All news in category “Vendor and Hyperscaler Watch”

🔧 Amazon SageMaker HyperPod now offers troubleshooting skills that deliver expert AI/ML cluster diagnostics into coding assistants like Claude Code, Cursor, and Kiro. These skills guide natural-language diagnostics for GPU hardware faults, NCCL communication issues, and performance bottlenecks across distributed clusters. They automate evidence collection via AWS Systems Manager and provide actionable recommendations without requiring infrastructure changes.

🚀 Google announces GKE standby buffers to complement active buffers, providing low-cost suspended node capacity that resumes faster than cold node provisioning. Standby buffers store node state to disk, releasing compute and memory costs while keeping persistent disk and IP charges, enabling near-instant scheduling with only a small single-digit percent overhead. Together, active and standby buffers reduce pod scheduling latency, replace manual balloon-pod workarounds, and help balance performance and cost for spiky workloads.

🚀Starting today, Amazon EC2 M8i and M8i‑flex instances are available in Asia Pacific (New Zealand). Powered by custom Intel Xeon 6 processors exclusive to AWS, these instances deliver the highest performance and fastest memory bandwidth among comparable Intel cloud processors. They provide up to 15% better price-performance and 2.5x more memory bandwidth versus prior Intel-based generations, plus up to 20% higher performance than M7i variants with specific workloads seeing larger gains. M8i‑flex targets common general-purpose sizes from large to 16xlarge, while M8i offers 13 SAP-certified sizes including bare metal and a new 96xlarge for large-scale applications.

🔧 Amazon SageMaker HyperPod now supports EFA-only network interfaces for cluster instance groups, allowing dedicated Elastic Fabric Adapter devices without attaching Elastic Network Adapters for IP networking. This reduces IP address consumption in VPC subnets and enables larger-scale distributed training clusters. To enable it, set efa-only in the ClusterNetworkInterface when creating or updating a HyperPod cluster via the API.

🛡️ AWS Direct Connect now supports Virtual Interface (VIF) Rate Limiters on dedicated connections to prevent a single VIF from consuming all bandwidth and causing congestion. You can cap bandwidth for up to 10 VIFs per dedicated connection with increments from 50 Mbps to 1.6 Tbps when using a link aggregation group. Rate limiting applies to both ingress and egress, and excess packets are dropped. New CloudWatch metrics include utilization as a percentage of configured capacity and dropped packet counts, and the feature is available in all supported commercial and China Regions via console, API, or SDK.

🟦 Amazon Bedrock customers can now monitor inference traffic to the bedrock-mantle endpoint using Amazon CloudWatch metrics, matching existing support for the bedrock-runtime endpoint and other AWS services. The bedrock-mantle endpoint supports OpenAI Responses and Chat Completions APIs as well as the Anthropic Messages API, enabling easy migration of OpenAI- or Anthropic-based applications to Bedrock. Metrics are published under the AWS/BedrockMantle namespace and include inference counts, token totals, and client error counts across account, project, model, and project-and-model granularities. These metrics are available in all Regions where the bedrock-mantle endpoint is offered and can be used to monitor production inference, set alarms, and plan capacity.

🚀 Amazon Bedrock now supports OpenAI GPT-5.5, GPT-5.4, and Codex for production use, offering the same AWS security, governance, and operational controls. GPT-5.5 delivers advanced capabilities for agentic coding, data analysis, and multi-step autonomous tasks on a next-generation inference engine. Codex is available via a dedicated App, CLI, and IDE integrations for Visual Studio Code, JetBrains, and Xcode, and can be configured to run through Bedrock with pricing aligned to OpenAI first-party rates.

📢 Amazon Simple Email Service (Amazon SES) now supports tenant-level suppression lists to isolate bounces and complaints per tenant. Previously, a single suppression list applied to all tenants in an account, which could cause one tenant's delivery issues to affect others. Administrators can choose suppression scope (TENANT or ACCOUNT) and reason filters (BOUNCE, COMPLAINT, or both). API operations with a TenantName parameter let you view and manage suppressed addresses.

🔧 Microsoft confirmed and mitigated an incident that prevented some users from setting up multi-factor authentication and accessing the My Sign-Ins site, where affected users encountered 504 Gateway Timeout errors. The company failed over to alternate infrastructure and monitored telemetry while evaluating further mitigations. Microsoft later restored the service, attributing the outage to a cache configuration change that caused high CPU and memory load during an EU traffic peak.

🛡️ Check Point warns that the primary cyber threat to the 2026 U.S. midterms is not vote tampering but a coordinated assault on trust through misinformation, lookalike news sites, and domain abuse. Attackers are cloning major media brands, registering thousands of election-themed domains, and exploiting leaked credentials to fuel phishing and impersonation. Security teams must prioritize brand protection, rapid takedown, and credential monitoring to mitigate politically motivated campaigns that exploit familiar operational vectors at greater scale.

🔒 At GTC Taipei during COMPUTEX 2026, NVIDIA highlighted its Vera BlueField-4 STX and DOCA innovations designed to secure enterprise AI infrastructure. Modern AI factories combine high-performance compute, distributed storage, Kubernetes, APIs, GPU farms, and sensitive data, creating new security needs. Check Point integrates its AI Factory Firewall with NVIDIA BlueField and DOCA to provide visibility, segmentation, runtime protections, and infrastructure-level policy enforcement across distributed AI environments.

🛡️ Amazon Inspector introduces the new Inspector VM Scanner for agent-based EC2 scanning, increasing detection coverage while lowering CPU usage on instances. The updated scanner expands findings to include software such as WordPress, Apache HTTP Server, Python packages, and Ruby gems, bringing agent-based coverage closer to agentless parity. Customers can enable the VM Scanner via the Inspector console or API, including organization-wide enablement for delegated administrators, with no additional IAM instance profile roles required.

📣 Amazon Simple Email Service (SES) introduced enhanced deliverability features that report inbox placement percentages and public email blocklist status. These insights use representative industry samples to show how many messages are routed to spam folders and provide visibility by sending domain and campaign. SES also offers pre-send content testing to estimate inbox placement at major mailbox providers and passive monitoring of blocklist activity. The new capabilities are available in all AWS commercial regions where SES is offered.

📣 AWS End User Messaging now supports RCS for Business messaging in 20 additional countries, bringing coverage to 22 countries including the United States and Canada. Businesses can use the existing SendTextMessage API with no code changes, and messages are delivered from a recognized business identity. Devices that do not support RCS automatically fall back to SMS, and RCS for Business is available in all Regions where AWS End User Messaging operates.

🩺 AWS HealthLake now natively supports payer compliance with the CMS Interoperability and Prior Authorization Final Rule (CMS-0057-F), requiring four standardized FHIR-based APIs by January 1, 2027. The release implements Patient Access, Provider Access, Payer-to-Payer, and Prior Authorization APIs using CARIN, DaVinci, and SMART standards, and adds metrics and consent integrations. It is HIPAA-eligible and available in multiple AWS Regions.

📅 Amazon Connect Customer now supports scheduling tasks up to 90 days in advance, enabling organizations to plan, route, and track long-running follow-up work. Examples include scheduling adjuster visits, parts availability checks, and repair completion follow-ups, with tasks routed to the appropriate team and carrying relevant claim context. Tasks can be scheduled via the StartTaskContact API, contact flows, or the agent workspace. This capability is available in all commercial and AWS GovCloud (US) regions where Amazon Connect Customer is offered.

📡 AWS Shield Advanced now provides DDoS attack flow logs that deliver packet-level visibility into traffic targeting Shield-protected resources. The logs capture source and destination IPs, ports, protocols, packet and byte counts, and source country details, and are published every five minutes during active attacks. Log data can be delivered to Amazon S3, Amazon CloudWatch Logs, or Amazon Data Firehose for forensic analysis, threat intelligence, and compliance. To use the feature, resources must be protected by Shield Advanced and log delivery must be configured; the feature is available in all regions where Shield Advanced operates.

🛡️ For the seventh consecutive time, Microsoft has been named a Leader in the 2026 Gartner® Magic Quadrant™ for Endpoint Protection, reflecting customer trust in Microsoft Defender. Defender provides industry-leading EDR backed by global threat intelligence and connects endpoints, identities, email, apps, cloud, and data to enable earlier detection and stronger prevention. Recent advancements include proactive attack disruption, custom telemetry, simplified onboarding, sovereign-ready protection, and agentic endpoint security for local AI agents.

🛡️ This Cloud CISO Perspectives post by Usman Chaudhary, Field CISO for Google Public Sector, outlines a pragmatic roadmap for public-sector CISOs to adopt AI-driven security. It emphasizes immediate quick wins in the first 90 days, tactical actions within six months, and strategic initiatives for months six to 12, combining internal automation, commercial AI capabilities, and vendor solutions like Gemini for Government. The guidance targets threat triage, talent augmentation, posture elevation, and governance to reduce toil and accelerate proactive defense.

🚀 AlloyDB for PostgreSQL introduces a Hot Standby HA architecture that keeps the standby node actively replaying WAL records, reducing failover time and preserving cache warmth. This change eliminates standby database startup delays and minimizes post-failover performance degradation, improving RTO and stabilizing application throughput. Hot Standby is rolling out for PostgreSQL 18 and will reach earlier versions in months, with no extra cost and retention of the 99.99% SLA.