Google Cloud

One of the ways threat actors keep up with the constantly evolving cyber defense landscape is by raising the level of sophistication of their attacks. This trend can be seen across many of our engagements, particularly when responding to China-nexus groups. These actors have demonstrated the ability to create custom malware ecosystems, identify and use zero-day vulnerabilities in security and other appliances, leverage proxy networks akin to botnets, target edge devices and platforms that traditionally lack endpoint detection and response, and employ custom obfuscators in their malware. They take these extra steps to evade detection, stifle analysis, and ultimately stay on systems for longer periods of time. 

However, not all successful attacks are highly complex and technical. Many times attackers will take advantage of the opportunities that are made available to them. This includes using credentials stolen in infostealer operations to gain initial access. Mandiant has seen such a rise in infostealer use that stolen credentials are now the second highest initial infection vector, making up 16% of our investigations. Other ways attackers are taking advantage of opportunities is by exploiting gaps and risks introduced in cloud migrations, and targeting unsecured data repositories to obtain credentials and other sensitive information. 

Today we released M-Trends 2025, the 16th edition of our annual report, to help organizations stay ahead of all types of attacks. We dive deep into several trends and share data and analysis from the frontlines of our incident response engagements to arm defenders with critical insights into the latest cyber threats.

Data and Trends

M-Trends 2025 data is based on more than 450,000 hours of Mandiant Consulting investigations. The metrics are based on investigations of targeted attack activity conducted between Jan. 1, 2024 and Dec. 31, 2024. Key findings in M-Trends 2025 include: 

• 55% of threat groups active in 2024 were financially motivated, which marks a steady increase, and 8% of threat groups were motivated by espionage.

• Exploits continue to be the most common initial infection vector (33%), and for the first time stolen credentials rose to the second most common in 2024 (16%).

• The top targeted industries include financial (17.4%), business and professional services (11.1%), high tech (10.6%), government (9.5%), and healthcare (9.3%).

• Global median dwell time rose to 11 days from 10 days in 2023. Global median dwell time was 26 days when external entities notified, 5 days when adversaries notified (notably in ransomware cases), and 10 days when organizations discovered malicious activity internally.

M-Trends 2025 dives deep into the aforementioned infostealer, cloud, and unsecured data repository trends, and several other topics, including:

• Democratic People’s Republic of Korea deploying citizens as remote IT contractors, using false identities to generate revenue and fund national interests.

• Iran-nexus threat actors ramping up cyber operations in 2024, notably targeting Israeli entities and using a variety of methods to improve intrusion success.

• Attackers targeting cloud-based stores of centralized authority, such as single sign-on portals, to gain broad access.

• Increased targeting of Web3 technologies such as cryptocurrencies and blockchains for theft, money laundering, and financing illicit activities.

Recommendations for Organizations

Each article in M-Trends 2025 offers critical recommendations for organizations to enhance their cybersecurity postures, with several of them being applicable to multiple trends. We advise that organizations:

• Implement a layered security approach that emphasizes sound fundamentals such as vulnerability management, least privilege, and hardening.

• Enforce FIDO2-compliant multi-factor authentication across all user accounts, especially privileged accounts.

• Invest in advanced detection technologies and develop robust incident response plans. 

• Improve logging and monitoring practices to identify suspicious activity and reduce dwell time. 

• Consider threat hunting exercises to proactively search for indicators of compromise.

• Implement strong security controls for cloud migrations and deployments. 

• Regularly assess and audit cloud environments for vulnerabilities and misconfigurations.

• Mitigate insider risk by practicing thorough vetting processes for employees (especially remote workers), monitoring for suspicious activity, and enforcing strict access controls.

• Keep up-to-date with the latest threat intelligence, adapt security strategies accordingly, and regularly review and update security policies and procedures to address evolving threats. 

Be Ready to Respond

The M-Trends mission has always been to equip security professionals with frontline insights into the latest evolving cyberattacks and to provide practical and actionable learnings for better organizational security.

Read the full M-Trends 2025 report today, and register for our M-Trends 2025 webinar series for a more in-depth look at the data, topics, and recommendations discussed in the report. The M-Trends 2025 Executive Edition is also available, featuring a high-level look at the data and trends, along with key recommendations.

Last year, Google Cloud and LangChain announced integrations that give generative AI developers access to a suite of LangChain Python packages. This allowed application developers to leverage Google Cloud’s database portfolio in their gen AI applications to drive the most value from their private data.

Today, we are expanding language support for our integrations to include Go, Java, and JavaScript.

Each package will have up to three LangChain integrations: 

1. Vector stores to enable semantic search for our databases 

2. Chat message history to enable chains to recall previous conversations

3. Document loader for loading documents from your enterprise data

Developers now have the flexibility to create intricate workflows and easily interchange underlying components (like a vector database) as needed to align with specific use cases. This technology unlocks a variety of applications, including personalized product recommendations, question answering, document search and synthesis, customer service automation, and more.

In this post, we’ll share more about the integrations – and code snippets to get started.

New language support

LangChain is known for its popular Python package; however, your team’s expertise and services may not be in Python. Java and Go are commonly used programming languages for production-grade and enterprise-scale applications. Developers may prefer Javascript and Typescript for the asynchronous programming support and compatibility with front-end frameworks like React and Vue.

In addition to Python developers, the LangChain developer community encompasses developers proficient in Java, JavaScript, and Go. It is an active and supportive community centered around the LangChain framework, which facilitates the development of applications powered by large language models (LLMs).

Google Cloud is dedicated to providing secure and easy to use database integrations for your Gen AI applications. Our integrations embed Google Cloud connectors that create secure connections, handle SSL certificates, and support IAM authorization and authentication. The integrations are optimized for PostgreSQL databases (AlloyDB for PostgreSQL, AlloyDB Omni, Cloud SQL for PostgreSQL) to ensure proper connection management, flexible tables schemas, and improved filtering. 

JavaScript Support

JavaScript developers can utilize LangChain.js, which provides tools and building blocks for developing applications leveraging LLMs. LangChain simplifies the process of connecting LLMs to external data sources and enables reasoning capabilities in applications. Other Google Cloud integrations, such as Gemini models, are available within LangChain.js, allowing seamless interaction with GCP resources.

Below are the integrations and their code snippets to get started.

Install the dependency:

Engine

Use this package with AlloyDB for PostgreSQL and AlloyDB Omni by customizing your Engine to connect your instance. You will need the AlloyDB Auth Proxy to make authorized, encrypted connections to AlloyDB instances.

Vector store

Chat message history

Loader

Java Support

For Java developers, there’s LangChain4j, a Java implementation of LangChain. This allows Java developers to build LLM-powered applications with a familiar ecosystem. In LangChain4j, you can also access the full array of VertexAI Gemini models.

*Note: Cloud SQL integrations will be released soon.

Below are the integrations and their code snippets to get started.

For Maven in pom.xml:

Engine

Embedding store

Document loader

Go support

LangchainGo is the Go programming language port of LangChain.

The LangChain framework was designed to support the development of sophisticated applications that connect language models to data sources and enable interaction with their environment. The most powerful and differentiated applications go beyond simply using a language model via an API; they are data-aware and agentic.

Last year Google’s SDKs were added as providers for LangChainGo; this makes it possible to use the capabilities of the LangChain framework with Google’s Gemini models as LLM providers.

We now have AlloyDB and Cloud SQL for PostgreSQL support in LangchainGo.

Below are the integrations and their code snippets to get started.

Install the dependency

Engine

Vector Store

Chat message history

*Note code is shown for AlloyDB. See links for Cloud SQL for Postgres examples.

Get started

The LangChain Vector stores integration is available for Google Cloud databases with vector support, including  AlloyDB, Cloud SQL for PostgreSQL, Firestore, Memorystore for Redis, and Spanner.

The Document loaders and Memory integrations are available for all Google Cloud databases including AlloyDB, Cloud SQL for MySQL, PostgreSQL and SQL Server, Firestore, Datastore, Bigtable, Memorystore for Redis, El Carro for Oracle databases, and Spanner. Below are a few resources to get started.

Resources:

• Github

• Build LLM applications in Go

• Codelabs

CodeRabbit, a rapidly growing AI code review tool, is leveraging Google Cloud Run to cut code review time and bugs in half by safely and efficiently executing untrusted code. 

CodeRabbit improves code quality and automates code reviews by analyzing changes against the entire codebase and generating scripts for deeper analysis. It integrates with code hosting platforms to provide automated feedback on pull requests.

To safely execute untrusted code, CodeRabbit needed an execution environment that was scalable, cost-effective, and secure enough to analyse and run their customers’ code.

In this post, we’ll share how CodeRabbit built an AI code review agent with Google Cloud Run to scale dynamically and handle high volumes efficiently and securely.

CodeRabbit integrates directly with platforms like GitHub and GitLab, providing automated code reviews triggered by pull requests. Its integration with the foundational models doesn’t just analyze the changed files; it assesses the impact of those changes on the entire codebase. This requires a sophisticated system that can:

1. Clone the user’s repository.

2. Set up a build environment with necessary dependencies (think npm install, go mod download, etc.).

3. Run static analysis tools including 20+ linters and security scanners.

4. Execute AI-generated scripts. This is where things get really interesting. CodeRabbit’s AI agent creates shell scripts to navigate the code, search for specific patterns (using tools like cat, grep, and even ast-grep), and extract relevant information. It can even generate Python code for analysis.

5. Interact with external services. CodeRabbit can also perform actions by generating and executing curl commands, for example to interfacing with services like Slack, Jira and Linear.

This solution needs to be scalable, cost-effective, and above all, secure. The code being analyzed and executed is, by definition, untrusted. It could be incomplete, buggy, or even contain malicious intent.

The solution: Cloud Run

CodeRabbit’s architecture cleverly combines several technologies to create a robust and isolated execution environment:

1. Cloud Run services: CodeRabbit uses Cloud Run services as the foundation. Incoming webhook events (from GitHub, GitLab, etc.) are first handled by a lightweight Cloud Run service that performs billing and subscription checks. This service then pushes a task to Google Cloud Tasks.

2. Google Cloud tasks: This acts as a queue, decoupling the webhook handling from the actual code execution. This allows CodeRabbit to handle bursts of pull requests without overwhelming the system.

3. Cloud Run execution service: This is the heart of the system. A separate Cloud Run service pulls tasks from the Cloud Tasks queue. Each task represents a code review request. This service is configured with a 3600 second long request timeout and a concurrency of 8 requests per instance, allowing it to scale based on CPU utilization. This setup is crucial because code reviews are long-running operations, often taking 10-20 minutes to complete. The Execution Service uses an in-memory volume mount where the entire repository, build artifacts, and temporary files are stored.

4. Sandboxing: All Cloud Run instances are sandboxed with two layers of sandboxing and can be configured to have minimal IAM permissions via dedicated service identity. In addition, CodeRabbit is leveraging Cloud Run’s second generation execution environment, a microVM providing full Linux cgroup functionality. Within each Cloud Run instance, CodeRabbit uses Jailkit to create isolated processes and cgroups to further restrict the privileges of the jailed process.

Sandboxing is especially critical for CodeRabbit in scenarios where untrusted code must be executed, such as:

• Static analyzers that support custom, untrusted plugins (e.g., ESLint, Rubocop)

• LLM-generated verification scripts for deeper analysis of the entire codebase

• LLM-generated CLI actions, such as opening GitHub or Jira issues

• Python-based advanced analyses

CodeRabbit’s use of Cloud Run allows it to scale dynamically. During peak hours, CodeRabbit’s Agentic PR Reviewer service receives up to 10 requests/second served by over 200 Cloud Run instances. Each Cloud Run instance is fairly bulky and utilizes 8vCPUs and 32GiB memory. CodeRabbit sees high CPU utilization, significant network traffic (downloading repositories and dependencies), and high memory usage when powering their PR reviewer service with Cloud Run.

Try this on your own

CodeRabbit’s use of Google Cloud Run is a compelling example of how to build a secure, scalable, and cost-effective platform for running AI-powered code analysis. Their architecture provides a blueprint for developers tackling similar challenges, and their experience highlights the evolving capabilities of serverless technologies. We’re excited to see how their platform advances as Cloud Run continues to add new features.

Learn more about developing, deploying and hosting AI agents on Cloud Run, watch the “Build AI Agents on Cloud Run” Cloud Next ’25 session featuring CodeRabbit, and give CodeRabbit a try.

For years, data teams have relied on the BigQuery platform to power their analytics and unlock critical business insights. But building, managing, and troubleshooting the data pipelines that feed those insights can be a complex, time-consuming process, requiring specialized expertise and a lot of manual effort. Today, we’re excited to announce our vision, a major step forward in simplifying and accelerating data engineering with BigQuery data engineering agent.

These agents aren’t just assistive tools, but agentic solutions, designed to act as intelligent partners in your data workflows. They automate daunting tasks, collaborate with your team, and continuously learn and adapt, freeing you to focus on what matters most: extracting value from your data.

Why a data engineering agent?

The world of data is changing. Organizations are generating more data than ever before, and that data is coming from a wider variety of sources, in a multitude of formats. At the same time, businesses need to move faster, making quick, data-driven decisions to stay competitive.

This creates a challenge. Traditional data engineering approaches often involve:

• Tedious manual coding: Building and modifying pipelines can require writing and updating complex SQL queries, which is time-consuming and error-prone.

• Schema struggles: Mapping data from different sources to the right format can be time-intensive, especially as schemas evolve.

• Difficult troubleshooting: Diagnosing and fixing pipeline issues can involve lengthy sifting through logs and code, delaying critical insights.

• Siloed expertise: Building and maintaining pipelines often requires specialized skills, creating bottlenecks and limiting who can contribute.

The BigQuery data engineering agent aims to address these pain points head-on and accelerate the way data pipelines are built and managed.

Meet your new AI-powered data engineering team

Imagine a team of expert data engineers, available 24/7, ready to jump in and tackle the toilsome pipeline development, maintenance, and troubleshooting tasks, enabling your data team to scale and focus on higher-value work. We are announcing the data engineering agent as experimental.  

Here are a few ways how BigQuery data engineering agent will change the game:

1. Autonomous pipeline building and modification

Do you need a new pipeline to ingest, transform, and validate data? Simply describe your needs in natural language – the agent handles the rest. For example:

“Create a pipeline to load data from the ‘customer_orders’ bucket, standardize the date formats, remove duplicate entries based on order ID, and load it into a BigQuery table named ‘clean_orders’.”

The agent, leveraging its understanding of data engineering best practices and your specific environment and context, generates the necessary SQL code, builds the pipeline, and even creates basic unit tests. It’s not just about automation; it’s about intelligent, context-aware automation.

Need to update an existing pipeline? Just tell the agent what you want to change. It analyzes the existing code, proposes modifications, and even highlights potential impacts on downstream processes. You remain in control, reviewing and approving changes, but the agent handles the heavy lifting.

2. Proactive troubleshooting and optimization

Pipeline issues? The agent monitors your pipelines, identifies issues such as schema and data drift, and proposes fixes. It’s like having a dedicated expert constantly watching over your data infrastructure.

3. Bulk draft pipelines

A powerful use of the data engineering agent is to scale pipeline generation or modification using previously acquired context and knowledge. This allows users to quickly scale pipelines for different departments or use cases, with customizations as needed, using the command line and API for automation at scale. In the example below, the agent takes instructions from the command line and leverages domain-specific agent instructions to create bulk pipelines.

How it works: Intelligence under the hood

To handle the complexity that most organizations have to deal with, the agents rely on several key concepts:

• Hierarchical context: The agents draw on multiple sources of knowledge:

• Universal understanding of common data formats, SQL best practices, etc.

• Vertical-specific knowledge of industry conventions (e.g., data formats in healthcare or finance)

• Organizational awareness of your company’s or department’s specific business context, data structures, naming conventions, and security policies

• Data pipeline-specific understanding the details of source and target schemas, transformations, and dependencies

A collaborative, multi-agent environment

BigQuery data engineering agent are a part of a multi-agent environment, where specialized agents collaborate to achieve complex goals, working together and delegating tasks, much like a real-world data engineering team:

• An ingestion agent expertly handles data intake from various sources.

• A transformation agent crafts efficient and reliable data pipelines.

• A validation agent helps ensures data quality and consistency.

• A troubleshooting agent proactively identifies and resolves issues.

• A data quality agent, powered by Dataplex metadata, monitors data and proactively alerts on anomalies.

Our initial focus is on ingestion, transformation and troubleshooting tasks, but we plan to expand these initial capabilities to other critical data engineering tasks. 

Your workflow, your way

Whether you prefer working in the BigQuery Studio UI, crafting code in your favorite IDE, or managing pipelines through the command line, we want to meet you where you are. We are initially making data engineering agent available in BigQuery Studio’s pipeline editor and API/CLI, but we plan to expose it in other contexts.

Data engineering agent and your data workers

The world is only beginning to see the full potential of AI-powered agents in revolutionizing how data workers interact with and derive value from their data. With BigQuery data engineering agent, the roles of data engineers, data analysts and data scientists are expanding beyond their traditional boundaries, empowering these teams to achieve more, faster, and with greater confidence. These agents act as intelligent collaborators, streamlining workflows, automating tedious tasks, and unlocking new levels of productivity. Initially we are focusing on core data engineering tasks of promoting data from Bronze to Silver in a data lake and expanding from there.

Coupled with products like Dataplex, BigQuery ML, and Vertex AI, BigQuery data engineering agent is poised to transform the way organizations manage, process, and derive value from their data. By automating complex tasks, promoting collaboration, and empowering data workers of all skill levels, these agents are paving the way for a new era of data-driven innovation.

Ready to get started?

This is just the beginning of our journey to build a truly intelligent, autonomous data platform. We’re committed to continuously expanding the capabilities of data engineering agent, making them even more powerful and intuitive partners for all your data needs.

BigQuery data engineering agent will be available soon. We’re excited to see how it fits into your data engineering workflows and help you unlock the full potential of your data. Show your interest in getting access here.

The unprecedented growth and unique challenges of AI applications are driving fundamental architectural changes to Google’s next-generation global network. 

The AI era brings an explosive surge in demand for network capacity, with novel traffic patterns characteristic of large-scale model training and inference. Simultaneously, the critical need for unwavering reliability has reached new heights; in an AI-driven world, outages are simply not an option. Furthermore, the requirement for enhanced security and fine-grained control, including data sovereignty considerations, is paramount. Finally, the operational cost and complexity associated with scaling traditional network architectures necessitate a more innovative approach, pushing us beyond basic automation towards true autonomy.

As we discussed in this blog, we are meeting these challenges head-on by building the next generation of Google’s global network upon four key architectural principles: (1) exponential scalability, (2) beyond-9s reliability, (3) intent-driven programmability, and (4) autonomous networking. 

In this blog, let’s peel back the layers and see how the underlying technology makes these four principles a reality.

Exponential scalability with a multi-shard network

We embrace elastic horizontal scaling as a core architectural principle for Google’s global network through our multi-shard network. Instead of one monolithic network, we’ve built multiple independent shards. This provides several benefits:

• Horizontal scaling: When more capacity is needed, we can scale up by growing a shard, and scale out by adding more shards, overcoming the limits and complexity of vertical scale. This is akin to adding more independent networks, rather than trying to make a single network bigger and bigger.

• Independent planes: The separation of control, data, and management planes within each shard significantly limits the impact radius of any potential issue. A software bug or operational error (such as an incorrect configuration push) in one shard is far less likely to impact others, enhancing the network’s overall stability.

In the AI era, the WAN is the new LAN and the continent is the data center. This horizontal scaling approach, inspired by the design of our massive data center fabrics, allows Google’s global network to handle the unprecedented bandwidth demands of today’s AI workloads. This multi-shard network has been a key enabler for us to accommodate the average 7X WAN traffic growth between 2020 and 2025, and more importantly, an order of magnitude growth in peak traffic due to the bursty nature of ML traffic over the same period. 

Beyond-9s reliability: Architecting for resilience

In a world of always-on services, reliability is paramount. Google’s global network incorporates several key innovations to achieve beyond-9s availability, emphasizing diversity and independence at every layer of the stack to avoid “shared fate” (cascading failures) and minimize impact during failures.

• Multi-shard isolation: Each network shard has independent data, control, and management planes. We control what can enter and leave these shards to a cluster or edge. This prevents a bad state from a cluster poisoning all the shards at the same time. The sharded architecture inherently provides a degree of isolation. Furthermore, we apply a multi-vendor paradigm when deploying our network shards, thanks to years of development of open API and models (discussed later) that allows us to operationalize any vendor platform under the same network function. This multi-vendor approach protects our network shards from vulnerabilities introduced by third-party software or hardware.

• Region isolation: With this approach, regional cores keep traffic within their domains, and regional gateways enforce policies for traffic that’s entering or leaving. This limits the impact of regional events, effectively shielding the rest of the network.

• Protective ReRoute: Google’s global network implements a unique transport technique for shortening user-visible outages that complements routing repair, and it marks a radical shift in how we think about network reliability. In the conventional network model, hosts send packets, and routers handle them. With Protective ReRoute, hosts actively shift traffic flows across network paths to improve reliability and performance, intelligently detecting network path anomalies and promptly, automatically rerouting traffic to a healthy, alternative path, which can be in the same or alternative shard. The host reroutes traffic in round-trip time scales, i.e., O(RTT), by changing a few bits in the packet header that are used to compute the hash function to select a specific path among many equally viable paths. This host-initiated re-routing protects customer traffic beyond what traditional routing and traffic engineering can achieve, and is independent of the type of network, scale of network, or type of failure, thereby providing robust and deterministic recovery and performance. With Protective ReRoute in our network, we have observed up to a 93% reduction in cumulative outage minutes.

For a conceptual overview of these scalability and resilience innovations, check out this video:

Also, be sure to check out this demo to see the combined value of our multi-shard network and Protective ReRoute in action. Here, we emulate a network shard failure and show how the host promptly detects a path failure and routes the traffic over an alternative path in a different, healthy shard, providing near-instant recovery.

Intent-driven programmability for fine-grained network controls

To cater to our customers’ diverse and evolving needs, network agility and fine-grained programmability is crucial. Google’s global network allows for network controls to be precisely tailored to specific business requirements, encompassing regulatory compliance, digital sovereignty mandates, and unique application performance needs, down to the most granular network attributes. This programmability is made possible by:

• Software-defined networking (SDN) controllers: Google’s global network is fully intent-driven, with SDN everywhere. We use SDN controllers to manage network behavior hierarchically. Orion, our hierarchical and federated SDN control plane platform, propagates top-level intent through layers of network control applications, which then react by updating their internal state and generating intermediate intent for each network switch. This hierarchical propagation results in changes to the programmed flow state in network switches.

• Universal network model: Our universal network model, Multi-Abstraction-Layer Topology representation, or MALT, allows us to specify generic intent and business policy. Our control and management planes can then use these representations to implement these policies coherently across the network.

• Standardized API: Because we rely on the OpenConfig software layer, we can use multiple routing vendors interchangeably, making the network more robust. With vendor diversity, a bug or an issue in one vendor’s software or hardware doesn’t impact the whole network, and we have options when scaling our network.

This programmability enables us to implement business policies directly into the network fabric, offering granularity and the ability to isolate bandwidth for critical applications. Customers with specific regulatory requirements can also leverage this programmability to enforce their desired network path controls for their data in motion. 

Autonomous networking for the network powering AI

The sheer scale and complexity of a global network of our scale demands a shift from traditional automation to a more intelligent, autonomous approach that requires minimal human intervention. This is especially critical to avoid the substantial increase in operational expenses that come with network growth, and to flatten the cost curves for network planning, design and operations. Below are some examples where we apply AI/ML techniques to help today. We see opportunities to expand into many more use cases:

• Network incident response with a Gemini and Vertex AI agentic framework: We are using an agentic AI approach to shorten outage times by identifying and mitigating failures faster, and to perform more effective root-cause analysis. This is helping us reduce the mean-time to detect and mean-time to resolve network issues.
• Demand forecasting and capacity planning: We are using AutoML for accurate demand forecasting, and employing graph optimization to optimize our network capacity planning.

• Reinforcement learning for routing optimization: We tune routing metrics for specific objectives, such as network performance, with reinforcement learning.

Autonomous networking has allowed us to slash failure mitigation times from hours to minutes, improving our network’s resilience and customer experience. Check out this demo to see an example of our autonomous network in action!

Putting it all together

Google’s next-generation global network represents a paradigm shift in network architecture designed to power the AI era, embracing horizontal scalability through multi-sharding, architecting for resilience at every layer with regional isolation and Protective ReRoute, enabling fine-grained programmability with SDN, and adopting autonomous network operation powered by AI/ML. This helps Google’s global network provide the scale, reliability, performance, and security that today’s mission-critical services and AI/ML applications demand. This transformation of Google’s software-defined global backbone not only meets the formidable challenges of the AI era, but empowers our customers to innovate and thrive in this new landscape. Our next-generation network is designed to be the invisible, yet indispensable, force driving the future of technology and connectivity.

This deep dive only scratches the surface, but hopefully, provides a glimpse into the innovative technologies that underpin Google’s global network. As we continue to navigate the exciting challenges and opportunities of the AI era, Google’s global network is the bedrock upon which we build and deliver transformative experiences for users and customers worldwide. Stay tuned for more updates as Google’s global network continues to evolve!

At Google Cloud Next 25, we announced incredible ways for enterprises to build multi-agent ecosystems with Vertex AI and Google Cloud Databases – including better ways for agents to communicate with each other using Agent2Agent Protocol and Model Context Protocol (MCP). With the growing excitement around MCP for developers, we’re making it easy for MCP Toolbox for Databases (formerly Gen AI Toolbox for Databases) to access your enterprise data in databases. This is another step forward in providing secure and standardized ways to innovate with agentic applications. Let’s take a look.

MCP Toolbox for Databases (formerly Gen AI Toolbox for Databases)

MCP Toolbox for Databases (Toolbox) is an open-source MCP (Model Context Protocol) server that allows developers to connect gen AI agents to enterprise data easily and securely. MCP is an emerging open standard created by Anthropic for connecting AI systems with data sources through a standardized protocol, replacing fragmented integrations that require custom integrations.

Currently, Toolbox can be used to build tools for a large number of databases: AlloyDB for PostgreSQL (including AlloyDB Omni), Spanner, Cloud SQL for PostgreSQL, Cloud SQL for MySQL, Cloud SQL for SQL Server, and self-managed MySQL and PostgreSQL. Because it’s fully open-source, it includes contributions from third-party databases such as Neo4j and Dgraph. Toolbox offers simplified development with reduced boilerplate code, enhanced security through OAuth2 and OIDC, and end-to-end observability with OpenTelemetry integration. This enables you to develop tools easier, faster, and more securely by handling the complexities such as connection pooling, authentication, and more.

As an MCP server, Toolbox provides the additional scaffolding for implementing production-quality database tools and making them accessible to any client in the growing MCP ecosystem. This compatibility allows developers building agentic applications to leverage Toolbox and securely query a wide range of databases through a single, standardized protocol, simplifying development and enhancing interoperability.

MCP Toolbox for Databases supports Agent Development Kit (ADK)

At Next, we launched the Agent Development Kit (ADK), an open-source framework that simplifies the process of building sophisticated multi-agent systems while maintaining precise control over agent behavior. With ADK, you can build an AI agent in under 100 lines of intuitive code. With ADK, you can:

• Shape how your agents think, reason, and collaborate through deterministic guardrails and orchestration controls. 
• Interact with your agents in human-like conversations with ADK’s unique bidirectional audio and video streaming capabilities enabled with just a few lines of code. Check out the demo of an interactive agent from the opening keynote at NEXT 2025 built on the ADK here.
• Choose the model or deployment that works best for your needs. ADK works with your stack of choice – whether that’s your preferred top-tier model, deployment target, or integration with remote agents built on other frameworks. ADK also supports the Model Context Protocol (MCP), enabling secure, two-way connections between your data sources and AI agents.
• Deploy to production using the direct integration to Vertex AI Agent Engine. This clear and reliable path from development to enterprise-grade deployment eliminates the typical overhead associated with moving agents into production.

To get started, go to Vertex AI Agent Garden to explore a curated set of agent samples for common use cases like data science and customer service agents. Discover tools that can be easily used to build agents with ADK such as connecting agents to databases with the integrated MCP Toolbox for Databases.  You can access source code in GitHub samples that you can clone and start using to develop your own agents.

Adding LangGraph support

LangGraph gives you essential built-in support for persistence layer, implemented through checkpointers. This helps you build resilient, stateful agents that can reliably manage long-running tasks or resume after interruptions. 

To leverage powerful managed databases for storing this state, Google Cloud offers dedicated integration libraries. Developers can choose the following:

• The highly scalable AlloyDB for PostgreSQL using the AlloyDBSaver class from the langchain-google-alloydb-pg-python library, or opt for

• Cloud SQL for PostgreSQL utilizing the corresponding checkpointer implementation, PostgresSaver, within the langchain-google-cloud-sql-pg-python library.

Both offer robust mechanisms to seamlessly save and load agent execution states, allowing workflows to be reliably paused, resumed, and audited, backed by the manageability and performance of Google Cloud’s PostgreSQL offerings.

When you compile graph with a checkpointer, the checkpointer saves a checkpoint of the graph state at every super-step. Those checkpoints are saved to a thread, which can be accessed after graph execution. Because threads allow access to graph’s state after execution, several powerful capabilities including human-in-the-loop, memory, time travel, and fault-tolerance are all possible. 

Install the packages:

Learn more about langgraph checkpoint usage for AlloyDB here and Cloud SQL PG here.

Get started 

This Colab demonstrates a complete workflow for building and deploying a LangGraph Hotel Agent which can search, book and cancel hotels. This sample shows how to build and deploy an agent (model, tools, and reasoning) using the Vertex AI SDK and MCP Toolbox for Databases.

The demonstration will begin with agent development, integrating the MCP Toolbox for Databases to Search, Book, and Cancel hotels. It will then walk you through deploying the agent to Agent Engine and the MCP Toolbox to Cloud Run, and conclude by demonstrating how to connect these services remotely.

Here are some more resources to get started with Toolbox and MCP.

• MCP Quick Start

• How to connect to Toolbox from MCP Client 

• MCP Introduction documentation 

• MCP github org

• MCP spec

Editor’s note: Ping Xie is a Valkey maintainer on the Valkey Technical Steering Committee (TSC).

Memorystore, Google Cloud’s fully managed in-memory service for Valkey, Redis and Memcached, plays an increasingly important role in our customers’ deployments — in fact, over 90% of the top 100 Google Cloud customers use Memorystore. Today, we’re excited that the Memorystore for Valkey service is now generally available, a significant step forward for open-source in-memory data management on the cloud. With the GA, you can now run your production workloads on Memorystore for Valkey backed by a 99.99% availability SLA along with features such as Private Service Connect, multi-VPC access, cross-region replication, persistence, and many more.

When we launched the preview of Memorystore for Valkey in August 2024, hundreds of Google Cloud customers like Major League Baseball (MLB) and Bandai Namco Studios Inc. jumped in and deployed the service. In the last few months, they’ve provided us with invaluable feedback that has shaped the service we’re announcing today:

“At Major League Baseball, our use of Memorystore has been a key part in optimizing how we bring data to our fans. We are excited about the general availability of Memorystore for Valkey, a truly open-source alternative. We believe its inherent flexibility and the power of community-driven development will further enhance our speed, scalability, and real-time data processing capabilities, allowing us to better serve our fans, players, and operations.” – Rob Engel, Vice President of Software Engineering, Major League Baseball

“Bandai Namco Studios uses Memorystore to power the low-latency and high-scale performance essential for many of our titles. We’re excited about the GA launch of Memorystore for Valkey. Its speed, features, and truly open-source nature will empower us to enhance real-time gameplay and scale for our global player base. We look forward to leveraging Memorystore for Valkey’s capabilities to continue pushing the boundaries of gaming innovation.” – Motoo Fukuda, Technical Director at Bandai Namco Studios Inc.

What’s new at GA

At GA, Memorystore for Valkey is backed by a 99.99% SLA powered by Google’s advanced high availability and zonal placement algorithms, and ships with a comprehensive suite of enterprise-grade features such as:

• Support for Private Service Connect: Memorystore for Valkey is built on top of Private Service Connect, which allows customers to connect to up to 250 shards using just two IP addresses. Memorystore’s discovery endpoint being highly available ensures no single point of failure for your cluster.

• Zero-downtime scaling: Memorystore for Valkey offers zero downtime scaling (in and out) so your cluster can grow with your application’s needs, and so it’s cost-optimized for your  workloads. It supports cluster sizes ranging from 1 to 250 nodes.

• Integrated Google-built vector similarity search: Memorystore for Valkey supports ultra-low latency, in-memory vector search, and can perform vector search at single-digit millisecond latency on over a billion vectors, with greater than 99% recall.

  This performance is powered by Google’s vector search module, the official search module for the Valkey OSS project, which is integrated into Memorystore for Valkey. The module enables modern AI applications for gen AI use cases such as retrieval-augmented generation (RAG), recommendation systems, and semantic search. With hybrid search support, users can achieve more accurate and contextually relevant search results, leading to improved application performance and a better user experience.

• Managed backups: Access to built-in managed backups enables both automated and on-demand backups for migrations, disaster recovery, and compliance.

• Cross-region replication (CRR): Using CRR, you can achieve disaster recovery prepared-ness and low-latency reads across regions. At this time, in addition to the primary region, we support up to two secondary regions with clusters that in turn can have varying numbers of replicas. Memorystore for Valkey ensures both the data plane and control plane remain in sync across regions.

• Multi-VPC access: Memorystore for Valkey supports multiple client-side VPCs to connect to one Private Service Connection endpoint on the Valkey cluster. Using this technology, you can securely connect clients across multiple projects and VPCs.

• Persistence: Memorystore for Valkey offers both RDB-snapshot and AOF-logging based persistence to meet varying data durability requirements.

Memorystore for Valkey supports both Valkey 7.2, and our engine of choice, Valkey 8.0, which offers many enhancements over its predecessors:

• Exceptional performance: With asynchronous I/O improvements, Memorystore for Valkey 8.0 delivers better throughput and achieves up to 2x Queries Per Second(QPS) of Memorystore for Redis Cluster at microseconds latency, helping applications handle demanding internet-scale workloads with ease.

  While priced in-line with Memorystore for Redis Cluster, Memorystore for Valkey’s performance optimizations can lead to substantial cost savings by potentially requiring fewer nodes to handle the same workload.

• Optimized memory efficiency: Valkey 8.0’s optimized memory management delivers improved memory savings, reducing operational costs across various workloads.

• Enhanced reliability: Valkey 8.0 offers significantly more reliable scaling with Google-contributed features like automatic failover for empty shards and highly available migration states. Additionally, we also introduced migration states auto-reparing to further strengthen system resilience.

In addition, Memorystore for Valkey also provides other capabilities, such as  maintenance windows, single zone clusters, single shard clusters, no-cost inter-zone replication, etc. 

Our commitment to open source and customer trust

Following licensing updates to Redis OSS by Redis Inc. in March 2024, the open-source community established Valkey OSS as an alternative that’s supported by organizations including Google, Amazon, Snap and others.

We deeply value the trust you place in us. To ensure you continue to have access to powerful, open technology, we launched Memorystore for Valkey on Google Cloud. Unlike Redis, the Valkey OSS project is under the BSD 3-clause license and backed by the Linux Foundation. The momentum behind Valkey has been exhilarating. 

In addition to Memorystore for Valkey, we are also committed to supporting and delivering new capabilities for Memorystore for Redis Cluster and Memorystore for Redis. And when Memorystore for Redis customers are ready to adopt Valkey — for its price-performance, reliability and open-source nature — we offer full migration support. Memorystore for Valkey is fully compatible with Redis OSS 7.2 APIs and your favorite clients, making it easy to switch to open source. Further, you can reuse your Memorystore for Redis and Memorystore for Redis cluster committed use discounts (CUDs), smoothing the transition.

Try Memorystore for Valkey today

The best way to experience the power of Memorystore for Valkey is to try it out. Get started with the documentation or deploy your first Valkey instance. Don’t let having to self-manage Redis hold you back. Experience the simplicity and speed of Memorystore for Valkey today and see how it can power your applications, so you can focus on what matters: innovating and creating impactful applications for your business!

Spring is a great reminder to spring clean – an annual tradition that should extend not only to your household, but also to your virtual cloud infrastructure. Why not start with Google Cloud’s FinOps Hub? 

As Google Cloud customers have adopted the FinOps hub to guide their optimization initiatives, we started getting additional feedback from our business community. For example, while DevOps users have access to tools and utilization metrics to identify waste, business teams often lack clear insights into resource consumption, leading to a significant blind spot. The most recent State of FinOps 2025 Report reinforces this need, underscoring the importance of workload optimization and waste reduction as the #1 Top FinOps concern. It’s extremely difficult to optimize workloads or applications if customers cannot fully understand how much is even being used. Why purchase a committed use discount for compute cores that you might not even be fully using? 

Sometimes the easiest optimizations our customers can make are really just using more efficiently the resources they are actually paying for. That’s why, in 2025, we are focused on the deep clean of your optimization opportunities and have upgraded FinOps Hub to help you find, highlight, and eliminate wasted spend.

1. Find waste: FinOps Hub 2.0 now comes with new utilization insights to zero in on optimization opportunities. 

At Google Cloud Next 2025, we introduced FinOps Hub 2.0, focused exclusively on bringing utilization insights on your resources to the forefront so you can see what potential waste may exist and take action immediately. Waste can come in many forms: from a VM that is barely getting used at 5% (overprovisioned), to a GKE cluster that is actually running hot at 110% utilization and might fail (underprovisioned), to managed resources like Cloud Run instances that may not be optimally configured (suboptimal configuration) or, worse yet, a VM that might not ever have been used (idle). FinOps users can now quickly view the most expensive waste category in one, easy-to-understand heatmap by service or AppHub application. But FinOps Hub doesn’t just show you where there may be waste; it also includes more cost optimizations for Kubernetes Engine (GKE), Compute Engine (GCE), Cloud Run, and Cloud SQL to remedy the waste too.

2. Highlight waste: Gemini Cloud Assist supercharges FinOps Hub to summarize optimization insights and send opportunities to engineering.

But perhaps what really makes this a 2.0 release is that we supercharged the most time-consuming tasks on FinOps Hub with Gemini Cloud Assist. Our first launch of Gemini Cloud Assist, which helps create personalized cost reports and synthesize insights, has resulted in >100k FinOps hours saved by our customers annually (from January 2024 to January 2025). The power of Gemini Cloud Assist to supercharge and automate workflows is a huge benefit, so we applied that to FinOps Hub in two ways. First, FinOps can now see embedded optimization insights on the hub itself –similar to cost reports – so you don’t need to solve the “needle in the haystack” problem of optimization. Second, you can now use Gemini Cloud Assist to summarize and send top waste insights to your engineering teams to take action and remediate fast.

3. Eliminate waste: introducing a NEW IAM role permission for your tech solution owners to see & directly take action on these optimization opportunities.

Finally, perhaps our most exciting feature – and long overdue for FinOps – is that we are unlocking access to the Billing console for tech solution owners, so that these owners can get FinOps insights and Gemini Cloud Assist insights across all their projects, in a single pane. For example, if you want to give access to FinOps Hub or cost reports to an entire department that only uses a subset of projects for their infrastructure – without providing them with broader billing data access, but still allowing them to see all of their data in a single view – now you can, with multi-project views in the billing console. Multi-project views are enabled using the new Project Billing Costs Manager IAM role (or related granular permissions). These new permissions are currently in private preview so sign-up to get access. Now you can truly extend the power of FinOps tools across your organization with these new access controls.  

So take this Spring to try FinOps Hub 2.0 with Gemini Cloud Assist, and do some spring cleaning on your cloud infrastructure, because as the saying goes, “With clouds overgrown, like winter’s old grime, Spring clean your servers, save dollars and time.” – well at least that’s what they say according to Gemini.

Driven by generative AI innovations, the Business Intelligence (BI) landscape is undergoing significant transformation, as businesses look to bring data insights to their organization in new and intuitive ways, lowering traditional barriers that have often kept discoveries out of the hands of the broader organization.

We’re spearheading this trend with Gemini in Looker, which builds upon Looker’s history as a cloud-first BI tool underpinned by a semantic layer that aligns data and that changes how users interact with it: with intelligent, AI-powered BI powered by Google’s latest AI models. The convergence of AI and BI stands to democratize data insights across organizations, moving beyond traditional methods to make data exploration more intuitive and accessible.

Gemini in Looker lowers technical barriers to accessing information, enhancing collaboration, and accelerating the process of turning raw data into actionable insights. As we announced at Google Cloud Next 25, we are expanding access to Gemini in Looker, making it now available to all Looker platform users. In this post, we discuss its key features, underlying architecture, and its transformative potential for both data analysts and business users.

Using AI to enhance productivity and efficiency

We designed Gemini in Looker with a clear objective: to improve productivity for analysts and business users with AI. Gemini in Looker makes it easier to prepare data and semantic models for BI, and simplifies building dashboard visualizations and reports. Additionally, Gemini in Looker can help business users’ efficiency by improving their data literacy and fluency, enabling them to tell data stories in their presentations, and use natural language to go beyond the dashboard to get answers to their questions. The result is analysts can do their jobs faster and business users can tell data stories and get answers.

Gemini in Looker does this through a suite of gen-AI-powered capabilities that make analytics tasks and workflows easier: 

• Looker Conversational Analytics allows users to ask questions about their data in natural language, gaining instant, highly visual answers powered by AI and grounded in Looker’s semantic model. Data exploration is now as simple as chatting with your team’s data expert.

• Automatic Slide Generation exports Looker reports to Google Slides, as well as AI-generated summaries of charts and their key insights, to automate creating presentations. With Automatic Slide Generation, presentations stay current and relevant, as the slides are directly connected to the underlying reports, so that the data they present is always up-to-date.

• Formula Assistant simplifies the creation of calculated fields for ad-hoc analysis by allowing analysts to describe the desired calculation in natural language. The formula is automatically generated using AI, saving time and effort for analysts and report builders.

• LookML Assistant simplifies LookML code creation by letting users describe what they are looking to build in natural language and automatically creating the corresponding LookML measures and dimensions. This helps streamline the process of creating and maintaining governed data.

• Advanced Visualization Assistant creates customized data visualizations that users describe with natural language, while. Gemini in Looker creates the necessary JSON code configurations.

The semantic layer: The foundation of AI accuracy

A critical component of Looker’s AI architecture is the LookML semantic modeling layer, which in conjunction with LLMs like Gemini, provides the necessary context for the LLM to comprehend the data, and helps ensure centralized metric definitions, preventing inconsistencies that can derail AI models. Without a semantic layer, AI answers may be inaccurate, leading to unreliable results, lack of adoption, and wasted effort. Looker’s semantic model enables data governance integration, maintaining compliance and trust with existing controls, and evolves with your business, iteratively updating data sets and measures so that AI answers are accurate. According to our own internal tests, Looker’s semantic layer reduces data errors in gen AI natural language queries by as much as two thirds.

How Google protects your data and privacy

You can use Gemini in Looker knowing that your data is protected. Gemini prioritizes data privacy, and does not store customer prompts and outputs without permission. Critically, customer data, including prompts and generated output, is never used to train Google’s generative AI models.

Looker’s agentic AI architecture powers intelligent BI

Announced at Next 25, the Looker Conversational Analytics API serves as the agentic backend for Looker AI. It answers questions using a reasoning agent that uses multiple tools to answer analytical questions. It also uses conversation history to answer multi-turn questions and enable more efficient Looker queries, including the ability to open them in the Explore UI.

Looker’s AI architecture is designed for accuracy and quality, taking a multi-pronged approach to gen AI quality:

1. Agentic reasoning

2. A semantic layer foundation

3. A dynamic knowledge graph that provides context for Retrieval Augmented Generation (RAG)

4. Fine-tuned models for SQL and Python generation

This robust architecture enables Looker to move beyond simply answering “What?” questions to addressing more complex queries like “How does this compare?” “Why?” “What will happen?” and ultimately, “What should we do?”

Looker’s AI and BI roadmap

With Looker, we’re committed to converging AI and BI, and are working on a number of new offerings including:

• Code Interpreter for Conversational Analytics makes advanced analytics easy, enabling business users to perform complex tasks like forecasting and anomaly detection using natural language, without needing in-depth Python expertise. You can learn more about this new capability and sign up here for the Preview.

• The Conversational Analytics API lets developers integrate Conversational Analytics across multiple experiences, including customer applications, chat apps, Agentspace, and BigQuery. Sign up here for preview access to the Conversational Analytics API.

• Centralize and share your Looker agents with Agentspace, which offers centralized access, faster deployment, enhanced team collaboration, and secure governance.

• Automated semantic model generation with Gemini helps democratize LookML creation, boost developer productivity, and unlock data insights with multi-modal inputs. Gemini leverages diverse input types like natural language descriptions, SQL queries, and database schemas.

Embracing BI’s AI-powered future

Gemini in Looker is a significant milestone in the AI/BI revolution. By integrating the power of Google’s Gemini models with Looker’s robust data modeling and analytics capabilities, organizations can empower their analysts, enhance the productivity of their business users, and unlock deeper, more actionable insights from their data. Gemini in Looker is transforming how we understand and leverage data to make smarter, more informed decisions. The journey from asking “What?” to confidently determining “What next?” is now within reach, powered by Gemini in Looker. Learn more at https://cloud.google.com/looker, or click here to learn more about Gemini in Looker and how to enable it for your Looker deployment. You can also choose to enable Trusted Tester features to gain access to early features in development.

Google Cloud Next 25 took place this week and we’re all still buzzing! It was a jam-packed week in Las Vegas complete with interactive experiences, including more than 10 keynotes and spotlights, 700 sessions, and 350+ sponsoring partners joining us for an incredible Expo show. Attendees enjoyed hands-on learning across AI innovation, data cloud, modern infrastructure, security, Google Workspace, and more. 

At our opening keynote, we showcased cutting-edge product innovations across our AI-optimized platform and featured hundreds of customers and partners building with Google Cloud as well as five awesome demos. You can catch up on all the highlights in our 10-minute keynote recap.

Our developer keynote showed how AI is revolutionizing the developer workflow, and featured seven incredible demos on everything from building with Gemini to creating multi-agent systems.

Last year, we shared how customers were exploring the exciting potential of generative AI to transform the way they work. This year, we showcased how customers are getting real business value from Google AI, celebrating hundreds of customer stories across the event, including the amazing story of how The Sphere is using Google AI to enrich their fully immersive The Wizard of Oz experience.

It was a busy week, so we’ve prepared a summary of all the 228 announcements from Next ‘25 below:

AI and Multi-Agent Systems

Models: Building on Google DeepMind research, we announced the addition of a variety of first-party models, as well as new third-party models to Vertex AI Model Garden. 

1. Gemini 2.5 Pro is available in public preview on Vertex AI, AI Studio, and in the Gemini app. Gemini 2.5 Pro is engineered for maximum quality and tackling the most complex tasks demanding deep reasoning and coding expertise. It is ranked #1 on Chatbot Arena. 

2. Gemini 2.5 Flash — our low latency and most cost-efficient thinking model — is coming soon to Vertex AI, AI Studio, and in the Gemini app.

3. Imagen 3: Our highest quality text-to-image model now has improved image generation and inpainting capabilities for reconstructing missing or damaged portions of an image.

4. Chirp 3: Our audio generation and understanding model now includes Instant Custom Voice, a new way to create custom voices with just 10 seconds of audio input. 

5. Lyria: The industry’s first enterprise-ready, text-to-music model, transforms simple text prompts into 30-second music clips. 

6. Veo 2: Our advanced video generation model has new editing and camera control features to help customers refine and repurpose video content with precision. 

(Check out our fun keynote countdown clock made with Veo 2).

7. Meta’s Llama 4 is generally available on Vertex AI. 

8. Ai2 announced a partnership with Google Cloud to make its portfolio of open models available in Vertex AI Model Garden.

Together, this makes Vertex AI the only hyperscaler platform with generative media models across video, image, speech, and music. 

To build on your AI agent initiatives, we announced new advancements in Vertex AI to help you better  tune, deploy and manage the right model:

9. Vertex AI Dashboards: These help you monitor usage, throughput, latency, and troubleshoot errors, providing you with greater visibility and control.

10. Model Customization and Tuning: You can also manage custom training and tuning with your own data on top of foundational models in a secure manner across all first-party model families including Gemini, Imagen, Veo, embedding, and translation models, as well as open models like Gemma, Llama, and Mistral.

11. Vertex AI Model Optimizer: Automatically generate the highest quality response for each prompt based on your desired balance of quality and cost

12. 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.

13. Vertex AI Global Endpoint: Provides capacity-aware routing for our Gemini models across multiple regions, maintaining application responsiveness even during peak traffic or regional service fluctuations.

We also introduced new capabilities to help you build and manage multi-agent systems — regardless of which technology framework or model  you’ve chosen.

14. Agent Development Kit (ADK): This open-source framework simplifies the process of building sophisticated multi-agent systems while maintaining precise control over agent behavior. Agent Development Kit supports the Model Context Protocol (MCP) which provides a unified way for AI models to access and interact with various data sources and tools, rather than requiring custom integrations for each.

15. Agent2Agent (A2A) protocol: We’re proud to be the first hyperscaler to create an open Agent2Agent protocol to help enterprises support multi-agent ecosystems, so agents can communicate with each other, regardless of the underlying framework or model. More than 50 partners, including Accenture, Box, Deloitte, Salesforce, SAP, ServiceNow, and TCS are actively contributing to defining this protocol, representing a shared vision of multi-agent systems.

16. Agent Garden: This collection of ready-to-use samples and tools is directly accessible in ADK. Leverage pre-built agent patterns and components to accelerate your development process and learn from working examples.

17. Agent Engine: This fully managed agent runtime in Vertex AI helps you deploy your custom agents to production with built-in testing, release, and reliability at a global, secure scale.

18. Grounding with Google Maps¹: For agents that rely on geospatial context, you can now ground your agents with Google Maps, so they can provide responses with geospatial information tied to places in the U.S. 

19. Customer Engagement Suite: This latest version includes human-like voices; the ability to understand emotions so agents can adapt better during conversation; streaming video support so AI agents can interpret and respond to what they see in real-time through customer devices; and AI assistance to build agents in a no-code interface. 

We announced exciting enhancements to Google Agentspace to help scale the adoption of enterprise search and AI agents across the enterprise. Agentspace puts the latest Google foundation models, Google-quality search, powerful AI agents, and actionable enterprise knowledge in the hands of every employee. 

20. Integrated with Chrome Enterprise: Bringing Agentspace directly into Chrome helps employees easily and securely find information, including data and resources, right within their existing workflows. 

21. Agent Gallery: This provides employees a single view of available agents across the enterprise, including those from Google, internal teams, and partners — making agents easy to discover and use. 

22. Agent Designer: A no-code interface for creating custom agents that automate everyday work tasks or enhance knowledge. Agent Designer helps employees adapt agents to their individual workflows and needs, no matter their technical experience. 

23. Idea Generation agent: Helps employees innovate by autonomously developing novel ideas in any domain, then evaluating them to find the best solutions via a competitive system inspired by the scientific method.

24. Deep Research agent: Explores complex topics on the employee’s behalf, synthesizing information across internal and external sources into comprehensive, easy-to-read reports — all with a single prompt. 

We brought the best of Google DeepMind and Google Research together with new infrastructure and AI capabilities in Google Cloud, including:

25. AlphaFold 3: Developed by Google DeepMind and Isomorphic Labs, the new AlphaFold 3 High-Throughput Solution, available for non-commercial use and deployable via Google Cloud Cluster Toolkit, enables efficient batch processing of up to tens of thousands of protein sequences while minimizing cost through autoscaling infrastructure. 

26. WeatherNext AI models: Google DeepMind and Google Research WeatherNext models enable fast, accurate weather forecasting, and are now available in Vertex AI Model Garden, allowing organizations to customize and deploy them for various research and industry applications. 

AI infrastructure

We’re pleased to introduce new innovations throughout the AI Hypercomputer stack that collectively deliver more intelligence per dollar: 

27. Ironwood: Our 7th generation TPU joins our AI-optimized hardware portfolio to power thinking, inferential AI models at scale (coming later in 2025). Read more here.

28. Google Distributed Cloud (GDC): We have partnered with NVIDIA to bring Gemini to NVIDIA Blackwell systems, with Dell as a key partner, so Gemini can be used locally in air-gapped and connected environments. Read more here. 

29. Pathways on Cloud: Developed by Google DeepMind, Pathways is a distributed runtime that powers all of AI at Google, and is now available for the first time on Google Cloud. 

30. vLLM on TPU: We’re bringing vLLM to TPUs to make it easy to run inference on TPUs. Customers who have optimized PyTorch with vLLM can how run inference on TPUs without changing their software stack, and also serve on both TPUs and GPUs if needed.

31. Dynamic Workload Scheduler resource management and job scheduling platform now features support for Trillium, TPU v5e, A4 (NVIDIA B200), and A3 Ultra (NVIDIA H200) VMs in preview via Flex Start mode, with Calendar mode support for TPUs coming later this month. 

32. A4 and A4X VMs: We’ve significantly enhanced our GPU portfolio with the availability of A4 and A4X VMs powered by NVIDIA’s B200 and GB200 Blackwell GPUs, respectively, and A4X VMs are now in preview. We were the first cloud provider to offer both of these options. 

33. NVIDIA Vera Rubin GPUs: Google Cloud will be among the first to offer NVIDIA’s next-generation Vera Rubin GPUs, which offer up to 15 exaflops of FP4 inference performance per rack. 

34. Cluster Director (formerly Hypercompute Cluster) lets you deploy and manage a group of accelerators as a single unit with physically colocated VMs, targeted workload placement, advanced cluster maintenance controls, and topology-aware scheduling. New updates coming later this year include Cluster Director for Slurm, 3600 observability features, and job continuity capabilities. Register to join the preview.

Application Development

Developing on top of Google Cloud, and with Google Cloud tools, gets better every day. 

35. The new Application Design Center, now in preview, provides a visual, canvas-style approach to designing and modifying application templates, and lets you configure application templates for deployment, view infrastructure as code in-line, and collaborate with teammates on designs.

36. The new Cloud Hub service, in preview, is the central command center for your entire application landscape, providing insights into deployments, health and troubleshooting, resource optimization, maintenance, quotas and reservations, and support cases. Try Cloud Hub here.

37. App Hub, which models your applications as interconnected services and workloads for an application-centric view, is now integrated with over 20 Google Cloud products.

38. Application Monitoring, in public preview, supports automatically tagging telemetry (logs, metrics, and traces) with application context, application-aware alerts, and out-of-the-box application dashboards.

39. Cost Explorer, in private preview, provides visibility into granular application costs and utilization metrics, allowing you to identify efficiency opportunities; sign up here to try it out.

40. Gemini Code Assist agents can help with common developer tasks such as code migration, new feature implementation, code review, test generation, model testing, and documentation, and their progress can be tracked on the new Gemini Code Assist Kanban board. 

41. Gemini Code Assist is now available in Android Studio for professional developers who want AI coding assistance with enterprise security and privacy features.

42. Gemini Code Assist tools, now in preview, helps you access information from Google apps and tools from partners including Atlassian, Sentry, Snyk, and more. 

43. An App Prototyping agent in preview for Gemini Code Assist within the new Firebase Studio development environment turns your app ideas into fully functional prototypes, including the UI, backend code, and AI flows.

44. Gemini Cloud Assist is integrated with Application Design Center in preview to accelerate application infrastructure design and deployment.

45. Gemini Cloud Assist Investigations leverages data in your cloud environment to accelerate troubleshooting and issue resolution. Register for the private preview here.

46. Gemini Cloud Assist is now integrated across Google Cloud services including Storage Insights, Cloud Observability, Firebase, Database Center, Flow Analyzer, FinOps Hub, as well as security- and compliance-related services.

47. FinOps Hub 2.0 now includes waste insights and cost optimization opportunities from Gemini Cloud Assist.

48. The new Enterprise tier of the Google Developer Program is in limited preview, providing a safe and affordable way to explore Google Cloud and its AI products for a set monthly cost of $75/month per seat. Learn more here.

Compute 

Whatever your workload, there’s a Compute Engine virtual machine to help you run it at the price, performance and reliability levels you need. 

49. New C4D VMs built on AMD’s 5th Gen EPYC processors and paired with Google Titanium deliver impressive performance gains over prior generations— up to 30% vs C3D on the estimated SPECrate®2017_int_base benchmark. Currently in preview, try out C4D today.

50. C4 VMs built on the 6th generation Intel Granite Rapids CPUs feature the highest frequency of any Compute Engine VM — up to 4.2 GHz. 

51. C4 shapes with Titanium Local SSD offer improved performance for I/O-intensive workloads like databases and caching layers, achieving Local SSD latency reductions of up to 35%. 

52. C4 bare metal instances provide performance gains of up to 35% for general compute and up to 65% for ML recommendation workloads compared to the prior generation. 

53. New, larger C4 VM shapes scale up to 288 vCPU, with 2.2TB of high-performing DDR5 memory and larger cache sizes. Request preview access here.

Compute Engine also features a variety of specialized VM families and unique capabilities:

54. New H4D VMs for demanding HPC workloads are built on the 5th gen AMD EPYC CPUs, and offer the highest whole-node VM performance of more than 12,000 flops, the highest per-core performance, and the best memory bandwidth of more than 950 GB/s of our VM families. Sign up for the H4D preview.

55. M4 VMs are certified for business-critical, in-memory SAP HANA workloads ranging from 744GB to 3TB, and for SAP NetWeaver Application Server, and offer up to 65% better price-performance and 2.25x more SAP Application Performance Standard (SAPS) compared to the previous memory-optimized M3.

56. The Z3 storage-optimized family now features new Titanium SSDs and offers nine new smaller shapes, ranging from 3TB to 18TB per instance. The Z3 family also introducing new storage-optimized bare-metal instance which include up to 72TB of Titanium SSDs and direct access to the physical server CPUs. Now in preview, register your interest here. 

57. Nutanix Cloud Clusters (NC2) on Google Cloud let you run, manage, and operate apps, data, and AI across private and public clouds. Sign up for the public preview here. 

58. Google Cloud VMware Engine now comes in 18 additional node shapes, bringing the total number of node shapes across VMware Engine v1 and v2 to 26. 

59. Within the Titanium family, Titanium ML Adapter securely integrates NVIDIA ConnectX-7 network interface cards (NICs), providing 3.2 Tbps of non-blocking GPU-to-GPU bandwidth. 

60. Titanium offload processors now integrate our GPU clusters with the Jupiter data center fabric, for greater cluster scale. 

61. You can now manage managed instance groups (MIGs) as a single entity.

62. MIGs now support committed use discounts (CUDs) and reservation sharing with Vertex AI and Autopilot.

Containers & Kubernetes

The case for running on Google Kubernetes Engine (GKE) keeps on getting stronger, across an ever expanding class of workloads, most recently — AI.

63. GKE Inference Gateway offers intelligent scaling and load-balancing capabilities, helping you handle request scheduling and routing with gen AI model-aware scaling and load-balancing techniques. 

64. With GKE Inference Quickstart, you can choose an AI model and your desired performance, and GKE configures the right infrastructure, accelerators, and Kubernetes resources to match. 

65. In partnership with Anyscale, we announced RayTurbo on GKE, an optimized version of open-source Ray that delivers 4.5x faster data processing and requires 50% fewer nodes for serving. 

66. Cluster Director for GKE (formerly Hypercompute Cluster) is now generally available, letting you deploy and manage large clusters of accelerated VMs with compute, storage, and networking — all operating as a single unit. 

67. We announced performance improvements to GKE Autopilot, including faster pod scheduling, scaling reaction time, and capacity right-sizing.

68. Starting in Q3, Autopilot’s container-optimized compute platform will also be available to standard GKE clusters, without requiring a specific cluster configuration.

Customers

We shared hundreds of new customer stories across every industry and region, highlighting the ways they’re using Google Cloud to drive real impact. Here are some highlights:

69. Agoda, one of the world’s largest digital travel platforms, creates unique visuals and videos of travel destinations with Imagen and Veo on Vertex AI.

70. Bayer built an agent that uses predictive AI and advanced analytics to predict flu trends.

71. Bending Spoons integrated Imagen 3 into its Remini app to launch a popular new AI filter, processing an astounding 60 million photos per day.

72. Bloomberg Connects is using Gemini to explore new ways to help museums and other cultural institutions make their digital content accessible to more visitors.

73. Citi is using Vertex AI to rapidly deploy generative AI-powered productivity tools to more than 150,000 employees. 

74. DBS, a leading Asian financial services group, is using Customer Engagement Suite to reduce customer call handling times by 20%. 

75. Deutsche Bank built DB Lumina, a new Gemini-powered tool that can synthesize financial data and research, turning, for example, a report that’s hundreds of pages into a one-page brief, delivering it in a matter of seconds to traders and wealth managers.

76. Deutsche Telekom has announced an expanded strategic partnership with Google Cloud, focusing on cloud and AI integration to modernize Deutsche Telekom’s IT, networks, and business applications, including migrating its SAP landscape.

77. Dun & Bradstreet is using Security Command Center to centralize monitoring of AI security threats. 

78. Fanatics is partnering with Google Cloud to use AI technology to enhance every aspect of the fan journey. With Vertex AI Search for Commerce, Fanatics has developed an intelligent search ecosystem that understands and anticipates fan preferences, improves quality assurance and delivers intelligent customer service, and more.

79. Freshfields is using Gemini for Google Workspace and Google Cloud’s Vertex AI to enhance client services, including powering Freshfields’ Dynamic Due Diligence solution. 

80. Globo, Latin America’s largest media company, used Vertex AI Search to create a recommendations experience inside its streaming platform that more than doubled their click-through-play rate on videos.

81. Gordon Food Services is simplifying insight discovery and recommending next steps with Agentspace.

82. The Home Depot built Magic Apron, an agent that offers expert guidance 24/7, providing detailed how-to instructions, product recommendations, and review summaries to make home improvement easier.

83. Honeywell has incorporated Gemini into its product development. 

84. KPMG is building Google AI into in its newly formed KPMG Law firm and implementing Agentspace to enhance its own workplace operations.

85. L’Oreal is using Gemini, Imagen and Veo to accelerate creative ideation and production for marketing and product design, significantly speeding up workflows while maintaining ethical standards. 

86. Lloyds Banking Group has taken a significant step in its strategic transformation by migrating its major platforms to Google Cloud. The transition is unlocking new opportunities to innovate with AI, enhancing the customer experience. 

87. Lowe’s is revolutionizing product discovery with Vertex AI Search to generate dynamic product recommendations and address customers’ complex search queries.

88. Nevada Department of Employment, Training and Rehabilitation developed an Appeals AI Assistant powered by BigQuery and Vertex AI that synthesizes case data to help Appeals Referees make fair benefit approvals 4x faster.

89. Nokia built a coding tool to speed up app development with Gemini, enabling developers to create 5G applications faster. 

90. Nuro, an autonomous driving company, uses vector search in AlloyDB to identify challenging scenarios on the road.

91. Mercado Libre deployed Vertex AI Search across 150M items in 3 pilot countries that is helping their 100M customers find the products they love faster, already delivering millions of dollars in incremental revenue.

92. Papa Johns is using AI to transform the ordering and delivery experience for its global customers. With Google Cloud’s AI, data analytics, and machine learning capabilities, Papa Johns can anticipate customer needs and personalize their pizza experience, as well as provide a consistent customer experience both inside the restaurants and online. 

93. Reddit is using Gemini on Vertex AI to power “Reddit Answers,” Reddit’s AI-powered conversation platform. Additionally, Reddit is using Enterprise Search to improve its homepage experience.

94. Samsung is integrating Gemini on Google Cloud into Ballie, its newest AI home companion robot, enabling more personalized and intelligent interactions for users.

95. Seattle Children’s hospital is launching Pathway Assistant, a gen AI-powered agent with Gemini that improves clinicians’ access to complex information and the latest evidence-based best practices needed to treat patients. 

96. Government of Singapore uses Google Cloud Web Risk to protect their residents online.

97. The Wizard of Oz at The Sphere is an immersive experience that reconceptualizes the 1939 film classic through the magic of AI, bringing it to life on a whole new scale for the colossal 160,000-square-foot domed screen at The Sphere in Las Vegas. It’s a collaboration between Sphere Entertainment, Google DeepMind, Google Cloud, Hollywood production company Magnopus, and five others.

98. Spotify uses BigQuery to harness enormous amounts of data to deliver personalized experiences to over 675 million users worldwide.

99. Intuit is using Google Cloud’s Document AI and Gemini models to simplify tax preparation for millions of TurboTax consumers this tax season, ultimately saving time and reducing errors. 

100. United Wholesale Mortgage is using Google Cloud’s gen AI and data analytics to improve the mortgage process for 50,000 mortgage brokers and their clients, focusing on speed, efficiency, and personalized service. 

101. Verizon is using Google Cloud’s Customer Engagement Suite to enhance its customer service for more than 115 million connections with AI-powered tools, like the Personal Research Assistant. 

102. Vodafone used Vertex AI along with open-source tools and Google Cloud’s security foundation to establish an AI security governance layer. 

103. Wayfair updates product attributes 5x faster with Vertex AI.

104. WPP built Open as a platform powered by Google models that all of its employees worldwide can use to concept, produce, and measure campaigns.

For more customer stories, check out our latest blog with 601 real AI use cases. 

Databases

105. You can now search structured data in AlloyDB with Google Agentspace.

106. The next-generation of AlloyDB natural language lets you query structured data in AlloyDB securely and accurately, enabling natural language text modality in apps.

107. In AlloyDB, we’ve optimized SQL functionality spanning vector search and structured filters and joins, enabling faster vector search using the Scalable Nearest Neighbor (ScaNN) for AlloyDB index. 

108. AlloyDB AI includes three new AI models: one that improves the relevance of vector search results using cross attention reranking; a multimodal embeddings model that supports text, images, and videos, and a new Gemini Embedding text model. 

109. The new AlloyDB AI query engine lets developers use natural language expressions and constructs within SQL queries. Sign up for the preview of these AlloyDB features here. 

110. MCP Toolbox for Databases (formerly Gen AI Toolbox for Databases) now supports Model Context Protocol (MCP) enabling seamless connections between AI agents and enterprise databases.

111. Firestore with MongoDB compatibility, in preview, lets developers take advantage of MongoDB’s API portability along with Firestore’s multi-region replication with strong consistency, virtually unlimited scalability, a 99.999% SLA, and single-digit milliseconds read latency. Get started here today. 

112. The new Oracle Base Database Service offers a flexible and controllable way to run Oracle Databases in the cloud.

113. Oracle Exadata X11M is now GA, bringing the Oracle Exadata platform to Google Cloud and adding additional enterprise-ready capabilities, including customer managed encryption keys (CMEK). 

114. Database Migration Service (DMS) now supports SQL Server to PostgreSQL migrations for Cloud SQL and AlloyDB, allowing you to fully execute on your database modernization strategy.

115. Cloud SQL and AlloyDB are available on C4A instances, our Arm-based Google Axion Processors delivering higher price-performace and throughput. Learn more here. 

116. Database Center is now generally available and supports every database in our portfolio, providing a unified, AI-powered fleet management solution.

117. Spanner vector search is now generally available, designed to work with our SQL, Graph, Key-Value, and Full-Text Search modalities.

118. Graph Visualization for Spanner is now generally available, allowing users to visually explore valuable information from graph data.

119. Repeatable Read isolation (in preview), improved performance, and low-touch migration tooling to simplify moving workloads from MySQL to Spanner.

120. Aiven for AlloyDB Omni, a fully-managed AlloyDB Omni service from our partner Aiven that runs on AWS, Azure, and Google Cloud, is now generally available.

121. Bigtable continuous materialized views, in preview, simplifies real-time updates for applications that rely on immediate reporting and insights.

122. New Cassandra-compatible APIs and live-migration tooling for zero-downtime migrations from Cassandra to Bigtable and Spanner.

123. Memorystore for Valkey is now generally available, with support for 7.2 and 8.0 engine versions.

124. Firebase Data Connect is now GA, offering the reliability of Cloud SQL for PostgreSQL with instant GraphQL APIs and type-safe SDKs

Data analytics

We announced several new innovations with our autonomous data to AI platform powered by BigQuery, alongside our unified, trusted, and conversational Looker BI platform:

125. BigQuery pipelines, now GA, supports building data pipelines.

126. BigQuery data preparation, now GA, transforms and enriches data.

127. BigQuery anomaly detection, now in preview, maintains data quality and automates metadata generation.

128. Data science agent, now GA, is embedded within Google’s Colab notebook, provides intelligent model selection, enabling scalable training, and faster iteration. 

129. Looker conversational analytics, in preview, lets business users interact with data using natural language. 

130. The Looker Conversational Analytics API, now in preview, lets developers build and embed conversational analytics into applications and workflows. Sign up to gain access.

131. BigQuery knowledge engine, in preview, leverages Gemini to analyze schema relationships, table descriptions, and query histories to generate metadata on the fly, model data relationships, and recommend business glossary terms. 

132. BigQuery semantic search, is now GA, providing AI-powered data insights and across BigQuery, grounding AI and agents in business context. 

133. BigQuery’s contribution analysis feature, now GA, helps you pinpoint the key factors (or combinations of factors) responsible for the most significant changes in a metric. 

134. We added Gemini in BigQuery features into existing BigQuery pricing models across all BigQuery compute pricing options. Express your interest in trying out the new features.

135. BigQuery pipe syntax is GA, letting you apply operators in any order and as often as you need, and is compatible with most standard SQL operators.

Then, for data science and analyst teams, we added AI-driven data science and workflows as part of BigQuery notebook:

136. New intelligent SQL cells understand your data’s context and provide smart suggestions as you write code, and let you join data sources directly within your notebook. 

137. Native exploratory analysis and visualization capabilities in BigQuery make it easy to explore data, as well as add features to enable easier collaboration with colleagues. Data scientists can also schedule analyses to run and refresh insights periodically. 

138. The new BigQuery AI query engine lets data scientists process structured and unstructured data together with added real-world context, co-processing traditional SQL alongside Gemini to inject runtime access to real-world knowledge, linguistic understanding, and reasoning abilities.

139. Google Cloud for Apache Kafka, now GA, facilitates real-time data pipelines for event sourcing, model scoring, messaging and real-time analytics.

140. Apache Spark workloads within BigQuery, now in preview, execute serverlessly, that is, on a fully managed platform. 

141. New dataset-level insights in BigQuery data canvas, in preview, surface hidden relationships between tables and generate cross-table queries by integrating query usage analysis and metadata.

142. BigQuery ML includes the new AI.GENERATE_TABLE in preview to capture the output of LLM inference within SQL clauses. 

143. BigQuery ML now supports open-source and Anthropic’s Claude, Llama, and Mistral models, hosted on Vertex AI.

144. BigQuery vector search includes a new index type, now GA, based on Google’s ScaNN model that’s coupled with a CPU-optimized distance computation algorithm for scalable, faster and more cost-efficient processing. 

145. The preview of BigQuery ML’s pre-trained TimesFM model developed by Google Research simplifies time-series forecasting. 

146. We integrated new Google Maps Platform datasets directly into BigQuery, to make it easier for data analysts and decision makers to access insights. 

147. In addition, Earth Engine in BigQuery brings the best of Earth Engine’s geospatial raster data analytics directly into BigQuery. Learn more here.

148. GrowthLoop introduced its Compound Marketing Engine built on BigQuery with Growth Agents powered by Gemini, so marketing can build personalized audiences and journeys that drive rapidly compounding growth. 

149. Informatica expanded its services on Google Cloud to enable sophisticated analytical and AI governance use cases.

150. Fivetran introduced its Managed Data Lake Service for Cloud Storage with native integration with BigQuery metastore and automatic data conversion to open table formats like Apache Iceberg and Delta Lake

151. DBT is now integrated with BigQuery DataFrames and DBT Cloud is now on Google Cloud. 

152. Datadog introduced expanded monitoring capabilities for BigQuery, providing granular visibility into query performance, usage attribution, and data quality metrics. 

BigQuery’s autonomous data foundation provides governance, orchestration for diverse data workloads, and a commitment to flexibility via open formats. Announcements in this area include:

153. BigQuery makes unstructured data a first-class citizen with multimodal tables in preview, bringing rich, complex data types alongside structured data for unified storage and querying via the new ObjectRef data type. 

154. BigQuery governance in preview provides a single, unified view for data stewards and professionals to handle discovery, classification, curation, quality, usage, and sharing.

155. The new BigQuery universal catalog brings together a data catalog (formerly known as Dataplex Catalog) and a fully managed, serverless metastore, now generally available. 

156. BigQuery metastore, now GA, enable engine interoperability across BigQuery, Apache Spark, and Apache Flink engines, with support for the Iceberg Catalog. 

157. BigQuery business glossary, now GA, lets you define and administer company terms, identify data stewards for these terms, and attach them to data asset fields.

158. BigQuery continuous queries, now GA, enable instant analysis and actions on streaming data using SQL, regardless of its original format.

159. BigQuery tables for Apache Iceberg in preview, lets you connect your Iceberg data to SQL, Spark, AI and third-party engines. 

160. New advanced workload management capabilities, now GA, scale resources, manage workloads, and help ensure their cost-effectiveness. 

161. BigQuery spend commit, now GA, simplifies purchasing, unifying spend across BigQuery data processing engines, streaming, governance, and more.

162. BigQuery DataFrames now has AI code assist capabilities in preview, letting you use natural language prompts to generate or suggest code in SQL or Python, or to explain an existing SQL query. 

163. SQL translation assistance, now GA, is an AI-based translator that lets you create Gemini-enhanced rules to customize your SQL translations, to accelerate BigQuery migrations. 

164. Catalog metadata export, GA, enables bulk extract of catalog entries into Cloud Storage.

165. BigQuery can now perform automatic at-scale cataloging of BigLake and object tables, now GA. 

166. BigQuery managed disaster recovery is now GA, featuring automatic failover coordination, continuous near-real-time data replication to a secondary region, and fast, transparent recovery during outages. 

167. New workload management capabilities in preview include reservation-level fair sharing of slots, predictability in performance of reservations, and enhanced observability through reservation attribution in billing.

Looker, is adding a host of new conversational and visual capabilities, aimed at making BI accessible and useful to all users, accelerated by AI.

168. Gemini in Looker features are now available to all Looker platform users, including Conversational Analytics, Visualization Assistant, Formula Assistant, Automated Slide Generation, and LookML Code Assistant.

169. Code Interpreter for Conversational Analytics is in preview, allowing business users to perform forecasting and anomaly detection using natural language without needing deep Python expertise. Learn more and sign up for it here.

170. New Looker reports feature an intuitive drag-and-drop interface, granular design controls, a rich library of visualizations and templates, and real-time collaboration capabilities, now in the core Looker platform.

171. With Google Cloud’s acquisition of Spectacles.dev, developers can automate testing and validation of SQL and LookML changes using CI/CD practices.

Firebase

172. The new Firebase Studio, available to everyone in preview, is a cloud-based, agentic development environment powered by Gemini that includes everything developers need to create and publish production-quality full-stack AI apps quickly, all in one place. Gemini Code Assist agents are available via private preview.

173. Genkit, an open-souce framework for building AI-powered applications, using your preferred language, now has early support for Python and expanded support for Go. Try this template in Firebase Studio to build with Genkit.

174. Vertex AI in Firebase now includes support for the Live API for Gemini models, enabling more conversational interactions in apps such as allowing customers to ask audio questions and get responses.

175. Firebase Data Connect is now GA, offering the reliability of Cloud SQL for PostgreSQL with instant GraphQL APIs and type-safe SDKs.

176. Firebase App Hosting is also GA, providing an opinionated, git-centric hosting solution for modern, full-stack web apps. 

177. A new App Testing agent within Firebase App Distribution, also in preview, prepares mobile apps for production by generating, managing, and executing end-to-end tests. 

Google Cloud Consulting

Google Cloud Consulting introduced several new pre-packaged service offerings: 

178. Agentspace Accelerator provides a structured approach to connecting and deploying AI-powered search within organizations, so employees can easily gain access to relevant internal information and resources when they need it. 

179. Optimize with TPUs helps customers migrate workloads to our purpose-built AI chips, TPUs. 

180. Oracle on Google Cloud lets customers combine Oracle databases and applications with Google Cloud’s advanced platform and AI capabilities for enhanced database and network performance.

181. We expanded access to Delivery Navigator, a series of proven delivery methodologies and best practices to help with migrations and technology implementations to customers as well as partners, in preview. 

Networking

Google’s global network is the backbone of Google Cloud’s services, and relies on a long list of innovations.

182. Cloud WAN, a Cross-Cloud Network solution, is a fully managed, reliable, and secure enterprise backbone that makes Google’s global private network available to all Google Cloud customers. Cloud WAN delivers up to 40% improved network performance, while reducing total cost of ownership by up to 40%. Read more here. 

183. The new 400G Cloud Interconnect and Cross-Cloud Interconnect, available later this year, offers up to 4x more bandwidth than our 100G Cloud Interconnect and Cross-Cloud Interconnect, providing connectivity from on-premises or other cloud environments to Google Cloud. 

184. Build massive AI services with networking support for up to 30,000 GPUs per cluster in a non-blocking configuration, available in preview now.

185. Zero-Trust RDMA security helps you secure your high-performance GPU and TPU traffic with our RDMA firewall, featuring dynamic enforcement policies. Available later this year.

186. Get accelerated GPU-to-GPU communication, with up to 3.2Tbps of non-blocking GPU-to-GPU bandwidth with our high-throughput, low-latency RDMA networking, now generally available.

187. Leverage GKE Inference Gateway and Cloud Load Balancing alongside Model Armor, NVIDIA NeMo Guardrails, and Palo Alto Networks AI Runtime Security, all using Service Extensions.

188. Cloud Load Balancing has optimizations for LLM inference, letting you leverage NVIDIA GPU capacity across multiple cloud providers or on-prem infrastructure.

189. New Service Extensions plugins, powered by WebAssembly (Wasm), let you automate, extend, and customize your applications with plugin examples in Rust, C++, and Go. Support for Cloud Load Balancing is now generally available, and Cloud CDN support will follow later this year.

190. Cloud CDN‘s fast cache invalidation delivers static and dynamic content at global scale with improved performance, now in preview.

191. TLS 1.3 0-RTT in Cloud CDN boosts application performance for resumed connections, now in preview.

192. App Hub provides streamlined service discovery and management by automating service discovery and cataloging. 

193. App Hub service health enables resilient global services with network-driven cross-regional failover. Available later this year.

194. Later in 2025, you’ll be able to use Private Service Connect to publish multiple services within a single GKE cluster, making them natively accessible from non-peered GKE clusters, Cloud Run, or Service Mesh. 

Then, to help you secure your workloads, we introduced enhancements to protect distributed applications and internet-facing services against network attacks:

195. The new DNS Armor detects DNS-based data exfiltration attacks performed using DNS tunneling, domain generation algorithms (DGA) and other sophisticated techniques. Available in preview later this year.

196. New hierarchical policies for Cloud Armor let you enforce granular protection of your network architecture. 

197. There are new network types and firewall tags for Cloud NGFW hierarchical firewall policies, coming this quarter in preview.

198. Cloud NGFW adds new layer 7 domain filtering, allowing firewall administrators to monitor and control outbound web traffic to only allowed destinations. Coming later in 2025.

199. Inline network DLP for Secure Web Proxy and Application Load Balancer provides real-time protection for sensitive data-in-transit via integration with third-party (Symantec DLP) solutions using Service Extensions. In preview this quarter.

200. Network Security Integration, now generally available, helps you maintain consistent policies across hybrid and multi-cloud environments without changing your routing policies or network architecture. 

201. Imperva Application Security is integrated with Cloud Load Balancing, via Service Extensions, and is now available in the Google Cloud Marketplace.

Additional partner announcements 

We’ve always taken an open approach to AI, and the same is true for agentic AI. With updates this week at Next ‘25, we’re now infusing partners at every layer of our agentic AI stack to enable multi-agent ecosystems. Here’s a closer look:

202. Expert AI services: Our ecosystem of services partners — including Accenture, BCG, Capgemini, Cognizant, Deloitte, HCLTech, Infosys, KPMG, McKinsey, PwC, TCS, and Wipro — have actively contributed to the A2A protocol and will support its implementation. 

203. AI Agent Marketplace: We launched a new AI Agent Marketplace — a dedicated section within Google Cloud Marketplace that allows customers to browse, purchase, and manage AI agents built by partners including Accenture, BigCommerce, Deloitte Elastic, UiPath, Typeface, and VMware, with more launching soon. 

204. Power agents with all your enterprise data: We are partnering with NetApp, Oracle, SAP, Salesforce, and ServiceNow to allow agents to access data stored in these popular platforms.

205. Better field alignment and co-sell: We introduced new processes to better capture and share partners’ critical contributions with our sales team, including increased visibility into co-selling activities like workshops, assessments, and proofs-of-concept, as well as partner-delivered services.

206. More partner earnings: We are evolving incentives to help partners capitalize on the biggest opportunities, such as a 2x increase in partner funding for AI opportunities over the past year. We also introduced new AI-powered capabilities in Earnings Hub, our destination for tracking incentives and growth. 

207. We partnered with Adobe, the leader in creativity, to bring our advanced Imagen 3 and Veo 2 models to applications like Adobe Express.

208. Together with Salesforce’s Agentforce, we’re leading the digital labor revolution, driving massive gains in human augmentation, productivity, efficiency, and customer success.  

Security

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

209. Google Unified Security: This solution brings together our visibility, threat detection, AI powered security operations, continuous virtual red-teaming, the most trusted enterprise browser, and Mandiant expertise — in one converged security solution running on a planet-scale data fabric.

210. 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, along with a history of the agent’s evidence and decision making. 

211. Malware analysis agent: This agent investigates whether code is safe or harmful. It builds on Code Insight to analyze potentially malicious code, including the ability to create and execute scripts for deobfuscation. 

212. In Google Security Operations, new data pipeline management capabilities can help customers better manage scale, reduce costs, and satisfy compliance mandates. 

213. We also expanded our Risk Protection Program, which provides discounted cyber-insurance coverage based on cloud security posture, to welcome new program partners Beazley and Chubb, two of the world’s largest cyber-insurers.

214. New employee phishing protections in Chrome Enterprise Premium use Google Safe Browsing data to help protect employees against lookalike sites and portals attempting to capture credentials.

215. The Mandiant Retainer provides on-demand access to Mandiant experts. Customers now can redeem prepaid funds for investigations, education, and intelligence to boost their expertise and resilience.

216. Mandiant Consulting is also partnering with Rubrik and Cohesity to create a solution to minimize downtime and recovery costs after a cyberattack. 

Storage

Storage is a critical component for minimizing bottlenecks in both training and inference, and we introduced new innovations to help:

217. We expanded Hyperdisk Storage Pools to store up to 5 PiB of data in a single pool — a 5x increase from before. 

218. Hyperdisk Exapools is the biggest and fastest block storage in any public cloud, with exabytes of storage delivering terabytes per second of performance. 

219. Hyperdisk ML can now hydrate from Cloud Storage using GKE volume populator.

220. Rapid Storage is a new Cloud Storage zonal bucket with <1ms random read and write latency, and compared to other leading hyperscalers, 20x faster data access, 6 TB/s of throughput, and 5x lower latency for random reads and writes.

221. Anywhere Cache is a new strongly consistent cache that works seamlessly with existing regional buckets to cache data within a selected zone. Reduces latency up to 70% and 2.5TB/s accelerating AI workloads; maximizing goodput by keeping data close to GPU/TPUs. 

222. The new Google Cloud Managed Lustre high-performance, fully managed parallel file system built on DDN EXAScaler. This zonal storage solution provides PB scale <1ms latency, millions of IOPS, and TB/s of throughput for AI workloads.

223. Storage Intelligence, the industry’s first offering enabling customers to generate storage insights specific to their environment by querying object metadata at scale, uses LLMs to provide insights into data estates, as well as take actions on them. 

Startups

224. We announced a significant new partnership with the leading venture capital firm Lightspeed, which will make it easier for Lightspeed-backed startups to access technology and resources through the Google for Startups Cloud Program. This includes upwards of $150,000 in cloud credits for Lightspeed’s AI portfolio companies, on top of existing credits available to all qualified startups through the Google for Startups Cloud Program.

225. The new Startup Perks program provides early stage startups with preferred access to solutions from our partners like Datadog, Elastic, ElevenLabs, GitLab, MongoDB, NVIDIA, Weights & Biases, and more. 

226. Google for Startups Cloud Program members will receive an additional $10,000 in credits to use exclusively on Partner Models through Vertex AI Model Garden, so they can quickly start using both Gemini models and models from partners like Anthropic and Meta.

Google Workspace: AI-powered productivity

Gemini not only powers best-in-class AI capabilities as a model, but through its own products, like Google Workspace, which includes popular apps like Gmail, Docs, Drive and Meet. We announced a number of new Workspace innovations to further empower users with AI, including:

227. Help me Analyze: This powerful feature transforms Google Sheets into your personal business analyst, intelligently identifying insights from your data without the need for explicit prompting, empowering you to make data-driven decisions with ease.

228. Docs Audio Overview: With audio overviews in Docs, you can create high-quality, human-like audio read-outs or podcast-style summaries of your documents.

229. Google Workspace Flows: Workspace Flows helps you automate daily work and repetitive tasks like managing approvals, researching customers, organizing your email, summarizing your daily agenda, and much more.

There’s no place like home

And with that, we’ve come to the end of Next 25. We hope you’ve enjoyed your time in Las Vegas, and wish you safe travels. 

See you in Vegas next year for Google Cloud Next: April 22 – 24, 2026.

^{1. Grounding with Google Maps is currently available as an experimental release in the United States, providing access to only places data in the United States.}

Attending a tech conference like Google Cloud Next can feel like drinking from a firehose — all the news, all the sessions, and breakouts, all the learning and networking… But after a busy couple of days, watching the developer keynote makes it seem like there’s a method to the madness. A coherent picture starts to emerge from all the things that you’ve seen, pointing the way to all the cool things you can do when you get back to your desk.

This year, the developer keynote was hosted by the inimitable duo of Richard Seroter, Google Cloud Chief Evangelist, and Stephanie Wong, Head of Developer Skills and Community, plus a whole host of experts from around Google Cloud product, engineering, and developer advocacy teams. The keynote itself was organized around a noble, relatable goal: Use AI to help remodel AI Developer Experience Engineer Paige Bailey’s 1970s era kitchen. But how?

It all starts with a prompt

The generative AI experience starts by prompting a model with data and your intent. Paige was joined on stage by Logan Kilpatrick, Senior Product Manager at Google DeepMind. There, Logan and Paige prompted AI Studio to analyze Paige’s kitchen, supplying it with text descriptions, floor plans, and images. In return, it suggested cabinets, a cohesive design, color palette, and materials, relying on Gemini’s native image generation capabilities to bring its ideas to life. Then, to answer important questions on cost, especially for Paige’s area, they used Grounding with Google Search to pull in real-world material costs, local building codes and regulations, and other relevant information.

As Logan said, “From understanding videos, to native image generation, to grounding real information with Google Search – these are things that can only be built with Gemini.” 

New things that make this possible:

• Gemini 2.5 Pro is available in preview on Vertex AI and the Gemini app. 

• Gemini 2.5 Flash — our workhorse model optimized specifically for low latency and cost efficiency — is coming soon to Vertex AI, AI Studio, and the Gemini app.

From prompt to agent

We all know that a prompt is the heart of a generative AI query. “But what the heck is an agent?” asked Richard. “That’s the million-dollar question.” 

“An agent is a service that talks to an AI model to perform a goal-based operation using the tools and context it has,” Stephanie explained. And how do you go from prompt to agent? One way is to use Vertex AI, our comprehensive platform for building and managing AI applications and agents, and Agent Development Kit (ADK), an open-source framework for designing agents. ADK makes it easier than ever to get started with agents powered by Gemini models and Google AI tools.

Dr. Fran Hinkelman, Developer Relations Engineering Manager at Google Cloud, took the stage to show off ADK. An agent needs three things, Fran explained: 1) instructions to define your agent’s goal, 2) tools to enable them to perform, and 3) a model to handle the LLM’s tasks. 

Fran wrote the agent code using Python, and in a matter of minutes, deployed it, and got a professionally laid out PDF that outlined everything a builder might need to get started on a kitchen remodel. “What a massive time-saver,” Fran said.  

New things that make this possible: 

• Agent Development Kit (ADK) is our new open-source framework that simplifies the process of building agents and sophisticated multi-agent systems while maintaining precise control over agent behavior. With ADK, you can build an AI agent in under 100 lines of intuitive code.

• ADK support for Model Context Protocol (MCP), which creates a standardized structure and format for all the information an LLM needs to process a data request.

From one agent to many

It’s one thing to build an agent. It’s another to orchestrate a collection of agents — exactly the kind of thing you need for a complex process like remodeling a kitchen. To show you how, Dr. Abirami Sukumaran, Staff Developer Advocate at Google Cloud, used ADK to create a multi-agent ecosystem with three types of agents: 1) a construction proposal agent 2) a permits and compliance agent 3) an agent for ordering and delivering materials.

And when the multi-agent system was ready, she deployed it directly from ADK to Vertex AI Agent Engine, a fully managed agent runtime that supports many agent frameworks including ADK. 

It gets better: After deploying her agent, Abirami tested it out in Google Agentspace, a hub for sharing your own agents and those from third-parties. 

There was a problem, though. Midway through, the agent system appeared to fail. Abirami sprung into action, launching Gemini Cloud Assist Investigations, which used Logs Explorer to return relevant observations and hypotheses about the source of the problem. It even supplied a recommended code fix for the agents. Abirami examined the code, accepted it, redeployed her agents, and saved the day. 

This is really key. “It’s hard enough to build systems that orchestrate complex agents and services,” Abirami said. “Developers shouldn’t have to sit around debugging multiple dependencies — getting to the logs, going through the code, all of this can take a lot of time and resources that devs typically don’t have.”

New things that make this possible: 

• Vertex AI Agent Engine is a fully managed runtime in Vertex AI that helps you deploy your custom agents to production with built-in testing, release, and reliability at a global, secure scale. 
• Cloud Assist Investigations helps diagnose problems with infrastructure and even issues in the code.
• Agent2Agent (A2A) protocol: We’re proud to be the first hyperscaler to create an open protocol to help enterprises support multi-agent ecosystems, so agents can communicate with each other, regardless of the underlying technology.

Choose your own IDE and models 

“Have you heard of vibe coding?” i.e., agentic coding, asked our next presenter, Debi Cabrera, Senior Developer Advocate at Google Cloud. Essentially, people can prompt an agent with ideas as well as code to get to an effective programming output. People are doing it more and more using Windsurf, a popular new Integrated Development Environment (IDE), and she’s a fan.

Debi also showed using Gemini in Cursor and IntelliJ with Copilot, but you could also use Visual Studio Code, Tabnine, Cognition, or Aider. (She even wrote her prompts in Spanish, which Gemini handled sin problema). At the end of the day, “we’re enabling devs to use Gemini wherever it suits you best,” Debi said. 

Conversely, if you don’t want to use Gemini as your model, you can also use one of the more than 200 models in Vertex AI Model Garden, including Llama, Gemma 3, Anthropic, and Mistral, or open source models from Hugging Face.

“No matter what you use, we’re excited to see what you come up with!” 

New things that make this possible:

• Gemini 2.5 Pro is now available in Gemini Code Assist for individuals.

• Android Studio support for Gemini Code Assist is now available in preview. 

• Gemini in Firebase provides complete AI assistance in the new Firebase Studio.

In a field of dreams

Next up, presenters took a break from Paige’s kitchen remodel to tackle another high-value problem: how to throw a pitch.

With all the data that Major League Baseball processes with Google Cloud — 25 million data points per game — pitching technique is a problem that’s ripe for AI.

Jake DiBattista, winner of the recent Google Cloud x MLB Hackathon, started by analyzing a video of a great left-handed pitcher, Clayton Kershaw. He pre-processed the video using a computer vision library, and stored it in Google Cloud, using selections such as pitch type and game state to pull MLB data. Finally, after sending all this information to the Gemini API, he got his answer: Kershaw threw his signature curveball with nearly no deviation from his ideal.

Impressive, but how well does it work for those of us who aren’t pros? Jake created an “amateur mode” for less experienced players, and used a video of our host, Richard, throwing a pitch! After some prompt engineering to adapt from the professional model for Kershaw to an amateur model for Richard, the results were a little more prescriptive: He has potential, he just needs to tighten up his arm a little, and use more leg drive to maximize his power.

Jake shared the inspiration for his project: As a shot putter in college, he wanted to measure the accuracy of his throwing technique. How can you improve if you don’t know what you’re doing wrong – or right? Back then, having this kind of data would have been incredibly valuable for his development. 

But what’s truly amazing is that Jake built this fully customizable prompt generator for analyzing pitches in just one week. “This essentially worked out of the box,” Jake said. “I didn’t need to implement a custom model or build overly complex datasets.”

Get back to work 

Meanwhile, back at his day job, our next presenter Jeff Nelson, Developer Advocate at Google Cloud, took the stage with a clear goal: to turn raw data into a data application for use by sales managers. He started in BigQuery Notebook to build a forecast and wrote some SQL code. BigQuery loaded the results into a Python DataFrame, because Python makes it easy to use libraries to execute code over tables of any size.

But how can you actually use this agent to forecast sales? Jeff selected the Gemini Data Science Agent built into the Notebook, hit “Ask Agent,” and inputted a prompt that asked for a sales forecast from his table. The best part – from that point onward, all code was generated and executed by the Gemini Data Science Agent.

Plus, he pointed out that the agent used Spark for feature engineering, which is only possible because of our new Serverless Spark engine in BigQuery. Switching between SQL, Spark, and Python is easy, so you can use the right tool for the job.

To build the forecast itself, Jeff used a new Google foundation model, TimesFM, that’s accessible directly from BigQuery. Unlike traditional models, this one’s pre-trained and on massive times-series datasets, so you get forecasts by simply inputting data. “The forecast becomes a data app accessible to everyone,” Jeff said.

• New things that make this possible:  

• Serverless Spark engine in BigQuery 

• Specialized agents to support Data Engineers, Data Analysts, and Business Users – in preview.

• Data Science Agent integrated in BigQuery notebooks is coming soon. You can get started in Colab today.

Focus on the fun stuff 

As a developer, how would you like it if you could hand off boring things like creating technical design or product requirement docs? Scott Densmore, Senior Director of Engineering, closed out the demos to show us an incredible way to cut through tedious work: Gemini Code Assist and its new Kanban board. 

Code Assist can help you orchestrate agents in all aspects of the software development lifecycle, including with what Scott calls a “backpack” that holds all your engineering context. Using a technical design doc for a Java migration as an example, Scott created a comment and assigned it to Code Assist right from the Google doc. Instantly, the new task shows up on the Kanban board, ready to be tracked. Nor is this capability limited to Google Docs — you can also assign tasks directly from your chatrooms and bug trackers, or have Code Assist proactively find them for you.

Then, he took a tougher example: he asked Code Assist to create a prototype for a product requirement doc. He told Code Assist the changes he wanted, and hit repeat until he was happy with what he saw. Easy.

“Gemini Code Assist provides an extra pair of coding hands to help you create applications and remove repetitive and mundane tasks — so you can focus on the fun stuff.”

New things that make this possible:

• Gemini Code Assist Kanban board lets you interact with our agents, review the workplan that Gemini creates to complete the tasks, and track the progress of the various jobs/requests.

Pretty amazing, right? But don’t just take our word for it, for a true sense of all the magic that we demonstrated here, go ahead and rewatch the full developer keynote. We promise that it will be an hour well spent. 

At Google Cloud Next, we introduced H4D VMs, our latest machine type for high performance computing (HPC). Building upon existing HPC offerings, H4D VMs are designed to address the evolving needs of demanding workloads in industries such as manufacturing, weather forecasting, EDA, and healthcare and life sciences.

H4D VMs are powered by the 5th Generation AMD EPYC^TM Processors, offering improved whole-node VM performance of more than 12,000 gflops and improved memory bandwidth of more than 950 GB/s. H4D provides low-latency and 200 Gbps network bandwidth using Cloud Remote Direct Memory Access (RDMA) on Titanium, the first of our CPU-based VMs to do so. This powerful combination enables you to efficiently scale your HPC workloads and achieve insights faster.

For open-source High-Performance Linpack (OSS-HPL), a widely-used benchmark for measuring the floating-point computing power of supercomputers, H4D offers 1.8x higher performance per VM and 1.6x higher performance per core compared to C3D. Additionally, H4D offers 5.8x higher performance per VM and 1.7x higher performance per core compared to C2D.

For STREAM Triad, a benchmark to measure memory bandwidth, H4D offers 1.3x higher performance per VM and 1.4x higher performance per core compared to C3D. Additionally, H4D offers 3x higher performance per VM and 1.4x higher performance per core compared to C2D.

Improved HPC application performance

H4D VMs deliver strong compute performance and memory bandwidth, significantly outperforming previous generations of AMD-based VMs like C2D and C3D, allowing for faster simulations and analysis, and delivering significant performance gains (relative to a prior generation AMD-based HPC VM, C2D) across various HPC applications and benchmarks, as illustrated below:

• Manufacturing
  • CFD apps like Siemens^TM Simcenter STAR-CCM+^TM/HIMach show up to 3.6x improvement.
  • CFD apps like Ansys Fluent/f1_racecar_140 show up to 3.6x improvement.
  • FEA Explicit apps like Altair Radioss/T10m show up to 3.6x improvement.
  • CFD apps like OpenFoam/Motorbike_20m show up to 2.9x improvement. 
  • FEA Implicit apps like Ansys Mechanical/gearbox shows up to 2.7x improvement.
• Healthcare and life sciences:
  • Molecular Dynamics (GROMACS) shows up to 5x improvement.
• Weather forecasting
  • Industry standard benchmark WRFv4 shows up to 3.6x improvement.

“Our deep collaboration with Google Cloud powers the next generation of cloud-based HPC with the announcement of the new H4D VMs. Google Cloud has leveraged the architectural advances of our 5^th Gen AMD EPYC CPUs to create an offering that delivers impressive performance uplift compared to previous generations across a variety of HPC benchmarks. This will empower customers to achieve fast insights and accelerate their most demanding HPC workloads.” – Ram Peddibhotla, corporate vice president, Cloud Business, AMD

Faster HPC with Cloud RDMA on Titanium

H4D’s performance is made possible with Cloud RDMA, a new Titanium offload that’s available for the first time on these VMs. Cloud RDMA is specifically engineered to support HPC workloads that rely heavily on inter-node communication, such as computational fluid dynamics, weather modeling, molecular dynamics, and more. By offloading network processing, Cloud RDMA provides predictable, low-latency, high-bandwidth communication between compute nodes, thus minimizing host CPU bottlenecks. 

Under the hood, Cloud RDMA uses Google’s innovative Falcon hardware transport for reliable, low-latency communication over our Ethernet-based data center networks, effectively resolving the traditional challenges of RDMA over Ethernet while helping to ensure predictable, high performance at scale. 

Cloud RDMA over Falcon speeds up simulations by efficiently utilizing more computational resources. For example, for smaller CFD problems like OpenFoam/motorbike_20m and Simcenter Star-CCM+/HIMach10, which have limited inherent parallelism and are typically challenging to accelerate, H4D results in 3.4x and 1.9x speedup, respectively, on four VMs compared to TCP.

For larger models, Falcon also helps maintain strong scaling. Using 32 VMs, Falcon achieved a 2.8x speedup over TCP for GROMACS/Lignocellulose and a 1.3x speedup for WRFv4/Conus 2.5km.

Cluster management and scheduling capabilities

H4D VMs will support both Dynamic Workload Scheduler (DWS) and Cluster Director (formerly known as Hypercompute Cluster).

DWS helps schedule HPC workloads for optimal performance and cost-effectiveness, providing resource availability for time-sensitive simulations and flexible HPC jobs.

Cluster Director, which lets you deploy and scale a large, physically-colocated accelerator cluster as a single unit, is now extending its capabilities to HPC environments. Cluster Director simplifies deploying and managing complex HPC clusters on H4D VMs by allowing researchers to easily set up and run large-scale simulations.

VM sizes and regional availability

We offer H4D VMs in both standard and high-memory configurations to cater to diverse workload requirements. We also provide options with local SSD for workloads that demand high-speed storage, such as CPU-based seismic processing and structural mechanics applications (e.g., Abaqus, NASTRAN, Altair OptiStruct and Ansys Mechanical).

H4D VMs are currently available in us-central1-a (Iowa), and europe-west4-b (Netherlands), with additional regions in progress.

What our customers and partners are saying

Want to try it out?

We’re excited to see how H4D VMs will empower you to achieve faster results with your HPC workloads! Sign up for the preview by filling out this form.

For decades, businesses have wrestled with unlocking the true potential of their data for real-time operations. Bigtable, Google Cloud’s pioneering NoSQL database, has been the engine behind massive-scale, low-latency applications that operate at a global scale. It was purpose-built for the challenges faced in real-time applications, and remains a key piece of Google infrastructure, including YouTube and Ads. 

This week at Google Cloud Next, we announced continuous materialized views, an expansion of Bigtable’ SQL capabilities. Bigtable SQL and continuous materialized views enable users to build fully-managed, real-time application backends using familiar SQL syntax, including specialized features that preserve Bigtable’s flexible schema — a vital aspect of real-time applications. 

Whether you’re building streaming applications, real-time aggregations, or global AI analysis on a continuous data stream, Bigtable just got a whole lot easier — and much more powerful. 

Bigtable’s SQL interface, now generally available   

Bigtable recently transformed the developer experience by adding SQL support, now generally available. SQL support makes it easier for development teams to work with Bigtable’s flexibility and speed.

The Bigtable SQL interface enhances accessibility and streamlines application development by facilitating rapid troubleshooting and data analysis. This unlocks new use cases, like real-time dashboards utilizing distributed counting for instant metric retrieval and improved product search through K nearest neighbors (KNN) similarity search. A wide range of customers, spanning innovative AI startups to traditional financial institutions, are enthusiastic about Bigtable SQL’s potential to broaden developer access to Bigtable’s capabilities.

“Imagine coding with AI that understands your entire codebase. That’s Augment Code, an AI coding platform that gives you context in every feature. Bigtable’s robustness and scaling enable us to work with large code repositories. Its ease of use allowed us to build security features that safeguard our customers’ valuable intellectual property. As our engineering team grows, Bigtable SQL will make it easier to onboard new engineers who can immediately start to work with Bigtable’s fast access to structured, semi-structured, or unstructured data while using a familiar SQL interface” said Igor Ostrovsky, cofounder and CTO, Augment.

“Equifax leverages Bigtable within our proprietary data fabric for the high-performance storage of financial journals.  Our data pipeline team evaluated Bigtable’s SQL interface and found it to be a valuable tool for directly accessing our enterprise data assets and improved Bigtable’s ease of use for SQL-experienced teams. This means more of our team can work efficiently with Bigtable and we anticipate boosted productivity and better integration capabilities,” said Varadarajan Elangadu Raghunathan and Lakshmi Narayanan Veena Subramaniyam, vice-presidents, Data Fabric Decision Science.

Bigtable SQL has also been praised for offering a smooth migration path from databases with distributed key-value architectures and SQL-based query languages, including Cassandra (CQL) and HBase with Apache Phoenix.

“At Pega, we are building real-time decisioning applications that require very low latency query responses to make sure our clients get real-time data to drive their business. The new SQL interface in Bigtable is a compelling option for us as we look for alternatives to our existing database,” said Arjen van der Broek, principal product manager, Data and Integrations, Pega.

This week, Bigtable is also adding new preview functionalities to its SQL language including GROUP BYs and aggregations, an UNPACK transform for working with timestamped data, and structured row keys for working with data that is stored in a multi-part row key.

Continuous materialized views, now in preview

Bigtable SQL integrates with Bigtable’s recently introduced continuous materialized views (preview), offering a solution to traditional materialized view limitations like data staleness and maintenance complexity. This allows for real-time aggregation and analysis of data streams across applications such as media streaming, e-commerce, advertising, social media, and industrial monitoring.

Bigtable materialized views are fully managed and make updates incrementally without impacting your user queries from applications. Bigtable materialized views also support a rich SQL language including functions and aggregations. 

“With Bigtable’s new Materialized Views, we’ve unleashed the full potential of low-latency use cases for clients of our Customer Data Platform. By defining SQL-based aggregations/transformations at ingestion, we’ve eliminated the complexities and delays of ETL in our time series use cases. Moreover, using data transformations during ingestion, we’ve unlocked the ability for our AI applications to receive perfectly prepared data with minimal latencies,” said Sathish KS, Chief Technology Officer, Zeotap.

Ecosystem integrations 

To get useful real-time analytics, you often need to pull data from many sources and do so with very low latency. As Bigtable expands its SQL interface, it is also expanding its ecosystem compatibility making it easier to build end to end applications using simple connectors and SQL. 

Open-source Apache Kafka Bigtable Sink 
Customers often rely on Google Cloud Managed Service for Apache Kafka to build pipelines that stream data into Bigtable and other analytics systems. To help customers build high-performance data pipelines, the Bigtable team has open-sourced a new Bigtable Sink for Apache Kafka so you can send data from Kafka to Bigtable in milliseconds. 

Open-source Apache Flink Connector for Bigtable
Apache Flink is a stream-processing framework that lets you manipulate data in real time. With the recently launched Apache Flink to Bigtable Connector, you can construct a pipeline that lets you transform streaming data and write the outputs into Bigtable using both the high-level Apache Flink Table API and the more granular Datastream API. 

BigQuery Continuous Queries, now generally available  
BigQuery continuous queries are SQL statements that run continuously and can export the output rows produced into a Bigtable table. This feature is now generally available and can help you build a real-time analytics database using the combination of Bigtable and BigQuery. 

Continuous Queries are also available from BigQuery’s Python frameworks as a bigrames.streaming API, so Python developers can create fully-managed jobs that synchronize online datasets in Bigtable with offline datasets in BigQuery. 

“BigQuery continuous queries enables our application to use real-time stream processing and ML predictions by simply writing a SQL statement. It’s a great service that allows us to launch products quickly and easily,” said Shuntaro Kasai and Ryo Ueda, MLOps Engineers, DMM.com.

Bigtable CQL Client: Bigtable is now in preview and Cassandra-compatible

The Cassandra Query Language (CQL) is the query language of Apache Cassandra. With the launch of Bigtable CQL Client, developers can now migrate their applications to Bigtable with minimal to no code change, and enjoy the familiarity of CQL on enterprise-grade, high-performance Bigtable. Bigtable also supports common tools in the Cassandra ecosystem like the CQL shell (CQLsh), as well as Cassandra’s own data migration utilities which enable seamless migrations from Cassandra, with no downtime significantly reducing operational overhead.

Get started using the Bigtable CQL Client and migration utilities here.

Convergence: NoSQL’s embrace of SQL power

In this blog, we discussed a significant advancement that empowers developers to use SQL with Bigtable. You can easily get started with the flexible SQL language from any existing Bigtable cluster using Bigtable Studio and start to create materialized views on streams of data coming from Kafka and Flink. 

To learn more, check out the GoogleSQL for Bigtable overview.

As an object storage service, Google Cloud Storage is popular for its simplicity and scale, a big part of which is due to the stateless REST protocols that you can use to read and write data. But with the rise of AI and as more customers look to run data-intensive workloads, two major obstacles to using object storage are its higher latency and lack of file-oriented semantics. With the launch of Rapid Storage on Google Cloud, we’ve added a stateful gRPC-based streaming protocol that provides sub-millisecond read/write latency and the ability to easily append data to an object, while maintaining the high aggregate throughput and scale of object storage. In this post, we’ll share an architectural perspective into how and why we went with this approach, and the new types of workloads it unlocks.

It all comes back to Colossus, Google’s internal zonal cluster-level file system that underpins most (if not all) of our products. As we discussed in a recent blog post, Colossus supports our most demanding performance-focused products with sophisticated SSD placement techniques that deliver low latency and massive scale. 

Another key ingredient in Colossus’s performance is its stateful protocol — and with Rapid Storage, we’re bringing the power of the Colossus stateful protocol directly to Google Cloud customers.

When a Colossus client creates or reads a file, the client first opens the file and gets a handle, a collection of state that includes all the information about how that file is stored, including which disks the file’s data is stored on. Clients can use this handle when reading or writing to talk directly to the disks via an optimized RDMA-like network protocol, as we previously outlined in our Snap networking system paper.

Handles can also be used to support ultra-low latency durable appends, which is extremely useful for demanding database and streaming analytics applications. For example, Spanner and Bigtable both write transactions to a log file that requires durable storage and that is on the critical path for database mutations. Similarly, BigQuery supports streaming to a table while massively parallel batch jobs perform computations over recently ingested data. These applications open Colossus files in append mode, and the Colossus client running in the application uses the handle to write their database mutations and table data directly to disks over the network. To ensure the data is stored durably, Colossus replicates its data across several disks, performing writes in parallel and using a quorum technique to avoid waiting on stragglers.

The above image shows the steps that are taken to append data to a file.

1. The application opens the file in append mode. The Colossus Curator constructs a handle and sends it to the Colossus Client running in-process, which caches the handle.

2. The application issues a write call for an arbitrary-sized log entry to the Colossus Client.

3. The Colossus Client, using the disk addresses in the handle, writes the log entry in parallel to all the disks.

Rapid Storage builds on Colossus’s stateful protocol, leveraging gRPC-based streaming for the underlying transport. When performing low-latency reads and writes to Rapid Storage objects, the Cloud Storage client establishes a stream, providing the same request parameters used in Cloud Storage’s REST protocols, such as the bucket and object name. Further, all the time-consuming Cloud Storage operations such as user authorization and metadata accesses are front-loaded and performed at stream creation time, so subsequent read and write operations go directly to Colossus without any additional overhead, allowing for appendable writes and repeated ranged reads with sub-millisecond latency.

This Colossus architecture enables Rapid Storage to support 20 million requests per second in a single bucket — a scale that is extremely useful in a variety of AI/ML applications. For example, when pre-training a model, pre-processed, tokenized training data is fed into GPUs or TPUs, typically in large files that each contain thousands of tokens. But the data is rarely read sequentially, for example, because different random samples are read in different orders as the training progresses. With Rapid Storage’s stateful protocol, a stream can be established at the start of the training run before executing massively parallel ranged-reads at sub-millisecond speeds. This helps to ensure that accelerators aren’t blocked on storage latency.

Likewise, with appends, Rapid Storage takes advantage of Colossus’s stateful protocol to provide durable writes with sub-millisecond latency, and supports unlimited appends to a single object up to the object size limit.  A major challenge with stateful append protocols is how to handle cases where the client or server hangs or crashes. With Rapid Storage, the client receives a handle from Cloud Storage when creating the stream. If the stream gets interrupted but the client wants to continue reading or appending to the object, the client can re-establish a new stream using this handle, which streamlines this flow and minimizes any latency hiccups. It gets trickier when there is a problem on the client, and the application wants to continue appending to an object from a new client. To simplify this, Rapid Storage guarantees that only one gRPC stream can write to an object at a time; each new stream takes over ownership of the object, transactionally locking out any prior stream. Finally, each append operation includes the offset that’s being written to, ensuring that data correctness is always preserved even in the face of network partitions and replays.

In the above image, a new client takes over ownership of an object, locking out the previous owner. 

1. Initially, client 1 appends data to an object stored on three disks.

2. The application decides to fail over to client 2, which opens this object in append mode. The Colossus Curator transactionally locks out client 1 by increasing a version number on each object data replica.

3. Client 1 attempts to append more data to the object, but cannot because its ownership was tied to the old version number.

To make it as easy as possible to integrate Rapid Storage into your applications, we are also updating our SDKs to support gRPC streaming-based appends and expose a simple application-oriented API. Writing data using handles is a familiar concept in the filesystems world, so we’ve integrated Rapid Storage into Cloud Storage FUSE, which provides clients with file-like access to Cloud Storage buckets, for low-latency file-oriented workloads. Rapid Storage also natively enables Hierarchical Namespace as part of its zonal bucket type, providing enhanced performance, consistency, and folder-oriented APIs.

In short, Rapid Storage combines the sub-millisecond latency of block-like storage, the throughput of a parallel filesystem, and the scalability and ease of use of object storage, and it does all this in large part due to Colossus. Here are some interesting workloads we’ve seen our customers explore during the preview:

• AI/ML data preparation, training, and checkpointing

• Distributed database architecture optimization

• Batch and streaming analytics processing

• Video live-streaming and transcoding

• Logging and monitoring

Interested in trying Rapid Storage? Indicate your interest here or reach out through your Google Cloud representative. 

Visit us at Google Cloud Next and attend the breakout sessions “What’s new with Google Cloud’s Storage” (BRK2-025), “AI Hypercomputer: Mastering your Storage Infrastructure” (BRK2-020), and “Under the Iceberg: Simple, unified Cloud Storage for analytics data lakes” (BRK2-026) to learn more.

As organizations continue to prioritize cloud-first strategies to accelerate innovation and gain competitive advantage, legacy databases  remain a bottleneck by hindering modernization and stifling growth with unfriendly licensing, complex agreements, and rigid infrastructure. 

That’s why this week at Google Cloud Next, we’re announcing that Database Migration Service (DMS) is extending its comprehensive database modernization offering to support SQL Server to PostgreSQL migrations, enabling you to unlock the potential of open-source databases in the cloud and build modern, scalable, and cost-effective applications.

While holding great benefits, migrating from SQL Server to a modern, managed PostgreSQL offering like AlloyDB or Cloud SQL can be a highly complex task. Even though SQL Server and PostgreSQL both adhere to SQL standards, they still have fundamental differences in their architectures, data types, and procedural languages which require deep expertise in both technologies for a successful migration.

For example, SQL Server’s T-SQL syntax and built-in functions often require manual translation to PostgreSQL’s PL/pgSQL. Data type mappings can be intricate, as SQL Server’s DATETIME precision and NVARCHAR handling differ from PostgreSQL’s equivalents. 

Furthermore, features like SQL Server’s stored procedures, triggers, and functions often necessitate significant refactoring to align with PostgreSQL’s implementation. This requires deep knowledge in both database systems, as well as specific migration expertise that developers typically don’t possess, and it requires hours of painstaking work, even with the benefit of an automated conversion tool.

Simplifying database modernization with Database Migration Service 

DMS is a fully-managed, serverless cloud service that offers a complete set of capabilities to simplify database “lift and shift” migrations and database modernization journeys.

For modernization efforts, DMS offers an interactive experience that includes data migration, as well as schema and resident code conversion, all in the same powerful user interface. For data migration, it offers high-throughput database initial loads followed by low-latency change data capture to reduce downtime and minimize the impact on business critical applications.

Announcing SQL Server to PostgreSQL migration

The new SQL Server to PostgreSQL.migration experience supports the migration of both self-managed and cloud-managed SQL Server offerings to Cloud SQL for PostgreSQL and AlloyDB to accelerate your database modernization journey. Similar to the existing database modernization offerings, this new experience features a high-throughput initial load of the database followed by seamless change data capture (CDC) replication to synchronize the SQL Server source and PostgreSQL destination, all while the production application is up and running to ensure minimal business interruption.

For SQL Server schema and code conversion, DMS offers a fast, customizable algorithmic code conversion engine that automates the conversion of most of the database schema and code to the appropriate PostgreSQL dialect, leaving minimal manual conversion work for the user to complete.  

The algorithmic conversion engine maps the source database data types and SQL commands to the most suitable PostgreSQL ones, and even refactors complex source features which have no direct PostgreSQL equivalents to achieve the same functionality using available PostgreSQL capabilities. Algorithmic engines are extremely accurate, by nature, for the scenarios they are programmed for. However, they’re limited to just those scenarios, and in real-life usage some of the database code will consist of scenarios that can’t be anticipated.

For these situations, we’re pushing the boundaries of automated database modernization with the introduction of the Gemini automatic conversion engine. This new engine automatically augments the output of the algorithmic conversion, further automating the conversion tasks and reducing the amount of remaining manual work. It also provides a comprehensive conversion report, highlighting which parts of the code were enhanced, why they were changed, and how they were converted.

Instead of spending time researching suitable PostgreSQL features and fixing conversion issues, you can simply review the Gemini recommendations in the conversion report and mark the conversion as verified. Reviewing the completed conversions instead of having to research and fix issues can significantly reduce the manual migration effort and speed up the conversion process. 

To further empower SQL Server DBAs, DMS offers a Gemini conversion assist with targeted yet comprehensive SQL Server to PostgreSQL conversion training. Gemini analyzes both the source and the converted code and explains the conversion rationale, highlighting the chosen PostgreSQL features, why they were used, and how they compare to the SQL Server ones. It can then optimize the migrated code for better performance and automatically generate comprehensive comments, for better long-term maintainability. 

At Google Cloud, we’ve been working closely with customers looking to modernize their database estate. One of them is Wayfair LLC, an American online home store for furniture and decor.

“Google Cloud’s Database Migration Service simplifies the process of modernizing databases. Features like Change Data Capture to reduce downtime and AI-assisted code conversion help evolve our database usage more efficiently. This makes the migration process less manual and time-consuming, allowing teams to spend more time on development and less on infrastructure,” said Shashank Srivastava, software engineering manager, Data Foundations, Wayfair.

How to get started

To start your Gemini-powered SQL Server migration, navigate to the Database Migration page in the Google Cloud console, and follow these simple steps:

1. Create your source and destination connection profiles, which contain information about the source and destination databases. These connection profiles can later be used for additional migrations.

2. Create a conversion workspace that automatically converts your source schema and the code to a PostgreSQL schema and compatible SQL. Make sure you choose to enable the new Gemini-powered conversion workspace capabilities.

3. Review the converted schema objects and SQL code, and apply them to your destination Cloud SQL for PostgreSQL or AlloyDB for PostgreSQL instance.

4. Create a migration job and choose the conversion workspace and connection profiles previously created. 

5. Test your migration job and get started whenever you’re ready.

To learn more about how Database Migration Service can help you modernize your SQL Server databases, please review our DMS documentation and start your migration journey today.

Supporting customers where they want to be is a core value at Google Cloud, and a big part of the reason that we have partnered with Oracle — so that you can innovate faster with the best of Google and the best of Oracle. 

This week at Google Cloud Next, we announced significant expansions to our Oracle Database offerings, including the preview of Oracle Base Database Service for a flexible and controllable way to run Oracle databases in the cloud; general availability of Oracle Exadata X11M, bringing the latest generation of the Oracle Exadata platform to Google Cloud; and additional enterprise-ready capabilities including customer managed encryption keys (CMEK).

We are continuing to invest in global infrastructure for Oracle, with a total of 20 locations available in the coming months, adding Oracle Database@Google Cloud presence in Australia, Brazil, Canada, India, Italy, and Japan. 

These announcements follow our developments with Oracle since last July, when we launched Oracle Database@Google Cloud. This partnership enables customers to migrate and modernize their Oracle workloads and start taking advantage of Google’s industry-leading data and AI capabilities such as BigQuery, Vertex AI platform, and Gemini foundation models. 

Additional features provide customers with even more options in their modernization journey, such as the fully managed Oracle Autonomous Database Serverless. They can also benefit from increased reliability and resiliency features, such as cross-region disaster recovery and Oracle Maximum Availability Gold certification.

“Banco Actinver is committed to providing innovative financial solutions to our clients. By combining the security and performance of Oracle Database with Google Cloud’s data analytics and AI tools, we’re gaining deeper insights into market trends, enhancing our services, and delivering personalized experiences to our customers,” said Jorge Fernandez, CIO, Banco Actinver.

Innovative new capabilities

We’re expanding our offerings to empower customers with the flexibility to manage a diverse set of database workloads cost effectively.

• Oracle Base Database Service: The new Base Database Service delivers a highly controllable and customizable foundational database platform, built on Oracle Cloud Infrastructure (OCI) virtual machines and general-purpose infrastructure. It can empower businesses with the flexibility to manage a diverse range of database workloads directly.

• Enhanced Oracle Database Services: In addition to the availability of Exadata Cloud Service, Autonomous Database Service, Oracle Linux, and Oracle on Google Compute Engine (GCE) and Google Kubernetes Engine (GKE), we are pleased to share general availability of Oracle Exadata X11M. Oracle Database@Google Cloud now offers the latest generation of Oracle Exadata machines, the X11M, with enhanced performance and scalability for demanding database workloads. These new machines provide significant performance gains and increased capacity, enabling customers to run even the most intensive Oracle applications with ease. X11M will be available in all new regions.

Customers are embracing Oracle Database@Google Cloud, and to support their global needs, we’re expanding our footprint while maintaining the highest standards of application performance and reliability. 

• Expanding to 20 Oracle Database@Google Cloud Locations in the coming months: To further support the growing demand for Oracle workloads on Google Cloud, we are launching in more locations, including U.S. Central 1 (Iowa), North America-Northeast 1 (Montreal), North America-Northeast 2 (Toronto), Asia-Northeast 1 (Tokyo), Asia-Northeast 2 (Osaka), Asia-South 1 (Mumbai), Asia-South 2 (Delhi), South America-East 1 (Sao Paulo), Europe-West (Italy), Australia-Southeast2 (Melbourne), and Australia-Southeast1 (Sydney) — and additional zones in Ashburn, Frankfurt, London, Melbourne, and Italy. The new regions and expanded capacity are in addition to Google Cloud regions across U.S. East (Ashburn), U.S. West (Salt Lake City), U.K. South (London), and Germany Central (Frankfurt) that are available today.

• New Partner Cross-Cloud Interconnect availability:  Partner Cross-Cloud Interconnect for OCI is pleased to expand our global network offerings with new multicloud connectivity between Google and Oracle Cloud Infrastructure in Toronto and Zurich. This further complements our existing 11 regions already served, ensuring the lowest possible latency between both clouds while keeping traffic private and secure.

• Cross Region Disaster Recovery: Cross Region Disaster Recovery support for Oracle workloads on Oracle Autonomous Database ensures high availability and resilience, protecting against potential outages and providing continuous operation for critical applications.

• Enterprise-grade networking upgrades: Advanced networking upgrades enable enterprises to efficiently deploy their Oracle resources along with Google Cloud and share resources.

Industry-leading certifications and user experience

Google Cloud is committed to providing a seamless and efficient experience for Oracle customers, ensuring that managing and utilizing Oracle databases is straightforward and effective. We offer a combination of native Google Cloud tools and Oracle Cloud Infrastructure (OCI) interfaces, along with robust support for various applications and systems.

• Enhanced user experience: Google Cloud is committed to providing an easy-to-use experience for Oracle customers, offering a Google Cloud integrated user experience for application developers and routine database operations, alongside an OCI-native experience for advanced database management. This includes support for Shared VPC, APIs, SDKs, and Terraform.

• Application support: Google Cloud is pleased to announce the support for Oracle applications running on Google Cloud, ensuring compatibility and optimal performance, including Oracle E-Business Suite, Peoplesoft Enterprise, JD Edwards Enterprise One, Hyperion Financial Management, and Retail Merchandising.

• SAP and Oracle Capability: Oracle workloads on Google Compute Engine are now supported by SAP and Oracle, further validating Google Cloud as a trusted platform for running enterprise applications.

• Integration with Google Cloud Monitoring: Provides enterprises a unified monitoring and alerting mechanism across all their Google Cloud database services, now including Oracle Database.

• New support in Google Cloud Backup and DR: Our backup service now provides central, policy-based management for backup of Oracle workloads along with other Google Cloud services using secure backup vaults for data protection — isolating and protecting data from threats like ransomware and accidental deletion.

Google Cloud’s strengths make it the preferred hyperscaler for running mission-critical Oracle workloads. 

Get started right away from your Google Cloud Console or learn more here.