Nvidia Ramps GB300 Output and Sends Six-Chip Rubin Platform to TSMC

Nvidia Corp. said the manufacturing ramp for its latest GB300 rack-based architecture is running at full speed, with output back to roughly 1,000 racks a week after a recent transition. The company expects production volumes to rise further during the third quarter as additional capacity comes online. Chief Executive Officer Jensen Huang added that the next-generation Rubin platform—Nvidia’s third-generation NVLink Rackscale AI supercomputer—is already in fabrication. All six chips that underpin Rubin have taped out at Taiwan Semiconductor Manufacturing Co., the CEO said. Chief Financial Officer Colette Kress said Rubin remains on schedule for volume production next year, keeping the company on its target of releasing a major platform annually. Nvidia emphasized that the Rubin supply chain is “much more mature and fully scaled,” positioning the firm to sustain its rapid cadence of AI hardware launches.