Morgan Stanley stated,$NVIDIA (NVDA.US)$The next generation Rubin GPU has started preparation half a year ahead of schedule, and the expected launch time will be moved up from the first half of 2026 to the second half of 2025. With the use of 3nm technology, CPO (Common Packaging Optics), and HBM4 (6th generation high-bandwidth memory), the chip area of the new GPU generation will be twice as large as the previous Blackwell generation, benefiting the related industry chain companies.

Analyst Charlie Chan from Morgan Stanley stated in a report on December 2 that although the output of the Blackwell chip is still growing, considering the complexity of the new chip,$Taiwan Semiconductor (TSM.US)$and the supply chain is already preparing for the launch of Nvidia's next generation Rubin chip. The 3nm Rubin GPU is expected to enter the tape-out phase in the second half of 2025, about half a year earlier than previously scheduled.

Morgan Stanley indicates that due to the migration to 3nm process technology, CPO, and HBM4 adoption, Rubin will be a powerful chip. According to the supply chain survey, the chip size of Rubin will be nearly twice that of Blackwell, with Rubin chips potentially incorporating four compute chips, double that of Blackwell.

Analysts currently expect that due to the increase in chip size, TSMC will further expand its CoWoS (Chip on Wafer Substrate) capacity in 2026. The company expects to start issuing 2026 orders to equipment suppliers in mid-2025, depending on the sustainability of AI capital expenditures.

In terms of the supply chain, analysts believe$ASMPT (00522.HK)$There is a slight lag in Rubin's tcb tool certification. TCB (Thermo-Compression Bonding) is the process required for Rubin GPU's Chip on Wafer (CoW). K&S previously announced that it had received taiwan semiconductor's solder-free tcb order for CoW in November, while ASMPT stated that it is still in communication with taiwan semiconductor for CoW certification.

Morgan Stanley believes that there is still an opportunity for ASMPT to complete the certification by the end of this year, and the low-volume equipment procurement for the company may be achieved in the second half of 2025. Industry surveys indicate that TSMC's initial orders for non-solder TCB for CoW through K&S involve fewer tools, with fewer than 10 expected in 2025. As the Rubin GPU is expected to enter mass production in 2026, the order volume may increase in 2026.

Another structural opportunity brought by the new GPU is the extension of testing time. Analysts at Morgan Stanley indicate that currently Blackwell (three times longer testing time than Hopper) and MI355 (twice as long) both require longer testing times. Due to strong demand, the delivery time for the new testing equipment has been extended from 3 months to 6 months.$Advantest (6857.JP)$From 3 months to 6 months.

In the long term, some AI dedicated integrated circuits (ASICs), such as AWS's 3nm AI accelerator, may start burn-in testing. According to the supply chain survey, all final testing of Blackwell will continue at KYEC. It is expected that by 2025, the shipment volume of B200/300 (dual-chip version) could reach approximately 5 million units based on taiwan semiconductor's CoWoS-L capacity.

Editor/Rocky