TSMC Plans On Expanding Advanced Packaging Capacity To Cater NVIDIA’s High AI Chip Demand

Due to increasing demand from NVIDIA in light of the AI boom, TSMC plans to further expand its advanced packaging capabilities.

According to a DigiTimes report, the demand for chip-on-wafer-on-substrate (CoWoS) has increased at an unprecedented rate, and TSMC plans to expand its facilities to cater to the high need. TSMC has reportedly decided to supply an extra 10,000 CoWoS wafers for NVIDIA. It is expected that TSMC will increase production with an additional 1,000 to 2,000 wafers monthly. This will require advancements in existing facilities; hence TSMC has decided to take this step.

Here is what Trendforce has to say about the rapid increase in the advanced packaging demand:

Due to the generative AI trend initiated by ChatGPT, the demand for advanced packaging orders for TSMC has increased, forcing an increase in advanced packaging capacity. TSMC also pointed out that the demand for TSMC’s advanced packaging capacity far exceeds the existing capacity, and it is forced to increase production as quickly as possible. Chairman Mark Liu stated that the current investment in R&D focuses on two legs, namely 3D IC (chip stacking) and advanced packaging.

At present, three-quarters of TSMC’s R&D expenditure is used for advanced processes, and one quarter for mature and special processes, with advanced packaging falling under mature and special processes.

-TSMC Annual Shareholders Meeting

To tell you how big the AI boom has been for NVIDIA, multiple reports have suggested that the company might need help to keep up with the massive AI chip demand despite its best efforts. We reported in a previous coverage that the company has seen a ramp-up in orders for its latest GPUs, so it has started to increase production of its chips at TSMC.

The business sells customized versions of its A100 and H100 GPUs in China under the names A800 and H800 to get under US restrictions. However, these chips cost 40% more than they did in the US. China now sees such high demand for these AI GPUs that even the V100, a 2018 GPU launch and the first with Tensor Core architecture, is priced at around 10,000 US or 69,000 RMB. The NVIDIA A800 costs roughly 252,000 RMB or $36,500 USD.

Due to the massive influx in demand, the delivery cycle of the GPUs has also been drastically affected. Compared to the previous delivery cycle, which was one month, it has increased to three months and, in some cases, even six months. NVIDIA intends to work with its partners to address supply chain bottlenecks and assure quicker delivery to its clients since the enormous demand is anticipated to persist for a long time.

News Source: Tom's Hardware