Your job called—it wants better business news

Welcome to Morning Brew—the world’s most engaging business newsletter. Seriously, we mean it.

Morning Brew’s daily email keeps professionals informed on the business news that matters, but with a twist—think jokes, pop culture, quick writeups, and anything that makes traditionally dull news actually enjoyable.

It’s 100% free—so why not give it a shot? And if you decide you’d rather stick with dry, long-winded business news, you can always unsubscribe.

Check it out

Today:

• Nvidia Announces New AI Superchips

• Gemini Adds Canvas, Audio Overview

• Llama Hits 1 Billion Downloads

• NVIDIA Opens Robotics Training Data

• Amazon Slashes Trainium AI Chip Price

Nvidia Announces New AI Superchips

Nvidia unveiled the Blackwell Ultra GB300 and Vera Rubin AI chips, aiming to extend its lead in AI computing. The Blackwell Ultra offers faster processing and increased memory compared to its predecessor. Vera Rubin, coming in 2026, will triple performance, with Rubin Ultra doubling it further by 2027. Nvidia’s CEO emphasized that AI demand is growing exponentially, reinforcing the company's push for ever-more powerful chips to sustain the AI revolution.

Why It Matters:

1. Faster AI Processing – These chips boost AI model efficiency, making complex computations quicker and more scalable.

2. Increased Memory Capacity – Larger memory allows bigger AI models, improving accuracy and capability.

3. Industry Shift – Nvidia’s dominance sets the pace for AI hardware, influencing AI research and business adoption worldwide.

Gemini Adds Canvas, Audio Overview

Google's Gemini introduces Canvas, an interactive space for editing documents and coding in real time, and Audio Overview, which turns files into podcast-style discussions. Canvas enables seamless content creation and coding collaboration, while Audio Overview enhances learning by summarizing materials in engaging conversations. These features, available for Gemini subscribers, streamline workflows and boost productivity, making AI a more effective creative and educational tool.

Why It Matters:

1. Improved Human-AI Collaboration – Canvas enables real-time editing and feedback, making AI a more effective co-creator.

2. Enhanced AI Accessibility – Audio Overview transforms complex text into interactive audio, making information easier to absorb.

3. Advancing AI in Coding – Canvas simplifies code generation, debugging, and prototyping, accelerating software development.

Llama Hits 1 Billion Downloads

Meta’s open-source AI model, Llama, has surpassed 1 billion downloads, reinforcing the impact of open AI accessibility. Companies like Spotify use Llama for personalized music recommendations, while Fynopsis applies it to streamline M&A deals. Developers like Srimoyee Mukhopadhyay utilize Llama for cultural discovery apps. This milestone highlights Llama’s role in driving innovation across industries, proving that open-source AI fosters creativity, economic growth, and technological advancement.

Why It Matters:

1. Democratizing AI Access – Open-source models enable widespread AI adoption, fostering innovation globally.

2. Diverse Real-World Applications – From music recommendations to business deals, Llama enhances various industries.

3. Encouraging AI Transparency – Open models like Llama promote trust, security, and community-driven advancements.

🧠RESEARCH

DropletVideo: A Dataset and Approach to Explore Integral Spatio-Temporal Consistent Video Generation

DropletVideo is a dataset and model designed to improve video generation by ensuring smooth storytelling and consistent visuals across camera movements. It includes DropletVideo-10M, a dataset of 10 million videos with detailed captions. The model enhances coherence in generated videos, balancing plot progression with dynamic camera motion.

Being-0: A Humanoid Robotic Agent with Vision-Language Models and Modular Skills

Being-0 is a humanoid robot that combines a vision-language model with modular skills for better real-world performance. A Connector module translates language-based plans into actions, improving navigation and object handling. Unlike past approaches, Being-0 runs efficiently on low-cost hardware, excelling in complex indoor tasks.

DreamRenderer: Taming Multi-Instance Attribute Control in Large-Scale Text-to-Image Models

DreamRenderer is a method that improves control over multiple objects in AI-generated images. Built on the FLUX model, it prevents visual mix-ups by using specialized techniques to link text descriptions with specific image regions. Tests show a significant boost in accuracy and quality over existing models.

🛠️TOP TOOLS

GoEnhance - Create AI animated short in Minutes

Face26 - Convert your old, blurry, and low-quality photos into vivid, high-definition portraits, colored images, or animated photos.

BigSpeak AI - AI-Powered Voice Generation and Content Creation

Nightcafe AI - Create amazing artworks in seconds

Caveduck - A platform where users can create, customize, and interact with characters in different scenarios.

📲SOCIAL MEDIA

🗞️MORE NEWS

• NVIDIA released an open-source AI dataset to train robots and autonomous vehicles. The dataset includes 15 terabytes of robotics data and traffic scenarios. Researchers can use it for AI development, improving safety and performance in real-world applications.

• Amazon is aggressively discounting its AI chip, Trainium, offering it at 25% of Nvidia’s H100 chip price. This move aims to reduce dependence on Nvidia, boost AWS margins, and compete in the growing AI cloud computing market.

• Adobe has introduced AI agents to its Experience Platform, providing prebuilt agents to assist with audience optimization, content production, customer journey management, and more. The company also announced partnerships with AWS, Microsoft, SAP, and others to enhance AI model interoperability.

• Stability AI’s new model, Stable Virtual Camera, turns 2D images into realistic 3D videos. It offers customizable camera movements but struggles with complex scenes. The model is free for research. Stability AI faces financial and leadership challenges.

• Nvidia unveiled GR00T N1, an AI model for humanoid robots, at GTC 2025. It enables adaptable learning and reasoning, powering robots from companies like Boston Dynamics. Developers can customize it using human demonstrations and simulations.

• Google is launching TxGemma, a set of AI models for drug discovery, to improve research efficiency. These models analyze chemicals and proteins to predict drug properties. Researchers can access them through Google’s Health AI Developer Foundations program.

• Oracle has secured a deal to provide Singapore's defense ministry with "air-gapped" cloud computing and AI services, marking its first such agreement in Southeast Asia. The infrastructure will be isolated from the internet, offering encrypted AI-driven capabilities, including video analysis for facial recognition and vehicle identification.