The Nvidia story reads like a Silicon Valley fairy tale. The company, which already dominated the gaming graphics card market, was able to capitalize on the meteoric rise of generative artificial intelligence. Its GPUs, initially designed for video gameshave proven to be ideally suited to training AI models. This unique position has propelled Nvidia to become the world’s largest capitalization, with a valuation reaching $3.7 trillion at the start of 2025.
Project DIGITS: the democratization of AI power
Jensen Huang saved the most impressive surprise for the end of his keynote: the DIGITS Project. This compact personal supercomputer embodies Nvidia’s desire to democratize access to the computing power necessary for the development of AI. Equipped with the new GB10 Grace Blackwell superchip, it delivers computing power of a petaflop, while requiring just a standard electrical outlet.
The GB10 represents a major technological feat. This system on chip (SoC) combines a latest generation Blackwell GPU with a 20-core Grace CPU based on the Arm architecture. Collaboration with MediaTek, leader in Arm SoC designs, has optimized the energy efficiency and performance of this revolutionary chip.
With 128 GB of unified memory and up to 4 TB of NVMe storage, Project DIGITS allows developers to run AI models with up to 200 billion parameters. Even more impressive, connecting two units via NVIDIA ConnectX pushes that limit to 405 billion settings.
A complete software ecosystem for developers
There’s more to Project DIGITS than its raw power. Nvidia supports it with a complete software suite including DGX OS, development SDKs, orchestration tools and access to models from the NGC catalog. Developers can thus exploit the NeMo framework for fine-tuning models, accelerate data processing with RAPIDS libraries, or use common frameworks like PyTorch.
The integration of NVIDIA Blueprints and NIM microservices paves the way for the development of autonomous AI agents. These tools, available through Nvidia’s developer program, allow you to create and test AI applications in a controlled environment before their deployment in production. With a price of 3000 dollars, Nvidia wants to open the door to AI to the general public.
Cosmos: the physics AI revolution
The Cosmos platform marks Nvidia’s commitment to the promising field of physical AI. Jensen Huang compares this moment to the impact that large language models have had on generative AI. Cosmos combines generative models and a video processing pipeline optimized for robots and autonomous vehicles.
Cosmos’ ability to simulate and evaluate different future scenarios represents a major breakthrough for the industry. Models can ingest texts, images or videos to generate virtual states of the world, favoring realistic environments and the coherence of objects.
Mass adoption by the automotive industry
Toyota, the world’s leading automobile manufacturer, has chosen to build its next vehicles on the NVIDIA DRIVE AGX Orin platform. This strategic decision confirms Nvidia’s dominant position in the autonomous vehicle sector. The list of manufacturers adopting Nvidia solutions is growing: Mercedes-Benz, BYD, Lucid, Volvo Cars, and many others.
The integration of Cosmos with Nvidia’s Omniverse revolutionizes training of autonomous systems. The generation of synthetic data makes it possible to exponentially multiply learning scenarios, thus accelerating the development of safer and smarter vehicles.
A long-term vision for industrial AI
“ Physical AI will revolutionize the $50 trillion manufacturing and logistics industries “, says Jensen Huang. “ Everything that moves — from cars to factories to warehouses — will be robotized and powered by AI. » This ambitious vision explains Nvidia’s massive investments in platforms like Cosmos and the Omniverse. Suffice it to say that with this vision of the future and the general interest in AI, Nvidia has every chance of remaining the most valuable company in the world for years to come.
- Nvidia becomes the world’s largest capitalization with $3.7 trillion, driven by the explosion of AI
- Project DIGITS democratizes access to AI supercomputers with petaflop power starting at $3,000
- The Cosmos platform and massive adoption by the automotive industry confirm Nvidia’s leadership in physical AI
