This year's Google CloudNext event, which took place on April 9-11, featured many exciting innovations and services related to Google Cloud technologies. As we did last year at the Google Cloud Next '23 event, we have summarized these innovations for you below.

‍

Scale with AI-Optimized Infrastructure

• A3 Mega: Next month, a new GPU-based instance developed with NVIDIA and using H100 Tensor Core GPUs will be available and will serve demanding workloads by offering twice as much bandwidth per GPU as compared to A3 instances. In addition, the Confidential A3 model will provide enhanced protection for sensitive data and AI tasks during both training and inference processes.
• NVIDIA HGX B200 and NVIDIA GB200 NVL72: Google Cloud will launch the latest NVIDIA Blackwell platform in early 2025 in two versions:

HGX B200 specifically designed for intensive artificial intelligence, data analysis and HPC tasks, and GB200 NVL72 optimized for real-time big-language model inference and comprehensive training of trillion-parameter scale models.

• TPU v5p: The general availability of TPU v5p, Google's powerful AI accelerator for training and inference, with four times the processing power per pod compared to the previous model, has been announced. In addition, Google Kubernetes Engine (GKE) now supports TPU v5p, which highlights a significant 900% increase in GPU and TPU usage in GKE last year.
• AI-optimized storage options: Cloud Storage FUSE and advanced caching features in Cloud Storage FUSE and ParallelStore increase training speed and make it easier for data to be closer to the customer's TPU or GPU. In addition, Hyperdisk ML, a next-generation block storage service introduced in the preview phase, speeds up model loading times up to 3.7 times compared to typical alternatives.
• New options for Dynamic Workload Scheduler: Calendar mode for uptime assurance and flexible startup for optimized economy will help customers deliver efficient resource management for the deployment of complex training and inferencing jobs.

‍

Google Distributed Cloud (GDC) empowers organizations to choose the environment, configuration, and controls tailored to their specific requirements.

New features announced by GDC:

• NVIDIA GPUs at GDC: NVIDIA GPUs are brought to GDC for both connected and air-gapped configurations. Each of these will support new GPU-based instances to run AI models efficiently.
  
• GKE at GDC: The same GKE technology used by leading AI companies in Google Cloud will also be available at the GDC.
  
• Support AI Models: Several open AI models, such as Gemma and Llama 2, are being validated at GDC to work in both air-gapped and connected edge environments.
  
• AlloyDB Omni for Vector Search: AlloyDB Omni for Vector Search was introduced to enable search and information retrieval in the GDC, serving private and sensitive data with remarkably low latency.
  
• Sovereign Cloud: GDC can be provided with a fully air-gapped setup with local operations to meet the most stringent regulatory demands; offering full sustainability and management options by Google or a selected partner. Users have full control with the flexibility to adapt instantly to legislative changes.

‍

Optimization is important for all cloud-based tasks, from web servers to containerized microservices, while not every workload involves AI.

• Google Axion: The first dedicated Arm-based CPU designed for the data center delivers up to 50% better performance and up to 60% better energy efficiency compared to the current generation of x86-based instances.

‍

In addition, two new machine series in general purpose VM portfolios, N4 and C4, native-bare-metal machine shapes in the C3 family of machines, general availability of Hyperdisk Advanced Storage Pools, and many similar things have been announced.

‍

Create Agents with Vertex AI

Vertex AI, an enterprise AI platform, is built on a high-end infrastructure and serves as the only unified platform that facilitates the discovery, customization, scaling, deployment, and management of productive AI models. With more than 130 models available, including Gemini, Claude 3, Gemma, Llama 2 and Mistral, the expanded reach offers customers a unique choice in model selection.

‍

• Gemini 1.5 Pro: Gemini 1.5 Pro, currently in public preview, has the world's largest context window, supporting up to 1 million tokens. In addition, the platform offers audio file processing capabilities that allow processing extensive data streams, including 1-hour videos, 11-hour audio files, large codebases, or more than 700,000 words.
  
• Claude 3: Anthropic's advanced models, the Claude 3 Sonnet and Claude 3 Haiku, are now generally available on Vertex AI with the help of a close partnership, while the Claude 3 Opus is expected to be available in the coming weeks.
  
• CodeGemma: A series of lightweight, state-of-the-art open models developed using the same research and technology as Gemini models, Gemma offers an enhanced version specifically designed for code generation and assistance, called CodeGemma, which is now accessible via Vertex AI.
  
• Imagen 2.0: The latest text-to-image technology offers a variety of image creation features that are tailored to meet specific brand needs, including text-to-live image functionality that enables the creation of animated images such as gifs, equipped with security filters and digital watermarks. In addition, advanced photo editing features such as inpainting and outpainting are now generally available, significantly increasing editing capabilities.
  
• Digitale Watermarkning: Powered by Google DeepMind's Synthid, this feature is now generally available for AI-generated images produced by Imagen 2.0.

‍

Once you've selected and adjusted and grounded the right model, Vertex AI can help you deploy, manage, and monitor models. Based on this, MLOps capabilities were announced:

• Prompt management tools: These tools allow you to collaborate on built-in prompts with notes and status, track changes over time, and compare the quality of responses from different prompts.
  
• Automatic side-by-side (AutoSxS): Now available to the general public, AutoSxS helps users understand the accuracy of the assessment by providing explanations and accuracy scores on why one response outperforms the other.
  
• Rapid Evaluation feature: Now in preview, this feature helps customers quickly evaluate models in a smaller dataset while revamping prompt design.

‍

Finally, Vertex AI Agent Builder combines basic models, Google Search, and developer tools to facilitate agent creation and deployment. It offers a code-free agent builder console with robust provisioning, orchestration and augmentation capabilities, enabling the rapid development of various generations of AI agents integrated with Google Search and organization data.

‍

‍

‍

Accelerate Development

Gemini Code Assist, an enterprise AI code assistance solution, has made significant improvements among Google developers; in the completion time of common tasks 40% faster and to write new code 55% reduction in time spent showed.

Gemini Code Assist facilitates integration with custom codebases, providing accessibility from a variety of locations, such as on-premises, GitHub, GitLab, Bitbucket, or many other platforms. Important announcements were also made on this topic to further enhance the industry leadership of developers:

• Gemini 1.5 Pro in Gemini Code Assist: Now in private preview, this upgrade is revolutionizing coding for even the largest projects by introducing a massive context window of 1 million tokens. Gemini Code Assist now offers more accurate code suggestions, deeper insights, and streamlined workflows.
  
• Gemini Cloud Assist: This provides AI assistance throughout your application lifecycle, making it easier to design, secure, run, troubleshoot, and optimize your application's performance and costs.

‍

Unleash the Potential of Artificial Intelligence with Data

Google Cloud enables the integration of artificial intelligence with enterprise data, ensuring data privacy and security. Recent developments in AI-Ready Data Cloud include LLM integration and vector matching capabilities across databases such as AlloyDB and BigQuery; data teams are being empowered to use Gemini models for multimodal analysis and advanced reasoning; operations in a variety of industries such as healthcare, supply chain management are being improved.

New enhancements announced to help organizations build data agents:

• Gemini in BigQuery: Gemini in BigQuery leverages artificial intelligence to assist data teams in data preparation, discovery, analysis, and governance. In addition, the new BigQuery data canvas allows the creation and execution of data pipelines, offering a notepad-like interface with natural language and built-in visualizations, both accessible in preview.
• Gemini in Databases: This makes it easy to move data from legacy systems safely and securely, such as converting your database to a modern cloud database like AlloyDB.
  
• Vector indexing: Using vector indexing directly in BigQuery and AlloyDB, new query capabilities allow you to leverage artificial intelligence over where your data is stored, enabling real-time and accurate responses.
  
• Gemini in the Looker: New features that facilitate seamless integration of data agents with workflows are being introduced in the preview phase. Advanced artificial intelligence capabilities also integrate with Google Workspace, enabling interaction with business data.

‍

Improve Your Cybersecurity Stance with AI-Powered Capabilities

• Gemini in Threat Intelligence: Using natural language, it provides in-depth insights into threat actor behaviors. With Gemini, larger samples of potentially malicious code can be analyzed, which allows examining module interactions and revealing the true purpose of the code.
  
  
• Gemini in Security Operations: A new assisted research feature turns natural language into detections, summarizes event data, suggests actions to take, and leads users across the platform via conversational chat.

‍

Increase Productivity with Google Workspace

Google Workspace, the world's leading productivity suite, serves more than 3 billion users and 10 million paying customers across individuals and organizations. In the past year, significant improvements have been made, including notable adoption rates among enterprise users, including the integration of Gemini for Workspace, an AI-powered tool built into Gmail, Docs, Sheets, and other tools.

• Google Vids: The new AI-powered video creation app for business combines video, writing, production and editing help in one tool. Creates editable storyboards, combines drafts using suggested scenes from ready-made media, and offers voiceover options, all with a user-friendly interface and secure collaboration capabilities in a Workspace environment.
  
• AI Meetings in Messaging add-on: The cost of this collaboration tool is only $10 per user per month, with “take notes for me”, chat summarization, and real-time translation in 69 languages (equal to 4,600 language pairs).
  
• New AI Security add-on: Workspace administrators now have the ability to automatically classify and protect sensitive files and data using privacy-protecting AI models and DataLoss Prevention controls tailored to their organization. Priced at $10 per user per month, the AI Security plugin is compatible with most Workspace plans.

‍

Take Advantage of Artificial Intelligence Agents

• Client Agents: They can understand customer needs, know about products, facilitate appropriate interaction and ultimately develop seamless interaction with your business. Effective customer representatives work using many methods, including text, voice, and others across a variety of channels such as web, mobile, call center, and point-of-sale.
  
• Laundry agents: It helps all your employees be more productive and work better together. Employee representatives can streamline processes, manage repetitive tasks, answer employee questions, as well as organize and translate critical communications.
  
• Creative agents: It can serve as the best designer and production team by working on visuals and slides and exploring concepts with you. It provides the most powerful platform for creating creative representatives and creates representatives for many clients, marketing teams, audio and video production teams, and all creative people who need help.

‍

A Leading Artificial Intelligence Ecosystem

To adopt GENAI widely, customers need an enterprise AI platform, an open platform that provides the broadest set of end-to-end capabilities, highly optimized for cost and performance, offers options, easy to integrate with existing systems, and supported. Par un grande ecosystema, Google se designar a plataforma de intelligencia artificiale para adresar estas problemas. Google Cloud is the only major cloud provider offering both first-party and extensible, partner-enabled solutions across every layer of the AI stack. Hvala na innovacije od Google Cloud se i nós partnerne, vi offre una choice di infrastruttura, chipsets, modelli, dataloluzioni e strumenti AI per aiutare in grado di gEnai applicazioni e creare il valore aziendente con questo tecnologia excitante.