TLDR IT 2026-10-23

OpenAI and Google Cloud are partnering to bring OpenAI frontier models and Codex agentic workflows into Google Cloud's Model Garden. Customers will be able to purchase the models through the Google Cloud Marketplace and draw down GCP commitments, with general availability planned in the coming weeks. Microsoft remains OpenAI's primary cloud partner.

Google Cloud and NVIDIA announced the general availability of infrastructure powered by the NVIDIA Vera Rubin architecture, featuring A5X instances designed for large-scale reinforcement learning and agentic inference. Additionally, Google Cloud plans to offer NVIDIA Confidential Computing at rack scale on bare metal instances and expand confidential capabilities across virtual machine offerings.

DeepSeek reportedly raised approximately $15 billion in one of China's largest funding rounds, with Tencent and battery manufacturer CATL among the backers. The funding strengthens its push to compete with OpenAI and Anthropic as the company prepares for a potential IPO and faces growing US restrictions over security concerns.

This comparison examines forecasts from 12 research and financial firms, explaining how different definitions produce vastly different estimates of AI infrastructure spending. It connects those projections to hyperscaler budgets, power availability, and the revenue needed to justify the buildout.

GitHub has placed copies of 99.99999% of its repositories in Azure as agent-driven traffic and data center capacity constraints accelerate its infrastructure migration. This engineering breakdown explains how moving write coordination from entire repositories to individual references reduces contention, delivering 3–4x higher throughput in internal tests.

Can you answer "why did users do that?" without filing a ticket? Is your A/B testing fast and rigorous enough? How long does it take you to root-cause a metric drop? Get the Datadog tech audit checklist to find gaps in your product stack before they slow your next launch. Download the checklist

Anthropic has announced a research preview of Claude Duet, a new proactive agent app designed for professionals and small business owners. Built around Claude Opus 5.5, the app operates independently in the background via a virtual machine with its own browser to manage tasks, send daily briefs, and monitor connected apps. The application includes robust security frameworks and safety classifiers to safeguard sensitive user information before performing actions on their behalf.

ServiceNow and Nvidia have launched Arc, a personal long-running AI agent built for enterprise use. Designed to run locally on Nvidia RTX Spark devices or within Nvidia OpenShell, Arc assists with daily tasks using productivity tools like email and Microsoft Teams. The companies emphasize a multi-level security framework that utilizes ServiceNow's Super Intelligence Control Tower, Nvidia OpenShell, and AI Passport to enforce runtime and hardware-layer policies and manage agent permissions.

Microsoft has announced that it will retire its native SMS and voice authentication for Entra ID on February 1, 2027, for most customers, with the deadline extended to July 1, 2027, for Global Administrators and external users. Starting October 30, 2026, organizations will be able to choose a third-party telephony provider through the Microsoft Security Store. Microsoft recommends that organizations transition to phishing-resistant passkeys instead of relying on telephony-based authentication.

Crusoe and Araico plan to build AI cloud infrastructure powered by Paraguay's Itaipú hydroelectric dam, with an initial deployment targeted for mid-2027. The project has a path to 106 MW of IT capacity and will serve both US customers and demand across Latin America.

The company advises security teams to measure mitigation timelines in hours, investigate from the earliest known exploitation date, and invest in signature-independent detections.

Through October 26, GitHub is rolling out enforcement that sets eligible, unconfigured Copilot features according to an organization's global default policy while preserving explicit administrator choices.