The latest official data from South Korea, the world’s largest memory production country, indicates a turnaround in October’s memory exports, marking the first growth in nearly 16 months. Meanwhile, the latest report from the Taiwan Semiconductor Industry Association (TSIA) anticipates a quarter-on-quarter growth of 9.1% in Taiwan’s memory and other manufacturing output for this season, leading the semiconductor manufacturing industry.

These signals suggest a recovery in the memory market. Key players in Taiwan’s memory industry, such as Nanya Technology and Winbond, are expected to see significant operational momentum.

South Korea stands as the world’s largest memory manufacturing country, with Samsung and SK Hynix ranking as the top two global DRAM manufacturers, collectively holding a market share of approximately 70%. They also play a significant role in NAND Flash storage memory, with Samsung leading the global market and SK Hynix being among the top five manufacturers. The shift in South Korea’s memory export figures to growth signifies a stabilization in global memory demand, making it a crucial indicator for the industry’s recovery.

The South Korean Ministry of Trade, Industry, and Energy announced on November 14th that in October, South Korea’s memory exports increased by 1% compared to the same period last year, a significant improvement from the 18% YoY decline in September.

Notably, the export rebound was led by a 12.2% growth in semiconductor packaging, and the YoY decline rate in DRAM exports also narrowed to a single digit, marking the first time in over a year.

The TSIA also cited data from the Industrial Technology Research Institute (ITRI) International Division yesterday, estimating an 8% quarterly increase in Taiwan’s foundry production value in the fourth quarter. Additionally, the memory and other manufacturing sectors are expected to see a 9.1% quarterly increase, leading all semiconductor manufacturing industries. This underscores the ongoing recovery in the memory industry.

Taiwanese memory manufacturers are optimistic about the future market outlook. Nanya Technology’s General Manager, Pei-Ing Lee, analyzing the demand side, Lee notes that the server market displayed relative weakness in the first three quarters of the year. However, with the increasing demand in cloud services, a recovery is foreseen in the fourth quarter. In the mobile phone market, there are signs of a rebound in demand in the Chinese market after experiencing a period of weakness for at least a year. Additionally, the PC market is experiencing a positive impact from the rising demand for DDR5, leading to a comparatively healthy performance.

Winbond Electronics is also expressing positive sentiment towards the upcoming market developments. Since the fourth quarter of last year, Winbond Electronics has been adjusting production in response to market weakness, initially announcing a 40% production cut. As demand improves, the recent reduction in production has been reduced to 20%, and the shipment volume is expected to increase in the fourth quarter. Prices are anticipated to remain stable or experience a slight decline, with overall operations not expected to be worse than the third quarter, offering a hopeful outlook for slight improvement.

Winbond Electronics anticipates that if terminal demand remains similar to or consistent with this year in 2024, with manufacturers having depleted their inventories, purchasing power will rebound.

The company’s analysis suggests that this trend will be observed across various markets, including PCs, laptops, and smartphones. Additionally, the server domain is expected to return to growth as well.

TrendForce reports indicate a price increase for both DRAM and NAND Flash starting in the fourth quarter. DRAM prices, for instance, are projected to see a quarterly surge of about 3-8%.

(Photo Credit: Link)

Read more

• Q4 DRAM Contract Prices Set to Rise, with Estimated Quarterly Increase of 3-8%, Says TrendForce

Despite the U.S. sanction on the semiconductor industry in China, China is still positively looking for further development.

Following our discussion on the shifts of Chinese wafer fabs in the previous article (China’s Wafer Fabs Hits 44 with Future Expansion 32, Mainly Targeting on The Mature Process), this article focuses on the application of 8-inch and 12-inch wafers, as well as provides detailed account of Chinese foundries’ strategic positioning in the landscape.

From a cost perspective, producing a 12-inch wafer incurs approximately 50% more cost than an 8-inch wafer. However, the chip output from a 12-inch wafer is nearly triple that of an 8-inch wafer, leading to a cost reduction of roughly 30% per chip. As manufacturing processes improve and yields increase, the cost of 12-inch wafers is expected to further decline in the future.

In terms of applications, a clear distinction emerges between 12-inch and 8-inch wafers. The versatility of 12-inch wafers is evident, covering a broad range of practical applications. As depicted in the table below, 8-inch wafers are primarily utilized for mature and specialized processes, focusing on the 0.13-90nm range.

8-inch wafers’ downstream applications are concentrated in industrial, mobile, and automotive sectors, encompassing power devices, power management chips, non-volatile memory, MEMS, display driver ICs, and fingerprint recognition chips, among others. The surging demand for power devices in automotive electronics and industrial applications has been a key driver for recent production expansions.

Currently, the market demand for 8-inch wafers remains robust. However, there is a trend of decreasing 8-inch production lines in the market. This shift is primarily due to the industry’s mainstream adoption of 12-inch wafers. Given the significant capital required to establish 12-inch wafer fabs (often exceeding billions of US dollars), many foundries are reevaluating their 8-inch wafer production lines. These lines face challenges such as outdated equipment (mostly sourced from the second-hand market), upgrades difficulties, and lower returns compared to 12-inch wafers. Consequently, an increasing number of major companies are transitioning their focus from 8-inch to 12-inch wafers.

While the quantity of 8-inch wafer fabs is far less than that of 12-inch wafer fabs, their presence is significant. According to data from SEMI, China has maintained rapid development in 8-inch wafers. It is projected that by 2026, China’s market share in 8-inch wafers will increase to 22%, with a monthly production capacity reaching 1.7 million wafers, ranking first globally. By the end of 2025, companies including Huahong, Sien, Silan, Yangdong Microelectronic, GTA Semiconductor, SMEIIC, Zkjx, Hwdz, and Eaerkey are expected to establish a total of nine new 8-inch wafer fabs.

China’s Semiconductor Focus: 33% Mature Process Capacity by 2027

Based on product requirements, we can categorize semiconductor processes into specialty processes and logic processes. Logic processes further divide into mature processes (28nm and above) and advanced processes (nodes below 28nm, primarily 16/14nm and below).

Considering the current scenario, challenges in advanced process technology and high expenditures have confined major players in advanced processes to Intel, TSMC, and Samsung. This year, Samsung and TSMC announced the mass production of 3nm processes, marking the most advanced nodes currently available. Examining China’s situation, the semiconductor industry, having started relatively late, is currently focusing on mature and specialty processes due to factors like equipment and material limitations and changes in the international landscape.

It’s worth noting that, apart from Chinese wafer fabs intensively researching mature processes, many major companies have started to reverse their focus on mature processes in the past two years. Companies such as TSMC, Samsung, Intel, UMC, and GlobalFoundries are actively expanding their mature process capacities.

Among them, UMC’s bet on mature capacities is unprecedented. It became the world’s first wafer foundry to announce leaving research and development of advanced processes. Since 2018, UMC has strategically focused on improving the company’s return on investment, particularly targeting processes of 28nm and above.

According to TrendForce, the compound annual growth rate of global semiconductor foundry capacities from 2021 to 2024 is expected to reach 11%. 28nm capacity is expected to be 1.3 times that of 2022 by 2024, making it the most actively expanded node in mature processes. It is anticipated that more applications of specialty processes will transition to 28nm. Moreover, from 2021 to 2024, the global capacity of mature processes (28nm and above) is expected to maintain a stable share of over 75%. This indicates the potential and significance of positioning in the mature process and specialty process markets.

TrendForce predicts that, with the expansion of mature process capacities below 28nm, mature process capacities are expected to account for 70% of the top ten foundries’ capacities by 2027. China is expected to hold 33% of mature process capacities in 2027, with the possibility of continuous upward adjustments.

(Image: SMIC)

On November 13, NVIDIA unveiled the AI computing platform HGX H200, featuring the Hopper architecture, equipped with H200 Tensor Core GPU and high-end memory to handle the vast amounts of data generated by AI and high-performance computing.

This marks an upgrade from the previous generation H100, with a 1.4x increase in memory bandwidth and a 1.8x increase in capacity, enhancing its capabilities for processing intensive generative AI tasks.

The internal memory changes in H200 represent a significant upgrade, as it adopts the HBM3e for the first time. This results in a notable increase in GPU memory bandwidth, soaring from 3.35TB per second in H100 to 4.8TB per second.

The total memory capacity also sees a substantial boost, rising from 80GB in H100 to 141GB. When compared to H100, these enhancements nearly double the inference speed for the Llama 2 model.

H200 is designed to be compatible with systems that already support H100, according to NVIDIA. The company states that cloud service providers can seamlessly integrate H200 into their product portfolios without the need for any modifications.

This implies that NVIDIA’s server manufacturing partners, including ASRock, ASUS, Dell, Eviden, GIGABYTE, HPE, Ingrasys, Lenovo, Quanta Cloud, Supermicro, Wistron, and Wiwynn, have the flexibility to replace existing processors with H200.

The initial shipments of H200 are expected in the second quarter of 2024, with cloud service giants such as Amazon, Google, Microsoft, and Oracle anticipated to be among the first to adopt H200.

What is HBM?

“The integration of faster and more extensive HBM memory serves to accelerate performance across computationally demanding tasks including generative AI models and [high-performance computing] applications while optimizing GPU utilization and efficiency,” said Ian Buck, the Vice President of High-Performance Computing Products at NVIDIA.

What is HBM? HBM refers to stacking DRAM layers like building blocks and encapsulating them through advanced packaging. This approach increases density while maintaining or even reducing the overall volume, leading to improved storage efficiency.

TrendForce reported that the HBM market’s dominant product for 2023 is HBM2e, employed by the NVIDIA A100/A800, AMD MI200, and most CSPs’ (Cloud Service Providers) self-developed accelerator chips.

As the demand for AI accelerator chips evolves, in 2023, the mainstream demand is projected to shift from HBM2e to HBM3, with estimated proportions of approximately 50% and 39%, respectively.

As the production of acceleration chips utilizing HBM3 increases gradually, the market demand in 2024 is expected to significantly transition to HBM3, surpassing HBM2e directly. The estimated proportion for 2024 is around 60%.

Since Manufacturers plan to introduce new HBM3e products in 2024, HBM3 and HBM3e are expected to become mainstream in the market next year.

TrendForce clarifies that the so-called HBM3 in the current market should be subdivided into two categories based on speed. One category includes HBM3 running at speeds between 5.6 to 6.4 Gbps, while the other features the 8 Gbps HBM3e, which also goes by several names including HBM3P, HBM3A, HBM3+, and HBM3 Gen2.

HBM3e will be stacked with 24Gb mono dies, and under the 8-layer (8Hi) foundation, the capacity of a single HBM3e will jump to 24GB.

According to the TrendForce’s previous news release, the three major manufacturers currently leading the HBM competition – SK hynix, Samsung, and Micron – have the following progress updates.

SK hynix and Samsung began their efforts with HBM3, which is used in NVIDIA’s H100/H800 and AMD’s MI300 series products. These two manufacturers are expected to sample HBM3e in Q1 2024 previously. Meanwhile, Micron chose to skip HBM3 and directly develop HBM3e.

However, according to the latest TrendForce survey, as of the end of July this year, Micron has already provided NVIDIA with HBM3e verification, while SK hynix did so in mid-August, and Samsung in early October.

(Image: Nvidia)

According to Economic Daily News’ report, the DDR3 market is experiencing a surge in prices, with an almost 10% short-term increase. Contract prices for this quarter are expected to rise by 10% to 15%, with a continued optimistic outlook for the first quarter of next year.

Market analysts point out that global leaders like Samsung, SK Hynix, and Micron currently dominate standard DRAM and NAND Flash. Taiwan-based companies are unable to compete with them in chip production, except for module companies which have the potential to capitalize on the rebound of the DRAM and NAND market due to their advantage of low-cost inventory.

On the other hand, DDR3 production remains primarily under Taiwan’s purview, and the price surge is significant, providing substantial assistance to Taiwanese memory industry players.

Industry sources indicate that the main impetus behind the current memory market upswing stems from the continual reduction in production by major international manufacturers. With tightening capacities, they are strictly controlling shipments, displaying their determination to support the price increases.

Besides, companies like Samsung, SK Hynix, and Micron are actively venturing into AI applications, shifting their main capacity to produce high-bandwidth memory, DDR5, and other advanced areas. This shift has created room in the DDR3 market. Recent replenishment of inventories on the client end, along with a surge in orders for consumer electronics, has been witnessed.

TrendForce indicated that DDR3 prices have been steadily rising since September. The cumulative increase for DDR3 4Gb has reached nearly 10%, while DDR3 2Gb has shown a cumulative rise of 14%; regarding contract prices, TrendForce expects a robust increase of 10% to 15% for this quarter, with a continued strength forecasted for the first quarter of next year, likely to rise another 5% to 10%.

DDR3-related companies are optimistic about market developments. Etron believes that with inventory digestion coming to an end, “the cyclical bottom is over,” and they are gradually heading towards the dawn of recovery. They hold a positive outlook for significant growth in the global DRAM market next year.

ESMT, on the other hand, continues to advance the progress of their in-house products. They are continuously developing new products in DRAM, Flash, and MCP, expanding into different markets. This includes accelerating the development of 19-nanometer DRAM, achieving mass production of 28-nanometer NAND Flash, and researching niche memory products for automotive applications.

Winbond’s General Manager, Pei-Ming Chen, points out that their operations for this quarter will be better than the third quarter and they have a positive view of the DRAM market for next year.

Read more

• Q4 Mobile DRAM and NAND Flash Contract Prices Forecast to Rise, Upward Trend May Continue into 1Q24, Says TrendForce

(Photo credit: Unsplash)

Rumors swirl around AMD’s upcoming chip architecture, codenamed “Prometheus,” featuring the Zen 5C core. As reported by TechNews, the chip is poised to leverage both TSMC’s 3nm and Samsung’s 4nm processes simultaneously, marking a shift in the competitive landscape from process nodes, yield, and cost to factors like capacity, ecosystem, and geopolitics, are all depends on customer considerations.

Examining yields, TSMC claims an estimated 80% yield for its 4nm process, while Samsung has surged from 50% to an impressive 75%, aligning with TSMC’s standards and raising the likelihood of chip customers returning. Speculation abounds that major players such as Qualcomm and Nvidia may reconsider their suppliers, with industry sources suggesting Samsung’s 4nm capacity is roughly half of TSMC’s.

Revegnus, a reputable X(formerly Twitter) source, unveiled information from high-level Apple meetings, indicating a 63% yield for TSMC’s 3nm process but at double the price of the 4nm process. In the 4nm realm, Samsung’s yield mirrors TSMC’s, with Samsung showing a faster-than-expected yield recovery.

Consequently, with Samsung’s significant improvements in yield and capacity, coupled with TSMC’s decision to raise prices, major clients may explore secondary suppliers to diversify outsourcing orders, factoring in considerations such as cost and geopolitics. Recent reports suggest Samsung is in final negotiations for a 4nm collaboration with AMD, planning to shift some 4nm processor orders from TSMC to Samsung.

Beyond AMD, the Tensor G3 processor in Google’s Pixel 8 series this year adopts Samsung’s 4nm process. Samsung’s new fabs in Taylor, Texas, sees its inaugural customer in its Galaxy smartphones, producing Exynos processors.

Furthermore, Samsung announced that U.S.-based AI solution provider Groq will entrust the company to manufacture next-generation AI chips using the 4nm process, slated to commence production in 2025, marking the first order for the new Texas plant.

Regarding TSMC’s 4nm clients, alongside longstanding partners like Apple, Nvidia, Qualcomm, MediaTek, AMD, and Intel, indications propose a potential transition to TSMC’s 4nm process for Tensor G4, while Tensor G5 will be produced using TSMC’s 3nm process. Ending the current collaboration with Samsung, TSMC’s chip manufacturing debut is anticipated to be delayed until 2025.

Last year, rumors circulated about Tesla, the electric vehicle giant, shifting orders for the 5th generation self-driving chip, Hardware 5 (HW 5.0), to TSMC. This decision was prompted by Samsung’s lagging 4nm process yield at that time. However, with Samsung’s improved yield, industry inclination leans towards splitting orders between the two companies.