A series of supply chain signals released by NVIDIA at the conference in Berlin are reshaping the landscape of the Asia-Pacific semiconductor back-end industry. Bank of America Securities believes that the core bottleneck in AI infrastructure construction has shifted from chip design to packaging and testing capacity, while the accelerated deployment of co-packaging optics (CPO) technology has opened up new growth avenues.

According to Zuifeng Trading Desk, a research report released by Bank of America Securities on October 22 indicated that NVIDIA confirmed at the GTC conference in Berlin on October 21 that Vera Rubin GPU and CPU have entered full mass production, and announced deepened cooperation with several key final testing manufacturers (OSAT), covering advanced packaging, testing, CPO, and substrates among other aspects. Among these, NVIDIA made significant capacity commitments to ASE / SPIL, KYEC, and Amkor, involving a total capital scale of over 4 billion US dollars.

The aforementioned cooperation arrangements directly address the most prominent contradiction in the current AI supply chain: insufficient back-end production capacity. Bank of America Securities pointed out that NVIDIA is accelerating the provision of equipment for consignment to OSAT manufacturers to boost production capacity, and at the same time, it is speeding up the deployment of CPO. The demand for related equipment is expected to see a turning point in growth in 2027.

NVIDIA makes a major move to secure backend production capacity

NVIDIA disclosed significant details of its collaborations at the event in Berlin. According to a report by Bank of America Securities, NVIDIA has reached agreements with ASE / SPIL, aiming to increase the production capacity of their optical engines ( OE ) by 15 times by mid-2027 and double it again by the end of 2028; it has also signed a prepayment agreement worth $1.1 billion with KYEC to support the construction of new testing facilities in the United States (expected to start production in April 2027) and to expand the testing operations in Singapore (expected to begin mass production in February 2027); furthermore, NVIDIA has signed a $3 billion agreement with Amkor with the goal of increasing the production capacity of their intermediate layers ( interposer ) in South Korea by over 300% by the end of 2028.

Bank of America Securities believes that the core logic behind the aforementioned arrangement lies in "customer fund protection and capacity expansion." NVIDIA transfers the risk of capacity expansion to itself through advance payments and consignment sales of equipment, thereby accelerating the ramp-up pace for OSAT manufacturers. The report estimates that if the capacity expansion of ASE / SPIL's optical engines progresses smoothly, this single business alone could contribute approximately $1 billion in revenue to ASE by 2028.

Test Phase: Wafer Probes and System-Level Testing Become New Foci

In the testing field, Bank of America Securities pointed out that NVIDIA's goal is to double its testing capacity, but the additional opportunities are mainly concentrated in wafer-level probe testing ( end wafer probing ) and system-level testing ( SLT ), rather than final testing ( final testing ). The latter already has sufficient capacity for the mass production of Vera Rubin and Mellanox network switching chips.

The key factor driving the rapid growth in testing requirements is the significant extension of the testing time for individual chips, rather than a simple increase in the number of tests. The report takes a cautious attitude towards KYEC's investments in the United States, noting that its new facilities in the US initially added only about 100 testing machines, and its business scale in Singapore is also relatively limited, mainly serving HPC customers who wish to reduce geopolitical risks.

CPO Acceleration: Device requirements are expected to see a turning point in 2027

Optical packaging ( CPO ) is another important topic of this GTC. Bank of America Securities stated that NVIDIA's plan to accelerate the deployment of CPO aligns highly with supply chain feedback – the demand for related devices is expected to grow tenfold by 2027, with an inflection point anticipated to occur in the fourth quarter of 2027.

NVIDIA also announced that it will diversify the packaging source of Vera CPU from Amkor / TSMC to ASE, and extend the role of ASE / SPIL from GPU board assembly to CPO process. Bank of America Securities believes that this strategic adjustment not only expands the business boundaries of ASE, but also creates new market space for tool manufacturers with the capability to supply CPO equipment. In addition to CPO, the demand for 2.5D packaging (such as CoWoS, FoCoS, EMIB-T) and 3D packaging (such as SoIC, WoW) related equipment also benefits from the capacity expansion needs of multiple ASIC / GPU / CPU / network switching chip customers.

~~~~~~~~~~~~~~~~~~~~~~~

The above wonderful content is from the Chasing Wind Trading Platform.

For more detailed interpretations, including real-time analysis and frontline research, please join [Chasing the Wind Trading Desk ▪ Annual Membership].