AI server solution provider Supermicro, at its earnings call last week, revealed that the liquid cooling industry has been facing shortages for critical components, which further hinders its shipments. According to the latest report by the Economic Daily News, sources from the supply chain indicate that the critical component Supermicro referred to is the “quick coupling.”

The report notes that Taiwanese companies, including Global Tek, Fositek and Lotes, are accelerating their pace on sample verification.

Thanks to the soaring demand, the market is seeing a frenzy for quick couplings, with buyers willing to pay premium prices to secure capacity, the report suggests. The report notes that the price of quick couplings has surged from the original USD 40 to USD 60 per unit, with eager buyers claiming to pay for more as long as the components are available.

Citing industry experts, the report explains that the liquid-cooling system in AI servers consists of six key components: cooling distribution units (CDUs), cold plates, cabinets, fan walls, coolant distribution manifolds (CDMs), and quick couplings.

Among these, quick couplings are responsible for connecting the coolant flow between the cold plate and the CDU with great speed. In addition, as they are also prone to leakage, their quality would be especially crucial for liquid-cooling systems to operate smoothly.

The reason for the shortage, the report notes, is primarily due to the fact that a single AI server cabinet requires hundreds of male and female threads for quick couplings. As major cloud service providers (CSPs) are purchasing AI servers in large quantities, the demand for quick couplings has surged drastically, even to hundreds of times of the cabinets themselves.

It is worth noting that currently, the market for liquid-cooling quick couplings in AI servers is dominated by seven companies, including two Chinese firms. However, as the sanctions implemented in the U.S.-China tech war restrict the momentum of the Chinese companies, the market demand could not be satisfy. Also, the patents related to the components have created high barriers for other suppliers to enter the market.

Taiwanese companies, including Global Tek, Fositek and Lotes, therefore, are actively investing in the business and accelerating their sample submission, eyeing for the opportunities to receive order transfers.

Global Tek, for example, is working with partners at its Wuxi plant in China and its Taoyuan plant in Taiwan, with samples being tested. The company anticipates revenue contribution as early as the fourth quarter of this year or early next year.

Fositek, supported by its parent company Asia Vital Components, is currently focused on developing quick couplings and has already submitted samples for customer certification. Lotes, on the other hand, expects to see progress by the end of the third quarter.

Read more

• [News] Supermicro’s Financial Results Reveal Profit Shortfall
• [News] Supermicro President Charles Liang Bullish on Direct Liquid Cooling, Targeting a 15% Market Share within Next Year 

(Photo credit: Supermicro)

Amid the advancement of emerging applications such as Artificial Intelligence (AI), High-Performance Computing (HPC), data center, and autonomous vehicle, fan-out panel-level packaging (FOPLP) technology has successfully garnered industry attention due to its advantages in significantly improving computing power, reducing latency, and increasing bandwidth. As a result, more manufacturers are entering this field.

Recently, semiconductor equipment manufacturer ACM Research Shanghai introduced the Ultra ECP ap-p panel-level electroplating equipment for FOPLP.

Prior to this, ACM has launched the Ultra C vac-p negative pressure cleaning equipment designed for fan-out panel-level packaging applications, signaling that ACM has successfully entered the high-growth FOPLP market.

It is worth mentioning that since the second quarter of this year, chip manufacturers like AMD have actively approached TSMC and professional OSAT (Outsourced Semiconductor Assembly and Test) companies to promote chip packaging using FOPLP technology, further increasing industry focus on FOPLP.

In fact, advanced packaging has an increasing demand for low-latency, high-bandwidth, and cost-effective semiconductor chips, and FOPLP offers high bandwidth and high-density chip interconnects, making it a technology with higher potential.

FOPLP is a process performed on larger square substrates, allowing multiple chips, passive components, and interconnects to be integrated into a single package on a panel, offering greater flexibility, scalability, and cost effectiveness.

By redistributing chips on larger rectangular panels, FOPLP largely reduces the costs of packaging large GPU and high-density, high-bandwidth memory (HBM).

It is reported that the utilization rate of traditional silicon wafers is less than 85%, while that of panels exceeds 95%. The effective area of a 600×600 mm panel is 5.7 times that of a 300 mm traditional silicon wafer, with overall panel cost expected to decrease by 66%.

The increase in area utilization leads to higher capacity, greater flexibility in AI chip design, and significant cost reduction.

Currently, major players in the FOPLP advanced packaging field include Powertech Technology, ASE Group, SPIL, TSMC, Innolux, JSnepes, and Samsung Electro-Mechanics.

TrendForce points out that FOPLP technology presents advantages and disadvantages, facing both opportunities and challenges. Its main strengths are lower unit cost and larger package size, but as its technology and equipment systems are still developing, the commercialization process is highly uncertain.

It is estimated that the mass production timeline for FOPLP in consumer IC and AI GPU may fall between the second half of 2024 to 2026, and 2027-2028, respectively.

Read more

• [News] Innolux Targets to Begin FOPLP Mass Production by Year-End
• [News] TSMC Reportedly Forms a Team on FOPLP Development, with Mini Line on the Road 

(Photo credit: ACMR)

According to a report by Nikkei, Japanese chip manufacturer Rapidus plans to establish a fully automated production line using robots and AI in northern Japan to produce 2nm chips for advanced AI applications, with mass production anticipated as early as 2027.

Reportedly, Rapidus claims that automated production will significantly accelerate production times, reducing chip delivery time to just one-third of that of its competitors. The company’s fab is expected to complete its external structure by October, with EUV lithography system set to arrive in December.

Compared to other companies already operating fabs, building a fully automated plant could give Rapidus a significant advantage. While the front-end of chip manufacturing are already highly automated, the back-end processes, such as packaging and testing, remain labor-intensive.

Rapidus CEO Atsuyoshi Koike stated that this approach will deliver higher performance and faster turnaround times for the same 2nm products compared to other competing chipmakers.

Per a report from Tom’s Hardware, Rapidus is currently two years behind TSMC and Samsung, both of which are expected to begin 2nm chip production in 2025. If Rapidus can deliver chips faster without compromising on price or quality, it may secure a place in the market.

Despite the optimistic outlook, Rapidus faces operational challenges. The company revealed that it will need JPY 2 trillion (approximately USD 14 billion) to begin operation in 2025, and at least JPY 5 trillion in total for the start of mass production.

Although Rapidus has received JPY 920 billion in subsidies from the Japanese government, private companies remain hesitant to invest due to the company’s lack of track record.

Atsuyoshi Koike added that, given the current situation, it is difficult for Rapidus to secure private financing. The company is discussing ways to make financing easier, such as implementing a government loan guarantee system.

Read more

• [News] Japanese Government Reportedly Plans to Offer Guaranteed Loans to Chip Startup Rapidus
• [News] Rapidus Partners RISC-V Player Esperanto, Targeting Low-Power AI Chip

(Photo credit: Rapidus)

According to a report from the South China Morning Post, the U.S. export controls, which are restricting China’s access to advanced chips and technology, have intensified China’s efforts to replace global semiconductor manufacturing equipment. However, industry sources have indicated that China still faces significant bottlenecks in this area.

The report mentions that Chinese semiconductor equipment companies like NAURA and AMEC are leading efforts to encourage local foundries to adopt domestic equipment.

Notably, sources cited in the same report also reveal that there is an unwritten rule among Chinese semiconductor fabs that locally-made tools should account for 70% of their production lines.

Per a report by TrendForce, Chinese manufacturers have achieved a self-sufficiency rate of 15% or higher in materials for mature processes, such as silicon wafers, photomasks, photoresists, electronic gases, and wet chemicals. However, items with a self-sufficiency rate still below 15% include photolithography equipment, photomasks, and EDA.

AMEC’s chairman and CEO, Gerald Yin Zhiyao, stated that China is expected to achieve a basic level of self-sufficiency in chip production equipment by this summer, something that was unimaginable just a few years ago.

He acknowledged that while there are still gaps in quality and reliability, China’s semiconductor supply chain can indeed achieve self-sufficiency. This, he suggested, is further evidence that U.S. export controls may have accelerated the development of China’s chip industry.

However, the report also pointed out that China remains constrained in one critical area: lithography technology, which is subject to the most stringent export controls.

Dutch company ASML is the sole supplier of Extreme Ultraviolet (EUV) systems, essential for producing advanced chips, and is also the main supplier of Deep Ultraviolet (DUV) systems needed for mature process chips.

President of foundry China Resources Microelectronics, Li Hong, stated that in 2023, only 1.2% of the lithography systems used by Chinese foundries was purchased from local suppliers.

In the second quarter of this year, ASML’s shipments to Chinese customers totaled EUR 2.35 billion, accounting for nearly half of its global sales. This indicates that China continues to rely heavily on ASML’s equipment in the legacy nodes, which is not subject to U.S. sanctions.

Paul Triolo, senior vice-president for China and technology policy lead at the U.S. consulting firm Albright Stonebridge Group, noted that the significant purchases of DUV lithography systems from ASML by Chinese companies highlight that SMEE, a major Chinese lithography equipment manufacturer, still lags behind ASML in reliably producing lithography systems for 28nm and below processes.

However, lithography technology is not the only bottleneck China faces. Li Hong also noted that the local supply ratios for ion implantation and inspection and metrology systems is only 1.4% and 2.4%, respectively.

As per Chinese customs data, the value of ion implantation systems imported by China in 2023 increased by 20% year-on-year to USD 1.3 billion.

A research report by Guohai Securities indicates as well that Chinese fabs rely heavily on metrology systems from companies like KLA, Applied Materials, and Japan’s Hitachi.

KLA reportedly holds a 50% global market share in inspection and metrology equipment.

An industry source cited in the report mentioned that the local supply ratio in the inspection and metrology sector is relatively low, with local substitution primarily occurring in lower-end products.

Read more

• [News] China Lists Semiconductor Self-Reliance Challenges, Excluding Lithography Equipment
• [News] ASML Former CEO Predicts Chip War Between China and the US May Persist for Decades

(Photo credit: ASML)

According to sources cited in a report from Commercial Times, in response to the global increase in chip orders and rapid AI development, TSMC is actively seeking available land to keep its most advanced process technologies in Taiwan.

Currently, TSMC has already planned three 2nm fabs at the Nanzih Technology Industrial Park in Kaohsiung, southern Taiwan.

Regarding the need for additional land to accommodate facilities for more advanced nodes beyond 2nm, the report notes that the Kaohsiung City Government has been proactively preparing by evaluating land availability, as well as water and electricity supply, for TSMC’s next-generation advanced technology production, specifically targeting the A14 (14 angstrom) process.

Yet, regarding the matter, TSMC has remained discreet and declined to comment on market rumors regarding the progress of expansion.

Reportedly, the Nanzih Park site has the capacity to accommodate up to five fabs for TSMC, and there are rumors that its fourth and fifth fabs are likely to focus on A14 process, although TSMC has yet to confirm this.

TSMC’s first 2nm process fab in Nanzih is expected to begin mass production in 2025. Per sources cited by the report, the node will be used in high-performance computing (HPC), smartphones, electric vehicles, and autonomous driving applications.

Earlier, concerns were raised about the progress of TSMC’s CoWoS advanced packaging plant due to the discovery of cultural heritage sites at the Chiayi Science Park.

However, sources cited by Commercial Times have pointed out that while there have been some delays due to cultural heritage issues, TSMC’s adjustment plan has been approved. The company will adjust its working procedures in order to proceed with construction according to the original schedule, with no changes to the completion timeline.

Read more

• [News] Samsung Expected to Unveil its 1nm Plan in June, Advancing it to 2026
• [News] Decipher TSMC’s Calm Take on High-NA EUV Lithography Machines: Who May Have the Last Laugh in the Angstrom Era?

(Photo credit: TSMC)