Good morning, AI enthusiast.

NVIDIA revealed a new blueprint aimed at helping nations create their own sovereign AI capabilities. The strategy provides the full-stack infrastructure and tools needed for a new era of regional industrial innovation.

This marks a shift from simply supplying hardware to becoming a foundational partner in national AI development. But does this model empower true digital sovereignty, or does it create a new form of dependence on NVIDIA's ecosystem?

In today’s AI recap:

• NVIDIA's blueprint for sovereign AI

• Google DeepMind’s new cyclone forecaster

• The startup aiming to automate office jobs

• Shrinking LLMs with quantum-inspired tech

NVIDIA's New Industrial Revolution

The Recap: At its GTC Paris keynote, NVIDIA CEO Jensen Huang unveiled a comprehensive strategy to empower Europe to build its own sovereign AI and industrial future. This blueprint moves beyond just hardware, providing the full-stack infrastructure and tools for a new wave of regional innovation.

Unpacked:

• To supercharge manufacturing, NVIDIA is launching the world's first industrial AI cloud in Germany, enabling companies to simulate and optimize factories at scale.

• The company is accelerating the development of sovereign AI models by partnering with European builders to create LLMs tailored for local languages and cultures.

• Pushing into the future of computing, the NVIDIA CUDA-Q platform for hybrid quantum-classical computing is now live on European supercomputers, advancing hybrid AI research.

Bottom line: NVIDIA is positioning itself not just as a component supplier, but as a foundational partner for national AI strategies. This approach enables countries and companies to generate their own economic value from AI, rather than just consuming it.

Google's AI Storm Chaser

The Recap: Google DeepMind has launched Weather Lab, an experimental AI that predicts cyclone paths and intensity with remarkable accuracy, and is already being tested by the U.S. National Hurricane Center.

Unpacked:

• The new model beats leading physics-based systems, improving five-day track forecasts by an average of 140 km and generating predictions in about one minute instead of hours.

• It overcomes a major forecasting challenge by simultaneously modeling both a storm's track and its intensity, a task that typically requires separate global and regional models.

• The partnership with the U.S. National Hurricane Center is a major validation, moving the AI from a research paper into an experimental tool for operational forecasters this hurricane season.

Bottom line: This marks a significant step in applying AI to solve complex, real-world scientific problems with life-or-death implications. As these models improve, they could enable earlier, more accurate warnings that help protect vulnerable communities worldwide.

Where AI Experts Share Their Best Work

Join our Free AI Automation Community

Join our FREE community AI Automation Mastery — where entrepreneurs, AI builders, and AI agency owners share templates, solve problems together, and learn from each other's wins (and mistakes).

What makes our community different:

• Real peer support from people building actual AI businesses

• Complete access to download our automation library of battle-tested n8n templates

• Collaborate and problem-solve with AI experts when you get stuck

Dive into our course materials, collaborate with experienced builders, and turn automation challenges into shared wins. Join here (completely free).

The Incredible Shrinking AI

The Recap: Spanish startup Multiverse Computing has secured $215 million for its quantum-inspired compression technology, which it claims can shrink large language models by up to 95% and slash inference costs by over 50%.

Unpacked:

• The company’s CompactifAI tech is built on tensor networks, a method that uses classical computers to mimic quantum computing principles for data compression, drawing on the co-founder's pioneering work in the field.

• Multiverse focuses on compressing popular open-source LLMs like Llama and Mistral, offering them through AWS or for on-premise licensing, but does not support proprietary models from providers like OpenAI.

• The compression is so effective that it enables powerful models to run on PCs and phones, cars, and even small devices like a Raspberry Pi, promising performance that is 4x to 12x faster than uncompressed versions.

Bottom line: This approach makes high-performance AI far more accessible by drastically reducing hardware and energy requirements. Such efficiency could unlock a new wave of AI applications that run locally on edge devices, independent of the cloud.

The Shortlist

Mattel partnered with OpenAI to integrate AI into its iconic toy brands like Barbie and Hot Wheels, with the first AI-powered products expected by the end of 2025.

Meta faces scrutiny over its new AI app, which features a public "Discover" feed filled with users' sensitive chatbot conversations, apparently shared without them realizing the chats were public.

Advocacy groups filed a complaint with the FTC, alleging that "therapist bots" from Meta and Character.AI are practicing medicine without a license and making false claims about their credentials.

Signing off,

David, Lucas, Mitchell — The Recap editorial team