Nvidia’s GTC 2026 keynote from Jensen Huang could mark a major turning point in the AI hardware race. The company is preparing to transition from its Blackwell AI platform to a new architecture known as Vera Rubin. The move represents a significant technological leap that could reshape the next phase of artificial intelligence infrastructure. Rubin is expected to feature a 1.6nm semiconductor process (A16) combined with HBM4 high-bandwidth memory, designed to support what Nvidia calls “Agentic AI.” Unlike traditional generative AI systems, agentic AI focuses on autonomous systems capable of reasoning, decision-making and real-time action. However, the aggressive roadmap also introduces new risks. Accelerating the chip cycle could potentially devalue existing Blackwell-based AI clusters that enterprises and hyperscalers are currently deploying. Key tensions investors are watching at GTC: • The shift from Generative AI to Agentic AI • Whether Nvidia can maintain 70%+ gross margins...