Nvidia CEO Jensen Huang denied reports that the company’s next-generation Rubin Ultra AI platform has been delayed, confirming at a private meeting with Morgan Stanley that the platform will be released as planned and product shipments are moving ahead normally. His comments reduced worries among investors and suppliers, giving confidence to Taiwanese companies including Delta Electronics, Lite-On, Poroway, and CSOT, which are expected to supply parts for the platform. The Rubin platform is a big improvement over Nvidia’s current Blackwell platform, designed to lower the cost of AI inference by ten times while providing up to 50 petaflops of NVFP4 computing power. Rubin Ultra will further improve AI performance by using 800VDC power technology and CPO optical communication. Delta Electronics said its work with customers on 800VDC is progressing well, with some customers already buying products, though larger orders will depend on future demand. Lite-On expects strong growth in 800VDC sales next year as more AI data centers adopt the technology. CPO technology is expected to enter small-scale production in the second half of this year for testing by cloud service providers, and if successful, will be widely used in Rubin Ultra and Nvidia’s future Feynman platform.