Overview

• The 'Blackwell Ultra' chip, set for release in late 2025, and 'Vera Rubin,' arriving in 2026, aim to reduce operational costs and boost computational efficiency for AI applications.
• Nvidia's new software platform 'Dynamo' will optimize server utilization, making inference calculations up to 25 times faster on Blackwell chips.
• Jensen Huang emphasized the need for 100 times more AI computing power due to advancements in reasoning-based AI models, which require significantly greater resources.
• The company introduced the concept of 'AI factories,' predicting industries will operate both physical and digital production hubs in the future.
• Despite innovations, Nvidia faces growing competition from AMD and in-house chip development by major tech firms, with its stock dropping 3.43% following the announcements.