hbm

The Information investigated the smuggling chain of high-end storage chips in South Korea and interviewed a dealer in Shenzhen in his 20s. He originally worked at a chip company and transitioned to the chip black market business by leveraging his network. Having been in the industry for only three years, he is now dressed in luxury brands and busy connecting buyers and sellers every day. The United States will begin to restrict exports of advanced high-bandwidth memory (HBM) to China starting at the end of 2024, but the rules have loopholes: HBM exports are restricted when sold separately, but if packaged together with processors, it may be judged based on the processing power of the processor to determine if it crosses the regulatory line.A smuggler in his 40s controls a chip design company that specializes in designing cheap and lower-performance server processors. These chips are rarely purchased but are used as a pretext for procuring high-end HBM from South Korea. Packaging factories place HBM from Samsung and SK Hynix on the same substrate as these processors and declare exports under the category of low-performance processors. After the goods enter mainland China via Hong Kong, workshops reheat and remove the HBM, while the cheap processors are discarded, and any mishandling during disassembly could damage the memory.A stack of HBM3E in the Chinese black market was priced at only a few hundred dollars a year ago, but it has now risen to about 4,000 dollars. An AI chip typically requires 6 to 8 stacks, costing between 24,000 to 32,000 dollars just for the memory. Some of the memory ultimately flows to Huawei customers. Huawei's next-generation Ascend 960DT is scheduled to launch in the first quarter of 2027, but there is a shortage of domestic HBM supply, and Huawei has asked some customers to procure memory on their own. With the increase in imitators, the aforementioned smuggler responsible for procuring chips from South Korea plans to sell his controlling stake in the company next year, fearing that the spread of the route will attract strict scrutiny from regulatory authorities in South Korea and the United States.

On the 31st, SK Group Chairman Choi Tae-won had lunch with OpenAI CEO Sam Altman. Industry insiders revealed that the two sides discussed cooperation in the field of artificial intelligence business.Although the specific content of the talks has not been made public, it is speculated that they explored avenues for collaboration in areas such as high bandwidth memory (HBM) supply and the development of next-generation artificial intelligence infrastructure. SK Group has maintained a close cooperative relationship with OpenAI, including being a key partner in its large-scale artificial intelligence infrastructure project "Stargate."

SK Hynix will announce its Q3 2026 financial report after the U.S. stock market closes on Tuesday, along with a new shareholder return plan. The market is focused on the demand for high bandwidth memory (HBM), profitability, and whether a large special dividend will be issued. Other memory manufacturers such as Samsung Electronics and Kioxia Japan will also gradually announce their performance.The demand for AI infrastructure is driving an upward trend in the memory chip market. KB Securities analysts estimate that more than 70% of the memory that the two major Korean memory chip manufacturers plan to produce next year is covered by five-year supply agreements signed with large technology companies. The market is beginning to worry about the return on AI capital expenditures and financing costs, with investors paying more attention to order quality, the sustainability of future demand, and the pace of capacity expansion.Korean media previously reported that SK Hynix is considering issuing a special dividend of about 16 trillion won (approximately 12 billion USD); other reports suggest that the new shareholder return could reach as high as 60 trillion won (approximately 45 billion USD). These amounts have not been officially confirmed by the company. SK Hynix had previously announced a buyback and cancellation of treasury shares worth 40 trillion won and plans to use more than 50% of its free cash flow from 2025 to 2027 for shareholder returns.

RBC Capital Markets released a research report stating that the demand for artificial intelligence continues to exceed supply, and the rise in high bandwidth memory (HBM) prices is expected to support the overall average selling price of DRAM. Memory chip manufacturer SK Hynix will continue to benefit from the memory chip upcycle. The report indicates that HBM prices could rise by 80% to 100% by 2027. RBC believes that the DRAM upcycle, which has already lasted for 14 quarters, will continue for another five quarters, with supply tightness expected to persist until 2027.RBC expects SK Hynix's third-quarter performance to slightly exceed market consensus, with revenue projected to reach 101.2 trillion Korean won and adjusted earnings per American Depositary Share to be 8,917 won (approximately $6.72). The DRAM contract price is expected to increase by 20% quarter-on-quarter in the third quarter, with a potential further increase of 15% in the fourth quarter. RBC reiterated its "Outperform" rating with a target price of $200, about 40% higher than the latest closing price. In the U.S. stock market, SK Hynix's stock price surged over 4% on Monday, closing at $142.62.Nomura Securities recently stated that the shortage of memory chips will continue until 2028, maintaining "Buy" ratings for Samsung Electronics and SK Hynix, with target prices of 670,000 won and 4.7 million won, respectively. KB Securities in Korea stated on Monday that the price-to-earnings ratios of both companies have dropped to about 3 times. This Wednesday, SK Hynix will release its third-quarter financial report, and according to a FnGuide survey, analysts expect revenue of 98.2 trillion won and operating profit of 77 trillion won. According to media reports, the company is considering distributing a special dividend of about 16 trillion won.

Samsung Electronics and SK Hynix plan to start mass production and delivery of 8-layer HBM4 in the second quarter of next year, supplying it to NVIDIA.According to informed sources, NVIDIA will equip the 8-layer HBM4 based on the scale of the Rubin Ultra NVL rack, and it is expected that the proportion of 8-layer HBM4 will significantly increase starting from the middle of next year.

According to the "Science and Technology Innovation Board Daily," South Korean memory chip companies SK Hynix and Samsung Electronics will announce their performance for the third quarter of 2026 this week, with SK Hynix set to release on Wednesday and Samsung Electronics on Thursday. The report states that with the rising demand for AI chips and increasing prices for memory chips, both companies are expected to deliver strong results, with a combined operating profit for the third quarter projected to reach 184.4 trillion won, approximately 139.1 billion dollars.Samsung Electronics' preliminary results released on October 8 show that third-quarter revenue increased by 127% year-on-year, reaching 195 trillion won; operating profit was 107.4 trillion won, approximately 80.17 billion dollars, compared to 12.17 trillion won in the same period last year, an increase of nearly 8 times. Analysts surveyed by FnGuide expect SK Hynix's third-quarter revenue to be 98.2 trillion won, with operating profit at 77 trillion won. In 2025, Samsung and SK Hynix's operating profits are projected to be 43.6 trillion won and 47.2 trillion won, respectively, totaling 90.8 trillion won. If these predictions come true, Samsung will become the first South Korean company to exceed 100 trillion won in quarterly operating profit.As of the market close on October 23, the stock prices of Samsung and SK Hynix have fallen by 36% and 52%, respectively, since the end of June. The report indicates that investors view the impressive financial reports as expected, with focus areas including HBM4 supply, long-term agreements, and shareholder returns. TrendForce predicts that the average selling price of HBM will increase by 121% next year compared to this year, and the production capacity of Samsung and SK Hynix for next year is already sold out. In the second-quarter financial report, Samsung disclosed long-term agreements with five data center customers, while SK Hynix finalized multi-year contracts with over ten customers. Samsung announced a shareholder return plan of 90 trillion to 110 trillion won this year and will determine a cash dividend of about 30 trillion won for the third quarter by the end of this month. SK Hynix announced a 40 trillion won buyback plan in August and halted the evaluation for Solidigm's IPO. Kim Dong-won, head of KB Securities, stated that despite the significant stock price correction, the profit expectations for both companies have been raised, and the key catalyst for stock revaluation is expected to be the sustainability of profit growth and whether shareholder returns can be further increased.

According to the Science and Technology Innovation Board Daily, Nvidia CEO Jensen Huang stated during a media interview at the GTC conference in Berlin on Thursday that the likelihood of a memory supply surplus by 2029 is low. He mentioned that one great aspect of memory is that there is always more that can be used, and he noted that HBM will continue to be applied in the coming years.Jensen Huang also stated that Nvidia is the first company in the world to introduce LPDDR memory, used in mobile phones, into data centers, which doubles its energy efficiency compared to any current global solution, thus it will continue to use this type of memory extensively. He mentioned last Thursday that customer demand for AI chip orders continues to exceed the company's supply capacity, with new orders pouring in almost weekly, if not daily.In the past week, storage chip stocks faced sell-offs due to market concerns over supply surplus and weak demand. A well-known media outlet in South Korea published a column last week, citing two experts who predicted that by 2029, there will be a significant increase in semiconductor supply, and the storage chip market may turn into a red ocean. On Thursday, the Nasdaq Composite Index closed down 2.6%, at 26008.86 points; SanDisk fell 8.51%, SK Hynix fell 8.07%, Seagate Technology fell 7.86%, Western Digital fell 6.43%, and Micron Technology fell 6.23%.

According to a report from ZDNet Korea on October 20, the AI chip company Groq, which has been effectively acquired by NVIDIA, is advancing the introduction of DRAM 3D stacked packaging in its next-generation chips to replace the fast but lower-capacity SRAM solution. Groq was founded in 2016 by former Google TPU designer Jonathan Ross, and it was reported that the company was acquired by NVIDIA in a roundabout manner last year.The report states that Groq is evaluating a solution called 3DS, which vertically connects DRAM chips through silicon vias for high-capacity packaging. 3D packaging allows multiple bare chips to be stacked vertically and operate as a single chip, reducing area and improving power efficiency compared to planar solutions. Both 3DS and HBM use silicon vias, but HBM focuses on transmission speed, while 3DS maintains existing DRAM interfaces and focuses on capacity expansion. Anonymous industry insiders indicated that Groq plans to adopt 3D packaging in its next-generation chips, stacking DRAM instead of SRAM; another individual noted that market demand is shifting towards larger capacity chips, prompting Groq to choose DRAM, which has a capacity advantage.

According to a report by Reuters, Micron's Taoyuan union in Taiwan announced on October 20 that it plans to initiate a strike this Thursday to protest the dispute over employee bonuses and profit-sharing plans, and to support the Taichung union, which is currently in mediation negotiations with the company.Currently, the mediation between the Taoyuan union and Micron broke down in September, and the former received authorization from its members to strike in early October. The Taichung union represents Micron's largest operational department in Taiwan and is still negotiating with the company, having not yet voted on whether to strike. Micron has approximately 15,000 employees in Taiwan, primarily producing key components for AI servers, DRAM, and high-bandwidth memory (HBM). The Taoyuan union did not specify the duration of the strike, and Taiwan's regional leader Lai Ching-te called on both sides to negotiate sincerely last week.

Micron Technology Vice President Li Zhe stated at the Korea Semiconductor Exhibition SEDEX 2026 held at COEX in Seoul that the company is developing Edge Inference Memory (EIM) for physical AI. The industry expects this product to enter mass production in the second half of 2028. Micron is researching the use of EIM in LPDDR6 to meet the demand for low-power DRAM in applications such as autonomous driving and humanoid robots, and is in discussions with major customers regarding related products.According to a report by the Science and Technology Innovation Board Daily, EIM is positioned between AI chips and memory, utilizing a method of stacking LPDDR on logic chips to process inference data at high speed, although the computing architecture has not been fully disclosed. Li Zhe mentioned that current robotic AI models have about 1 billion to 2 billion parameters, with low bandwidth requirements; however, as more advanced visual language models emerge, LPDDR alone will not suffice, and HBM will eventually be used for humanoid robots. Micron anticipates that future high-performance humanoid robots will require approximately 200 GB of DRAM and 4 TB of storage, while current service robots require about 12 GB of DRAM and 256 GB of storage.Micron CEO Sanjay Mehrotra stated that the storage capacity of a single humanoid robot is about 10 times that of an ordinary L2+ passenger vehicle. The storage industry may enter a period of sustained demand growth for several decades around 2030, driven by the need to synchronize data from hundreds of sensors and run large models locally for millisecond-level control. Counterpoint Research predicts that global DRAM demand for humanoid robots will increase from approximately 8,300 Tbit this year to about 172,000 Tbit by 2030.

NVIDIA has requested Samsung Electronics, SK Hynix, and Micron Technology to use TSMC's 3nm process for the production of NVHBM, while the core bare chips will utilize sixth-generation DRAM technology. Samsung Electronics has decided to accept NVIDIA's technical requirements but will not comment on matters involving customers.

Samsung Electronics will adopt TSMC's 3-nanometer process to produce the next generation of custom HBM chip bare dies for NVIDIA. This is the first time Samsung has collaborated with TSMC in the HBM manufacturing stage.NVIDIA has requested Samsung, SK Hynix, and Micron Technology to use TSMC's process to manufacture NVHBM bare dies and to utilize sixth-generation DRAM technology. Samsung plans to continue independently producing standard HBM and custom products for other customers.

According to DealSite, storage chip manufacturer SK Hynix is expanding the mass production of the sixth-generation high bandwidth memory HBM4. However, some industry insiders believe that the expansion speed is slower than expected, and some equipment orders planned for this year may be postponed until next year.Industry insiders stated that some equipment originally scheduled for delivery in the second half of this year has shown signs of delay, and subsequent order plans have not yet been clarified. An anonymous source indicated that some equipment originally planned for this year may be pushed to next year, and there is no data confirming whether the delivered equipment is operating at full capacity. The investment plans related to HBM4 for next year have not been specifically communicated to some partners, and this year's orders have largely been completed, but there are no clear signals of increased or decreased investment. Reports suggest that the direct correlation between the equipment investment schedule and product quality issues has not yet been confirmed.Liu Henggen, a researcher at Daxin Securities, believes that concerns about major customers' HBM4 quality issues leading to performance below expectations are excessive. There may have been some issues in the backend process, but they cannot be regarded as insurmountable risks, and it is not yet time to discuss bad compensation. There are also industry claims that, against the backdrop of rising prices for general memory, some HBM investments may shift towards general memory. SK Hynix plans to prioritize the introduction of DRAM production equipment into the first cleanroom of the Yongin cluster Y1 wafer fab, and the industry expects that the initial production line will adopt the sixth-generation 10-nanometer 1c process, with the new capacity being prioritized for general DRAM. Some opinions also suggest that changes in the schedule of certain backend equipment are insufficient to determine that HBM4 mass production is hindered, as SK Hynix has already supplied HBM4 16-layer samples to major customers.

The account pequityresearch stated that Bloomberg's research institution, Bloomberg Intelligence, predicts that the high bandwidth memory market will reach $64.1 billion by 2030, up from the previous forecast of $26 billion.The forecast cited by this account also shows that Bloomberg Intelligence expects the prices of HBM3E and HBM4 to rise by 30% to 45% in 2027, with the average price of high bandwidth memory increasing from $15.46 in 2026 to $34.94 in 2030.

TrendForce reported on October 12, citing Chosun Biz, that as AI packaging accommodates more computing and storage chips, the intermediary layer connecting computing chips and HBM is expected to significantly increase in size, making next-generation HBM cooling more important. Samsung Electronics is preparing multiple technologies to address the cooling challenges of HBM4E, which is scheduled to begin mass production next year.Currently, the area of the AI intermediary layer is about 5.5 times the maximum exposure area of 858 square millimeters of an EUV lithography machine. Semiconductor equipment industry insiders expect that the intermediary layer carrying HBM could expand to 9 times, 12 times that area, and ultimately about 40 times. Reasons include limitations in process miniaturization and the popularity of chiplet-style 2.5D packaging, where multiple computing chiplets and HBM stacks are placed on the same intermediary layer, making it difficult for a single advanced process chip to meet AI computing power demands.Samsung is considering hybrid copper bonding and thermal path blocks. The former replaces the micro bumps between stacked DRAM chips with direct copper-to-copper connections, which can reduce inter-layer gaps, make the stack thinner, and improve heat dissipation, a technology Samsung showcased at NVIDIA GTC in March 2026. The thermal path block establishes a dedicated thermal path in the D2D PHY area responsible for high-speed transmission between HBM base chips and GPUs, which Samsung is validating with HBM4E and plans to implement in HBM5. According to ETNews, SK Hynix's iHBM places high thermal conductivity but electrically insulating cooling materials around the heat-concentrated D2D PHY, which is expected to reduce thermal resistance by about 30%.

According to South Korean media reports, Samsung Electronics has signed long-term agreements for storage supply for next year, covering approximately 75% to 80%, including high bandwidth memory. The reports indicate that Samsung Electronics has finalized most contracts with major tech companies such as NVIDIA, Google, and Microsoft.SK Hynix and Micron are also accelerating the locking of supply volumes and signing a series of large-scale long-term contracts with major customers. Samsung Electro-Mechanics is reportedly close to signing a long-term agreement worth 700 billion Korean won with Taiwanese power electronics company Delta Electronics for the supply of multilayer ceramic capacitors.

According to a report by Reuters, reprinted by Nikkei Asia, the union representing employees at Micron Technology's Taoyuan plant in Taiwan announced on Wednesday that it has obtained authorization for a strike, with the focus of the dispute being the company's bonus plan. The union stated that 1,994 members voted in favor of the strike, accounting for 99% of the voters, while details such as the timing of the strike are still under discussion. Micron has approximately 15,000 employees in Taiwan, with independent unions in Taoyuan and Taichung, and the two unions together represent over 80% of Micron's employees in Taiwan. Micron Taiwan is the company's largest manufacturing base.Taiwan is a key production site for Micron's DRAM and high-bandwidth memory, with related chips being important components for AI servers. Mediation with the Taoyuan union did not reach an agreement in September, and the Taichung union is still in negotiations. In September, Micron announced that it would distribute bonuses for the fiscal year 2026 to over 60,000 employees worldwide, stating that the company had experienced an "extraordinary year," with Taiwanese employees receiving a cash bonus of 1 million New Taiwan Dollars, approximately 31,400 US dollars. The Taoyuan union stated that the plan did not address the request for a permanent profit-sharing mechanism and criticized Micron for announcing the bonuses without first reaching an agreement with the union during the mediation period.

AMD Chairwoman and CEO Lisa Su arrived at Samsung Electronics' Seocho office building in Seoul at around 4:44 PM on October 7, meeting with Vice Chairman Choi Yong-hyun, head of the Device Solutions Division, approximately seven months after her last visit to South Korea. Jack Huynh, head of AMD's Computing and Graphics Group, and Vice President Joseph Macri accompanied her. Lisa Su stated that both sides would discuss a broader partnership and mentioned that Samsung has many partners; when asked if she would meet with Lee Jae-yong, she said it was inconvenient to disclose.Before visiting Samsung, Lisa Su attended AMD Player Day in Seoul, referring to Samsung and SK Hynix as core supply chain partners and stating that they are planning a long-term supply chain due to rising demand for artificial intelligence. She mentioned that the next-generation MI455 and Helios systems have begun shipping and both use HBM4, supported by two companies, and that the HBM collaboration will not stop at one generation. Reports indicate that the Instinct MI455X features 432GB of HBM4, an increase of 50% compared to the previous generation MI355X's 288GB of HBM3E. She noted that memory supply is very tight and that they are optimizing usage with customers and asking partners to expand production. TrendForce predicted on the 29th of last month that the limited advanced process capacity shared between HBM and regular DRAM may continue to lead to tight supply into next year.Reports state that both parties signed a memorandum of understanding in March this year centered on the supply of HBM4 and next-generation DDR5, designating Samsung as the main supplier of HBM4 for the MI455X and discussing Samsung's contract manufacturing of AMD's next-generation products. Choi Yong-hyun stated at that time that Samsung can provide memory, foundry, and packaging simultaneously.

According to the Korea Economic Daily, Samsung Electronics' next-generation high-bandwidth memory "12-layer HBM4E" product has passed quality verification by Nvidia and major hyperscale data center operators.