TSMC previously announced in November 2021 that it plans to establish two wafer fabrication plants for the 7nm and 28nm processes in Kaohsiung, a southern city of Taiwan. Construction was set to begin in 2022, with official production expected to commence in 2024. However, following the announcement, there have been changes in the progress of the Kaohsiung plant development.

Firstly, there were reports of adjustments to the 7nm plant by the end of 2022, in response to weakened demand in the smartphone and PC markets. Subsequently, there were also reports of changes to the 28nm plant’s plans.

It wasn’t until TSMC’s Q1 2023 earnings conference that they officially announced the adjustment of the Kaohsiung 28nm plant’s construction plans, focusing on capacity enhancement for more advanced process technologies.

At the time, TSMC didn’t specify the exact advanced process that would be introduced, only emphasizing that the construction of the wafer fab would proceed as planned. This triggered market speculation that TSMC was likely to adopt the advanced 2nm process technology at the Kaohsiung plant, in response to the rapidly growing demand in the artificial intelligence market.

This week, TSMC confirmed that the Kaohsiung plant will adopt the 2nm process technology. TSMC stated that the construction of the wafer fab in Kaohsiung will proceed as usual, but the previous expansion plans will be adjusted to accommodate the production of the 2nm advanced process technology, in response to strong market demand for advanced processes.

As for the specific details and contents of the plant development, they have not been further disclosed at this stage. According to TSMC’s plans, mass production of the 2nm process is expected to begin in 2025, with production bases including the previously announced Hsinchu and Taichung facilities, as well as the newly announced Kaohsiung facility.

(Photo credit: TSMC)

SK Hynix, a semiconductor leader, announced a significant breakthrough in NAND flash technology on August 9 at the Flash Memory Summit in California. They introduced a pioneering 321-layer 1Tb TLC 4D NAND flash memory, marking a notable advancement in NAND capabilities.

As the first company in the industry to surpass 300 layers in NAND development, SK Hynix plans to refine the 321-layer NAND for mass production by mid-2025. This leap builds upon their expertise gained from 238-layer NAND production, showcasing their commitment to pushing technological boundaries.

The efficiency of the 321-layer NAND is 59% higher than its 238-layer predecessor, attributed to enhanced data storage unit stacking. This translates to increased storage capacity and improved chip output per wafer.

In response to the growing demand for high-performance memory solutions in the AI market, SK Hynix unveiled next-gen NAND products, including PCIe 5-equipped enterprise SSDs and UFS 4.0 solutions, catering to the increasing need for advanced storage.

SK Hynix’s dedication to innovation is further emphasized by its ongoing development of PCI 6.0 and UFS 5.0 products, solidifying its market leadership position. Meanwhile, VP of SK Hynix NAND Flash Development, affirmed their commitment to leading the NAND technology domain with the development of fifth-gen 321-layer 4D NAND, tailored to the AI era.

(Source: https://mp.weixin.qq.com/s/OJ3_hw2jQuXT5YGiX0K8-Q)

DRAM Spot Market
Compared with last week, there are signs that the decline in DRAM spot prices is easing, but there has not been a noticeable increase in overall transaction volume. The spot market still has an abundant supply of DDR4 products mainly due to the influx of rebelled chips from decommissioned server DRAM modules. As a result, DDR4 products are under greater pressure to sell compared with DDR5 products. Currently, consumer 1Gx8 DDR4 chips are priced around US$0.88-0.9, while those from CXMT are priced around US$1. DDR4 chips from other suppliers are mostly being sold to customers in the industrial equipment segment, resulting in higher spot prices of around US$1.4. DDR5 products have relatively stable spot prices as they belong to the new generation and their supply has not been affected by recalled products. The average spot price of DDR5 2Gx8 chips from suppliers is now above US$4. In sum, DRAM spot prices are not showing signs of a rebound in the near future, mainly due to sluggish demand and an excess supply of DDR4 products. The average spot price of mainstream chips (DDR4 1Gx8 2666MT/s) dropped by 0.41% from US$1.468 last week to US$1.462 this week.

NAND Spot Market
Peak season effects are yet to be seen from the demand end, where NAND Flash suppliers’ successive increase of quotations has not ramped up stocking dynamics as some module houses have long raised their inventory, and the retail market is experiencing a poor level of actual transactions without aggressive price negotiations. Prices are maintained at a consolidation this week. 512Gb TLC wafer spots have risen by 1.06% this week, arriving at US$1.436.

Leading semiconductor companies TSMC, Robert Bosch GmbH, Infineon, and NXP Semiconductors have jointly to invest in the European Semiconductor Manufacturing Company (ESMC) GmbH, situated in Dresden, Germany. This strategic move aims to bolster the region’s semiconductor manufacturing capabilities, particularly catering to the burgeoning automotive and industrial sectors. The establishment of ESMC marks a significant stride towards the realization of a 300mm fabrication facility, pending the final decision on public funding, as part of the European Chips Act framework.

The planned fab is expected to have a monthly production capacity of 40,000 300mm (12-inch) wafers on TSMC’s 28/22 nanometer planar CMOS and 16/12 nanometer FinFET process technology, further strengthening Europe’s semiconductor manufacturing ecosystem with advanced FinFET transistor technology and creating about 2,000 direct high-tech professional jobs. ESMC aims to begin construction of the fab in the second half of 2024 with production targeted to begin by the end of 2027.

The prospective joint venture will see TSMC holding a substantial 70% ownership stake, while Bosch, Infineon, and NXP will each possess a 10% equity share, contingent upon regulatory approvals and meeting specific conditions. Total investments exceeding 10 billion euros are anticipated. Operational oversight of the fabrication facility will reside under TSMC’s purview.

However, industry analysts at TrendForce have highlighted potential challenges that lie ahead for TSMC’s groundbreaking endeavor. One such challenge pertains to the looming labor shortage issue in TSMC’s US fabrication facility, which is projected to reverberate globally. Moreover, navigating the intricacies of implementing subsidy policies in accordance with the European chip legislation and anticipated administrative procedures is expected to introduce a layer of complexity to the venture.

(Photo credit: TSMC)

Charlie Boyle, Vice President of NVIDIA’s DGX Systems, recently addressed the issue of limited GPU production at the company.

Boyle clarified that the current GPU shortage is not a result of NVIDIA misjudging demand or constraints in Taiwan Semiconductor Manufacturing Company’s (TSMC) wafer production. The primary bottleneck for GPUs lies in the packaging process.

It’s worth noting that the NVIDIA A100 and H100 GPUs are currently manufactured by TSMC using their advanced CoWoS (Chip-on-Wafer-on-Substrate) packaging technology. TSMC has indicated that it may take up to a year and a half, including the completion of additional wafer fabs and expansion of existing facilities, to normalize the backlog of packaging orders.

Furthermore, due to the significant strain on TSMC’s CoWoS capacity, there have been reports of overflow of NVIDIA GPU packaging orders to other manufacturers.

Sources familiar with the matter have revealed that NVIDIA is in discussions with potential alternative suppliers, including Samsung, as secondary suppliers for the 2.5D packaging of NVIDIA’s A100 and H100 GPUs. Other potential suppliers include Amkor and the Siliconware Precision Industries Co., Ltd. (SPIL), a subsidiary of ASE Technology Holding.

In December 2022, Samsung established its Advanced Packaging (AVP) division to seize opportunities in high-end packaging and testing. Sources suggest that if NVIDIA approves of Samsung’s 2.5D packaging process yield, a portion of AI GPU packaging orders may be placed with Samsung.

TrendForce’s research in June this year indicated that driven by strong demand for high-end AI chips and High-Bandwidth Memory (HBM), TSMC’s CoWoS monthly capacity could reach 12,000 units by the end of 2023. Particularly, demand from NVIDIA for A100 and H100 GPUs in AI servers has led to nearly a 50% increase in CoWoS capacity compared to the beginning of the year. Coupled with the growth in demand for high-end AI chips from companies like AMD and Google, the second half of the year is expected to witness tighter CoWoS capacity. This robust demand is projected to continue into 2024, with advanced packaging capacity potentially growing by 30-40% if the necessary equipment is in place.

(Photo credit: NVIDIA)