At SC26, DataDirect Networks introduced software-defined EXAScaler deployment, intended to give enterprise AI customers greater hardware flexibility when assembling AI infrastructure. The company also highlighted joint offerings with Super Micro Computer and said DDN’s Managed Lustre customers on Google Cloud achieved 70% better training performance and 40% higher GPU efficiency. SiliconANGLE AI reported that AI infrastructure sales cycles have shortened to as little as 30 days, although the performance figures and sales-cycle examples were presented by DDN executives.
Google says Agent Anomaly Detection has entered Private Preview with support for monitoring agents outside Google Cloud when their traces and logs are sent to a Google Cloud project. Eligible agents can include ADK for Python applications running on AWS Lambda, Microsoft Foundry, or on-premises infrastructure. In vendor-provided tests, the service flagged catalog scraping, resource exhaustion, and unauthorized refunds.
Google Developers Blog published a guide to authenticating custom interfaces that call the Gemini Enterprise StreamAssist API. It covers Google Identity, Workforce Identity Federation, Workload Identity Federation, Federated Connector OAuth, and on-demand authorization for custom ADK agents. The package includes architecture details, runnable code, interactive test interfaces, CLI scripts, and curl commands.
Google Developers Blog published a guide for connecting Gemini Enterprise agents to internal StreamableHTTP MCP servers hosted on Google Cloud Run. The reference architecture combines Egress Agent Gateway, Private Service Connect and FQDN-scoped Cloud DNS with dual service and user authentication, and includes a Python server, middleware and deployment scripts. Google says internal-only ingress blocks default public Gemini Enterprise calls with an HTTP 404 at Google’s edge.
Gemini CLI released v0.71.0-preview.0 with a change to parse PowerShell commands asynchronously, intended to prevent the interface spinner from freezing. The release notes also list a change to push a development worker to a private artifact registry.
36Kr reports that some developers are renting physical Mac mini systems in the United States or Australia to access Claude Code, while others are moving Codex or Claude Code environments to overseas cloud hosts for unattended automation. The article attributes these workarounds to tighter Claude enforcement in unsupported regions and anecdotes of accounts being disabled without refunds, but provides no Anthropic statement or data establishing their prevalence. It also discusses Google Cloud credits and recommends commercial APIs, adequately resourced virtual machines, and remote development workflows.
Google AI on DEV says the on-demand “Build Agents with Claude Platform on Google Cloud” technical session is now available, with registration required. The session covers file handling, sandboxed code execution, custom skills, long-running task orchestration, tool use, production guardrails, and two deployment paths for Claude on Google Cloud. Google AI on DEV presents the Anthropic-managed path as using Google Cloud IAM, billing, and commitments, while the Google-managed path is described as supporting regional data residency, VPC Service Controls, and native integrations.
At SC26 in Chicago, Google Cloud presented an agentic HPC strategy combining purpose-built hardware, open-source orchestration through Slurm and Kubernetes, and collaborative AI agents. Google Cloud said the agents can help researchers navigate literature, debug code, narrow simulation parameters and generate design prototypes. The company also said NOAA's forecasting platform has moved from on-premises operation to Google Cloud, while the National Hurricane Center used WeatherNext to predict Hurricane Melissa's path and intensity in October 2025.
Google Developers Blog says MaxText v0.2.5 introduces Streaming DiLoCo, which lets independently located learner groups train across datacenters while reducing cross-slice communication through periodic global updates and fine-grained parameter streaming. In tests using Qwen3-32B on two TPU v5p-32 slices, Google reported about 64–65 times less cross-slice traffic than data parallelism; the worked example says the 10 Gbit/s link required roughly 1.3 Gbit/s for Streaming DiLoCo. The feature is available through MaxText v0.2.5 on PyPI, its Docker images and the GitHub main branch.
An AI Agent Clinic episode features Google Cloud engineer Laxmi Harikumar and developer Miles Libbey demonstrating memory and context management in Scribant, a voice-to-action agent built by Libbey with TypeScript and Agent Development Kit. The demonstration compares flat prompting, vector RAG, knowledge graphs, and Memory Bank in the Gemini platform. It distinguishes RAG for knowledge-base searches from memory for keeping user facts current, then applies a 40-word rolling window, a rolling summary, and Memory Bank to different time spans. In one test of about 200 spoken words, the 40-word window used about 86% fewer classifier tokens than resending the full speech stream. The rolling summary used 92 tokens, while Memory Bank stored about 24 tokens of facts compared with about 1,400 tokens to ingest the full conversation. These figures come from a single demo, not standardized benchmarks; 40 words is a suggested starting point to tune for users’ speaking patterns.
NVIDIA reported record fiscal 2027 third-quarter revenue of $115.844 billion and GAAP net income of $65.287 billion, driven by Blackwell Ultra and Vera Rubin. Data Center revenue reached $107.61 billion. TechNews reported that the shipped Vera Rubin NVL72 platform contributed nearly $20 billion in incremental quarterly revenue, a figure not independently confirmed in the supplied material. A separate report said NVIDIA shares fell after the results and that the decline also affected neocloud companies CoreWeave and IREN, but provided no detailed explanation for the market reaction.
TestingCatalog reported that “gemini-4-neon” appeared in the Gemini web app on November 16 and in Google Cloud’s SKU catalog on November 18, alongside an Antigravity CLI Pro orchestrator variant. QbitAI and 36Kr also reported the web-app identifier and a related Neon subscription tier, but Google had not announced the model or made it publicly available. Its model class, capabilities, pricing, and release timing remain unconfirmed.
A translated compilation of The Verge's Decoder podcast examines Anthropic's S-1 disclosure ahead of its reported IPO, focusing on revenue growth, inference margins, compute commitments and cloud dependencies. SemiAnalysis analyst Joey Brookhart also discusses a large reported loss tied to non-cash accounting and the competitive pressure to keep developing frontier models. The figures and filing details come from a secondary Chinese compilation rather than direct examination of the S-1 in the supplied source.
Google Cloud published 10 test results for its seventh-generation Ironwood TPU, covering frontier pretraining and what it described as the first MLPerf benchmark for agentic reinforcement learning. The company reported that 640 Ironwood TPUs reached a target model's coding-benchmark accuracy in 18.02 minutes while performing code generation and learning, with training-chip waiting time below 3%.
Papa John's is asking customers to submit unedited pizza photos and feedback for an AI quality-control campaign intended to check whether pizzas are properly cooked and cut. Eligible Papa Rewards customers must opt into text messages and place an app or website order worth at least $15. After their photo is confirmed, they can receive a free large one-topping pizza, provided they spend at least $5 on a later order. TechSpot reported that the promotion, launched on November 9 and scheduled to run through November 20, collected thousands of photos within days. The company said customer submissions could help monitor more franchised restaurants than the few hundred equipped with in-store quality-control cameras.
MLPerf Training v6.1 introduced NVIDIA’s Vera Rubin architecture and the first agentic reinforcement-learning benchmark, with Google Cloud, NVIDIA and other partners submitting results. NVIDIA said Vera Rubin NVL72 achieved the fastest training time on six of seven workloads and up to twice the performance of an optimized GB300 NVL72 baseline, while a 10,240-GPU GB300 cluster posted the fastest absolute DeepSeek-V3 671B result. Google Cloud reported that 8,192 Ironwood TPUs completed DeepSeek-V3 pretraining in about 102 seconds and that 640 Ironwood chips completed the Qwen3.5-397B agentic RL test in 18.02 minutes with 76.5% accuracy; these performance comparisons remain vendor-reported.
Cassava Technologies and Graphiant reportedly plan to offer quantum-safe networks across Africa, beginning with South African customers in the first half of 2027. The post-quantum service would encrypt data, identify the countries it traverses and restrict where AI agents can send data, according to Liquid C2 chief commercial officer Vinay Hiralall. Neither company has publicly confirmed the plan, and The Next Web did not independently verify the timeline or capabilities.
Google Developers Blog published a step-by-step guide for configuring Agent Anomaly Detection on the Gemini Platform. It covers telemetry setup, testing an ordinary session and an out-of-scope session, reviewing findings, and sending alerts through Cloud Monitoring and Security Command Center. The guide also outlines prerequisites for the Private Preview and a rollout process for production agents.
Chinese AI infrastructure firms are moving toward public markets, according to TMTPost. Infinigence AI announced a CNY 2.3 billion Series C round, while DataCanvas filed for a Hong Kong listing on September 29 and SiliconFlow began its Hong Kong IPO process in late June. TMTPost also cited media reports of a confidential Hong Kong filing by Infinigence AI, which was not publicly confirmed in the supplied report. China Academy of Information and Communications Technology data show daily token consumption in China reached about 500 trillion in June 2026, nearly four times the December 2025 level. However, disclosed figures show weak margins in token-related businesses: SiliconFlow’s 2025 revenue rose 653.2% to CNY 55.33 million while its blended gross margin fell to negative 24%, and its public-cloud segment, which includes its Serverless API business, had a gross margin of negative 119%.
Aikido Security disclosed a campaign dubbed wfx-loot in which attackers used 4,837 compromised GitHub accounts to open probe pull requests across 61,764 repositories, then abused write access to modify GitHub Actions workflows, add npm-triggered shell scripts or create new workflow files. At least 1,835 repositories reportedly executed the payload, exfiltrating GitHub tokens, AWS, Google Cloud and Microsoft Azure credentials, plus OpenAI and Anthropic API keys; the malware can also create AWS long-term access keys and Azure federated credentials, though researchers said the scale of those persistence operations is unconfirmed. StepSecurity reported that attackers reverted their repository edits and deleted workflow execution logs, usually within a minute.
Google is tailoring Kubernetes clusters to specific agentic workloads, citing the need for greater compute and memory efficiency, according to Google Cloud engineers speaking with SiliconANGLE. Its Agent Substrate project runs AI agents in sandboxes directly on Kubernetes, adds network guardrails for agent-generated code, and was donated to the Cloud Native Computing Foundation before reaching version 1.0.
Meta’s Muse Wonder, Muse Spark 1.3, and Muse Ember are now publicly available through Google Cloud’s Model Garden, according to Google AI on DEV. Google Cloud says the models can be deployed with enterprise-grade security and serverless scale, though those benefits are vendor claims and no benchmarks were provided.
Google Cloud’s 2027 cybersecurity forecast says AI agents could help security operations centers collect and analyze incident data, prioritize alerts, and summarize investigations. It warns that excessive permissions, shared service accounts, prompt injection, shadow agents, software supply-chain attacks, and theft of AI assets could widen the attack surface. The report recommends independent agent identities, restricted access, operation logging, red-team testing, and human approval for high-risk actions.
ElevenLabs unveiled ElevenAI, a preview workplace assistant for dictation, meeting notes, cross-meeting queries, live call assistance, and follow-up actions, with TechCrunch reporting that access requires early-access signup. The company also previewed enterprise fine-tuning for its speech models and open-weight LLMs, an Interaction Engine that selects models by request complexity, and ElevenAgents features for memory and pre- or post-conversation routines. It opened the ElevenLabs Marketplace for agents and applications across healthcare, financial services, insurance, education, and retail.
The Information reports that some AI inference startups are beginning to purchase their own AI chips as they struggle to secure enough cloud computing capacity for customers. Others are sourcing capacity from dozens of smaller neoclouds, though a recent hack of Fireworks highlights the security risks of this approach.
Insilico Medicine announced profitability in the first half of 2026, according to a 36Kr report. The report says 35 pipeline assets developed with its generative-AI platform, Pharma.AI, reached preclinical candidate compound status, while disclosed business-development cooperation involving its AI platform and pipeline totaled US$11 billion. In the interview, co-CEO Ren Feng said 70%–80% of the company’s code is written by AI and that AI now has substantial weight in drug-development decisions. He also said automated laboratories still require manual validation and calibration before experiments can be run, and that current technology is insufficient for a fully autonomous research loop.
Google Cloud launched Gemini for Health Care, an AI platform designed to connect fragmented information across hospitals, insurers, and other health systems. Google Cloud healthcare leader Aashima Gupta said Gemini can connect medical records, clinical guidelines, and business systems while reducing administrative work. The company also says it is building privacy, security, and HIPAA compliance into the platform.
Google Cloud launched Gemini for healthcare in US preview on November 10. The AI agent works across medical records, billing, clinical guidelines and insurance systems, with skills for tasks such as summarizing records, preparing for appointments and checking insurance benefits. Epic and Oracle Health connectors currently provide read-only access, while clinicians and insurance professionals retain final decision-making authority. The platform also provides model selection, source citations, audit records and human oversight. Launch customers include Blue Cross Blue Shield of Massachusetts, Hartford HealthCare, HCA Healthcare, Highmark Health, Northwell Health, Optum and Stanford School of Medicine. Google Cloud has not published quantified time or cost savings from these deployments.
Anthropic said in a filing that it had made new commitments with Microsoft and Google to spend roughly $10 billion more on cloud-server rentals through 2033 than it had previously disclosed. The company also said a $34.9 billion server-leasing agreement with Lambda could be cancelled if Lambda failed to build a data center by the contractual deadline, though Anthropic currently had no intention of exercising that option.
Google Cloud launched Gemini for healthcare, a product intended to unify clinical data and automate workflows. The company says it can help reduce administrative burnout.
What problem it solves
Go applications need client libraries for interacting with Google Cloud Platform APIs. The project provides packages for those services and supports common authentication flows for local development and Google Cloud Platform environments.
Key features
Google Cloud API client packages — The repository contains Go packages for Google Cloud Platform services, including storage,…
What problem it solves
Developers and enterprise teams need a terminal-based agent that can understand and modify code, automate operational tasks, and integrate additional capabilities into development workflows. Gemini CLI also needs clear access requirements and authentication options for organizations using Gemini Code Assist or supported Google Cloud services.
Key features
Code understanding …
What problem it solves
Accessing Google Workspace APIs from the command line can require repetitive request construction, manual pagination, and custom tooling for agents. The project addresses this by generating its command surface from Google Discovery Service and producing structured JSON output across Workspace services.
Key features
Dynamic Workspace command surface — The CLI reads Google's D…
What problem it solves
The repository provides reusable Agent Skills for working with Google products and technologies. It covers areas including Google Cloud, AI/ML, infrastructure, databases and analytics, application development, security, identity, and advertising.
Key features
Broad Google technology coverage — Skills are organized across product and workflow areas, including Google Cloud, AI…