英伟达下一代Rubin芯片已流片

Core Insights - Nvidia's CFO Colette Kress announced the progress of the Rubin architecture chips, which are set to enter the market with six chips already taped out [2] - Nvidia's data center revenue for Q2 FY2026 includes contributions from the Blackwell architecture, with GB200 and GB300 still being shipped in Q3 [2] - Nvidia anticipates a capital expenditure of $3 trillion to $4 trillion in data center infrastructure over the next five years, driven by the demand for AI and cloud services [3] Group 1 - The Rubin architecture chips are preparing for market entry, with significant demand already identified [2] - Nvidia's transition from the Blackwell architecture has been smooth, with ongoing high demand for computational power, particularly from inference models [2] - The company expects to maintain a GAAP gross margin of 73.3% for Q3 FY2026, reflecting effective product transitions and cost management [4] Group 2 - Nvidia's capital expenditure forecast highlights the importance of evolving computing platforms and the role of AI factories in the global industry [3] - The lifespan and replacement cycle of deployed chips remain stable, with many clients continuing to use the previous Hopper architecture due to its high performance [3] - Nvidia's supply chain management has been strengthened through long-term partnerships, ensuring the ability to keep pace with product iterations [4]