Though previous rumors suggested that there might be a downward revision of its HBM capacity target due to delayed progress on 12-Hi HBM3e, Samsung’s officials have stated that it will enhance its semiconductor packaging facilities in South Chungcheong Province to increase the production of HBM, according to a report by the Korea Herald.

Samsung aims to complete the new facilities by December 2027, which will feature advanced packaging lines for HBM chips, the Korea Herald report indicates.

However, it is worth noting that the capacity is not built from scratch. According to the report, Samsung will repurpose an underutilized liquid crystal display plant, previously owned by Samsung Display, into a semiconductor fabrication facility. The plant is said to be located in Cheonan, approximately 85 kilometers south of Seoul.

According to the report, Samsung anticipates that the upgraded facilities in Cheonan will help the company regain its competitive edge in the global semiconductor market.

The current HBM leader, SK hynix, is reportedly investing in advanced chip packaging as well, as it aims to capture more demand for HBM in the AI boom.

According to a previous report by Bloomberg, Lee Kang-Wook, currently leading SK Hynix’s packaging research and development, stated that the company is investing over USD 1 billion in South Korea to expand and enhance the final steps of its chip manufacturing process.

On the other hand, Micron stated last year that in response to the growing demand in the AI market, it will continue to invest in advanced processes and packaging technologies to produce HBM products. Micron Taiwan is reportedly the only Micron facility globally with advanced packaging capabilities.

This August, Taiwanese panel maker AUO announced that it will sell three idled manufacturing facilities in Tainan, Southern Taiwan as Micron emerged as the buyer. Micron is also mulling HBM expansion in Malaysia and the U.S.

Read more

• [News] SK hynix and Samsung Reportedly Step up Focus on HBM4 and CXL amid Rising Chinese Competition
• [News] Samsung Reportedly Starts Developing Custom HBM4 for Big CSPs, Eyeing Mass Production by 2025

(Photo credit: Samsung)

As Samsung has implied earlier that it plans to collaborate with TSMC on HBM4, the memory giant seems to take a key step forward. A report by South Korean media outlet Maeli Business Newspaper discloses that the memory giant has already begun developing “Custom HBM4,” a next-gen high-bandwidth memory tailored specifically for Microsoft and Meta.

Samsung aims to begin mass production of HBM4 promptly upon completing development by the end of 2025, the report suggests.

Industrial sources cited by the report state that Samsung is establishing a dedicated production line for HBM4, which is now in the “pilot production” phase, where small-scale, trial manufacturing takes place ahead of mass production.

Citing sources familiar with the situation, the report further notes that Samsung is actively working on HBM4 designed for Microsoft and Meta. Both tech heavyweights have their own AI chips, namely, Microsoft’s Maia 100 and Meta’s Artemis.

As major tech companies make strides in scaling up their own AI data centers, there is a strong demand to cut costs associated with chip purchases. Therefore, many design and utilize their own AI accelerators while also buying AI chips from NVIDIA and AMD, according to Maeli.

Samsung, with its memory division and an LSI division capable of designing computational chips, is ideally positioned as a top partner for these major tech companies, according to Maeli.

Though the specifics of the HBM4 product that Samsung will supply to these companies remain undisclosed, Samsung did reveal the general specifications in February, according to Maeli.

Its data transmission speed, or bandwidth, reportedly reaches 2 terabytes per second (TB), marking a 66% increase over HBM3E, while its capacity has risen 33% to 48 gigabytes (GB), up from 36GB.

The report explains that unlike previous memory products, HBM4 is also referred to as “Computer-in-Memory (CIM)” due to its advanced capabilities, which go beyond standard memory functions to include specialized operations aligned with customer requirements.

Up to HBM3, the focus is said to be mainly on managing heat generation and maximizing speed, according to a semiconductor industrial official cited by the report. With HBM4, however, the emphasis will shift toward integrating AI computation capabilities (NPU) and specific data processing functions (IP) to meet evolving needs, the report says.

Read more

• [News] SK hynix and Samsung Reportedly Step up Focus on HBM4 and CXL amid Rising Chinese Competition
• [News] SK hynix Reportedly to Tape-out HBM4 in October, Paving the Way for NVIDIA’s Rubin

(Photo credit: Samsung)

According to TrendForce’s latest memory spot price trend report, regarding DRAM, the decline in DDR5’s spot prices has been relatively modest, while DDR4 and DDR3 products experience more significant pressure. As for NAND flash, spot prices could continue to fall also due to the possible expansions of wafer provision from suppliers at the end of the year. Details are as follows:

DRAM Spot Price:

Regarding DDR5 products, the decline in their spot prices has been relatively modest, but the trading momentum remains sluggish. As for DDR4 and DDR3 products, their spot prices continue to drop due to experiencing more significant pressure. Looking at DDR4’s future spot price trend, demand has been rapidly shifting towards platforms that primarily adopt DDR5. Consequently, clearing existing DDR4 inventories in the spot market is challenging, and the downward price trend is expected to persist for a considerable period. The average spot price of mainstream chips (i.e., DDR4 1Gx8 2666MT/s) has fallen by 2.33% from US$1.887 last week to US$1.843 this week.

NAND Flash Spot Price:

Transactions have been sluggish after the conclusion to the peak period of purchase momentum, and market rumors of production cuts among suppliers have proven to be ineffective towards halting the deterioration of spot prices that could continue to fall also due to the possible expansions of wafer provision from suppliers at the end of the year. Spot prices of 512Gb TLC wafers have dropped by 2.66% this week, arriving at US$2.380.

As South Korean memory giants Samsung and SK hynix announced their third quarter financial reports, posting a 112% and 94% year-over-year revenue growth, respectively, the threat from increasing output of Chinese rivals such as CXMT, which drives prices down, has reportedly prompted them to significantly cut back on legacy memory chip production, according to the report by the Korea Economic Daily.

According to the report, China’s ChangXin Memory Technologies (CXMT) has been ramping up the production of older chips like DDR4 and LPDDR4X, resulting in severe price pressure in legacy products.

CXMT has expanded its monthly DRAM production capacity from 40,000 wafer sheets in 2020 to 160,000 sheets. This capacity is expected to reach 200,000 sheets by year-end and 300,000 by the close of 2025, the report said.

SK hynix to Reduce DDR4 Production to 20% of Total DRAM Output

Industry sources cited by the report noted that in a recent investor relations session with Goldman Sachs, SK hynix suggested that it plans to reduce DDR4 DRAM production to 20% of its total DRAM output by the end of the year, down from 30% in September and 40% in June.

On the other hand, according to the report, in an earnings call with analysts on last week, Kim Jae-joon, executive vice president of Samsung’s device solutions (DS) division, confirmed plans to reduce production of legacy DRAM and NAND flash chips, aligning with industry expectations that chipmakers are scaling back on conventional memory output.

HBM and eSSD Emerge as the New Focus

Instead, both memory giants highlighted in their earnings call that they would shift their focus to highly profitable premium products like HBM and enterprise solid-drivers (eSSDs).

These adjustments by Samsung and SK hynix align with strong server DRAM demand driven by major tech firms like Google and China’s Baidu investing in server infrastructure, while PC DRAM sales have remained stagnant, according to the Korea Economic Daily.

According to SK hynix, as generative AI is developing into a multi-modal1 form and global big tech companies continue to invest to develop artificial general intelligence (AGI), the demand of memory for AI servers such as HBM and eSSD has grown noticeably this year. SK hynix predicts that this trend will continue next year.

According to the Korea Economic Daily, anticipating a prolonged global over supply, SK is accelerating the upgrade of its older DRAM lines in Wuxi, China, to advanced lines for producing fourth-generation 10-nanometer 1a DRAM.

While maintaining steady NAND flash production, in the meantime, SK is increasing the operation rate at its eSSD facility in Dalian, China, to nearly full capacity, according to sources cited by the report.

On the other hand, Samsung noted that in 2025, the company plans to expand the sales of HBM3E and the portion of high-end products such as DDR5 modules with 128GB density or higher for servers and LPDDR5X for mobile, PC, servers, and so on.

Read more

• [News] SK hynix and Samsung Reportedly Step up Focus on HBM4 and CXL amid Rising Chinese Competition

(Photo credit: SK hynix)

While introducing the industry’s first 48GB 16-high HBM3E at SK AI Summit in Seoul today, South Korean memory giant SK hynix has reportedly seen strong demand for its next-gen HBM. According to reports by Reuters and South Korean media outlet ZDNet, NVIDIA CEO Jensen Huang requested SK hynix to accelerate the supply of HBM4 by six months.

The information was disclosed by SK Group Chairman Chey Tae-won earlier today at the SK AI Summit, according to the reports. In October, the company said that it planned to deliver the chips to customers in the second half of 2025, according to the reports.

When asked by ZDNet about HBM4’s accelerated timetable, SK hynix President Kwak Noh-Jung responded by saying “We will give it a try.”

A spokesperson for SK hynix cited by Reuters noted that this new timeline is quicker than their original target, but did not provide additional details.

According to ZDNet, NVIDIA CEO Jensen Huang also made his appearance in a video interview at the Summit, stating that by collaborating with SK hynix, NVIDIA has been able to achieve progress beyond Moore’s Law, and the company will continue to need more of SK hynix’s HBM in the future.

According to the third-quarter financial report released by SK hynix in late October, the company posted record-breaking figures, including revenues of 17.5731 trillion won, an operating profit of 7.03 trillion won (with an operating margin of 40%), and a net profit of 5.7534 trillion won (with a net margin of 33%) for the third quarter of this year.

In particular, HBM sales showed excellent growth, up more than 70% from the previous quarter and more than 330% from the same period last year.

SK hynix is indeed making strides in its HBM, as it started mass production of the world’s first 12-layer HBM3E product with 36GB in September. It has also been developing 48GB 16-high HBM3E in a bid to secure technological stability and plans to provide samples to customers early next year, according to the company’s press release.

On the other hand, according to another report by Business Korea, Kim Jae-jun, Vice President of the Memory Business Division, stated In the earnings call that the company is mass-producing and selling both HBM3E 8-stack and 12-stack products, and have completed key stages of the quality testing process for a major customer. Though Kim did not specify the identity of the major customer, industry analysts suggest it is likely NVIDIA.

To shorten the technology gap with SK hynix, Samsung is reportedly planning to produce the next-generation HBM4 products in the latter half of next year.

Read more

• [News] SK hynix Introduces World’s First 16-High HBM3E, Providing Samples in Early 2025
• [News] SK hynix Reportedly to Tape-out HBM4 in October, Paving the Way for NVIDIA’s Rubin

(Photo credit: NVIDIA)