To capture the booming demand of AI processors, memory heavyweights have been aggressively expanding HBM (High Bandwidth Memory) capacity, as well as striving to improve its yield and competitiveness. The latest development would be Micron’s reported new plant in Hiroshima Prefecture, Japan.

The fab, targeting to produce chips and HBM as early as 2027, is reported to manufacture DRAM with the most advanced “1γ” (gamma; 11-12 nanometers) process, using extreme ultraviolet (EUV) lithography equipment in the meantime.

Why is HBM such a hot topic, and why is it so important?

HBM: Solution to High Performance Computing; Perfectly Fitted for AI Chips

By applying 3D stacking technology, which enables multiple layers of chips to be stacked on top of each other, HBM’s TSVs (through-silicon vias) process allows for more memory chips to be packed into a smaller space, thus shortening the distance data needs to travel. This makes HBM perfectly fitted to high-performance computing applications, which requires fast data speed. Additionally, replacing GDDR SDRAM or DDR SDRAM with HBM will help control energy consumption.

Thus, it would not be surprising that AMD, the GPU heavyweight, collaborated with memory leader SK hynix to develop HBM in 2013. In 2015, AMD launched the world’s first high-end consumer GPU with HBM, named Fiji. While in 2016, NVIDIA introduced P100, its first AI server GPU with HBM.

Entering the Era of HBM3e

Years after the first AI server GPU with HBM was launched, NVIDIA has now incorporated HBM3e (the 5^th generation HBM) in its Blackwell B100/ Hopper H200 models. The GPU giant’s GB200 and B100, which will also adopt HBM3e, are on the way, expected to be launched in 2H24.

The current HBM3 supply for NVIDIA’s H100 is primarily met by SK hynix. In March, it has reportedly started mass production of HBM3e, and secured the order to NVIDIA. In May, yield details regarding HBM3e have been revealed for the first time. According to Financial Times, SK hynix has achieved the target yield of nearly 80%.

On the other hand, Samsung made it into NVIDIA’s supply chain with its 1Znm HBM3 products in late 2023, while received AMD MI300 certification by 1Q24. In March, Korean media Alphabiz reported that Samsung may exclusively supply its 12-layer HBM3e to NVIDIA as early as September. However, rumors have it that it failed the test with NVIDIA, though Samsung denied the claims, noting that testing proceeds smoothly and as planned.

According to Korea Joongang Daily, Micron has roused itself to catch up in the heated competition of HBM3e. Following the mass production in February, it has recently secured an order from NVIDIA for H200.

Regarding the demand, TrendForce notes that HBM3e may become the market mainstream for 2024, which is expected to account for 35% of advanced process wafer input by the end of 2024.

HBM4 Coming Soon? Major Players Gear up for Rising Demand

As for the higher-spec HBM4, TrendForce expects its potential launch in 2026. With the push for higher computational performance, HBM4 is set to expand from the current 12-layer (12hi) to 16-layer (16hi) stacks. HBM4 12hi products are set for a 2026 launch, with 16hi in 2027.

The Big Three have all revealed product roadmaps for HBM4. SK hynix, according to reports from Wccftech and TheElec, stated to commence large-scale production of HBM4 in 2026. The chip will, reportedly, be the first chip from SK hynix made through its 10-nm class Gen 6 (1c) DRAM.

As the current market leader in HBM, SK hynix shows its ambition in capacity expansion as well as industrial collaboration. According to Nikkei News, it is considering expanding the investment to Japan and the US to increase HBM production and meet customer demand.

In April, it disclosed details regarding the collaboration with TSMC, of which SK hynix plans to adopt TSMC’s advanced logic process (possibly CoWoS) for HBM4’s base die so additional functionality can be packed into limited space.

Samsung, on the other hand, claimed to introduce HBM4 in 2025, according to Korea Economic Daily. The memory heavyweight stated at CES 2024 that its HBM chip production volume will increase 2.5 times compared to last year and is projected to double again next year. In order to embrace the booming demands, the company spent KRW 10.5 billion to acquire the plant and equipment of Samsung Display located in Tianan City, South Korea, for HBM capacity expansion. It also plans to invest KRW 700 billion to 1 trillion in building new packaging lines.

Meanwhile, Micron anticipates launching 12-layer and 16-layer HBM4 with capacities of 36GB to 48GB between 2026 and 2027. After 2028, HBM4e will be introduced, pushing the maximum bandwidth beyond 2+ TB/s and increasing stack capacity to 48GB to 64GB.

Look back at history. As the market demand for AI chips keeps its momentum, GPU companies tend to diversify their sources, while memory giants vie for their favor by improving yields and product competitiveness.

In the era of HBM3, the supply for NVIDIA’s H100 solution is primarily met by SK hynix at first. Afterwards, Samsung’s entry into NVIDIA’s supply chain with its 1Znm HBM3 products in late 2023, though initially minor, signifies its breakthrough in this segment. This trend of diversifying suppliers may continue in HBM4. Who would be able to claim the lion’s share in the next-gen HBM market? Time will tell sooner or later.

Read more

• [News] Micron Reportedly Set to Build New DRAM Plant in Hiroshima, Japan, Operational Expected by End of 2027
• [News] Overview of Expansion Plans by HBM Giants

(Photo credit: Samsung)

On May 30th, Taiwanese Minister of Ministry of Economic Affairs, J.W. Kuo, proposed a crucial industry policy. According to a report from China Times, the first step is to take Taiwan’s manufacturing parks global, with the initial site planned for Kyushu, in conjunction with TSMC’s Kumamoto fab, to create a semiconductor industrial park.

Additionally, the ministry reportedly plans to invite the world’s top 100 companies to set up sales offices in Taiwan as well as offering tax incentives, to attract around 400 million consumers from Japan, South Korea, the Philippines, Vietnam, and other regions.

Kuo further emphasized the importance of the government’s proactive stance on taking small and medium-sized enterprises (SMEs) and their supply chains overseas. The Ministry of Economic Affairs is said to be planning to utilize state-owned enterprises or establish a development service company to help eliminate obstacles to overseas investment. Meanwhile, it also plans to establish an overseas one-stop service window to expedite the setup of plants by companies.

He then pointed out that TSMC already has two fabs in Kumamoto, Japan. The plan, as per the same report, is to set up a semiconductor industrial park in Kyushu, bringing Taiwan’s supply chain to Japan. This park will not only serve TSMC but also local Japanese companies.

Looking ahead, the industrial parks will primarily follow TSMC, expanding to Japan, the USA, and Germany, with plans to relocate 10 to 15% of the supply chain capacity.

Kuo expressed the vision of considering Japan, South Korea, the Philippines, and Vietnam, all within a three-hour flight radius, as Taiwan’s domestic market. Additionally, within four years, the goal is to attract 500 Michelin-starred restaurants, along with popular performances and medical beauty services, to draw consumers from neighboring Asian countries to Taiwan.

Read more

• [News] Export Crisis Averted, TSMC’s Nanjing Plant Reportedly Granted Indefinite Exemption
• [Insights] Trendforce: Foundry Capacity Market Share of Advanced Process to Decline in Taiwan, Korea until 2027, While US on the Rise
  

(Photo credit: TSMC)

According to a report from Liberty Times, Taiwanese foundry UMC stated yesterday that the company’s operations in the second quarter would see a slight increase compared to the first quarter, and the second half of the year would be better than the first half.

With UMC’s technology and processes, the company estimated that it can capture about 10-20% of the AI foundry market share, which is expected to drive future business growth.

• Edge Computing to Become Widespread in 4 Years

At UMC’s shareholder meeting yesterday, Co-General Manager Jason Wang stated that semiconductor applications are becoming increasingly diverse and important. The most promising growth areas include autonomous vehicles, AI servers, and AI PCs.

UMC sees significant growth potential in high-speed transmission and power management. In high-performance computing (HPC), UMC will focus on back-end integration, including interposers and advanced 3D IC packaging.

Jason Wang pointed out that AI is currently in early stages, and thus more focused on building the infrastructure for high-speed computing. However, once the infrastructure is complete, the market will gradually expand to the widespread adoption of edge computing, which he estimates will take about four years.

UMC plans to position itself early, developing technologies that align with customer applications. UMC is optimistic about the market prospects and has high expectations for the future, Wang noted.

UMC’s CFO Chitung Liu stated that while UMC does not have advanced processes for producing HPC chips in the AI field, it has made significant progress in edge computing and related process technologies. With UMC’s technology, processes, and capacity, it is estimated that the company can still capture a 10-20% share of the AI foundry market, which is considerable and will be a major driver of future operational growth.

UMC currently produces CoWoS advanced packaging-related silicon interposers at its Singapore plant, with monthly capacity doubling to 6,000 wafers this year. UMC will continue to invest according to market conditions, according to Liu.

• Effects of U.S.-China Trade War Order Transferring Take At Least Six Month to be Seen

Regarding the benefits of diversification amid the U.S.-China trade war, Liu mentioned that it takes time for customers to redesign and transition orders. It can take at least six to nine months in the short term for the effects to be seen, and up to one to two years in the long term for orders to be successfully transitioned.

Currently, there are no significant effects from diversification in the short term. However, UMC’s production is diversified across regions including Singapore, Japan, China, and Taiwan, with collaborations with Intel in Arizona, USA. Thus, UMC can meet customer needs regardless of where they choose to manufacture, Liu explained.

Liu reiterated the stance from last month’s briefing, stating that the situation in the first half of the year has improved from the economic downturn, and second-quarter revenue is expected to see a slight increase compared to the previous quarter. He hopes for better performance in the second half of the year.

In terms of application markets, the short-term performance of the automotive and industrial sectors appears weak, but growth is expected in the medium to long term. On the other hand, prospects for the communication and consumer sectors in the second half of the year are expected to be better than the first half.

Read more

• [News] UMC’s Global and Local Orders Surge Amid U.S. Tariffs on China and Networking Market Recovery
• [News] UMC Ventures into US Production of 12nm Mature Chips through Collaboration with Intel

(Photo credit: UMC)

According to a report from Bloomberg, US officials have slowed down the issuance of licenses for chip manufacturers like NVIDIA and AMD to export large quantities of AI accelerators to the Middle East. Meanwhile, officials are conducting a national security assessment of AI developments in the region.

As per Bloomberg’s report citing sources, it is still unclear how long the assessment will take and what exactly constitutes a large-scale export. They said that officials are particularly concerned about large-scale sales because countries like the UAE and Saudi Arabia are looking to import significant quantities of chips for AI data centers.

AI accelerators can help data centers process the massive amounts of information required for developing AI chatbots and other tools. They have become essential equipment for companies and governments seeking to build AI infrastructure.

Reportedly, sources have revealed that slowing down exports is intended to give Washington time to formulate a comprehensive strategy on how advanced chips should be deployed overseas. Some of these sources mentioned that this includes negotiating who will manage and secure the facilities used to train AI models.

The US Department of Commerce stated in a statement that “protecting national security” is the top priority.

“With regards to the most cutting edge technologies, we conduct extensive due diligence through an interagency process, thoroughly reviewing license applications from applicants who intend to ship these advanced technologies around the world,” a representative for the department said. “As always, we remain committed to working with our partners in the Middle East and around the world to safeguard our technological ecosystem.”

Addressing national security concerns, earlier this month, the U.S. government has reportedly revoked the licenses of Intel and Qualcomm to supply semiconductor chips used in laptops and handsets to Huawei. According to Reuters citing sources, some companies received notices on May 7th, and the revocation of the licenses took immediate effect.

Read more

• [News] U.S. Imposes Further Sanctions, Revoking Intel and Qualcomm’s License to Supply Chips to Huawei
• [News] US Considers Sanctioning Huawei-Linked Chinese Chip Network, Huawei-Related Enterprises Blacklisted

(Photo credit: NVIDIA)

Recently, rumors surfaced that price drops in the silicon carbide (SiC) wafer market. So, what is the actual market situation for SiC wafer?

Currently, most companies in the supply chain agree on the fact that the price of SiC wafer is on the decline. For instance, Xu Xiulan, the chairwoman of GlobalWafers, publicly stated that the global release of 6-inch SiC wafer production capacity, coupled with a temporary slowdown in demand for electric vehicles, is putting downward pressure on SiC wafer prices in 2024. On May 9, 2024, SICC highlighted two internal reasons for the price drop in an investor relations report: technological advancements and scale effect, which brought down wafer costs.

Indeed, changes in technology and production capacity of SiC wafer have become increasingly evident since 2H23.

Technically, more than ten China-base companies have entered the sample delivery and small-batch production stages for 8-inch SiC wafer in addition to international ones, including SemiSiC, JSG, SICC, GZSC, Synlight Crystal, Tankeblue, KY Semiconductor, Hunan San’an Semiconductor, Hypersics, Taisic Materials, Heligenius, Cengol Semi,and GlobalWafers.

Regarding scale effect, while SiC wafer manufacturers’ early invested projects are now reaching the investment return phase, not a few wafer companies have been shifting their production focus to 8-inch wafer.

For example, JSG’s project for an annual production of 250,000 6-inch and 50,000 8-inch SiC wafers officially signed and started in November 2023; Cengol’s 8-inch SiC processing line got ready and went into small-batch production in February 2024; GZSC’s 8-inch SiC single crystal and wafer project was established in Jinan, Shandong, in June 2023, with full production expected by 2025; KY Semiconductor signed a strategic cooperation agreement with Russian company N in March 2024 to work on the “8-inch SiC Perfect Seed Crystal” project.

The decline in SiC wafer prices is an inevitable trend, towards which most companies hold a positive attitude. As CGEE stated, the expansion of market space and improvement in yield levels will unavoidably cause price adjustments during the competition, which will place stress on related companies in the short term.

However, for the entire supply chain, the advantages from the improvement in yield and decrease in prices outbalance disadvantages. That means cost reduction will invigorate more downstream applications, and thereby enable the industry to maintain a sound growth rate as a whole.

As a part of SiC wafer market, SICC also pointed out that currently, the price of SiC wafer is much higher than that of Si wafer. As such, the decrease in SiC wafer prices will help expand downstream applications and promote the penetration and adoption of SiC technology and materials, fostering overall growth.

Read more

• [News] Continuous Breakthroughs in SiC Substrates, Global 8-Inch Fab to Reach 11

(Photo credit: JSG)