Google Cloud It has opened Your Next 25 event in Las Vegas With their Last advances in computing and Ia for companies. The first is Google WANa global new generation network assisted by AI and with a practically nil latency. It offers support for essential services of Google and covers more than three million kilometers of fiber optics, distributed throughout more than 200 countries and territories. In addition, it is used to train the Gemini model.
This global network, from now on available to companies around the world, is optimized to maximize applications performance. It is up to 40% faster and reduces the total cost of ownership also by 40%.
Among the main ads made in Cloud Next are TPU Ironwoodthe seventh generation of Google TPU, ten times more powerful than the TPU of the previous generation. The Ironwood TPU has more than 9,000 chips by Pod. Each of these, in addition, has a calculation power of 42.5 Exaflops. Google Cloud will also be the first cloud provider to offer the two new Nvidia New Generation GPU in its infrastructure.
Improved models
The company has also commented that Google Gemini 2.5 models have joined Vartex AI Two weeks ago in the testing phase, and soon he will also come to seex ai gemini 2.5 flash. In addition, Google Cloud has announced improvements in Google’s models. Alphafold 3 and Weathernext models have also arrived to Vertex AI Model Garden, which already has more than 200 models.
Among them the image update 3 to generate higher quality images and an improvement in the image reconstruction function. Also Chirp 3, which allows you to generate personalized voices from several audio files, as well as Lyria, the first text conversion model in music suitable for company applications, and I see 2, an advanced video generation model and an edition platform with inpainting and outpainting functions, as well as cinematographic controls and interpolation.
From now on, VERTEX AI allows you to base Google Maps the responses of agents that need geographical information, in order to achieve more precise and updated results. It also allows additional third -party sources. It will also have an agent development kit (ADK), an Open Source work environment that simplifies the construction process of a multiagente system. In addition, it allows accurately to control the behavior of the agents.
With Adk, users can create an AFF -intuitive agent, with less than 100 lines of code. In addition, it is compatible with the Model Protocol (MCP), allowing developers to have facilities to link the available tools.
On the other hand, Google Cloud has created an agent2agent open protocolintended to offer support to multiagente ecosystems, facilitating communication between them, regardless of the technology they use. In their definition, more than 50 Google Cloud partners have actively participated.
Among them Accenture, Box, Deloitte, Salesforce, SAP, Servicenow and TCS. Agent Garden is a repository of samples and tools ready to use, which can be accessed directly from ADK. It allows to connect agents to more than a hundred predefined connectors, to personalized APIS, to integration workflows or data stored in the cloud systems of an organization. Google Cloud has also announced that your customers can create agents of AI directly about their data stored in NetApp.
Google Cloud shows improvements in agentspace
Google Cloud has also shown Improvements in Agentspaceas the integration with Chrome Enterprise to searches from the Google browser search bar. Agent Gallery, on the other hand, offers a unified view of all agents available in the organization. Both those of Google and the inmates or those of partners.
Agent Designer is an interface that allows you to generate personalized agents to automate daily tasks. It does not need code to use and facilitate the adaptation of agents to workflows and needs of each employee, regardless of their technical knowledge.
Agent Idea Generation is responsible for classifying ideas based on criteria defined by the employee, and agent deep Research explores complex issues at the request of the user and presents results in a report. On the other hand, Google Agentspace search engine offers ready -to -use functions that unify access to all data, safely and sensitively to each user’s permits.
New functions for Google Workspace and security with the help of AI
Workspace now has the Help Me Analyze function, which incorporates a kind of personal business analyst to the application sheets. Among other things, it takes conclusions from the data without the need for explicit instructions, and facilitates data -based decision making.
Documents will have audio summaries to generate summaries of documents in read or podcast style, high quality and natural voices. As for Google Workspace Flows, it helps automate daily and repetitive tasks. Among them the approval management, customer research, email organization or the preparation of summaries of the daily agenda.
The solution Google Unified Security (GUS)a novelty in protection presented in Google Cloud Next, uses Gemini to improve efficiency in all aspects of the experience of cybersecurity professionals. This solution creates a unified network of safety data, scalable and with search options, which covers the entire surface that can be attacked. It offers visibility and detection capacity, as well as response in networks, endpoints, cloud and applications. It also automatically enriches Google’s advanced security data.
A new alert classification agent It is responsible for conducting dynamic investigations in the name of users. Analyze the context of each alert, collect the relevant information and issue a verdict, with a history of evidence and decision process. Finally, the malware analysis agent Investigate source code to decide whether it is safe or can be dangerous. Use the Code Insight tool to analyze potentially malicious code and can create and execute scripts for refusal purposes.
