduces:

• Ironwood TPUs: The seventh-generation TPU, Ironwood represents Google’s largest and most powerful TPU to date with 10x improvement from the most recent high-performance TPU.
• Cluster Director: Cluster Director helps companies deploy a large number of accelerators. These are managed as a single, unified unit of compute.

This builds on the Google commitment to delivering AI hardware optionality to customers, including NVIDIA GPU-based offerings:

• A4 and A4X VMs: Google has significantly enhanced the GPU portfolio. A4 and A4X VMs are now available. They are powered by NVIDIA’s groundbreaking B200 and GB200 Blackwell GPUs.
• NVIDIA Vera Rubin GPUs: Google Cloud will be among the first to offer NVIDIA’s Vera Rubin GPUs.

We’re introducing groundbreaking storage innovations:

• Hyperdisk Exapools: Offering the highest aggregate performance and capacity block storage of any hyperscaler.
• Anywhere Cache: Intelligently keeps data close to accelerators, reducing storage latency by up to 70% and significantly accelerating training times.
• Rapid Storage: Our first zonal object storage solution. It offers an impressive 5x lower latency for random reads and writes.

Software is the key to orchestrating and simplifying access to this powerful hardware. Today, we’re introducing three significant enhancements for AI inference:

• Google Kubernetes Engine (GKE) Inferencing: New inference capabilities in GKE include gen AI-aware scaling and load-balancing features.
• Pathways Availability: Google’s own distributed ML runtime, Pathways, is now available for the first time to Cloud customers. Developed by Google DeepMind, Pathways enables state-of-the-art, multi-host inferencing for dynamic scaling with exceptional performance at optimal cost.
• vLLM Availability: We’re bringing vLLM to TPUs. Customers who have optimized PyTorch with vLLM for GPUs can easily and cost-effectively run their workloads on TPUs.
• All of these AI Hypercomputer hardware and software enhancements work together. They enable Google to deliver intelligence and useful AI output.

Not everyone has been able to benefit from these advancements. Historically, organizations with strict regulatory, sovereignty, latency, and data volume issues have been able to access the latest AI technology. They must keep their data on-premises. We have partnered with NVIDIA to bring Gemini to NVIDIA Blackwell systems. Dell is a key partner. This allows for local use in air-gapped and connected environments. This complements our GDC air-gapped product, which is now authorized for U.S. Government Secret and Top Secret levels. Gemini is available on it. It provides the highest levels of security and compliance.

Google’s Leading Models: Bringing the Best of Google DeepMind to Cloud Customers

We build upon the groundbreaking research of Google DeepMind. We’re delivering rapid innovation across a diverse spectrum of first-party models. Each model is designed to meet the unique needs of various customers.

Gemini, the most capable family of AI models, has been at the forefront of this innovation. Gemini 2.5 models are thinking models. Google brought Gemini 2.5 Pro to Vertex AI in public preview. Pro is optimized for precision.

Coming to Vertex AI is Gemini 2.5 Flash – our workhorse model optimized specifically for low latency and cost efficiency. Real-time summaries or quick access to documents are needed in these situations. Gemini 2.5 Flash adjusts the depth of reasoning based on the complexity of prompts. You can also control performance based on customers’ budgets. These new features make AI easier to use. These features enable our customers to build AI that solves complex problems.

Beyond Gemini, we have an incredible suite of generative media models. These models are driving new levels of efficiency. They enhance creativity and boost customer engagement.

Google is the only company to offer models across all modalities. These include images, voice, music, and video.

Thomas kurian

All of these are available today on Vertex AI. These creative tools are delivering real-world impact for customers like Agoda. Agoda is a leading digital travel platform. It creates unique and captivating visuals and videos of travel destinations using Imagen and Veo on Vertex AI. This enhances customer engagement and drives bookings.

Google is announcing groundbreaking advancements:

• Imagen 3: Our highest quality text-to-image model now has improved image generation. It also has enhanced inpainting capabilities for reconstructing missing or damaged portions of an image. This update significantly elevates the quality of object removal, delivering a more natural and seamless editing experience. 
• Chirp 3: Our groundbreaking audio generation model now offers a new method. You can create custom voices with just 10 seconds of audio input. This enables enterprises to personalize call centers. They can also develop accessible content and establish unique brand voices. These transcription features separate and identify individual speakers in multi-speaker recordings. This significantly improves the clarity of transcriptions, enhancing usability for applications like meeting summaries, podcast analysis, and multi-party call recordings.
• Lyria: The industry’s first enterprise-ready, text-to-music model can transform simple text prompts into 30-second music clips. This produces high-fidelity audio, capturing subtle nuances and delivering rich, detailed compositions across a range of musical genres. With Lyria, organizations can create soundtracks for marketing campaigns and product launches. They can also design immersive in-store experiences, podcasts, and other digital content.
• Veo 2: Veo 2 is the Google industry-leading video generation model. This transforms Veo on Vertex AI from a generation tool to a comprehensive video creation and editing platform. Advancements include: inpainting to remove unwanted background images, logos or distractions. This allows professional edits without manual retouching. Outpainting is used to extend the frame of existing video footage. This extension helps customers optimize video for different screen sizes and platforms. Sophisticated cinematic techniques are employed to provide shot composition, camera angles, and pacing. Interpolation lets companies define the beginning and end of a video sequence. Veo seamlessly generates the connecting frames.

At Google, we bring the best of Google DeepMind and Google Research together with new infrastructure and AI capabilities in Google Cloud, including:

• AlphaFold 3 was developed by Google DeepMind and Isomorphic Labs. It can predict the structure and interactions of all of life’s molecules with unprecedented accuracy. The new AlphaFold 3 High-Throughput Solution is available for non-commercial use and deployable via Google Cloud Cluster Toolkit. This solution enables batch processing of up to tens of thousands of protein sequences. It minimizes cost with autoscaling infrastructure. 
• WeatherNext AI models, from Google DeepMind and Google Research, enable fast, accurate weather forecasting. They are now available in Vertex AI Model Garden.

Vertex AI: The Comprehensive Platform for AI Innovation

These are the models available on Vertex AI.

Vertex AI is empowering companies to gain significant efficiencies by automating and accelerating routine, mission-critical processes.

We are also announcing new advancements in Vertex AI to improve your ability to manage your AI initiatives:

• Vertex AI Dashboards: These help you monitor usage, throughput, latency, and troubleshoot errors, providing you with greater visibility and control.
• Training and Tuning capabilities: You can manage custom training and tuning with your own data. This is done on top of foundational models. You can do this securely across all first-party model families. These include Gemini, Imagen, Veo, embedding and translation models. There are also open models like Gemma, Llama, and Mistral.
• Vertex AI Model Optimizer: This capability uses Google’s unique understanding of Gemini. It automatically directs your query to the most performant model and tools. This is based on your quality, speed, and cost preferences. 
• Live API: To enable truly conversational interactions, Live API offers streaming audio and video directly into Gemini. Now your agents can process and respond to rich media in real time, opening new possibilities for immersive, multimodal applications.

Expanding Vertex to Enable a Multi-Agent Ecosystem

We believe Vertex is the most open developer AI platform in the cloud. It is the only one delivering multi-agent solutions. These solutions empower multiple AI agents to work together. Agents are intelligent systems that exhibit reasoning, planning and memory capabilities. This is all to accomplish tasks on your behalf and under your supervision. Agents are set to become increasingly important in the workforce. They will collaborate with employees to drive efficiencies. They will also enhance decision-making and accelerate innovation.

• Agent Development Kit (ADK): Our new AI Agent Development Kit (ADK) is an open-source framework. It simplifies the process of building sophisticated multi-agent systems. It maintains precise control over agent behavior. With ADK, you can build an AI agent in under 100 lines of intuitive code. ADK supports Model Control Protocol (MCP) to allow developers plugin available tools. ADK also connects seamlessly with APIs managed by Apigee to leverage your own business APIs
• Agent2Agent (A2A) protocol: We’re proud to be the first hyperscaler to create an open Agent2Agent (A2A) protocol. This protocol helps enterprises support multi-agent ecosystems.
• Agent Garden: Agent Garden is a collection of ready-to-use samples and tools, directly accessible  in ADK. It enables you to connect your agents to over 100 pre-built connectors. You can also use your custom APIs, integration workflows, or data stored within your cloud systems, like BigQuery and AlloyDB.
• Interoperability:  With Vertex AI, you can seamlessly manage agents built on multiple agent frameworks, including LangGraph and Crew AI. We are also announcing that you can build AI agents directly on your existing NetApp data, no data duplication required. You can choose any method to build agents. We can assist you in connecting them to leading application and enterprise data companies.

Google Agentspace: Empowering Every Employee with AI

We’re empowering enterprises to put AI agents in the hands of every employee with Google Agentspace.

Agentspace combines Google-quality enterprise search with conversational AI. It includes Gemini and third-party agents to empower employees. They can find and synthesize information from within their organizations. They can also converse with AI agents and take action with their enterprise applications. It delivers a broad set of tools. These tools include pre-built connectors to search and transact with documents, databases, and SaaS applications. Additionally, there is advanced security and compliance to protect your data and IP. 

Today, employees can use Agentspace to access expert Google-built AI agents. These include NotebookLM, which is already used by more than 100,000 businesses. It allows you to upload multiple source materials. These materials include PDFs, Google Docs, websites, and YouTube videos. You can then summarize the content, ask questions about the materials, and format responses in a specific way.

Google is announcing enhancements to Agentspace:

• Chrome Enterprise: Agentspace is now integrated with Chrome Enterprise. This allows employees to search and access all their enterprise resources directly from the search box in Chrome. This integration streamlines workflows and boosts productivity.
• Agent Gallery: This provides employees a single view of available agents across the enterprise. This includes agents from Google, internal teams, and partners.
• Agent Designer: A no-code interface for creating custom agents automates everyday work tasks. Agent Designer helps employees adapt agents to their individual workflows and needs, no matter their technical experience. 
• Idea Generation agent: This agent uses a tournament-style framework. It effectively ranks ideas based on employee-defined criteria. The agent can help employees refine or generate new ideas.
• Deep Research agent: This agent explores complex topics on employees’ behalf. It provides them with findings in a comprehensive, easy-to-read report. 

Google Workspace: AI-Powered Productivity

Gemini is not only powering best-in-class AI capabilities as a model. It is also supercharging our own products, like Google Workspace. This includes popular apps like Gmail, Docs, Drive, and Meet. Workspace’s AI features have been enhancing employee collaboration for more than a decade. Google is announcing a number of Workspace innovations to further empower users with AI, including:

• Help Me Analyze: This powerful feature turns Google Sheets into your personal business analyst. It intelligently identifies insights from your data without the need for explicit prompting.
• Docs Audio Overview: With audio overviews in Docs, you can create high-quality audio. These can be human-like or podcast-style summaries of your documents.
• Google Workspace Flows: Workspace Flows helps you automate daily work and repetitive tasks. This includes managing approvals, researching customers, organizing your email, and summarizing your daily agenda.

High-Impact Agents: Delivering Tangible Business Results

Across our AI portfolio, we’re witnessing a surge in the creation of highly-advanced AI agents. Organizations are pushing the boundaries. They are developing agents that not only excel in coding, data, and security. These agents also revolutionize customer service and the creative process. Here are five categories of agents where we are already seeing business impact:

Customer Agents empower your customers to quickly find answers and the right products. They can synthesize and reason across all types of multi-modal information, including text, audio, images, and video. They communicate and engage naturally with human-like speech and dialog. They connect across enterprise applications on behalf of the user and can be used anywhere.

We have already introduced Vertex AI Search for Healthcare and Retail. It allows doctors, nurses, and providers to rapidly search and analyze diverse patient data. Retailers can also add product discovery to their websites powered by Google Search.

Google Cloud’s own pre-built Customer Engagement Suite is transforming customer service. Grounded in a company’s data, it provides out-of-the-box functionality to build agents. These agents can be deployed across web, mobile, call center, in-store, and with third-party telephony and CRM systems. These unique capabilities have led to rapid growth in conversational AI agent usage.

Today, we’re announcing the next-generation of our Customer Engagement Suite. It will include human-like, high-definition voices. The suite will have the ability to understand emotions so agents can adapt better during conversations. It will offer streaming video support so virtual agents can interpret and respond in real-time through customer devices. Additionally, there will be AI assistance to build custom agents in a no-code interface.  

We are enhancing conversational customer experiences beyond the call center. We offer purpose-built vertical agents that address specific industry use cases. These include Food Ordering, Automotive, and Retail. Examples of these agents in-action include:

• Wendy’s AI drive-through ordering system handles 60,000 orders daily.
• MercedesBenz provides conversational search and navigation in its new CLA car series.
• The Home Depot built Magic Apron, an agent that offers expert home improvement guidance 24/7.

Creative Agents are being used to supercharge creative teams, including those in media production, marketing, advertising, design. In some cases, agents are augmenting creative teams to enable content production at massive scale. In others, they are helping reimagine how stories can be told for a new generation of audiences. Google is using this technology. Marketing teams are directing its application to build the Fall Pixel phone ad campaign. A few other examples include:

• WPP, which built Open for its 120,000 employees to concept, produce and measure campaigns.
• Monks. Flow is using Google AI to help localize creative for campaigns.
• The Brandtech Group built Pencil, a gen AI platform for brands to create ads, predict performance, and optimize active campaigns.
• Mondelez is aiming for a 25% return on investment. They are using gen AI for content and creative development. This includes their brands like Oreo and Cadbury.

Google is thrilled to partner with Adobe. Adobe is the leader in creativity. The partnership brings advanced Imagen 3 and Veo 2 models to applications like Adobe Express.

Data Agents enable data teams to effectively manage data and business teams to activate it. Our data platform BigQuery has 5x more customers than the two leading independent data cloud companies. With BigQuery, you can activate all your data for AI. You can combine structured and unstructured data. Additionally, you can work with open formats like Apache Iceberg directly integrated into BigQuery. You can also use BigQuery to access data in any storage system, any SaaS application or on any cloud.

Google is announcing specialized agents for every member of the data team:

• For data engineering teams, we are delivering support to build data pipelines. We help perform data preparation like transformation and enrichment. We also maintain data quality with anomaly detection. Additionally, we automate metadata generation. 
• For data science teams, our AI agent acts as a comprehensive coding partner in your data science notebook. It automates feature engineering. It provides intelligent model selection. It enables scalable training and faster iteration.
• For data analysts and business users, our Conversational Analytics Agent performs powerful, trustworthy analyses – entirely in natural language. You can embed this agent inline in your own web mobile application. 

Customers are seeing benefits with Google Data Agents. Mattel, for example, can analyze sentiment and consumer preferences in real time. Using BigQuery, Spotify harnesses enormous amounts of data. It delivers personalized experiences to over 675 million users worldwide. Additionally, Unilever reaches millions of retailers in emerging markets. It processes 75,000 orders daily. Bayer built an agent that combines Google search trends and internal data to forecast flu trends, improving public health outcomes. And public sector organizations like the State of Nevada are using agents to speed up benefit claims.

Coding Agents: At Google, AI is powering tools across our software development life cycle, including tools that help developers code. Today, more than 25% of new code at Google is generated by AI.

Gemini’s fast performance, large context window, and advanced reasoning capabilities make it exceptionally well-suited for coding assistance. Google offers Gemini Code Assist in Google Cloud, Android Studio, Firebase Studio, and your favorite IDE. Our enterprise version understands your code base, standards, and conventions. We’re announcing new Code Assist agents to help with everything from modernizing code to assisting with the full software development lifecycle:

• Kanban board: Developers can interact with our agents on a Kanban board. 
• Partner Integration: Code Assist also has integrations from partners such as Atlassian, Sentry, Snyk, with many more coming soon. Outside of Google, Gemini is also available for your development needs in Aider, Cursor, GitHub Copilot, Replit, Tabnine, and Windsurf.

Security Agents can dramatically increase the speed and effectiveness of security analysts. AI integration in our security products is a key factor. This is why organizations worldwide are making Google part of their security team.

We offer critical cyber defense capabilities for today’s challenging threat environment, and today, we’re introducing a number of new innovations:

• Google Unified Security: This solution brings together our visibility, threat detection, and AI powered security operations. It includes continuous virtual red-teaming and the most trusted enterprise browser. It incorporates Mandiant expertise. All these are in one converged security solution running on a planet-scale data fabric.
• Alert triage agent: This agent performs dynamic investigations on behalf of users. It analyzes the context of each alert, gathers relevant information and renders a verdict on the alert. It also provides a history of the agent’s evidence and decision making. 
• Malware analysis agent: This agent investigates whether code is safe or harmful. It builds on Code Insight to analyze potentially malicious code.

Commitment to Openness and Partnership

Realizing the full potential of gen AI requires an enterprise AI platform. This platform offers a broad, practical set of capabilities. It should be optimized for both cost and performance. This platform must also be open, seamlessly integrated with existing systems, and be supported by a strong partner ecosystem. 

At Google Cloud, the commitment to interoperability enables you to:

• Leverage your existing IT environment by connecting with other clouds and applications. You can secure cross-cloud networking using your existing security platforms with Cross Cloud Interconnect. You can support federated identity with Microsoft EntraID. Use Google Cloud’s BigQuery, Spanner, and AlloyDB without moving from Amazon or Azure.
• Access hundreds of the world’s leading ISVs. They integrate with Google AI. You can easily deploy them from the Google Cloud Marketplace.
• Utilize our extensive network of services partners. These include Accenture, Capgemini, Deloitte, HCLTech, KPMG, TCS, Wipro, and others.
• Implement sovereign clouds, which Google enabled with local partners to meet international regulations. They offer Google Cloud Sovereign AI services in the public cloud. They are also available in sovereign cloud and distributed clouds. Additionally, they provide these services with Google Workspace. 

This is an unprecedented time to be working with these transformative technologies. Google delivers world-class infrastructure, models, platforms and agents.

Google has been incorporating machine learning into our products for more than 20 years. Our investment in AI is deeply rooted in our core mission. This mission is to organize the world’s information and make it universally accessible and useful.

Now, if you can’t come to the cloud. Google cloud will bring Google AI to you.

Jensen Huang, CEO, NVIDIA

Google Cloud AI. Google Cloud Next 25 Opening Keynote: The new way to cloud’ Thomas Kurian, CEO, Google. (2025, April).

Google Cloud AI. Google Cloud Next 25 Opening Keynote: The new way to cloud’ Sundar Pichai, CEO, Google and Alphabet. (2025, April).

Google Cloud AI. Google Cloud Next 25 Opening Keynote: The new way to cloud’ Amin Vahdat, VP, GM, ML. (2025, April).

Google Cloud AI. Google Cloud Next 25 Opening Keynote: The new way to cloud’ Gabe Weiss, Developer Advocacy Manager. (2025, April).

Google Cloud AI. Google Cloud Next 25 Opening Keynote: The new way to cloud’ Lisa O’Malley, Sr. Director, Product Management, Applied AI. (2025, April).

Google Cloud AI. Google Cloud Next 25 Opening Keynote: The new way to cloud’ Pali Bhat, Product Officer. (2025, April).

Google Cloud AI. Google Cloud Next 25 Opening Keynote: The new way to cloud’ Brad Calder, VP & GM, Google Cloud. (2025, April).

Google Cloud AI. Google Cloud Next 25 Opening Keynote: The new way to cloud’ Sandra Joyce, Google Threat Intelligence. (2025, April).

Google Cloud AI. Google Cloud Next 25 Opening Keynote: The new way to cloud’ Payal Chakravarty, Director Product Management,Google Security. (2025, April).

Google Cloud AI. Google Cloud Next 25 Opening Keynote: The new way to cloud’ Nav Jagpal, Principal Engineer, Google Security. (2025, April).

Google Cloud AI. Google Cloud Next 25 Opening Keynote: The new way to cloud’ Jensen Huang, CEO, NVIDIA. (2025, April).