Top Tech Stories

OpenAI & NVIDIA Forge $100B Partnership for AI Supercomputing Infrastructure

OpenAI and Nvidia have announced a landmark partnership to build OpenAI’s next-generation AI infrastructure, deploying at least 10 gigawatts of Nvidia systems. As part of the agreement, Nvidia will make a progressive investment of up to $100 billion as each gigawatt of infrastructure is completed. This strategic move highlights the trend of AI leaders securing massive computing capacity through diverse hardware and cloud partnerships. In a related deal, OpenAI has also entered a multiyear agreement with AMD to supply up to 6 gigawatts of its MI450 GPUs, which includes an option for OpenAI to acquire a stake of approximately 10% in AMD.

Sources:

Generative Video Heats Up: Google Veo 3.1 Adds Audio, OpenAI’s Sora 2 Hits Azure

The race for advanced AI video generation is accelerating. Google has released Veo 3.1, an update to its video model that now includes synchronized audio generation for all video outputs, creating more immersive and complete experiences.

Meanwhile, OpenAI’s next-generation video model, Sora 2, is now available in public preview on Microsoft’s Azure AI Foundry. This marks the first time the Sora 2 API is accessible on a major cloud platform, signaling a move toward commercializing generative video tools. The API is priced at $0.10 per second of generated video. The new version also includes storyboard tools for greater creative control, allowing free users to generate videos up to 15 seconds and Pro users up to 25 seconds.

Sources:

Machine Learning Models Accelerate Drug Discovery and Cancer Research

Researchers are achieving significant breakthroughs by applying machine learning to scientific discovery. A new ML model from Vanderbilt University aims to solve the “generalizability gap” in drug discovery, where models often fail with new chemical structures. By focusing on the interaction space between proteins and drug molecules, the model learns the transferable principles of molecular binding.

In other research, a 27-billion parameter model from Google’s DeepMind, Gemma C2S-Scale, has successfully predicted a two-drug combination that could make “cold” tumors more responsive to immune therapies. This demonstrates that specialized, mid-sized models can be highly effective for specific scientific challenges.

Sources:

Anthropic Launches Cost-Effective Claude Haiku 4.5 and ‘Skills’ for Real-Time AI

Anthropic has launched Claude Haiku 4.5, a new compact and cost-efficient AI model optimized for real-time, low-latency applications like customer service and chat assistants. For developers, the pricing is set at $1 per million input tokens and $5 per million output tokens. Subscribers to Claude’s web and app platforms will receive access at no additional charge. Concurrently, Anthropic introduced “Claude Skills,” a feature allowing users to upload specific instructions and resources, enabling the AI to perform specialized tasks with greater accuracy and modular expertise.

Sources:

Waymo’s Self-Driving Cars Target 2026 for Fully Autonomous Rides in the UK

Waymo, an Alphabet subsidiary, has announced plans to bring its autonomous vehicle services to London, aiming for fully driverless rides to be publicly available in 2026. The initial rollout will feature vehicles with human safety drivers. The UK government is actively developing regulations to facilitate driverless cars, with a goal of starting regulated public trials by spring 2026. Waymo’s expansion will require investment in local infrastructure, including depots and charging stations. The Royal National Institute of Blind People has endorsed the initiative, highlighting its potential to improve mobility for visually impaired individuals.

Sources:

Microsoft Enters Text-to-Image Race with In-House MAI-Image-1 Model

Microsoft has unveiled its first internally developed text-to-image model, MAI-Image-1. This new generative AI tool is designed to produce photorealistic and diverse visuals, positioning it as a direct competitor to other leading image generation models. MAI-Image-1 is being integrated into Microsoft products like Copilot and Bing Image Creator to enhance creative capabilities for all users. Microsoft emphasized the high-quality training data and feedback from creative professionals as key to its performance.

Sources:

Google Cloud Expands Gemini to Enterprise for Internal Data Analysis

On the enterprise front, Google has launched Gemini Enterprise, a conversational AI platform enabling employees to interact with their company’s internal data and applications. The tool is designed to streamline data analysis, content planning, and reporting. Early adopters of Gemini Enterprise include prominent companies like Gap, Figma, and Klarna.

Sources:

Microsoft’s October 2025 Patch Tuesday Fixes 172 Flaws, Including Two Actively Exploited Zero-Days

Microsoft’s October 2025 Patch Tuesday addresses 172 security vulnerabilities across its product suite. The update includes fixes for four zero-day flaws, two of which are being actively exploited. The patches primarily target elevation of privilege and remote code execution vulnerabilities. Critical fixes include a privilege escalation flaw in the Windows Remote Access Connection Manager (CVE-2025-59230) and remote code execution flaws in Microsoft Office and Excel (CVE-2025-59234 and CVE-2025-59236). The update also patches critical flaws in Azure Container Instances and Compute Gallery and a significant vulnerability in the Windows Server Update Service (WSUS) that could be used in supply-chain attacks.

Sources:

Oracle and Microsoft Announce Massive AI Cloud Infrastructure Expansions

Oracle is partnering with AMD and Nvidia to enhance its Oracle Cloud Infrastructure (OCI) for growing AI demands. Oracle will deploy 50,000 AMD Instinct MI450 Series GPUs on OCI, starting in Q3 2026. In a separate deal, Nscale will supply Microsoft with approximately 200,000 NVIDIA GB300 GPUs for hyperscale AI infrastructure in Europe and the U.S., providing sovereign AI solutions for European customers. Oracle also introduced OCI Zettascale10, a large-scale AI supercomputer with Nvidia, which will be a key part of the Stargate infrastructure initiative with OpenAI.

Sources:

Kubernetes v1.35 Development Reaches Enhancements Freeze Milestone

The development of Kubernetes version 1.35 has reached its Enhancements Freeze milestone on October 16th, 2025. This deadline locks in the features and improvements planned for the upcoming release. The release cycle, which began on September 15, is scheduled for final release on December 17, 2025. The next major deadline is the Code and Test Freeze on November 6, 2025.

Sources:

Quantum Computing Breakthrough: New Error Correction Method Slashes Errors by 100x

A major breakthrough in quantum computing could reduce errors by up to 100 times, significantly advancing the timeline for stable quantum computers. The results, published in Nature, are based on tests on a simulated neutral-atom quantum computer. This advancement tackles the critical challenge of high error rates, a primary obstacle in developing fault-tolerant quantum machines, and could lead to smaller, faster quantum computers.

Cisco Launches Project CodeGuard to Secure AI-Generated Code

Cisco has announced Project CodeGuard, an open-source framework designed to improve the security of software code written by artificial intelligence. As AI becomes more prevalent in software development, this initiative addresses the associated security concerns. Project CodeGuard provides tools and methodologies to help developers identify and mitigate vulnerabilities in AI-generated code, promoting more secure coding practices.

Sources:

2025 Shenyang Robot Conference Kicks Off in Northeastern China

The 2025 Shenyang Robot Conference commenced on Friday in Shenyang, the capital of northeast China’s Liaoning Province. The two-day event serves as a platform for showcasing the latest advancements in robotics and intelligent systems.

Sources: