Nvidia CEO Jensen Huang took the stage at Computex 2026 to unveil a trio of AI models targeting distinct sectors, from autonomous vehicles to enterprise efficiency. The announcements included the Alpamayo 2 Super, described as the company's most powerful open AI model for robotaxis, and the Nemotron 3 Ultra, an enterprise-grade model. Additionally, the company introduced Cosmos 3, an open physical AI foundation model designed to help robots and autonomous cars understand the real world with limited training data.
The Alpamayo 2 Super could accelerate innovation in autonomous vehicles by fostering collaboration and potentially transforming urban mobility, according to the company. Meanwhile, Cosmos 3 is aimed at helping physical systems better predict real-world environments. These releases signal Nvidia's deepening commitment to open AI in robotics and self-driving technology.
Alongside the software announcements, Huang outlined the DGX Spark roadmap for laptops and desktop PCs, spanning three generations. The first-generation RTX Spark platform was detailed, with future iterations codenamed Rubin (featuring LPDDR6 memory) and Rosa Feynman. The hardware roadmaps suggest Nvidia is extending its AI compute capabilities beyond data centers into consumer devices.
The Nemotron 3 Ultra model may shift investor focus toward traditional tech applications rather than crypto ventures, analysts noted. These launches collectively position Nvidia to capture growth across autonomous driving, enterprise AI, and edge computing — potentially making its technology more accessible to developers and researchers.
By open-sourcing key models like Alpamayo 2 Super and Cosmos 3, Nvidia is betting on community-driven innovation. However, competitors like Tesla and AMD are also advancing their own autonomous driving and AI hardware platforms, which could limit Nvidia's market share in specialized niches.