According to a report from The Chosun Daily, major tech companies like NVIDIA are considering using Intel’s foundry services (IFS) as an alternative due to TSMC’s packaging capacity shortages. While packaging is a core competency for semiconductor foundries, Samsung, which is facing difficulties in attracting customers, will need to exert maximum effort to secure orders.

It’s highlighted by the report that the demand for AI accelerators is growing rapidly, but TSMC’s AI chip production capacity is unable to keep up.

Furthermore, per The Chosun Daily citing sources, major clients like NVIDIA and Apple have secured TSMC’s 3nm advanced process capacity, pushing order backlogs into 2026.

Therefore, major tech companies seeking alternatives are turning to Intel’s IFS. Since Intel’s Foveros is said to be comparable to TSMC’s CoWoS-S, it has made Intel’s advanced packaging a viable option to ensure supply.

TSMC and Intel offer advanced packaging services to customers under the names CoWoS and Foveros, respectively. Both CoWoS and Foveros are advanced packaging technologies that connect two or more semiconductor chips on a wafer and then place them onto a packaging substrate.

Reportedly, in addition to NVIDIA and Microsoft, Amazon and Cisco are considering outsourcing to Intel Foundry to reduce their dependence on TSMC.

The report further emphasizes that this market trend is likely to cause anxiety for Samsung, which competes with TSMC and Intel in advanced processes.

Recently, Samsung’s former clients, Google and Qualcomm, have chosen TSMC, while Intel, seen as a latecomer, is catching up by securing advanced packaging orders from major tech companies, potentially narrowing the gap with Samsung.

Kim Hak-sung, head of Hanyang Institute of Smart Semiconductor, said that packaging is a technology that critically influences customer acquisition in the AI semiconductor era, where various types of chips are interconnected.

He noted that although there may not be a substantial technical capabilities difference between Samsung Electronics and Intel, as mass production experience allows the process to stabilize and become more appealing to customers.

Kim eventually addressed that to stay competitive, Samsung needs to focus on capturing the volumes that TSMC cannot accommodate, positioning itself ahead of Intel.

Read more

• [News] TSMC Reportedly to Raise 3nm & 5nm Prices Soon, Looking to Maintain Long-Term Profit Margins 
• [News] Major Clients Reportedly Fully Allocate TSMC’s Production Capacity Until 2026, 3nm Process in High Demand
  

(Photo credit: Samsung)

As the demand for memory chips used in AI remains strong, prompting major memory companies to accelerate their pace on HBM3e and HBM4 qualification, SK hynix CEO Kwak Noh-jung stated on August 7 that driven by the high demand for memory chips like high-bandwidth memory (HBM), the market is expected to stay robust until the first half of 2025, according to a report by the Korea Economic Daily.

However, Kwak noted that the momentum beyond 2H25 “remains to be seen,” indicating that the company needs to study market conditions and the situation of supply and demand before making comments further. SK hynix clarified that was not an indication of a possible downturn.

According to the analysis by TrendForce, HBM’s share of total DRAM bit capacity is estimated to rise from 2% in 2023 to 5% in 2024 and surpass 10% by 2025. In terms of market value, HBM is projected to account for more than 20% of the total DRAM market value starting in 2024, potentially exceeding 30% by 2025.

SK hynix, as the current HBM market leader, said earlier in its earnings call in July that its HBM3e shipment is expected to surpass that of HBM3 in the third quarter, with HBM3e accounting for more than half of the total HBM shipments in 2024. In addition, it expects to begin supplying 12-layer HBM3e products to customers in the fourth quarter.

The report notes that for now, the company’s major focus would be on the sixth-generation HBM chips, HBM4, which is under development in collaboration with foundry giant TSMC. Its 12-layer HBM4 is expected to be launched in the second half of next year, according to the report.

Samsung, on the other hand, had been working since last year to become a supplier of NVIDIA’s HBM3 and HBM3e. In late July, it is said that Samsung’s HBM3 has passed NVIDIA’s qualification, and would be used in the AI giant’s H20, which has been developed for the Chinese market in compliance with U.S. export controls. On August 6, the company denied rumors that its 8-layer HBM3e chips had cleared NVIDIA’s tests.

Notably, per a previous report from the South Korean newspaper Korea Joongang Daily, following Micron’s initiation of mass production of HBM3e in February 2024, it has recently secured an order from NVIDIA for the H200 AI GPU.

Read more

• [News] SK hynix, TSMC, and NVIDIA Reportedly Forge Alliance to Develop Next-Generation HBM
• [News] SK hynix Secures up to USD 450 Million Funding for Indiana Packaging Facility under CHIPS Act

(Photo credit: SK hynix)

AI industry has been driving semiconductor industry to advance forward. Benefited from the surge in AI-driven demand for advanced process chip, foundry industry is experiencing a gradual turnaround, while demands for consumer chip and automotive chip have not yet fully recovered, and competition remains fierce in the mature process chip sector, representing a stark contrast within the wafer foundry industry.

• Financial Results Indicate Strong AI Development but Weak Automotive Terminal Demands

Recently, several major foundries released their Q2 financial reports and shared outlook on future market conditions.

For the second quarter ending June 30, TSMC reported consolidated revenue of approximately USD 20.82 billion, up 32.8% YoY and 10.3% QoQ, which was attributed to strong demand for its 3nm and 5nm technologies.

As per the financial report, revenue from advanced technologies (7nm and below) accounted for 67% of TSMC’s total wafer revenue in 2Q24. In terms of application areas, HPC has replaced mobile business as the core driver of the company’s growth, contributing 52% of revenue.

Additionally, although TSMC’s automotive electronics revenue grew 5% QoQ, the company warned of a potential downturn in the automotive market this year.

UMC reported Q2 revenue of TWD 56.8 billion, up 4% QoQ. UMC expected customer inventories in the communications, consumer electronics, and computer sectors to return to seasonal levels as usual in the second half of this year, and to reach healthy levels by the end of the year.

However, demand in the automotive end market remains weak, which may extend the period of inventory adjustment, with healthy levels anticipated only by the first quarter of next year.

On August 6, GlobalFoundries released its latest financial report.

In the second quarter of this year, the company achieved revenue of USD 1.63 billion, a year-on-year decrease of 12% and a quarter-on-quarter increase of 5%. Net profit was USD 155 million, a year-on-year decrease of 35% and a quarter-on-quarter increase of 16%.

Industry sources cited by the report from WeChat account DRAMeXchange believe that during the pandemic, customers in sectors such as IoT, mobile device, and data center accumulated high inventory, which impacted GlobalFoundries’ revenue.

Moreover, the company is experiencing a cyclical downturn due to soft demands in the automotive, industrial, and other sectors.

• Advanced Processes Continue to Thrive while Mature Processes Face Intense Competition

The adoption of AI generative models keeps on the rise, driving high demand for AI chip. In this context, advanced processes have been well-received, leading to price increase and production expansion.

TrendForce’s survey in June showed that TSMC is seeing full capacity utilization in its 5/4nm and 3nm nodes due to strong demand from AI applications, new PC platforms, HPC applications, and high-end smartphones.

Its capacity utilization is expected to exceed 100% in the second half of the year, with visibility extending into 2025. Given cost pressures from overseas expansion and rising electricity prices, TSMC plans to raise prices for its advanced processes, which are experiencing strong demand.

TSMC is seeing full capacity utilization in its 5/4nm and 3nm nodes due to strong demand from AI applications, new PC platforms, HPC applications, and high-end smartphones. Its capacity utilization is expected to exceed 100% in the second half of the year, with visibility extending into 2025.

Given cost pressures from overseas expansion and rising electricity prices, TSMC plans to raise prices for its advanced processes, which are experiencing strong demand.

As per other sources cited by the same report, TSMC informed customers of a price increase for 5/3nm process products in 2024 at the beginning of this year.

In late July, TSMC notified several customers that due to rising costs, prices for 5/3nm process products will increase again starting January 2025, and the increase will range from 3-8%, depending on the tape-out plan, product, and partnership.

Meanwhile, the surge in demand for advanced packaging driven by AI will also lead to higher CoWoS prices.

To seize the significant opportunities brought by AI, many companies are actively investing in advanced processes. Currently, the 3nm process is the most advanced in the industry.

Meanwhile, TSMC, Samsung, Intel, and Rapidus are vigorously promoting the construction of 2nm fabs. Previously, TSMC and Samsung intended to produce 2nm chip at scale in 2025, while Rapidus planed to start trial production in 2025.

Following 2nm, 1nm chip will be the next goal for these fabs. According to their plans, the industry is likely to see the mass production of 1nm chip from 2027 to 2030.

Unlike the rising prices and volume in advanced process chip, mature process chip faces some uncertainty due to weaker-than-expected recovery in end-user demand, and sees more intense competition among manufacturers.

TrendForce’s survey reveals that the capacity utilization rates of PSMC and Vanguard is expected to improve more than anticipated in the second half of the year. However, overall demand for mature processes remains weak, with average capacity utilization still around 70–80%—indicating no significant shortages.

TrendForce further pointed out that in 2024, concerns over global inflation and weak recovery in end-demand may result in inconsistent momentum in replenishing inventory. Many foundries might offer price incentives to attract customers and boost capacity utilization, leading to a decline in overall ASP.

Furthermore, a significant amount of new capacity is expected to come online in 2025, including TSMC JASM, PSMC P5, SMIC’s new Beijing/Shanghai plants, HHGrace Fab9, HLMC Fab10, and Nexchip N1A3.

This increase in mature process capacity could intensify competition and impact future pricing negotiations.

Read more

• [News] Intel and NVIDIA’s New Platform Orders Rolling Out, TSMC Unaffected by Market Turbulence
• Foundry Market Shows Mixed Conditions with Price Adjustments for Specific Chinese Processes and Anticipated Price Increases for Taiwanese Advanced Processes, Says TrendForce

(Photo credit: TSMC)

Per a report from Reuters, Intel is said to be receiving the second new High-NA EUV equipment from ASML, costing EUR 350 million (~USD 383 million).

According to Intel’s earnings call on August 1, CEO Pat Gelsinger stated that Intel began receiving the first large equipment in December, and the installation process would take several months, which is expected to bring about a new generation of more powerful computer chip.

Gelsinger noted during the call that the second High-NA equipment is about to enter the facility in Oregon. Due to the poor stock performance following Intel’s earnings report, this statement did not attract much attention.

Previously, a senior executive from ASML once mentioned in July that the company already begun shipping the second High NA equipment to an unnamed customer, but would only record revenue for the first set this year. However, there are still some uncertainties regarding when the customer will adopt this equipment.

ASML has already received orders for over ten High-NA equipment from customers including TSMC, Samsung, Intel, Micron, and SK Hynix. Intel plans to use this technology for mass production by 2027, and TSMC is also set to receive the equipment this year, the time to put into production has not been disclosed, though.

ASML executive Christophe Fouquet stated on July 17 that DRAM memory chip manufacturers, which could refer to Samsung, SK Hynix, or Micron, are expected to start using High-NA equipment by 2025 or 2026.

Read more

• [News] ASML Slightly Beats Q2 Earnings Forecasts, with Full-Year Outlook Unchanged
• [News] Price for ASML’s Hyper-NA EUV Rumored to Double, Causing TSMC, Samsung and Intel to Hesitate

(Photo credit: ASML)

With the chip war between the two great powers heating up, the U.S. is reportedly mulling new measures to limit China’s access to AI memory. As the restrictions might be imposed as early as late August, rumor has it that Chinese tech giants like Huawei and Baidu, along with other startups, are stockpiling high bandwidth memory (HBM) semiconductors from Samsung Electronics, according to the latest report by Reuters.

Citing sources familiar with the matter, the report notes that these companies have increased their purchases of AI-capable semiconductors since early this year. One source states that in accordance with this trend, China contributed to around 30% of Samsung’s HBM revenue in 1H24.

Regarding the details of the potential restrictions, sources cited by Reuters said that the U.S. authority is anticipated to establish guidelines for restricting access to HBM chips. While the U.S. Department of Commerce declined to comment, it did state last week that the government is continually evaluating the evolving threat landscape and updating export controls.

The Big Three in the memory sector, Samsung, SK hynix and Micron, are all working on their 4^th generation (HBM3) and 5^th generation (HBM3e) products, while closely cooperating with AI giants such as NVIDIA and AMD in developing AI accelerators.

Reuters notes that the surging HBM demand from China recently has primarily focused on HBM2e, which is two generations behind HBM3e. However, as the capacities of other manufacturers are already fully booked by other American AI companies, China has turned to Samsung for its HBM demand.

Sources cited by Reuters also indicate that a wide range of businesses, from satellite manufacturers to tech firms like Tencent, have been purchasing these HBM chips. Meanwhile, Huawei has been using Samsung HBM2e to produce its advanced Ascend AI chip, according to one of the sources. It is also reported that Chinese memory giant ChangXin Memory Technologies (CXMT) has started mass production of HBM2.

Samsung and SK hynix declined to comment, neither did Micron, Baidu, Huawei and Tencent respond to requests for comment, Reuters notes.

Read more

• [News] US Reportedly Weighs Stricter Limits on AI Memory Access for China
• [News] U.S. Reportedly Enhancing Chip Export Restrictions to China, while NVIDIA’s H20 Might Get Banned

(Photo credit: Samsung)