
HBM price increase again, storage in urgent situation!

HBM memory prices are rising, exacerbating the storage shortage issue. Maingear CEO Wallace Santos stated that the memory shortage problem will last for several years. South Korean media reported that HBM3E prices have increased by about 20% due to a surge in orders from companies like NVIDIA, and memory companies are focusing on expanding HBM4 production capacity. Analysts say that the price increase of HBM3E is rare and expect prices to drop after the HBM4 market opens next year, but demand for HBM3E will continue to grow
Today, Maingear CEO Wallace Santos stated during a conference call with the media: "Regarding the current memory shortage situation, I was initially told that this was just a two to five month issue, and that production lines are gradually recovering, everything will be fine."
"As the owner of an original equipment manufacturer, I am prepared, and we are doing everything we can without inventory. However, based on the information we currently have, the memory shortage issue will last for several years," Santos continued.
Especially after Korean media reported that HBM prices would rise, the storage situation remains critical.
HBM3E Price, Up About 20%
According to Korean media reports, as NVIDIA is expected to officially begin exporting the fifth-generation high-bandwidth memory (HBM) HBM3E-equipped H200 artificial intelligence (AI) chips to China, memory semiconductor companies such as Samsung Electronics and SK Hynix have raised the supply prices for HBM3E for next year. Many companies developing their own AI accelerators, including NVIDIA, Google, and Amazon, have placed orders, leading to a surge in order volume.
"This price increase is also influenced by memory semiconductor companies focusing on expanding the production capacity of the sixth-generation HBM (HBM4), with expectations of significant growth in demand for HBM4 next year, resulting in limited capacity for these companies to meet the demand for HBM3E," the Korean media added.
Industry insiders revealed on the 24th that several memory semiconductor companies, including Samsung Electronics and SK Hynix, have raised the supply price of HBM3E by nearly 20%. Analysts noted that it is rare for HBM3E prices to increase, as they typically decline before the release of the next-generation products. It is widely believed that companies like Google and Amazon have increased their order volume for HBM3E next year, and NVIDIA, previously the largest customer of HBM3E, has also participated in this price increase. Both Google and Amazon design their own AI accelerators.
An industry insider stated: "This year, HBM3E has established its dominant position in the HBM market, but prices are expected to decline once the HBM4 market opens next year." He also pointed out: "However, as global large technology companies, aside from NVIDIA, release AI accelerators equipped with HBM3E next year, the demand for HBM3E continues to grow. From the perspective of memory semiconductor companies, they cannot afford to give up expanding HBM4 production capacity, so supply cannot meet demand, leading to a price premium of about 20%."
With NVIDIA's H200 AI chip approved for export to China, the demand for HBM3E seems to exceed expectations. Each H200 chip from NVIDIA is equipped with six HBM3E memory units. Reuters reported on the 22nd (local time) that "NVIDIA plans to use existing inventory to fulfill the first batch of orders, with an expected total shipment of 5,000 to 10,000 chip modules (approximately 40,000 to 80,000 H200 chips)," adding that "NVIDIA has informed Chinese customers of its plans to expand chip production capacity, with new orders expected to be accepted starting in the second quarter of next year." Global tech giants like Google and Amazon also need HBM3E. This is because Google's TPU (Tensor Processing Unit) and Amazon's Trainium are equipped with HBM3E and will start shipping next year. The HBM capacity of these two products will increase by nearly 20% to 30% compared to the previous generation. It is reported that Google's seventh-generation TPU can accommodate 8 HBM3E units per device, while Amazon's Trainium3 can accommodate 4 HBM3E units.
As Samsung Electronics and SK Hynix are expected to focus on producing the next generation HBM4, the demand for HBM3E has exceeded supply. KB Securities analyst Kim Dong-won stated, "It is expected that HBM4 will account for 55% of HBM market revenue next year, while HBM3E will account for 45%, and starting from the third quarter of next year, HBM4 is expected to quickly absorb the demand for HBM3E."
With the surge in HBM demand and the continuous rise in prices of mainstream products like DRAM, market expectations for the profits of Samsung Electronics and SK Hynix next year have also been raised. According to financial data company FnGuide, Samsung Electronics has raised its full-year operating profit forecast for next year from 76.6544 trillion won a month ago to 85.4387 trillion won, while SK Hynix's operating profit forecast for the same period has also been raised from 71.4037 trillion won to 76.1434 trillion won.
The high demand for HBM has made industry players the big winners.
The Three Storage Giants Are Raking It In
This wave of HBM frenzy has allowed Micron to stage a comeback. The high demand for memory has driven Micron Technology's stock price up, with expectations of a 168% increase by 2025.
Micron is one of three companies providing high-bandwidth memory for artificial intelligence applications. For example, AMD extensively uses Micron's memory in its latest AI chips.
Micron Technology reported that for the quarter ending November 27, its revenue was $13.64 billion, a year-on-year increase of 56.6%, surpassing the previous quarter's growth rate of 46%, mainly due to the booming development of artificial intelligence driving market demand. According to Generally Accepted Accounting Principles (GAAP), the company's profit was $5.24 billion, a year-on-year increase of 231%. The profit margin rose from 28.3% in the previous quarter to 38.4% in the current quarter.
President and CEO Sanjay Mehrotra stated, "In the first fiscal quarter, Micron achieved record revenue and significant profit margin growth at both the company level and across all business segments."
To this end, Micron plans to maximize the capacity of its fabs by increasing the output of high-density process nodes and investing in new cleanroom space to enhance capacity. Specifically, additional cleanroom space is needed to meet the growing demand for HBM, but the construction cycles for cleanrooms in different regions are all being extended "We are adjusting the construction schedule for our first wafer fab in Idaho, which is now expected to achieve its first wafer output by mid-2027, earlier than our previous expectation of the second half of 2027. Earlier this year, we announced plans to build a second wafer fab in Idaho, which will begin construction in 2026 and be operational by the end of 2028.
"We plan to break ground on our first wafer fab in New York in early 2026, with the expectation that this fab will provide supply in 2030 and beyond... In Singapore, our HBM advanced packaging facility is progressing as planned and is expected to make significant contributions to our HBM supply in 2027."
The memory capacity for 2026 is already sold out, including HBM4. Micron predicts that the HBM market size will grow at a compound annual growth rate of about 40%, increasing from approximately $35 billion in 2025 to about $100 billion in 2028. This forecast for the HBM market size in 2028 will exceed the size of the entire DRAM market in 2024.
Micron believes it is in the best competitive position in history and is one of the largest drivers of artificial intelligence in the semiconductor industry.
In addition to Micron, the performance of the Korean semiconductor giants is also particularly impressive.
In the past three months, the earnings outlook for major domestic semiconductor companies, including Samsung Electronics and SK Hynix, has significantly improved. With expectations rising, investors are increasingly focused on whether the recent upward trend in stock prices can be sustained.
According to data released by financial information company FnGuide on the 24th, as of the 22nd, the annual operating profit forecast for Samsung Electronics this year (market average forecast) is 39.1491 trillion won, an increase of 30% from the forecast three months ago (30.1454 trillion won). During the same period, SK Hynix's annual operating profit forecast has also been raised by 12%, from 38.2883 trillion won to 42.7712 trillion won.
The upward revision of next year's performance expectations is even greater. The annual operating profit forecast for Samsung Electronics next year has been generally raised to 85.4387 trillion won, soaring 94% from 44.1092 trillion won three months ago. The market even expects its operating profit to exceed 100 trillion won. Among various securities firms, Hanwha Securities provided the highest forecast, predicting that Samsung Electronics' operating profit next year will reach 112.8194 trillion won.
SK Hynix's operating profit forecast for next year has also been raised by 66% within three months, from 45.9060 trillion won to 76.1434 trillion won. iM Securities predicts that SK Hynix's operating profit next year will reach 93.8430 trillion won, the highest forecast among domestic securities firms.
Securities firms point out that the increase in DRAM prices and the growing demand for artificial intelligence driving the increase in HBM (high bandwidth memory) shipments are key factors for performance improvement next year. Hanwha Securities researcher Kim Rok-ho stated, "The recent trend of rising memory prices is likely to continue, and Micron's expectations for memory prices in the first quarter of next year are also likely to be raised." "He also added: 'There is still room for upward adjustment in the performance expectations of Samsung Electronics and SK Hynix.'"
Future Asset Securities researcher Kim Yong-kun assessed: 'It is expected that global capital expenditure by large technology companies will reach $460 billion this year and expand to $600 billion next year.' He added: 'The server upgrade cycle introduced during the first investment cycle from 2016 to 2022 is likely to truly begin soon, which is also a positive factor.'
Because of this, a new concern has emerged globally, which is that everyone is focusing on HBM, posing challenges to PCs and mobile devices. For example, earlier this month, Micron announced that it would stop selling memory and other components directly to consumers to ensure the supply of artificial intelligence chips and data centers.
Other Storage Buyers Face Increasing Difficulties
According to consulting firm IDC, the memory market is at an unprecedented turning point, with demand far exceeding supply. In their view, this situation is different for an industry that has long been characterized by cycles of boom and bust.
The rapid expansion of artificial intelligence infrastructure and workloads has put immense pressure on the memory ecosystem. These AI workloads require a large amount of memory, and part of the reason for the memory shortage is that manufacturers have shifted capacity from consumer electronics to profitable memory solutions to support the development of artificial intelligence.
"The major memory manufacturers have not expanded the capacity of traditional DRAM and NAND used in smartphones, personal computers, and other consumer electronics, but have shifted production to memory used in AI data centers, such as high bandwidth memory (HBM) and high-capacity DDR5. This has limited the supply of general memory modules and driven up the prices of all memory modules," IDC further pointed out.
IDC stated that the single system memory required for AI servers and enterprise environments is much higher than that for consumer-grade devices, so the deployment of AI is consuming a disproportionate amount of global memory resources, leading to a memory shortage as suppliers prioritize fulfilling orders from hyperscale data centers and original equipment manufacturers (OEMs) building AI servers. This dynamic has resulted in a reduction of DRAM available for consumer-grade devices, exacerbating the already tight market price pressure.
However, this is not merely a cyclical shortage caused by supply and demand mismatch, but rather a potential permanent strategic reallocation of global silicon wafer capacity.
IDC pointed out that for decades, the production of DRAM and NAND flash for smartphones and personal computers has been the main driver of silicon wafer capacity. Today, this pattern has reversed. The enormous demand for HBM from hyperscale data centers such as Microsoft, Google, Meta, and Amazon has forced the three major memory manufacturers (Samsung Electronics, SK Hynix, and Micron Technology) to redirect limited cleanroom space and capital expenditures toward more profitable enterprise-grade components.
"This is a zero-sum game: every wafer allocated to HBM stacks for NVIDIA GPUs means one less wafer for mid-range smartphone LPDDR5X memory modules or consumer-grade laptop solid-state drives (SSDs)." "IDC emphasizes. Therefore, IDC expects that the supply growth of DRAM and NAND will be below historical averages in 2026, with year-on-year increases of 16% and 17%, respectively.
In IDC's view, the result of this supply-demand imbalance is twofold: in recent months, the prices of DRAM and NAND/SSD have surged significantly, while the supply of these components is limited, forcing device manufacturers to cope with a rapidly changing situation.
They believe that in 2026, the global smartphone market, especially Android smartphone manufacturers, will face threats. The trend of making flagship features available in affordable smartphones, which has popularized configurations in the smartphone industry over the past decade, is now reversing.
The cost structure of smartphones largely depends on the memory used. For mid-range models, memory costs can account for 15% to 20% of the total bill of materials, while for high-end flagship models, this proportion is about 10% to 15%. As memory prices continue to soar, OEM manufacturers may have to significantly raise prices, reduce configurations, or do both.
The impact of shortages is highly asymmetric, creating winners and losers in terms of supply chain resilience and vertical integration.
Manufacturers targeting the low-end market may suffer severe blows. Companies like TCL, Transsion, Realme, Xiaomi, Lenovo, OPPO, vivo, Honor, and Huawei have business models built on thin margins. Rising costs will significantly squeeze their profit margins, leaving them with no choice but to pass on costs (or part of the costs) to end users.
In the high-end market, although Apple and Samsung face pressure, their structural advantages allow them to lock in memory supply 12-24 months in advance. On the other hand, new flagship models in 2026 may not upgrade memory, with Pro models still maintaining a configuration of 12GB instead of upgrading to 16GB. Additionally, existing models are unlikely to see significant price drops after the release of new models.
The Logic of Storage Has Changed
Bloomberg pointed out in an article that typically, the most concerning period for the memory chip industry is now. Rapid growth, huge profits, and predictions of a bright future often herald disaster.
Two types of chips—NAND flash memory for storing information and DRAM chips that provide a computing environment for complex processors from companies like NVIDIA and Intel—are crucial in the computing field, encompassing everything from the largest data centers to the smartphones we rely on. They have even penetrated into automobiles and consumer electronics.
However, despite the indispensability of these chips, their manufacturers have long faced cycles of oversupply or undersupply. A company's chips can directly replace another company's chips, making them a commodity. Coupled with the rapid fluctuations in short-term demand and the significant time and cost required to build factories, this creates an oversupply situation that drives prices below production costs Due to the belief among other manufacturers that the memory chip market is overly competitive and difficult to survive in, there are currently only four major memory chip manufacturers. It is well known that even Intel, the creator of the modern memory industry, and the leading chip production center in Taiwan have given up.
Nevertheless, if your understanding of the memory industry is limited to recent years, it has now become a hot field. SK Hynix and Micron Technology have both achieved record highs in revenue and market capitalization, and Wall Street predicts that their growth momentum will further accelerate from this point. Kioxia Holdings, focused on NAND flash memory, has seen its stock price rise more than sixfold since its listing in December 2024. Even Samsung Electronics, which has experienced technical setbacks, achieved a rapid rebound in 2025.
So, what exactly has changed all of this?
Chip manufacturers believe that the surge in spending in the field of artificial intelligence computing has brought about a compounding effect. AI software requires massive amounts of data, which has created unprecedented demand. AI needs to access this data quickly, necessitating chips equipped with high-speed interfaces, which are expensive. The difficulty in manufacturing these chips occupies a large amount of production line space, thereby limiting the memory capacity available for smartphones and laptops.
Despite Micron Technology's stock price rising more than twice this year, about 88% of analysts still recommend buying its stock. SK Hynix's stock is even more favored, mainly due to analysts' recommendations. Both companies are confident about their recent prospects, and external forecasts corroborate this optimism. Today, their total net profits have exceeded their revenue levels from not long ago. From a financial perspective, their strength has reached unprecedented heights.
So, what iceberg lies behind this bright outlook? The shadow of an artificial intelligence bubble bursting looms over the entire tech industry, not just the memory industry. More specifically, while the unique fusion of technological advancements has created a favorable environment, there is also a potential trend that has exacerbated the fate of SK Hynix and Micron Technology.
Since the 1990s, Samsung, which has dominated the industry, has been slow to respond to the demands of AI device manufacturers, particularly NVIDIA. Technical issues mean that its usual strategies are no longer effective. In previous cycles, this Korean company would quickly ramp up production using its strong financial resources. However, this time, there has been no transition from shortage to abundance in supply, which means prices remain high.
Considering that building a factory from scratch typically takes more than a year to become operational, unless AI demand collapses, it is unlikely that there will be an oversupply of memory chips by 2026. However, anyone hoping for continued growth and profitability from SK Hynix and Micron Technology should closely monitor Samsung's technological progress.
Bloomberg analysts believe that this situation is different from the past, and I should not rely on history as a guide, but it is worth noting that in 2023, SK Hynix and Micron together incurred a net loss of over $10 billion. In this industry, the situation is ever-changing Risk Warning and Disclaimer
The market carries risks, and investment should be approached with caution. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Investment based on this is at one's own risk
