Microsoft has announced the general availability of Anthropic’s Claude models within Azure AI Foundry, running natively on Nvidia’s GB300 Blackwell Ultra infrastructure. This deployment utilizes the GB300 NVL72 rack-scale system, which integrates 72 liquid-cooled GPUs and 36 Grace CPUs to handle intensive AI workloads. High-speed Quantum-X800 InfiniBand networking supports these models, enabling the creation of complex agentic systems that operate across different business domains.

Source

Microsoft has released the public preview of Azure Linux 4.0, a free, Fedora-based distribution specifically optimized for Azure virtual machines and VM Scale Sets. Originally developed as an internal project called CBL-Mariner, the operating system now powers major infrastructures like LinkedIn and Databricks. Unlike general-purpose distributions, this version is a stripped-down, text-based OS designed to minimize attack surfaces by excluding graphical interfaces and unnecessary tools.

Source

The Azure Developer CLI (azd) has introduced a new `azd tool` command group to manage development prerequisites like language SDKs, Docker, and the Bicep CLI. This feature allows for the discovery, installation, and upgrading of essential tools directly from the terminal to prevent template deployment failures. Additionally, a new cross-platform command runner called `azd exec` enables scripts to run with full environment context and Key Vault secret resolution.

Source

The Cloud Infrastructure Service Providers Europe (CISPE) has challenged Broadcom’s claims that VMware Cloud Foundation (VCF) can serve as a foundation for sovereign cloud infrastructure. CISPE argues that the proprietary nature of VCF creates a dangerous over-reliance on a foreign vendor that maintains unilateral control over the software. Because Broadcom is a US-based entity, its products remain subject to American export regulations and legal frameworks regardless of where the data centers are physically located.

Source

Nationwide Building Society is expanding its partnership with Broadcom to implement VMware Cloud Foundation as the cornerstone of its new private cloud platform. This integrated solution combines compute, storage, networking, and security to help the organization build a custom, resilient infrastructure. The transition supports a broader digital transformation strategy aimed at delivering scalable services to millions of customers.

Source

A sophisticated supply chain attack known as Miasma has compromised over 20 npm packages associated with the Leo Platform and RStreams ecosystems. The campaign began after attackers gained access to a maintainer account and published malicious updates in a fully automated operation lasting only seconds. This malware specifically targets developer workstations and continuous integration runners to harvest sensitive access data.

Source

Salesforce has introduced Help Agent, a prepackaged solution built on the Agentforce platform designed for rapid deployment of autonomous customer service. This tool allows organizations to connect AI agents to company knowledge bases and communication channels like web, text, and voice within minutes. Unlike traditional chatbots, these agents utilize the Atlas Reasoning Engine to plan and execute multi-step workflows independently.

Source

The European Commission has reached a preliminary conclusion to designate Amazon Web Services and Microsoft Azure as gatekeepers under the Digital Markets Act. Although these cloud providers do not meet the quantitative thresholds for designation, their vast user bases and market dominance have drawn regulatory scrutiny. If the designation becomes final, both companies will face strict obligations regarding interoperability, data access, and customer lock-in restrictions.

Source

The House Energy and Commerce Committee is considering the Ratepayer Protection Act to address the massive electricity demand from artificial intelligence. This bipartisan legislation aims to prevent residential households and small businesses from subsidizing the infrastructure required for large-scale computing facilities. The bill would require state utilities to implement a large load standard, ensuring that data center developers pay for the specific grid upgrades they necessitate.

Source

VMware Cloud Foundation now includes the vSphere Kubernetes Service, a CNCF-certified runtime that allows for the management of virtual machines and containers from a single control plane. This architectural integration eliminates infrastructure silos by allowing administrators to provision Kubernetes clusters using the same platform they use for traditional virtual machines. The service maintains close alignment with upstream Kubernetes releases, ensuring compatibility with standard tools like Helm and Argo CD without imposing proprietary constraints.

Source

DigiCert and Google Cloud have partnered to introduce independent trust validation for confidential computing environments. This collaboration applies Public Key Infrastructure principles to cloud infrastructure to provide cryptographic verification of system integrity. The service acts as a neutral third-party root of trust to ensure that cloud-hosted workloads remain authentic and untampered.

Source

AWS has launched Lambda MicroVMs, a serverless compute primitive designed to run isolated Linux environments for significantly longer durations than standard functions. Built on Firecracker virtualization technology, these units extend the maximum execution time from 15 minutes to eight hours. This new capability allows developers to run complex tasks like vulnerability scanning, CI/CD pipelines, and AI-generated code execution within a secure hardware-virtualized boundary.

Source

The 67th Top500 list reveals that China has overtaken the United States with its new Lineshine supercomputer, the first to exceed 2.2 exaflops in performance. This system, also known as Ling Chen, utilizes entirely domestic hardware developments to bypass international export restrictions on high-end components. It marks a significant return to the public rankings for China, which had largely stopped sharing official performance data for its leading systems since 2019.

Source

Nvidia has introduced a new reference design for AI data centers that utilizes a closed-loop liquid cooling system to eliminate external water consumption. This infrastructure employs a recirculated mixture of water and propylene glycol to dissipate heat from high-density GPU clusters. By transitioning away from traditional evaporative cooling methods, the company aims to mitigate the environmental impact of massive AI factories.

Source

Microsoft has announced the deprecation of the Azure DevOps issuer for workload identity federation service connections, with a final retirement date set for July 1, 2027. This change is part of a broader initiative to standardize on the Microsoft Entra issuer to improve security and consistency across Azure services. The transition affects service connections in the Azure public cloud that utilize single-tenant applications or managed identities for secretless authentication.

Source

Amazon has acknowledged that its proprietary artificial intelligence models currently lag behind the industry-leading frontier systems developed by OpenAI and Anthropic. Peter DeSantis, the executive overseeing Amazon’s AI and chip divisions, stated that the company’s models have not yet met the requirements for the most demanding enterprise workloads. Despite this gap, the company expects to be in the conversation regarding leading frontier models within the next year.

Source

Salesforce has disabled the Klue Battlecards app integration after a security breach at the competitive intelligence firm exposed customer data. The incident originated from a compromised legacy credential within Klue's infrastructure, which allowed an extortion group known as Icarus to push malicious code. This code collected OAuth tokens that customers use to connect Klue to their own environments, effectively bypassing standard authentication measures like passwords or multi-factor authentication.

Source

Amazon is exploring a significant shift in its hardware strategy by considering the sale of its custom Trainium AI chips to third-party data centers. Currently, these processors are exclusive to the Amazon Web Services (AWS) cloud environment, where they are used by organizations like OpenAI and Anthropic for AI training and inference. This potential move marks a direct challenge to Nvidia's dominance in the semiconductor market by offering a proprietary alternative for high-performance AI infrastructure.

Source

Amazon has launched a managed web search tool for Bedrock AgentCore to provide AI agents with real-time information. This feature addresses the limitation of static training data by allowing agents to access current web content, such as stock prices or recent news. The service is fully managed, meaning administrators do not need to provision external search APIs or manage third-party credentials.

Source